Content uploaded by James J. Corbett
Author content
All content in this area was uploaded by James J. Corbett on Aug 25, 2015
Content may be subject to copyright.
Third IMO Greenhouse Gas Study 2014
Safe, secure and efficient
shipping on clean oceans
Third IMO
Greenhouse Gas
Study 2014
www.imo.org
IMO is the specialized agency of the United Nations with responsibility for ensuring that
lives at sea are not put at risk and that the environment is not polluted by international
shipping. The Convention establishing IMO was adopted in 1948 and IMO first met in 1959.
IMO’s 170 member States use IMO to develop and maintain a comprehensive regulatory
framework for shipping. IMO has adopted more than 50 binding treaty instruments,
covering safety, environmental concerns, legal matters, technical co-operation, maritime
security and the efficiency of shipping. IMO’s main Conventions are applicable to almost
100% of all merchant ships engaged in international trade.
The sixty-seventh session of IMO’s Marine Environment Protection Committee (MEPC)
approved the Third IMO GHG Study 2014, providing updated estimates for GHG emissions
from ships. According to current estimates presented in this study, international shipping
emitted 796 million tonnes of CO
2
in 2012, which accounts for no more than about 2.2% of
the total emission volume for that year. By contrast, in 2007, before the global economic
downturn, international shipping is estimated to have emitted 885 million tonnes of CO
2
which represented 2.8% of the global emissions of CO
2
for that year. These percentages
are all the more significant when considering that shipping is the principal carrier of world
trade, carrying as much as 90% by volume, and therefore providing a vital service to global
economic development and prosperity.
These updated emissions estimates are considered necessary, in general, to provide a
better foundation for future work by IMO to address GHG emissions from international
shipping especially as the Business as Usual scenarios, depending on future economic
and energy developments, forecast a growth in CO
2
emissions for international maritime
transport of between 50% to 250% in the period up to 2050. Sea transport is fuel-efficient
and without these updated figures it would be difficult to provide a meaningful baseline to
illustrate the steadily on-going improvement in fuel efficiency due to improved hull design,
more effective diesel engines and propulsion systems and more effective utilization of
individual ships resulting from the introduction of mandatory technical and operational
measures for ships from 1 January 2013.
For more information, please contact:
International Maritime Organization
4 Albert Embankment, London SE1 7SR, United Kingdom
Tel: +44 (0)20 7735 7611 • Fax: +44 (0)20 7587 3210 • Email: info@imo.org
24696 GHG 2014.indd 1 4/9/15 5:21 PM
Third IMO GHG Study 2014
Executive Summary and
Final Report
Published in 2015 by the
INTERNATIONAL MARITIME ORGANIZATION
4 Albert Embankment, London SE1 7SR
www.imo.org
Printed by Micropress Printers, Suffolk, UK
Copyright © International Maritime Organization 2015
All rights reserved.
No part of this publication may be reproduced,
stored in a retrieval system, or transmitted in any form or by any means,
without prior permission in writing from the
International Maritime Organization.
The views and conclusions drawn in this report are those of the authors of the report.
This publication has been prepared from official documents of IMO, and every effort
has been made to eliminate errors and reproduce the original text(s) faithfully.
Foreword by the Secretary-General,
Mr Koji Sekimizu
In recognition of the magnitude of the climate change challenge and the
importance of global action to address it, we, at IMO, for some time now, have
been energetically pursuing the development and implementation of measures to
address greenhouse gas (GHG) emissions from international shipping.
According to current estimates presented in this Third IMO GHG Study 2014,
international shipping emitted 796 million tonnes of CO
2
in 2012, which accounts
for no more than about 2.2% of the total emission volume for that year. By contrast,
in 2007, before the global economic downturn, international shipping is estimated to have emitted 885 million
tonnes of CO
2
, which represented 2.8% of the global emissions of CO
2
for that year. These percentages are
all the more significant when considering that shipping is the principal carrier of world trade, carrying as much
as 90% by volume and therefore providing a vital service to global economic development and prosperity.
In 2011, IMO adopted a suite of technical and operational measures which together provide an energy-
efficiency framework for ships. These mandatory measures entered into force as a ‘package’ on 1 January
2013, under Annex VI of the International Convention for the Prevention of Pollution from Ships (the MARPOL
Convention). These measures address ship types responsible for approximately 85% of CO
2
emissions from
international shipping and, together, they represent the first-ever, mandatory global regime for CO
2
emission
reduction in an entire industry sector.
Without reference to the findings of this Third IMO GHG Study 2014, it would be extremely difficult for
IMO to demonstrate the steady and ongoing improvement in ships’ energy efficiencies resulting from the
global introduction of the mandatory technical and operational measures. Furthermore, the study findings
demonstrate that IMO is best placed, as the competent global regulatory body, to continue to develop both
an authoritative and robust greenhouse gas emissions control regime that is relevant for international shipping
while also matching overall expectations for climate change abatement.
That said, the mid-range forecasted scenarios presented in this Third IMO GHG Study 2014 show that, by
2050, CO
2
emissions from international shipping could grow by between 50% and 250%, depending on
future economic growth and energy developments. Therefore, if we are to succeed in further enhancing the
sector’s energy efficiency, which is already the most energy-efficient mode of mass transport of cargo, the
international community must deliver realistic and pragmatic solutions, both from a technical standpoint and
a political perspective. I believe that 2015 will be a crucial year for progress on difficult and complex matters
in the world’s climate change negotiations, culminating in the international conference to be convened in
Paris in December 2015, which should identify the way forward for all sectors. IMO will bring the findings
of the Study to the attention of Parties to the United Nations Framework Convention on Climate Change
(UNFCCC) and I am confident that, in the light of the progress made by the Organization, both in gathering
relevant information and in supporting implementation of the package of mandatory technical and operational
measures, we have a positive message to convey to the global community.
The Study constitutes, without any doubt, a significant scientific work. It was undertaken on a global scale by
a consortium of world-renowned scientific experts under the auspices of IMO, and I would like to congratulate
all the experts involved for the comprehensive and rigorous research work they carried out.
On behalf of the Organization, I also applaud and extend my wholehearted thanks to the Steering Committee
of twenty IMO Member Governments for their dedication and support in overseeing this important Study
for the Organization, that is, Belgium, Brazil, Canada, Chile, China, Finland, India, Islamic Republic of Iran,
Japan, Malaysia, the Marshall Islands, the Netherlands, Nigeria, Norway, the Republic of Korea, the Russian
Federation, South Africa, Uganda, the United Kingdom and the United States. I would also like to express
profound appreciation to the Governments of Australia, Denmark, Finland, Germany, Japan, the Netherlands,
Norway, Sweden and the United Kingdom and to the European Commission for their financial contributions,
without which the Study would not have been possible.
I trust that the Third IMO GHG Study 2014 will become the paramount reference for the Organization’s
Marine Environment Protection Committee as it continues its consideration of further appropriate measures
as part of a robust regime to regulate international shipping emissions at the global level.
Contents
Preface ..................................................................... xi
List of abbreviations and acronyms................................................ xiii
Key definitions ............................................................... xv
List of figures ................................................................ xvii
List of tables ................................................................. xxv
Executive Summary ......................................................... 1
Key findings from the Third IMO GHG Study 2014 ................................... 1
Aim and objective of the study ................................................... 4
Structure of the study and scope of work . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
Summary of Section 1: Inventories of CO
2
emissions from international shipping 2007–2012 ... 6
2012 fuel consumption and CO
2
emissions by ship type............................. 6
2007–2012 fuel consumption by bottom-up and top-down methods:
Third IMO GHG Study 2014 and Second IMO GHG Study 2009 ...................... 8
2007–2012 trends in CO
2
emissions and drivers of emissions ......................... 11
Summary of Section 2: Inventories of emissions of GHGs and other
relevant substances from international shipping 2007–2012 ............................ 15
Summary of Section 3: Scenarios for shipping emissions 2012–2050 ...................... 18
Maritime transport demand projections ......................................... 18
Maritime emissions projections ............................................... 20
Summary of the data and methods used (Sections 1, 2 and 3) ........................... 22
Key assumptions and method details ........................................... 22
Inventory estimation methods overview (Sections 1 and 2) ........................... 23
Scenario estimation method overview (Section 3).................................. 26
1 Inventories of CO
2
emissions from international shipping 2007–2012......... 27
1.1 Top-down CO
2
inventory calculation method ................................... 27
1.1.1 Introduction ....................................................... 27
1.1.2 Methods for review of IEA data ......................................... 27
1.1.3 Top-down fuel consumption results ...................................... 28
1.2 Bottom-up CO
2
inventory calculation method . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 32
1.2.1 Overall bottom-up approach ........................................... 32
1.2.2 Summary of data and method input revisions............................... 33
1.2.3 Aggregation of ship types and sizes ...................................... 34
1.2.4 Estimating activity using AIS data........................................ 35
1.2.5 Ship technical data .................................................. 36
1.2.6 Sources and alignment/coverage of data sources ............................ 37
1.2.7 Bottom-up fuel and emissions estimation .................................. 38
1.2.8 Classification of international and domestic fuel............................. 38
1.3 Inventories of CO
2
emissions calculated using both the top-down
and bottom-up methods .................................................... 41
1.3.1 CO
2
emissions and fuel consumption by ship type ........................... 41
1.3.2 CO
2
and fuel consumption for multiple years 2007–2012 ...................... 46
Page
vi Third IMO GHG Study 2014
1.3.3 Trends in emissions and drivers of emissions 2007–2012 ...................... 48
1.3.4 Variability between ships of a similar type and size
and the impact of slow steaming ........................................ 53
1.3.5 Shipping’s CO
2
e emissions............................................. 56
1.3.6 Shipping as a share of global emissions ................................... 57
1.4 Quality assurance and control of top-down and bottom-up inventories................ 60
1.4.1 Top-down QA/QC .................................................. 60
1.4.2 Top-down QA/QC efforts specific to this study ............................. 62
1.4.3 Bottom-up QA/QC .................................................. 63
1.4.4 Comparison of top-down and bottom-up inventories ......................... 82
1.5 Analysis of the uncertainty of the top-down and bottom-up CO
2
inventories ........... 85
1.5.1 Top-down inventory uncertainty analysis .................................. 85
1.5.2 Bottom-up inventory uncertainty analysis ................................. 87
1.6 Comparison of the CO
2
inventories in this study
to the Second IMO GHG Study 2009 inventories................................. 89
2 Inventories of emissions of GHGs and other relevant substances
from international shipping 2007–2012 .................................... 95
2.1 Top-down other relevant substances inventory calculation method ................... 95
2.1.1 Method for combustion emissions ....................................... 95
2.1.2 Methane slip ....................................................... 96
2.1.3 Method for estimation for non-combustion emissions ........................ 96
2.2 Bottom-up other relevant substances emissions calculation method .................. 102
2.2.1 Method ........................................................... 102
2.2.2 Main engine(s) ...................................................... 102
2.2.3 Auxiliary engines .................................................... 102
2.2.4 Boilers............................................................ 103
2.2.5 Operating modes ................................................... 103
2.2.6 Non-combustion emissions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 103
2.2.7 Combustion emissions factors .......................................... 104
2.3 Other relevant substances emissions inventories for 2007–2012 ..................... 112
2.3.1 Top-down fuel inventories ............................................. 113
2.3.2 Top-down GHG inventories ........................................... 113
2.3.3 Top-down air pollutant inventories ...................................... 114
2.3.4 Bottom-up fuel inventories ............................................ 116
2.3.5 Bottom-up GHG inventories ........................................... 116
2.3.6 Bottom-up air pollutant inventories ...................................... 117
2.4 Quality assurance and quality control of other relevant substances
emissions inventories ...................................................... 118
2.4.1 QA/QC of bottom-up emissions factors ................................... 118
2.4.2 QA/QC of top-down emissions factors.................................... 120
2.4.3 Comparison of top-down and bottom-up inventories ......................... 120
2.5 Other relevant substances emissions inventory uncertainty analysis .................. 123
2.6 Other relevant substances emissions inventory comparison
against Second IMO GHG Study 2009 ......................................... 123
3 Scenarios for shipping emissions 2012–2050 ............................... 127
3.1 Introduction ............................................................. 127
3.1.1 Similarities with and differences from Second IMO GHG Study 2009 ............ 127
Page
vii
3.1.2 Outline ........................................................... 128
3.2 Methods and data ......................................................... 128
3.2.1 The emissions projection model ........................................ 128
3.2.2 Base scenarios...................................................... 129
3.2.3 Transport demand projections .......................................... 131
3.2.4 Fleet productivity ................................................... 132
3.2.5 Ship size development................................................ 133
3.2.6 EEDI, SEEMP and autonomous improvements in efficiency .................... 134
3.2.7 Fuel mix: market- and regulation-driven changes . . . . . . . . . . . . . . . . . . . . . . . . . . . . 135
3.2.8 Emissions factors .................................................... 136
3.3 Results ................................................................. 137
3.3.1 Transport demand ................................................... 137
3.3.2 Projected CO
2
emissions .............................................. 139
3.3.3 Results for other substances............................................ 143
3.3.4 Sensitivity to productivity and speed assumptions ........................... 144
3.3.5 Uncertainty ........................................................ 145
3.4 Main results ............................................................. 145
Bibliography for Main Report and Annexes.................................... 147
Annex 1 Details for Section 1.2:
bottom-up method............................................... 153
IHSF technical data and method for populating missing data ............................ 153
IHSF operational data .......................................................... 154
Estimating ship activity over the course of a year using AIS data ......................... 159
Sources and spatial and temporal coverage ...................................... 159
Pre-processing AIS data ........................................................ 160
Multi-MMSI merging........................................................... 163
Extrapolating ship annual profile to generate complete annual operational profiles........... 165
Assumptions for auxiliary and boiler power demands.................................. 167
Assumptions for main and auxiliary fuel type ........................................ 173
Assumptions for hull fouling and deterioration over time ............................... 173
Assumptions for the impact of weather on fuel consumption ............................ 173
Activity and fleet data merger ................................................... 174
Bottom-up model calculation procedure ........................................... 175
Powering subroutine: Power_at_op ............................................ 176
Emissions subroutine: Emissions_at_op.......................................... 177
Aggregation by ship type and size ............................................. 178
Fleet estimate assembly......................................................... 178
Annex 2 Details for Section 1.3:
inventory results ................................................. 179
2011 Detailed Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 180
2010 Detailed Results .......................................................... 183
2009 Detailed Results .......................................................... 186
2008 Detailed Results .......................................................... 189
2007 Detailed Results .......................................................... 192
Page
viii Third IMO GHG Study 2014
Annex 3 Details for Section 1.4:
bottom-up QA/QC ............................................... 195
Activity estimate quality of spatial coverage ......................................... 195
Activity estimates temporal coverage QA/QC........................................ 200
Fleet technical data quality ...................................................... 205
Noon report data for activity and fuel consumption quality assurance .................... 206
Description of noon report data............................................... 206
Method of processing noon report data in preparation for comparison
against bottom-up model output .............................................. 207
Results of noon report and bottom-up output quality assurance of activity estimate
and fuel consumption (all years) ............................................... 208
Annex 4 Details for Section 1.5.1:
top-down uncertainty analysis .................................... 229
Organization of top-down uncertainty analysis ...................................... 229
Ongoing data quality efforts related to uncertainty in fuel sales.......................... 229
Review of EIA accuracy analyses (estimation of percentage error) ........................ 229
IEA sources of uncertainties that can be quantified for this work ......................... 231
Estimates of potential adjustment to top-down statistics................................ 231
Export-import misallocation .................................................. 231
Transfers category reporting .................................................. 233
Data accuracy ............................................................ 233
Results of top-down uncertainty analysis ........................................... 234
Uncertainty in top-down allocations of international and domestic shipping ................ 236
Annex 5 Details for Section 1.5.2:
bottom-up inventory uncertainty analysis.......................... 237
Sources of uncertainty in the Second IMO GHG Study 2009 ............................ 237
Overview of sources of uncertainty in current work................................... 237
Uncertainty in the emissions from a ship in one hour.................................. 239
Estimate of uncertainty of the input parameters ................................... 239
Uncertainty in the aggregation of hourly emissions into annual emissions ................ 242
Uncertainty in the aggregation of a fleet of ships’ emissions .......................... 244
Annex 6 Details for Section 2:
other GHG emissions and relevant substances ..................... 247
Emissions factors.............................................................. 247
Method for selecting/developing baseline and actual emissions factors .................. 247
Baseline emissions factors ................................................... 248
SFOCs ................................................................. 250
Fuel correction factors – NO
x
, SO
x
, PM, N
2
O .................................... 251
Annex 7 Details for Section 3 .............................................. 255
The emissions projection model .................................................. 255
Analysis of historical transport work data........................................... 256
Introduction.............................................................. 256
Methodology............................................................. 256
Results ................................................................. 257
Sensitivities .............................................................. 259
Page
ix
Fleet productivity projections .................................................... 259
Historical ship productivity .................................................. 260
Ship productivity projection.................................................. 263
Remarks/caveats .......................................................... 264
Ship size projections ........................................................... 264
Container ships ........................................................... 266
Oil tankers............................................................... 269
Dry bulk carriers .......................................................... 272
Liquefied gas carriers ....................................................... 275
Regulatory and autonomous efficiency improvements ................................. 279
EEDI and SEEMP .......................................................... 279
Fuel mix .................................................................... 283
Market and regulatory drivers................................................. 283
Fuel mix scenarios used in emissions projection model ............................. 285
Emissions factors.............................................................. 287
Emissions factors .......................................................... 287
CO
2
................................................................. 287
CH
4
................................................................. 288
N
2
O ................................................................. 288
HFC ................................................................. 288
PFC ................................................................. 289
SF
6
................................................................. 289
NO
x
................................................................. 289
NMVOC ................................................................ 290
CO ................................................................. 290
PM ................................................................. 291
SO
2
................................................................. 291
Detailed results............................................................... 291
Page
Preface
This study of greenhouse gas emissions from ships (hereafter the Third IMO GHG Study 2014) was commissioned
as an update of the International Maritime Organization’s (IMO) Second IMO GHG Study 2009. The updated
study has been prepared on behalf of IMO by an international consortium led by the University College
London (UCL) Energy Institute. The Third IMO GHG Study 2014 was carried out in partnership with the
organizations and individuals listed below.
Consortium members, organizations and key individuals
Organization Location Key individual(s)
UCL Energy Institute UK Dr. Tristan Smith
Eoin O’Keeffe
Lucy Aldous
Sophie Parker
Carlo Raucci
Michael Traut
(visiting researcher)
Energy & Environmental Research Associates (EERA) USA Dr. James J. Corbett
Dr. James J. Winebrake
Finnish Meteorological Institute (FMI) Finland Dr. Jukka-Pekka Jalkanen
Lasse Johansson
Starcrest USA Bruce Anderson
Archana Agrawal
Steve Ettinger
Civic Exchange Hong Kong, China Simon Ng
Ocean Policy Research Foundation (OPRF) Japan Shinichi Hanayama
CE Delft The Netherlands Dr. Jasper Faber
Dagmar Nelissen
Maarten ‘t Hoen
Tau Scientific UK Professor David Lee
exactEarth Canada Simon Chesworth
Emergent Ventures India Ahutosh Pandey
The consortium thanks the Steering Committee of the Third IMO GHG Study 2014 for its helpful review and
comments.
The consortium acknowledges and thanks the following organizations for their invaluable data contributions
to this study: exactEarth, IHS Maritime, Marine Traffic, Carbon Positive, Kystverket, Gerabulk, V.Ships and
Shell. In the course of its efforts, the consortium gratefully received input and comments from the International
Energy Agency (IEA), the International Association of Independent Tanker Owners (INTERTANKO), the
International Chamber of Shipping (ICS), the World Shipping Council (WSC), the Port of Los Angeles, the Port
of Long Beach, the Port Authority of New York & New Jersey, the Environmental Protection Department of
the HKSAR Government and the Marine Department of the HKSAR Government.
The views and conclusions expressed in this report are those of the authors.
xii Third IMO GHG Study 2014
The recommended citation for this work is: Third IMO GHG Study 2014; International Maritime Organization
(IMO) London, UK, April 2015; Smith, T. W. P.; Jalkanen, J. P.; Anderson, B. A.; Corbett, J. J.; Faber, J.; Hanayama,
S.; O’Keeffe, E.; Parker, S.; Johansson, L.; Aldous, L.; Raucci, C.; Traut, M.; Ettinger, S.; Nelissen, D.; Lee, D. S.;
Ng, S.; Agrawal, A.; Winebrake, J. J.; Hoen, M.; Chesworth, S.; Pandey, A.
Approval of the Third IMO GHG Study 2014
The Marine Environment Protection Committee, at its sixty-seventh session (October 2014), approved the
Third IMO GHG Study 2014.
Consortium members:
Data partners:
List of abbreviations
and acronyms
AIS Automatic Identification System
AR5 Fifth Assessment Report of IPCC
BAU business as usual
BSFC brake-specific fuel consumption
DG ENV Directorate-General for the Environment (European Commission)
DOE Department of Energy (US)
dwt deadweight tonnage
ECA emission control area
EEDI Energy Efficiency Design Index
EEZ Exclusive Economic Zone
EF emissions factor
EIA Energy Information Administration
EPA (US) Environmental Protection Agency
FCF fuel correction factors
FPSO floating production storage and offloading
GDP gross domestic product
GHG greenhouse gas
gt gross tonnage
GWP global warming potential (GWP100 represents the 100-year GWP)
HCFC hydrochlorofluorocarbon
HFC hydrofluorocarbon
HFO heavy fuel oil
HSD high-speed diesel (engine)
IAM integrated assessment models
IEA International Energy Agency
IFO intermediate fuel oil
IHSF IHS Fairplay
IMarEST Institute of Marine Engineering, Science and Technology
IMO International Maritime Organization
IPCC Intergovernmental Panel on Climate Change
LNG liquefied natural gas
LRIT long-range identification and tracking (of ships)
MACCs marginal abatement cost curves
MCR maximum continuous revolution
MDO marine diesel oil
xiv Third IMO GHG Study 2014
MEPC Marine Environment Protection Committee (IMO)
MGO marine gas oil
MMSI Maritime Mobile Service Identity
MSD medium-speed diesel (engine)
nmi nautical mile
NMVOC non-methane volatile organic compounds
PFC perfluorocarbon
PM particulate matter
QA quality assurance
QC quality control
RCP representative concentration pathways
S-AIS Satellite-based Automatic Identification System
SEEMP Ship Energy Efficiency Management Plan
SFOC specific fuel oil consumption
SSD slow-speed diesel (engine)
SSP shared socioeconomic pathway
UNEP United Nations Environment Programme
UNFCCC United Nations Framework Convention on Climate Change
VOC volatile organic compounds
Key definitions
International shipping: shipping between ports of different countries, as opposed to domestic shipping.
International shipping excludes military and fishing vessels. By this definition, the same ship may frequently
be engaged in both international and domestic shipping operations. This is consistent with the IPCC 2006
Guidelines (Second IMO GHG Study 2009).
International marine bunker fuel: “[…] fuel quantities delivered to ships of all flags that are engaged in
international navigation. The international navigation may take place at sea, on inland lakes and waterways,
and in coastal waters. Consumption by ships engaged in domestic navigation is excluded. The domestic/
international split is determined on the basis of port of departure and port of arrival, and not by the flag or
nationality of the ship. Consumption by fishing vessels and by military forces is also excluded and included in
residential, services and agriculture” (IEA website: http://www.iea.org/aboutus/glossary/i/).
Domestic shipping: shipping between ports of the same country, as opposed to international shipping.
Domestic shipping excludes military and fishing vessels. By this definition, the same ship may frequently be
engaged in both international and domestic shipping operations. This definition is consistent with the IPCC
2006 Guidelines (Second IMO GHG Study 2009).
Domestic navigation fuel: fuel delivered to vessels of all flags not engaged in international navigation (see the
definition for international marine bunker fuel above). The domestic/international split should be determined
on the basis of port of departure and port of arrival and not by the flag or nationality of the ship. Note that this
may include journeys of considerable length between two ports in the same country (e.g. San Francisco to
Honolulu). Fuel used for ocean, coastal and inland fishing and military consumption is excluded (http://www.
iea.org/media/training/presentations/statisticsmarch/StatisticsofNonOECDCountries.pdf).
Fishing fuel: fuel used for inland, coastal and deep-sea fishing. It covers fuel delivered to ships of all flags
that have refuelled in the country (including international fishing) as well as energy used in the fishing
industry (ISIC Division 03). Before 2007, fishing was included with agriculture/forestry and this may
continue to be the case for some countries (http://www.iea.org/media/training/presentations/statisticsmarch/
StatisticsofNonOECDCountries.pdf).
Tonne: a metric system unit of mass equal to 1,000 kilograms (2,204.6 pounds) or 1 megagram (1 Mg). To
avoid confusion with the smaller “short ton” and the slightly larger “long ton”, the tonne is also known as a
“metric ton”; in this report, the tonne is distinguished by its spelling.
Ton: a non-metric unit of mass considered to represent 907 kilograms (2,000 pounds), also sometimes called
“short ton”. In the United Kingdom the ton is defined as 1016 kilograms (2,240 pounds), also called “long
ton”. In this report, ton is used to imply “short ton” (907 kg) where the source cited used this term, and in
calculations based on these sources (e.g. Section 2.1.3 on refrigerants, halogenated hydrocarbons and other
non-combustion emissions).
Page
List of figures
Figure 1: Bottom-up CO
2
emissions from international shipping by ship type 2012 .......... 6
Figure 2: Summary graph of annual fuel consumption broken down by ship type
and machinery component (main, auxiliary and boiler) 2012 ................... 7
Figure 3: CO
2
emissions by ship type (international shipping only)
calculated using the bottom-up method for all years (2007–2012) ............... 8
Figure 4: Summary graph of annual fuel use by all ships, estimated using
the top-down and bottom-up methods, showing Second IMO GHG Study 2009
estimates and uncertainty ranges. ....................................... 9
Figure 5: Summary graph of annual fuel use by international shipping,
estimated using the top-down and bottom-up methods, showing
Second IMO GHG Study 2009 estimates and uncertainty ranges ............... 9
Figure 6: Time series for trends in emissions and drivers of emissions
in the oil tanker fleet 2007–2012. All trends are indexed to their values
in 2007 .......................................................... 11
Figure 7: Time series for trends in emissions and drivers of emissions
in the container ship fleet 2007–2012. All trends are indexed to their values
in 2007 .......................................................... 12
Figure 8: Time series for trends in emissions and drivers of emissions
in the bulk carrier fleet 2007–2012. All trends are indexed to their values
in 2007 .......................................................... 12
Figure 9: Time series of bottom-up results for GHGs and other substances (all shipping).
The green bar represents the Second IMO GHG Study 2009 estimate ............ 16
Figure 10: Time series of bottom-up results for GHGs and other substances
(international shipping, domestic navigation and fishing) ...................... 17
Figure 11: Historical data to 2012 on global transport work for non-coal combined
bulk dry cargoes and other dry cargoes (billion tonne-miles)
coupled to projections driven by GDPs from SSP1 through to SSP5
by 2050 .......................................................... 18
Figure 12: Historical data to 2012 on global transport work for ship-transported coal
and liquid fossil fuels (billion tonne-miles) coupled to projections of coal and
energy demand driven by RCPs 2.6, 4.5, 6.0 and 8.5 by 2050 .................. 19
Figure 13: BAU projections of CO
2
emissions from international maritime transport
2012–2050 ........................................................ 20
Figure 14: Projections of CO
2
emissions from international maritime transport.
Bold lines are BAU scenarios. Thin lines represent either
greater efficiency improvement than BAU
or additional emissions controls or both .................................. 20
Figure 15: Projections of CO
2
emissions from international maritime transport under the
same demand projections. Larger improvements in efficiency have
a higher impact on CO
2
emissions than a larger share
of LNG in the fuel mix ............................................... 21
Figure 16: Geographical coverage in 2007 (top) and 2012 (bottom), coloured according
to the intensity of messages received per unit area.
This is a composite of both vessel activity and geographical coverage;
intensity is not solely indicative of vessel activity ............................ 24
Page
xviii Third IMO GHG Study 2014
Figure 17: Total noon-reported quarterly fuel consumption of the main engine,
compared with the bottom-up estimate over each quarter of 2012,
with a filter to select only days with high reliability observations
of the ship for 75% of the time or more .................................. 25
Figure 18: Oil products and products from other sources used in shipping
(international, domestic and fishing) 1971–2011 ............................ 28
Figure 19: IEA fuel oil sales in shipping 2007–2011 .................................. 29
Figure 20: IEA gas/diesel sales in shipping 2007–2011 ................................ 30
Figure 21: IEA natural gas sales in shipping 2007–2011 ............................... 30
Figure 22: Correlation between world GDP and international bunker fuel oil
during the recession ................................................. 31
Figure 23: Data assembly and method for Sections 1.2 and 2.2 ......................... 32
Figure 24: Chart showing the coverage of one of the merged AIS data sets used in this study
(2012, all sources, but no LRIT) ......................................... 33
Figure 25: Chart showing the coverage of one of the LRIT data sets used in this study (2012) ... 34
Figure 26: Venn diagram describing the sets of ships observed in the two main data types
used in the bottom-up method (IHSF and AIS) ............................. 37
Figure 27: Bottom-up CO
2
emissions from international shipping by ship type (2012) ......... 41
Figure 28: Summary graph of annual fuel consumption (2012), broken down
by ship type and machinery component (main, auxiliary and boiler) ............. 42
Figure 29: CO
2
emissions by ship type (international shipping only),
calculated using the bottom-up method for all years 2007–2012 ................ 46
Figure 30: Summary graph of annual fuel use by all ships,
estimated using the top-down and bottom-up methods ....................... 47
Figure 31: Summary graph of annual fuel use by international shipping,
estimated using the top-down and bottom-up methods ....................... 48
Figure 32: Average trends in the tanker sector 2007–2012, indexed to 2007 ................ 50
Figure 33: Average trends in the bulk carrier sector 2007–2012, indexed to 2007 ............ 50
Figure 34: Average trends in the container ship sector 2007–2012, indexed to 2007 .......... 50
Figure 35: Fleet total trends in the oil tanker sector (2007–2012), indexed to 2007 ........... 51
Figure 36: Fleet total trends in the bulk carrier sector (2007–2012), indexed to 2007 ......... 51
Figure 37: Fleet total trends in the container ship sector (2007–2012), indexed to 2007 ....... 51
Figure 38: Variability within ship size categories in the bulk ship fleet (2012).
Size category 1 is the smallest bulk carrier (0–9,999 dwt) and size category 6
is the largest (200,000+ dwt) .......................................... 52
Figure 39: Variability within ship size categories in the container ship fleet (2012).
Size category 1 is the smallest container ship (0–999 TEU) and size category 8
is the largest (14,500+ TEU) ........................................... 52
Figure 40: Variability within ship size categories in the tanker fleet (2012).
Size category 1 is the smallest oil tanker (0–9,999 dwt) and size category 8
is the largest (200,000+ dwt) .......................................... 53
Figure 41: Time series of bottom-up CO
2
e emissions estimates for a) total shipping and
b) international shipping .............................................. 57
Figure 42: Comparison of shipping with global totals: a) CO
2
emissions compared, where the
percentage indicates international shipping emissions of CO
2
as a percentage of
global CO
2
from fossil fuels; b) CO
2
e emissions compared, where the
percentage indicates international shipping emissions of CO
2
e as a
percentage of global CO
2
e from fossil fuels ................................ 59
Figure 43: OECD versus non-OECD data collection system ............................ 61
Figure 44: Comparison of IEA and EIA international marine bunker fuel oil statistics ......... 62
Figure 45: Confidence bands showing statistical difference between IEA and EIA data, 2000–2010 63
Page
List of figures xix
Figure 46: Geographical coverage in 2007 (top) and 2012 (bottom), coloured according
to the intensity of messages received per unit area. This is a composite of both
vessel activity and geographical coverage; intensity is not solely indicative
of vessel activity .................................................... 65
Figure 47: The average volume of AIS activity reports for a region reported by a vessel
for up to 300 randomly selected VLCCs (2007–2012) ........................ 66
Figure 48: Activity estimate quality assurance (2012) ................................. 67
Figure 49: Comparison of at-sea and at-port days, calculated using both the bottom-up
model output (y-axis) and noon report data (x-axis) (2012) ..................... 68
Figure 50: Comparison of average ship speed and average ship draught calculated
using both the bottom-up model output (y-axis) and noon report data (x-axis) (2012) . 69
Figure 51: Comparison of at-sea days and average ship speed, calculated using both
the bottom-up model output (y-axis) and noon report data (x-axis) (2009) ......... 70
Figure 52: General boiler operation profile (Mys´ków & Borkowski, 2012) .................. 72
Figure 53: Operational profile of an auxiliary boiler of a container vessel
during six months of operations (Mys´ków & Borkowski, 2012) .................. 73
Figure 54: Average noon-reported daily fuel consumption of the main and auxiliary engines
compared with the bottom-up estimate over each quarter of 2012 .............. 75
Figure 55: Total noon-reported quarterly fuel consumption of the main and auxiliary engines
compared with the bottom-up estimate over each quarter of 2012 .............. 76
Figure 56: Total noon-reported quarterly fuel consumption of the main engine
compared with the bottom-up estimate over each quarter of 2012,
with a filter to select only days with high reliability observations of the ship
for 75% of the time or more ........................................... 77
Figure 57: Total noon-reported quarterly fuel consumption of the main engine
compared with the bottom-up estimate over each quarter of 2007 .............. 77
Figure 58: Total noon-reported quarterly fuel consumption of the main engine
compared with the bottom-up estimate over each quarter of 2009 .............. 78
Figure 59: Total percentage of in-service time for which high-reliability activity estimates
are available from AIS ................................................ 80
Figure 60: Emissions weighted average of the total percentage of in-service time
for which high-reliability activity estimates are available from AIS ............... 80
Figure 61: Top-down and bottom-up comparison for a) all marine fuels
and b) international shipping ........................................... 82
Figure 62: Comparison of top-down fuel allocation with initial and updated bottom-up
fuel allocation (2007–2012) ............................................ 83
Figure 63: Adjusted marine fuel sales based on quantitative uncertainty results (2007–2011) ... 87
Figure 64: Summary of uncertainty on top-down and bottom-up fuel inventories
for a) all ships and b) international shipping ................................ 88
Figure 65: Top-down and bottom-up inventories for all ship fuels,
from the Third IMO GHG Study 2014 and
the Second IMO GHG Study 2009 ...................................... 89
Figure 66: Top-down and bottom-up inventories for international shipping fuels,
from the Third IMO GHG Study 2014 and the Second IMO GHG Study 2009 ..... 90
Figure 67: Crossplots of deadweight tonnes, gross tonnes and average installed
main engine power for the year 2007, as reported by
the Second IMO GHG Study 2009 (x-axis)
and the Third IMO GHG Study 2014 (y-axis) ............................... 91
Page
xx Third IMO GHG Study 2014
Figure 68: Crossplots for days at sea, average engine load (% MCR) and auxiliary engine
fuel use for the year 2007, as reported by the Second IMO GHG
Study 2009 (x-axis) and the Third IMO GHG Study 2014 (y-axis) ................ 92
Figure 69: Crossplots for average main engine daily fuel consumption and total vessel
daily fuel consumption for 2007, as reported by the Second IMO GHG Study
2009 (x-axis) and the Third IMO GHG Study 2014 (y-axis) .................... 92
Figure 70: Crossplots for main engine annual fuel consumption, total vessel annual fuel
consumption, aggregated vessel type annual fuel consumption and CO
2
for the year 2007, as reported by the Second IMO GHG Study
2009 (x-axis) and the Third IMO GHG Study 2014 (y-axis) .................... 93
Figure 71: Estimated refrigerant emissions of the global fleet 2007–2012 .................. 101
Figure 72: Impact of engine control tuning (ECT) to specific fuel oil consumption
during low load operation of MAN 6S80ME-C8.2. Standard tuning is shown by
the solid black line, part load optimization by the solid blue line
and low load tuning by the broken line (from MAN, 2012) .................... 109
Figure 73: Impact of engine load on brake-specific fuel consumption of various
selected SSD, MSD and HSD engines (emissions factors by engine type) .......... 110
Figure 74: Comparison of PM emissions factors reported in Second IMO GHG Study 2009
[blue diamond](Figure 7.7, based on data from Germanischer Lloyd)
with values from Jalkanen et al. (2012) [red square]
and Starcrest (2013) [green triangle] ..................................... 112
Figure 75: Time series of top-down results for a) CO
2
, b) CH
4
, c) N
2
O, d) SO
x
, e) NO
x
,
f) PM, g) CO, and h) NMVOC, delineated by international shipping,
domestic navigation and fishing ........................................ 121
Figure 76: Time series of bottom-up results for a) CO
2
, b) CH
4
, c) N
2
O, d) SO
x
, e) NO
x
,
f) PM, g) CO, and h) NMVOC, delineated by international shipping,
domestic navigation and fishing ........................................ 122
Figure 77: Time series of bottom-up results for a) CO
2
, b) CH
4
, c) N
2
O, d) SO
x
, e) NO
x
,
f) PM, g) CO, and h) NMVOC. The green bar represents the Second IMO GHG
Study 2009 estimate for comparison ..................................... 125
Figure 78: Schematic presentation of the emissions projection model .................... 129
Figure 79: Historical data on world coal and oil consumption, coal and oil transported
(upper panel), total (non-coal) bulk dry goods, other dry cargoes
and global GDP (lower panel). . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 132
Figure 80: Historical data to 2012 on global GDP (constant 2005 US$ billion/yr)
coupled with projections of GDP from SSP1 through to SSP5 by 2050 ........... 137
Figure 81: Historical data to 2012 on global consumption of coals and oil (EJ/yr)
coupled with projections from RCP2.6 through to RCP8.5 by 2050 .............. 138
Figure 82: Historical data to 2012 on global transport work for non-coal combined
bulk dry cargoes and other dry cargoes (billion tonne-miles)
coupled with projections driven by GDPs from SSP1
through to SSP5 by 2050 ............................................. 138
Figure 83: Historical data to 2012 on global transport work for ship-transported coal
and liquid fossil fuels (billion tonne-miles) coupled with projections of coal
and energy demand driven by RCPs 2.6, 4.5, 6.0 and 8.5 by 2050 .............. 139
Figure 84: CO
2
emissions projections ............................................ 141
Figure 85: Emissions projections for the BAU transport demand scenarios ................. 142
Figure 86: Output for demand scenarios under conditions of high LNG/extra ECA
and high efficiency .................................................. 142
Figure 87: Specific output for scenario 15 (RCP4.5, SSP3, low LNG/no additional ECA,
low efficiency) ..................................................... 143
Figure 88: Impact of productivity assumptions on emissions projections .................. 145
Page
List of figures xxi
Annex Figures
Figure 1: Stand-alone pre-processor program with a graphic user interface.
The pre-processor has been programmed with Java ......................... 161
Figure 2: Sea region definition illustration. GIS shapefile has been read by the pre-processor.
The resolution of the sea region mapping is 0.1 × 0.1 degrees .................. 161
Figure 3: NA-ECA polygons drawn with Google Earth 2014. . . . . . . . . . . . . . . . . . . . . . . . . . . 162
Figure 4: Geographical distribution of AIS messages processed by the pre-processor
for 2007 and 2012. All available AIS data sets have been combined.
Unit: total number of messages per grid cell with an area of 0.2 × 0.2 degrees.
Both plots are the same scale .......................................... 163
Figure 5: Top plot shows the reported speed for an indicative ship between 15 January 2011
and 31 January 2011, each message with an opacity of 50% so that density
is apparent. The lower plot shows the same ship with the speeds resampled.
A reliability of 0 indicates that there was no activity report
for that resampled hour .............................................. 164
Figure 6: Example plot of coverage indicated by source of data ........................ 165
Figure 7: Characterization of ship phases used in the extrapolation algorithm
for an example very large crude carrier (VLCC) in 2012. The top plot shows
the phase labels for each data point at given speeds (y-axis)
and the lower plot classifies the data into high and low
standard deviation of speed within a six-hour window ....................... 166
Figure 8: Illustration of the extrapolation process ................................... 167
Figure 9: Activity and ship technical and operational data merger process ................ 175
Figure 10: Geographical distribution of AIS messages processed by the pre-processor
for 2009–2012. All available AIS data sets (both satellite and terrestrial)
have been combined. Unit: total number of messages per grid cell
with an area of 0.2 × 0.2 degrees ....................................... 196
Figure 11: Repeat plots for 2008 and 2007 as for Figure 4 with the same scale ............. 197
Figure 12: Geographical distribution of LRIT messages processed by the pre-processor
for 2009–2012. Unit: total number of messages per grid cell
with an area of 0.2 × 0.2 degrees ....................................... 198
Figure 13: The average volume of AIS activity reports for a region reported by a ship
for up to 300 randomly selected VLCCs from 2007 to 2012 ................... 199
Figure 14: The average volume of AIS activity reports for a region reported by a ship
for up to 300 randomly selected Capesize bulk carriers from 2007 to 2012 ........ 200
Figure 15: Fraction of time intervals between consecutive messages that fall between
five and seven hours for each ship type and size category ..................... 201
Figure 16: Plots of difference in fraction of time spent at sea for all ships, with increasing
high-confidence AIS count over the year. For each ship in one of these 5%-wide bins,
the difference between fraction of time at sea between AIS and LRIT is calculated.
The mean of this difference per bin is plotted in red, and the standard deviation
of the difference in each bin is plotted in blue .............................. 203
Figure 17: Distribution of difference between resampled hourly speeds
and the reported speed within the hour sampled across 10 VLCCs in 2011.
The standard deviation was calculated as 0.75 nm/hr ........................ 204
Figure 18: Distribution of difference between reported speeds when the time difference
in reporting is within two hours (sampled as message from 105 minutes to
120 minutes from the original message) for all VLCCs captured in AIS.
The standard deviation of the sample was 1.85 nm/hr ........................ 204
Page
xxii Third IMO GHG Study 2014
Figure 19: Difference between average speed at sea for each ship size and type category.
Negative values indicate that LRIT data provides a lower estimate of speed
than the extrapolated AIS data ......................................... 205
Figure 20: Fuel oil shipping sales, export-import discrepancy
and statistical difference at world balance ................................. 232
Figure 21: Gas/diesel shipping sales, export-import discrepancy
and statistical difference at world balance ................................. 232
Figure 22: World natural gas shipping sales, export-import discrepancy
and statistical difference .............................................. 233
Figure 23: Stacked graph showing sum of fuel transfer balance
and export-import discrepancy ......................................... 234
Figure 24: Time series of adjustments due to primary and secondary sources of uncertainty ... 235
Figure 25: Adjusted marine fuel sales based on quantitative uncertainty results (2007–2011) ... 235
Figure 26: Bottom-up model with overview of QA/QC
and uncertainty characterization ........................................ 238
Figure 27: Relationship between speed over ground and speed through the water ........... 240
Figure 28: Comparison between draught estimated in the bottom-up model from AIS data
and reported in noon reports .......................................... 241
Figure 29: Estimation of the power law relating deadweight to resistance
for samples of different ship types ....................................... 242
Figure 30: Uncertainty around the annual emissions (x-axis is ‘00,000 tonnes of CO
2
;
y-axis is frequency) from a Monte Carlo simulation of an “average”
Panamax bulk carrier (60,000–99,999 dwt capacity) in 2007 .................. 243
Figure 31: Uncertainty around the annual emissions (x-axis is ‘00,000 tonnes of CO
2
;
y-axis is frequency) from a Monte Carlo simulation of an “average”
Panamax bulk carrier (60,000–99,999 dwt capacity) in 2012 .................. 244
Figure 32: Schematic presentation of the emissions projection model .................... 255
Figure 33: Transport work for all categories of cargo, provided by UNCTAD,
from 1970 to 2012 in billion tonne-miles; also illustrated with global GDP
(right-hand axis) in US$ billion (constant 2005 prices) ........................ 256
Figure 34: Ratios of TST (different subtypes) in billion tonne-miles
to historical global GDP in US$ (constant 2005 prices) and coal/oil consumption
(from BP Statistical Review) ............................................ 258
Figure 35: Historical and modelled growth curves to 2050 for ratios of total oil,
coal, total (non-coal) bulk dry goods and other dry cargoes ................... 258
Figure 36: The role of fleet productivity in the model structure ......................... 259
Figure 37: Historical fleet productivity (Stopford, 2009) ............................... 260
Figure 38: Productivity of oil tankers measured in thousand tonne-miles per dwt,
1970–2013 ........................................................ 261
Figure 39: Productivity of dry bulk carriers measured in thousand tonne-miles
(five main dry bulks) per dwt (all bulk carriers), 1970–2013 .................... 262
Figure 40: Productivity of container and liquefied gas ships
measured in thousand tonne-miles per dwt, 1999–2013 ...................... 263
Figure 41: First (upper chart) and second (lower chart) methodologies
to determine the number of ships per size category in 2050 ................... 265
Figure 42: Composition of global container fleet in the period 2002–2014
(beginning of year figures) ............................................. 267
Figure 43: Historical development of average ship size of cellular fleet ................... 269
Figure 44: Projected tanker fleet development 1992–2013
(projection for 2012 and 2013) ......................................... 270
Page
List of figures xxiii
Figure 45: Capacity distribution of tankers over size categories (1994–2013) ............... 271
Figure 46: Capacity distribution of bulker fleet over size categories
(1994–2013) ....................................................... 273
Figure 47: Development of average capacity of LNG carriers over the period 1970–2011
and corresponding linear trend ......................................... 276
Figure 48: Development of the average size of LPG carriers
in the period 1999–2012 ............................................. 278
Figure 49: Impact of the Poisson distribution on EEDI efficiency improvements ............. 281
Page
Page
List of tables
Table 1 – a) Shipping CO
2
emissions compared with global CO
2
(values in
million tonnes CO
2
); and b) Shipping GHGs (in CO
2
e)
compared with global GHGs (values in million tonnes CO
2
e)..................... 1
Table 2 – International, domestic and fishing CO
2
emissions 2007–2011,
using top-down method. ............................................... 10
Table 3 – International, domestic and fishing CO
2
emissions 2007–2012,
using bottom-up method................................................ 10
Table 4 – Relationship between slow steaming, engine load factor (power output)
and fuel consumption for 2007 and 2012 ................................... 14
Table 5 – Summary of the scenarios for future emissions from international shipping,
GHGs and other relevant substances ...................................... 22
Table 6 – AIS observation statistics of the fleet identified in the IHSF database
as in service in 2007 and 2012 ........................................... 23
Table 7 – Comparison of 2011 and 2012 marine fuels reporting to IEA ..................... 28
Table 8 – Comparison of Second IMO GHG Study 2009 top-down ship fuel consumption data
(million tonnes)....................................................... 29
Table 9 – Summary of IEA fuel sales data in shipping (million tonnes)...................... 31
Table 10 – IHSF vessel types and related vessel classes ................................. 35
Table 11 – Classification of ships in the bottom-up approach............................. 38
Table 12 – Summary of vessel types and sizes that can be expected to engage
in international shipping ................................................ 40
Table 13 – Summary of vessel types and sizes that can be expected to engage
in domestic shipping .................................................. 41
Table 14 – Tabular data for 2012 describing the fleet (international, domestic and fishing)
analysed using the bottom-up method ..................................... 43
Table 15 – International, domestic and fishing CO
2
emissions 2007–2011 (million tonnes),
using the top-down method ............................................. 46
Table 16 – International, domestic and fishing CO
2
emissions 2007–2012 (million tonnes),
using the bottom-up method ............................................ 47
Table 17 – Relationship between slow steaming, engine load factor (power output)
and fuel consumption for 2007 and 2012 ................................... 55
Table 18 – Bottom-up CO
2
e emissions estimates with climate-carbon feedbacks
from total shipping (thousand tonnes) ...................................... 56
Table 19 – Bottom-up CO
2
e emissions estimates with climate-carbon feedbacks
from international shipping (thousand tonnes) ................................ 56
Table 20 – Shipping CO
2
emissions compared with global CO
2
(values in million tonnes CO
2
) ........................................... 57
Table 21 – Shipping CH
4
emissions compared with global CH
4
(values in thousand tonnes CH
4
).......................................... 58
Table 22 – Shipping N
2
O emissions compared with global N
2
O
(values in thousand tonnes N
2
O) ......................................... 58
Table 23 – Shipping GHGs (in CO
2
e) compared with global GHGs
(values in million tonnes CO
2
e)........................................... 58
Page
xxvi Third IMO GHG Study 2014
Table 24 – Comparison of fuel sales data between IEA and EIA in international shipping
(million tonnes)....................................................... 62
Table 25 – Summary of the findings on the QA of the bottom-up method
estimated fuel consumption using noon report data ........................... 74
Table 26 – Observed, unobserved and active ship counts (2007–2012) ..................... 79
Table 27 – Statistics of the number of in-service ships observed on AIS
and of the average amount of time during the year for which a ship is observed ...... 81
Table 28 – International, domestic and fishing CO
2
emissions 2007–2011 (million tonnes),
using top-down method ................................................ 84
Table 29 – International, domestic and fishing CO
2
emissions 2007–2012 (million tonnes),
using bottom-up method ............................................... 84
Table 30 – Summary of average domestic tonnes of fuel consumption per year (2007–2012),
MMSI counts and correlations between domestic fuel use statistics................ 85
Table 31 – Upper range of top-down fuel consumption by vessel type (million tonnes).......... 86
Table 32 – Results of quantitative uncertainty analysis on top-down statistics (million tonnes) ..... 86
Table 33 – Summary of major differences between the Second IMO GHG Study 2009
and Third IMO GHG Study 2014 ......................................... 93
Table 34 – Emissions factors for top-down emissions from combustion of fuels ............... 96
Table 35 – Year-specific emissions factors for sulphur-dependent emissions (SO
x
and PM) ....... 96
Table 36 – Amounts of refrigerants carried by various types of ships (from DG ENV report) ...... 98
Table 37 – Annual loss of refrigerants from the global fleet during 2012.
Annual release of 40% total refrigerant carried is assumed
except for passenger-class vessels, where 20% refrigerant loss is assumed.
Ro-ro, pax, ro-pax and cruise vessels are calculated as passenger ships............. 100
Table 38 – Global warming potential of refrigerants commonly used in ships.
The GWP100 is described relative to CO
2
warming potential
(IPCC Fourth Assessment Report: Climate Change 2007)........................ 101
Table 39 – Annual emissions of refrigerants from the global fleet
and estimated shares of different refrigerants................................. 102
Table 40 – Vessel operating modes used in this study .................................. 103
Table 41 – NO
x
baseline emissions factors........................................... 105
Table 42 – SO
x
baseline emissions factors ........................................... 106
Table 43 – Annual fuel oil sulphur worldwide averages ................................. 106
Table 44 – PM baseline emissions factors ........................................... 106
Table 45 – CO baseline emissions factors ........................................... 107
Table 46 – CH
4
baseline emissions factors ........................................... 107
Table 47 – N
2
O baseline emissions factors........................................... 108
Table 48 – NMVOC baseline emissions factors ....................................... 108
Table 49 – Specific fuel oil consumption of marine diesel engines (ll values in g/kWh) .......... 109
Table 50 – Specific fuel oil consumption (SFOC
baseline
) of gas turbines, boiler
and auxiliary engines used in this study as the basis to estimate dependency
of SFOC as a function of load. Unit is grams of fuel used per power unit (g/kWh)
(IVL, 2004) .......................................................... 110
Table 51 – Annual fuel oil sulphur worldwide averages ................................. 111
Table 52 – Top-down fuel consumption inventory (million tonnes)......................... 113
Table 53 – Top-down CH
4
emissions estimates (tonnes) ................................. 113
Table 54 – Top-down N
2
O emissions estimates (tonnes)................................. 114
Table 55 – Top-down SO
x
emissions estimates (thousand tonnes as SO
2
) .................... 114
Page
List of tables xxvii
Table 56 – Top-down NO
x
emissions estimates (thousand tonnes as NO
2
) ................... 115
Table 57 – Top-down PM emissions estimates (thousand tonnes) .......................... 115
Table 58 – Top-down CO emissions estimates (thousand tonnes) .......................... 115
Table 59 – Top-down NMVOC emissions estimates (thousand tonnes) ...................... 116
Table 60 – Bottom-up fuel consumption estimates (million tonnes)......................... 116
Table 61 – Bottom-up CH
4
emissions estimates (tonnes)................................. 116
Table 62 – Bottom-up N
2
O emissions estimates (tonnes) ................................ 116
Table 63 – Bottom-up SO
x
emissions estimates (thousand tonnes as SO
2
).................... 117
Table 64 – Bottom-up NO
x
emissions estimates (thousand tonnes as NO
2
)................... 117
Table 65 – Bottom-up PM emissions estimates (thousand tonnes).......................... 117
Table 66 – Bottom-up CO emissions estimates (thousand tonnes).......................... 117
Table 67 – Bottom-up NMVOC emissions estimates (thousand tonnes)...................... 117
Table 68 – Comparison of emissions factors, Second IMO GHG Study 2009
and Third IMO GHG Study 2014 ......................................... 119
Table 69 – Descriptions and sources of representative concentration pathways ............... 130
Table 70 – Short narratives of shared socioeconomic pathways ........................... 130
Table 71 – Ship type productivity indices used in emissions projection model ................ 133
Table 72 – 2012 distribution and expected distribution 2050 of container
and LG carriers over bin sizes............................................ 134
Table 73 – 2012 distribution and expected distribution 2050 of oil/chemical tankers
and dry bulk carriers over bin sizes........................................ 134
Table 74 – Fuel mix scenarios used for emissions projection (mass %) ...................... 136
Table 75 – NO
x
emissions factors in 2012, 2030 and 2050 (g/g fuel) ....................... 136
Table 76 – HFCs used on board ships .............................................. 137
Table 77 – Overview of assumptions per scenario ..................................... 140
Table 78 – CO
2
emissions projections (million tonnes) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 141
Table 79 – Emissions of CO
2
and other substances in 2012, 2020 and 2050 (million tonnes) ...... 144
Annex Tables
Table 1 – Data gap-filling methods by IHSF ship technical field .......................... 153
Table 2 – IHSF ship groups and classes ............................................ 155
Table 3 – Cargo-carrying category: class, subclass and StatCode5 designations .............. 156
Table 4 – Ship class capacity bins ................................................ 158
Table 5 – Number of processed messages (in millions) in 2007–2012 for each terrestrial
and satellite data sets. EMSA LRIT data were used for QA/QC
of the bottom-up emissions estimation only ................................. 159
Table 6 – Ship-weighted auxiliary engine loads, by mode, for selected ship classes,
with VBP data ....................................................... 169
Table 7 – Ship-weighted auxiliary engine loads for selected ship classes, with FMI data ........ 169
Table 8 – Auxiliary engine loads, by ship class and mode............................... 170
Table 9 – Auxiliary boiler loads, by ship class and mode ............................... 171
Table 10 – Description of bottom-up model subroutines and calculation stages ............... 176
Table 11 – Mean number of messages by ship for LRIT ships used in the analysis.............. 200
Table 12 – AIS to LRIT ship mapping (number of ships) ................................. 201
Table 13 – Analysis of 2011 and 2012 observed cargo-carrying fleet ....................... 206
Table 14 – List of operators and their fleets (number of ships) used in this analysis ............. 206
Page
xxviii Third IMO GHG Study 2014
Table 15 – List of ship types (number of ships) used in this analysis ........................ 207
Table 16 – Results of quantitative uncertainty analysis on top-down statistics (million tonnes)..... 235
Table 17 – Characterization of uncertainty in bottom-up model........................... 238
Table 18 – Summary table of uncertainty characterizations used .......................... 242
Table 19 – Estimated parameters for the uncertainty in the inputs
to the annual emissions calculation........................................ 243
Table 20 – Estimated parameters for the uncertainty in the inputs
to the annual emissions calculation........................................ 244
Table 21 – Estimated characteristics of the uncertainty for individual ship type
and size categories .................................................... 245
Table 22 – Baseline emissions factors .............................................. 249
Table 23 – IMO Tier I and II SFOC assumptions for NO
x
baseline emissions factors ............ 250
Table 24 – EF-related SFOCs used to convert energy-based baseline emissions factors
to fuel-based ........................................................ 251
Table 25 – IMO annual average global sulphur contents ................................ 251
Table 26 – NO
x
FCFs – HFO global sulphur averages .................................. 251
Table 27 – NO
x
FCFs – MGO global sulphur averages .................................. 252
Table 28 – SO
x
FCFs – HFO global sulphur averages ................................... 252
Table 29 – SO
x
FCFs – MGO global sulphur averages .................................. 252
Table 30 – PM FCFs – HFO global sulphur averages ................................... 252
Table 31 – PM FCFs – MGO global sulphur averages................................... 253
Table 32 – Emissions factors for bottom-up emissions due to the combustion of fuels........... 253
Table 33 – Year-specific bottom-up emissions factors for SO
x
and PM ...................... 253
Table 34 – Ship type productivity indices used in emissions projection model ................ 263
Table 35 – Methodology applied for the projection of ship size distribution
of the different ship types differentiated in the study ........................... 266
Table 36 – 2012 distribution of container ships over the size categories
in terms of numbers ................................................... 266
Table 37 – Development of the distribution of container ships over size categories
(in terms of numbers) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 268
Table 38 – Size bins for tankers ................................................... 269
Table 39 – Development of the distribution of oil tankers over size categories
(in terms of capacity) .................................................. 272
Table 40 – Size bins for dry bulk carriers ............................................ 272
Table 41 – Development of the distribution of dry bulk ships (including combined carriers)
over size categories in terms of capacity .................................... 274
Table 42 – Distribution of global LNG fleet over size categories in terms of numbers in 2012 ..... 275
Table 43 – Development of distribution of global LNG fleet over size categories
in terms of numbers ................................................... 276
Table 44 – Distribution of LPG fleet, end of 2011 (nine size categories) ..................... 277
Table 45 – Distribution of 2012 LPG fleet in terms of numbers (three size categories) ........... 277
Table 46 – Development of distribution of global LPG fleet in terms of numbers .............. 278
Table 47 – Development of distribution (in terms of numbers of ships)
of the global gas carrier fleet ............................................ 278
Table 48 – Reduction factors (percentage) for EEDI relative to the EEDI reference line........... 279
Table 49 – Impact of the SFC on EEDI efficiency improvements........................... 280
Table 50 – Impact of the Poisson distribution on EEDI efficiency improvements ............... 281
Page
List of tables xxix
Table 51 – Assessment of potential reductions of CO2 emissions from shipping
by using known technology and practices (from Second IMO GHG Study 2009) ..... 282
Table 52 – Impact of EEDI on operational efficiency of new ships ......................... 283
Table 53 – Efficiency improvements in the period 2030–2050 ............................ 283
Table 54 – IMO sulphur requirements .............................................. 284
Table 55 – Emission control areas established in 2014 .................................. 284
Table 56 – IMO NO
x
limits ...................................................... 285
Table 57 – Main compliance options of regulations 13 and 14 of MARPOL Annex VI........... 286
Table 58 – Fuel mix scenarios used for emissions projection (% m/m) ...................... 287
Table 59 – CO
2
emissions factors (g/g fuel) .......................................... 287
Table 60 – CH
4
emissions factors (g/g fuel) .......................................... 288
Table 61 – N
2
O emissions factors (g/g fuel) .......................................... 288
Table 62 – HFCs used on board ships .............................................. 288
Table 63 – HFC emissions per ship (tonnes per year) ................................... 289
Table 64 – NO
x
emissions factors (g/g fuel) .......................................... 290
Table 65 – NMVOC emissions factors (g/g fuel) ....................................... 290
Table 66 – CO emissions factors (g/g fuel) ........................................... 290
Table 67 – PM emissions factors (g/g fuel) ........................................... 291
Table 68 – SO
2
emissions factors (g/g fuel)........................................... 291
Table 69 – Scenarios 1 and 9 (RCP8.5, SSP5, high efficiency)............................. 291
Table 70 – Scenarios 2 and 10 (RCP6.0, SSP1, high efficiency) ............................ 292
Table 71 – Scenarios 3 and 11 (RCP4.5, SSP3, high efficiency) ............................ 292
Table 72 – Scenarios 4 and 12 (RCP2.6, SSP4, high efficiency) ............................ 293
Table 73 – Scenarios 5 and 13 (RCP8.5, SSP5, low efficiency) ............................ 293
Table 74 – Scenarios 6 and 14 (RCP6.0, SSP1, low efficiency) ............................ 294
Table 75 – Scenarios 7 and 15 (RCP4.5, SSP3, low efficiency) ............................ 294
Table 76 – Scenarios 8 and 16 (RCP2.6, SSP4, low efficiency) ............................ 295
Page
Executive Summary
Key findings from the Third IMO GHG Study 2014
1 Shipping emissions during the period 2007–2012 and their significance relative to other
anthropogenic emissions
1.1 For the year 2012, total shipping emissions were approximately 938 million tonnes CO
2
and
961million tonnes CO
2
e for GHGs combining CO
2
, CH
4
and N
2
O. International shipping emissions for 2012
are estimated to be 796 million tonnes CO
2
and 816 million tonnes CO
2
e for GHGs combining CO
2
, CH
4
and
N
2
O. International shipping accounts for approximately 2.2% and 2.1% of global CO
2
and GHG emissions on
a CO
2
equivalent (CO
2
e) basis, respectively. Table 1 presents the full time series of shipping CO
2
and CO
2
e
emissions compared with global total CO
2
and CO
2
e emissions.
For the period 2007–2012, on average, shipping accounted for approximately 3.1% of annual global CO
2
and
approximately 2.8% of annual GHGs on a CO
2
e basis using 100-year global warming potential conversions
from the IPCC Fifth Asssessment Report (AR5). A multi-year average estimate for all shipping using bottom-up
totals for 2007–2012 is 1,015 million tonnes CO
2
and 1,036 million tonnes CO
2
e for GHGs combining CO
2
,
CH
4
and N
2
O. International shipping accounts for approximately 2.6% and 2.4% of CO
2
and GHGs on a
CO
2
e basis, respectively. A multi-year average estimate for international shipping using bottom-up totals for
2007–2012 is 846 million tonnes CO
2
and 866 million tonnes CO
2
e for GHGs combining CO
2
, CH
4
and N
2
O.
These multi-year CO
2
and CO
2
e comparisons are similar to, but slightly smaller than, the 3.3% and 2.7% of
global CO
2
emissions reported by the Second IMO GHG Study 2009 for total shipping and international
shipping in the year 2007, respectively.
Table 1 – a) Shipping CO
2
emissions compared with global CO
2
(values in million tonnes CO
2
); and
b) Shipping GHGs (in CO
2
e) compared with global GHGs (values in million tonnes CO
2
e)
Third IMO GHG Study 2014 CO
2
Year Global CO
2
1
Total shipping % of global International shipping % of global
2007 31,409 1,100 3.5% 885 2.8%
2008 32,204 1,135 3.5% 921 2.9%
2009 32,047 978 3.1% 855 2.7%
2010 33,612 915 2.7% 771 2.3%
2011 34,723 1,022 2.9% 850 2.4%
2012 35,640 938 2.6% 796 2.2%
Average 33,273 1,015 3.1% 846 2.6%
Third IMO GHG Study 2014 CO
2
e
Year Global CO
2
e
2
Total shipping % of global International shipping % of global
2007 34,881 1,121 3.2% 903 2.6%
2008 35,677 1,157 3.2% 940 2.6%
2009 35,519 998 2.8% 873 2.5%
2010 37,0 85 935 2.5% 790 2.1%
2011 38,196 1,045 2.7% 871 2.3%
2012 39,113 961 2.5% 816 2.1%
Average 36,745 1,036 2.8% 866 2.4%
1
Global comparator represents CO
2
from fossil fuel consumption and cement production, converted from Tg C y
–1
to million metric
tonnes CO
2
. Sources: Boden et al. 2013 for years 2007–2010; Peters et al. 2013 for years 2011–2012, as referenced in IPCC (2013).
2
Global comparator represents N
2
O from fossil fuels consumption and cement production. Source: IPCC (2013, Table 6.9).
2 Third IMO GHG Study 2014
1.2 This study estimates multi-year (2007–2012) average annual totals of 20.9 million and 11.3 million
tonnes for NO
x
(as NO
2
) and SO
x
(as SO
2
) from all shipping, respectively (corresponding to 6.3 million and
5.6 million tonnes converted to elemental weights for nitrogen and sulphur respectively). NO
x
and SO
x
play
indirect roles in tropospheric ozone formation and indirect aerosol warming at regional scales. Annually,
international shipping is estimated to produce approximately 18.6 million and 10.6 million tonnes of NO
x
(as
NO
2
) and SO
x
(as SO
2
) respectively; this converts to totals of 5.6 million and 5.3 million tonnes of NO
x
and SO
x
respectively (as elemental nitrogen and sulphur respectively). Global NO
x
and SO
x
emissions from all shipping
represent about 15% and 13% of global NO
x
and SO
x
from anthropogenic sources reported in the IPCC Fifth
Assessment Report (AR5), respectively; international shipping NO
x
and SO
x
represent approximately 13% and
12% of global NO
x
and SO
x
totals respectively.
1.3 Over the period 2007–2012, average annual fuel consumption ranged between approximately
247million and 325 million tonnes of fuel consumed by all ships within this study, reflecting top-down and
bottom-up methods respectively. Of that total, international shipping fuel consumption ranged on average
between approximately 201 million and 272 million tonnes per year, depending on whether consumption was
defined as fuel allocated to international voyages (top-down) or fuel used by ships engaged in international
shipping (bottom-up), respectively.
1.4 Correlated with fuel consumption, CO
2
emissions from shipping are estimated to range between
approximately 739 million and 795 million tonnes per year in top-down results, and to range between
approximately 915 million and 1135 million tonnes per year in bottom-up results. Both the top-down and
the bottom-up methods indicate limited growth in energy and CO
2
emissions from ships during 2007–2012,
as suggested both by the IEA data and the bottom-up model. Nitrous oxide (N
2
O) emission patterns over
2007–2012 are similar to the fuel consumption and CO
2
patterns, while methane (CH
4
) emissions from ships
increased due to increased activity associated with the transport of gaseous cargoes by liquefied gas tankers,
particularly over 2009–2012.
1.5 International shipping CO
2
estimates range between approximately 596 million and 649 million
tonnes calculated from top-down fuel statistics, and between approximately 771 million and 921 million
tonnes according to bottom-up results. International shipping is the dominant source of the total shipping
emissions of other GHGs: nitrous oxide (N
2
O) emissions from international shipping account for the majority
(approximately 85%) of total shipping N
2
O emissions, and methane (CH
4
) emissions from international ships
account for nearly all (approximately 99%) of total shipping emissions of CH
4
.
1.6 Refrigerant and air conditioning gas releases account for the majority of HFC (and HCFC) emissions
from ships. For older vessels, HCFCs (R-22) are still in service, whereas new vessels use HFCs (R134a/R404a).
Use of SF
6
and PFCs in ships is documented as rarely used in large enough quantities to be significant and is
not estimated in this report.
1.7 Refrigerant and air conditioning gas releases from shipping contribute an additional 15 million tons
(range 10.8 million–19.1 million tons) in CO
2
equivalent emissions. Inclusion of reefer container refrigerant
emissions yields 13.5 million tons (low) and 21.8 million tons (high) of CO
2
emissions.
1.8 Combustion emissions of SO
x
, NO
x
, PM, CO and NMVOCs are also correlated with fuel consumption
patterns, with some variability according to properties of combustion across engine types, fuel properties, etc.,
which affect emissions substances differently.
2 Resolution, quality and uncertainty of the emissions inventories
2.1 The bottom-up method used in this study applies a similar approach to the Second IMO GHG Study
2009 in order to estimate emissions from activity. However, instead of analysis carried out using ship type,
size and annual average activity, calculations of activity, fuel consumption (per engine) and emissions (per
GHG and pollutant substances) are performed for each in-service ship during each hour of each of the years
2007–2012, before aggregation to find the totals of each fleet and then of total shipping (international, domestic
and fishing) and international shipping. This removes any uncertainty attributable to the use of average values
and represents a substantial improvement in the resolution of shipping activity, energy demand and emissions
data.
2.2 This study clearly demonstrates the confidence that can be placed in the detailed findings of the
bottom-up method of analysis through both quality analysis and uncertainty analysis. Quality analysis includes
Executive Summary 3
rigorous testing of bottom-up results against noon reports and LRIT data. Uncertainty analysis quantifies, for
the first time, the uncertainties in the top-down and the bottom-up estimates.
2.3 These analyses show that high-quality inventories of shipping emissions can be produced through
the analysis of AIS data using models. Furthermore, the advancement in the state-of-the-art methods used in
this study provides insight and produces new knowledge and understanding of the drivers of emissions within
subsectors of shipping (ships of common type and size).
2.4 The quality analysis shows that the availability of improved data (particularly AIS data) since 2010 has
enabled the uncertainty of inventory estimates to be reduced (relative to previous years’ estimates). However,
uncertainties remain, particularly in the estimation of the total number of active ships and the allocation of
ships or ship voyages between domestic and international shipping.
2.5 For both the top-down and the bottom-up inventory estimates in this study, the uncertainties relative
to the best estimate are not symmetrical (the likelihood of an overestimate is not the same as that of an
underestimate). The top-down estimate is most likely to be an underestimate (for both total shipping and
international shipping), for reasons discussed in the main report. The bottom-up uncertainty analysis shows
that while the best estimate is higher than top-down totals, uncertainty is more likely to lower estimated values
from the best estimate (again, for both total shipping and international shipping).
2.6 There is an overlap between the estimated uncertainty ranges of the bottom-up and the top-down
estimates of fuel consumption in each year and for both total shipping and international shipping. This provides
evidence that the discrepancy between the top-down and the bottom-up best estimate value is resolvable
through the respective methods’ uncertainties.
2.7 Estimates of CO
2
emissions from the top-down and bottom-up methods converge over the period of
the study as the source data of both methods improve in quality. This provides increased confidence in the
quality of the methodologies and indicates the importance of improved AIS coverage from the increased use
of satellite and shore-based receivers to the accuracy of the bottom-up method.
2.8 All previous IMO GHG studies have preferred activity-based (bottom-up) inventories. In accordance
with IPCC guidance, the statements from the MEPC Expert Workshop and the Second IMO GHG Study 2009,
the Third IMO GHG Study 2014 consortium specifies the bottom-up best estimate as the consensus estimate
for all years’ emissions for GHGs and all pollutants.
3 Comparison of the inventories calculated in this study with the inventories of the Second
IMO GHG Study 2009
3.1 Best estimates for 2007 fuel use and CO
2
emissions in this study agree with the “consensus estimates”
of the Second IMO GHG Study 2009 as they are within approximately 5% and approximately 4%, respectively.
3.2 Differences with the Second IMO GHG Study 2009 can be attributed to improved activity data, better
precision of individual vessel estimation and aggregation and updated knowledge of technology, emissions
rates and vessel conditions. Quantification of uncertainties enables a fuller comparison of this study with
previous work and future studies.
3.3 The estimates in this study of non-CO
2
GHGs and some air pollutant substances differ substantially
from the 2009 results for the common year 2007. This study produces higher estimates of CH
4
and N
2
O than
the earlier study, higher by 43% and 40% respectively (approximate values). The new study estimates lower
emissions of SO
x
(approximately 30% lower) and approximately 40% of the CO emissions estimated in the
2009 study.
3.4 Estimates for NO
x
, PM and NMVOC in both studies are similar for 2007, within 10%, 11% and 3%
respectively (approximate values).
4 Fuel use trends and drivers in fuel use (2007–2012), in specific ship types
4.1 The total fuel consumption of shipping is dominated by three ship types: oil tankers, container ships and
bulk carriers. Consistently for all ship types, the main engines (propulsion) are the dominant fuel consumers.
4.2 Allocating top-down fuel consumption to international shipping can be done explicitly, according to
definitions for international marine bunkers. Allocating bottom-up fuel consumption to international shipping
4 Third IMO GHG Study 2014
required application of a heuristic approach. The Third IMO GHG Study 2014 used qualitative information
from AIS to designate larger passenger ferries (both passenger-only pax ferries and vehicle-and-passenger
ro-pax ferries) as international cargo transport vessels. Both methods are unable to fully evaluate global
domestic fuel consumption.
4.3 The three most significant sectors of the shipping industry from a CO
2
perspective (oil tankers,
container ships and bulk carriers) have experienced different trends over the period of this study (2007–2012).
All three contain latent emissions increases (suppressed by slow steaming and historically low activity and
productivity) that could return to activity levels that create emissions increases if the market dynamics that
informed those trends revert to their previous levels.
4.4 Fleet activity during the period 2007–2012 demonstrates widespread adoption of slow steaming.
The average reduction in at-sea speed relative to design speed was 12% and the average reduction in daily
fuel consumption was 27%. Many ship type and size categories exceeded this average. Reductions in daily
fuel consumption in some oil tanker size categories was approximately 50% and some container-ship size
categories reduced energy use by more than 70%. Generally, smaller ship size categories operated without
significant change over the period, also evidenced by more consistent fuel consumption and voyage speeds.
4.5 A reduction in speed and the associated reduction in fuel consumption do not relate to an equivalent
percentage increase in efficiency, because a greater number of ships (or more days at sea) are required to do
the same amount of transport work.
5 Future scenarios (2012–2050)
5.1 Maritime CO
2
emissions are projected to increase significantly in the coming decades. Depending on
future economic and energy developments, this study’s BAU scenarios project an increase by 50% to 250%
in the period to 2050. Further action on efficiency and emissions can mitigate the emissions growth, although
all scenarios but one project emissions in 2050 to be higher than in 2012.
5.2 Among the different cargo categories, demand for transport of unitized cargoes is projected to increase
most rapidly in all scenarios.
5.3 Emissions projections demonstrate that improvements in efficiency are important in mitigating
emissions increase. However, even modelled improvements with the greatest energy savings could not yield
a downward trend. Compared to regulatory or market-driven improvements in efficiency, changes in the fuel
mix have a limited impact on GHG emissions, assuming that fossil fuels remain dominant.
5.4 Most other emissions increase in parallel with CO
2
and fuel, with some notable exceptions. Methane
emissions are projected to increase rapidly (albeit from a low base) as the share of LNG in the fuel mix
increases. Emissions of nitrogen oxides increase at a lower rate than CO
2
emissions as a result of Tier II and
Tier III engines entering the fleet. Emissions of particulate matter show an absolute decrease until 2020, and
sulphurous oxides continue to decline through to 2050, mainly because of MARPOL Annex VI requirements
on the sulphur content of fuels.
Aim and objective of the study
This study provides IMO with a multi-year inventory and future scenarios for GHG and non-GHG emissions
from ships. The context for this work is:
• The IMO committees and their members require access to up-to-date information to support working
groups and policy decision-making. Five years have passed since the publication of the previous study
(Second IMO GHG Study 2009), which estimated emissions for 2007 and provided scenarios from
2007 to 2050. Furthermore, IPCC has updated its analysis of future scenarios for the global economy in
its AR5 (2013), including mitigation scenarios. IMO policy developments, including MARPOL Annex
VI amendments for EEDI and SEEMP, have also occurred since the 2009 study was undertaken. In this
context, the Third IMO GHG Study 2014 updates the previous work by producing yearly inventories
since 2007.
• Other studies published since the Second IMO GHG Study 2009 have indicated that one impact of
the global financial crisis may have been to modify previously reported trends, both in demand for
shipping and in the intensity of shipping emissions. This could produce significantly different recent-year
emissions than the previously forecasted scenarios, and may modify the long-run projections for 2050
ship emissions. In this context, the Third IMO GHG Study 2014 provides new projections informed
by important economic and technological changes since 2007.
• Since 2009, greater geographical coverage achieved via satellite technology/AIS receivers has
improved the quality of data available to characterize shipping activity beyond the state of practice
used in the Second IMO GHG Study 2009. These new data make possible more detailed methods
that can substantially improve the quality of bottom-up inventory estimates. Additionally, improved
understanding of marine fuel (bunker) statistics reported by nations has identified, but not quantified,
potential uncertainties in the accuracy of top-down inventory estimates from fuel sales to ships.
Improved bottom-up estimates can reconcile better the discrepancies between top-down and
bottom-up emissions observed in previous studies (including the Second IMO GHG Study 2009). In
this context, the Third IMO GHG Study 2014 represents the most detailed and comprehensive global
inventory of shipping emissions to date.
The scope and design of the Third IMO GHG Study 2014 responds directly to specific directives from the IMO
Secretariat that derived from the IMO Expert Workshop (2013) recommendations with regard to activity-based
(bottom-up) ship emissions estimation. These recommendations were:
• to consider direct vessel observations to the greatest extent possible;
• to use vessel-specific activity and technical details in a bottom-up inventory model;
• to use “to the best extent possible” actual vessel speed to obtain engine loads.
The IMO Expert Workshop recognized that “bottom-up estimates are far more detailed and are generally based
on ship activity levels by calculating the fuel consumption and emissions from individual ship movements”
and that “a more sophisticated bottom-up approach to develop emission estimates on a ship-by-ship basis”
would “require significant data to be inputted and may require additional time […] to complete”.
Structure of the study and scope of work
The Third IMO GHG Study 2014 report follows the structure of the terms of reference for the work, which
comprise three main sections:
Section 1: Inventories of CO
2
emissions from international shipping 2007–2012
This section deploys both a top-down (2007–2011) and a bottom-up (2007–2012) analysis of CO
2
emissions
from international shipping. The inventories are analysed and discussed with respect to the quality of methods
and data and to uncertainty of results. The discrepancies between the bottom-up and top-down inventories
are discussed. The Third IMO GHG Study 2014 inventory for 2007 is compared to the Second IMO GHG
Study 2009 inventory for the same year.
Section 2: Inventories of emissions of GHGs and other relevant substances from international
shipping 2007–2012
This section applies the top-down (2007–2011) and bottom-up (2007–2012) analysis from Section 1 in
combination with data describing the emissions factors and calculations inventories for non-CO
2
GHGs –
methane (CH
4
), nitrous oxide (N
2
O), HFCs and sulphur hexafluoride (SF
6
) – and relevant substances – oxides
of sulphur (SOx), oxides of nitrogen (NO
x
), particulate matter (PM), carbon monoxide (CO) and NMVOCs.
The quality of methods and data and uncertainty of the inventory results are discussed, and comparisons are
made between the top-down and bottom-up estimates in the Third IMO GHG Study 2014 and the results of
the Second IMO GHG Study 2009.
Section 3: Scenarios for shipping emissions 2012–2050
This section develops scenarios for future emissions for all GHGs and other relevant substances investigated
in Sections 1 and 2. Results reflect the incorporation of new base scenarios used in GHG projections for
non-shipping sectors and method advances, and incorporate fleet activity and emissions insights emerging
from the 2007–2012 estimates. Drivers of emissions trajectories are evaluated and sources of uncertainty in
the scenarios are discussed.
Executive Summary 5
emissions than the previously forecasted scenarios, and may modify the long-run projections for 2050
ship emissions. In this context, the Third IMO GHG Study 2014 provides new projections informed
by important economic and technological changes since 2007.
• Since 2009, greater geographical coverage achieved via satellite technology/AIS receivers has
improved the quality of data available to characterize shipping activity beyond the state of practice
used in the Second IMO GHG Study 2009. These new data make possible more detailed methods
that can substantially improve the quality of bottom-up inventory estimates. Additionally, improved
understanding of marine fuel (bunker) statistics reported by nations has identified, but not quantified,
potential uncertainties in the accuracy of top-down inventory estimates from fuel sales to ships.
Improved bottom-up estimates can reconcile better the discrepancies between top-down and
bottom-up emissions observed in previous studies (including the Second IMO GHG Study 2009). In
this context, the Third IMO GHG Study 2014 represents the most detailed and comprehensive global
inventory of shipping emissions to date.
The scope and design of the Third IMO GHG Study 2014 responds directly to specific directives from the IMO
Secretariat that derived from the IMO Expert Workshop (2013) recommendations with regard to activity-based
(bottom-up) ship emissions estimation. These recommendations were:
• to consider direct vessel observations to the greatest extent possible;
• to use vessel-specific activity and technical details in a bottom-up inventory model;
• to use “to the best extent possible” actual vessel speed to obtain engine loads.
The IMO Expert Workshop recognized that “bottom-up estimates are far more detailed and are generally based
on ship activity levels by calculating the fuel consumption and emissions from individual ship movements”
and that “a more sophisticated bottom-up approach to develop emission estimates on a ship-by-ship basis”
would “require significant data to be inputted and may require additional time […] to complete”.
Structure of the study and scope of work
The Third IMO GHG Study 2014 report follows the structure of the terms of reference for the work, which
comprise three main sections:
Section 1: Inventories of CO
2
emissions from international shipping 2007–2012
This section deploys both a top-down (2007–2011) and a bottom-up (2007–2012) analysis of CO
2
emissions
from international shipping. The inventories are analysed and discussed with respect to the quality of methods
and data and to uncertainty of results. The discrepancies between the bottom-up and top-down inventories
are discussed. The Third IMO GHG Study 2014 inventory for 2007 is compared to the Second IMO GHG
Study 2009 inventory for the same year.
Section 2: Inventories of emissions of GHGs and other relevant substances from international
shipping 2007–2012
This section applies the top-down (2007–2011) and bottom-up (2007–2012) analysis from Section 1 in
combination with data describing the emissions factors and calculations inventories for non-CO
2
GHGs –
methane (CH
4
), nitrous oxide (N
2
O), HFCs and sulphur hexafluoride (SF
6
) – and relevant substances – oxides
of sulphur (SOx), oxides of nitrogen (NO
x
), particulate matter (PM), carbon monoxide (CO) and NMVOCs.
The quality of methods and data and uncertainty of the inventory results are discussed, and comparisons are
made between the top-down and bottom-up estimates in the Third IMO GHG Study 2014 and the results of
the Second IMO GHG Study 2009.
Section 3: Scenarios for shipping emissions 2012–2050
This section develops scenarios for future emissions for all GHGs and other relevant substances investigated
in Sections 1 and 2. Results reflect the incorporation of new base scenarios used in GHG projections for
non-shipping sectors and method advances, and incorporate fleet activity and emissions insights emerging
from the 2007–2012 estimates. Drivers of emissions trajectories are evaluated and sources of uncertainty in
the scenarios are discussed.
6 Third IMO GHG Study 2014
Summary of Section 1: Inventories of CO
2
emissions from international shipping
2007–2012
2012 fuel consumption and CO
2
emissions by ship type
Figure 1 presents the CO
2
emissions by ship type for 2012, calculated using the bottom-up method. Equivalent
ship-type-specific results cannot be presented for the top-down method because the reported marine fuel
sales statistics are only available in three categories: international, domestic and fishing.
Figure 1: Bottom-up CO
2
emissions from international shipping by ship type 2012
Figure 2 shows the relative fuel consumption among vessel types in 2012 (both international and domestic
shipping), estimated using the bottom-up method. The figure also identifies the relative fuel consumption of
the main engine (predominantly for propulsion purposes), auxiliary engine (normally for electricity generation)
and the boilers (for steam generation). The total shipping fuel consumption is shown in 2012 to be dominated
by three ship types: oil tankers, bulk carriers and container ships. In each of those ship types, the main engine
consumes the majority of the fuel.
Figure 2: Summary graph of annual fuel consumption broken down by ship type
and machinery component (main, auxiliary and boiler) 2012
Executive Summary 7
Figure 2 shows the relative fuel consumption among vessel types in 2012 (both international and domestic
shipping), estimated using the bottom-up method. The figure also identifies the relative fuel consumption of
the main engine (predominantly for propulsion purposes), auxiliary engine (normally for electricity generation)
and the boilers (for steam generation). The total shipping fuel consumption is shown in 2012 to be dominated
by three ship types: oil tankers, bulk carriers and container ships. In each of those ship types, the main engine
consumes the majority of the fuel.
Figure 2: Summary graph of annual fuel consumption broken down by ship type
and machinery component (main, auxiliary and boiler) 2012
8 Third IMO GHG Study 2014
2007–2012 fuel consumption by bottom-up and top-down methods:
Third IMO GHG Study 2014 and Second IMO GHG Study 2009
Figure 3 shows the year-on-year trends for the total CO
2
emissions of each ship type, as estimated using the
bottom-up method. Figure 4 and Figure 5 show the associated total fuel consumption estimates for all years
of the study, from both the top-down and bottom-up methods. The total CO
2
emissions aggregated to the
lowest level of detail in the top-down analysis (international, domestic and fishing) are presented in Table 2
and Table 3.
Figure 3 presents results from the Third IMO GHG Study 2014 (all years). Figure 4 presents results from
both the Third IMO GHG Study 2014 (all years) and the Second IMO GHG Study 2009 (2007 results only).
The comparison of the estimates in 2007 shows that using both the top-down and the bottom-up analysis
methods, the results of the Third IMO GHG Study 2014 for the total fuel inventory and the international
shipping estimate are in close agreement with the findings from the Second IMO GHG Study 2009. Further
analysis and discussion of the comparison between the two studies is undertaken in Section 1.6 of this report.
Figure 3: CO
2
emissions by ship type (international shipping only) calculated using the bottom-up
method for all years (2007–2012)
In Figure 4 the vertical bar attached to the total fuel consumption estimate for each year and each method
represents the uncertainty in the estimates. For the bottom-up method, this error bar is derived from a Monte
Carlo simulation of the most important input parameters to the calculation. The most important sources of
uncertainty in the bottom-up method results are the number of days a ship spends at sea per year (attributable
to incomplete AIS coverage of a ship’s activity) and the number of ships that are active (in service) in a given
year (attributable to the discrepancy between the difference between the number of ships observed in the AIS
data and the number of ships described as in service in the IHSF database). The top-down estimates are also
uncertain, including observed discrepancies between global imports and exports of fuel oil and distillate oil,
observed transfer discrepancies among fuel products that can be blended into marine fuels, and potential for
misallocation of fuels between sectors of shipping (international, domestic and fishing). Neither the top-down
nor the bottom-up uncertainties are symmetric, showing that uncertainty in the top-down best estimate is
more likely to increase the estimate of fuel consumption from the best estimate, and that uncertainty in the
bottom-up best-estimate value is more likely to lower estimated values from the best estimate.
Differences between the bottom-up and the top-down best-estimate values in this study are consistent with
the differences observed in the Second IMO GHG Study 2009. This convergence of best estimates is important
because, in conjunction with the quality (Section 1.4) and uncertainty (Section 1.5) analyses, it provides
evidence that increasing confidence can be placed in both analytical approaches.
Figure 4: Summary graph of annual fuel use by all ships, estimated using the top-down and bottom-up
methods, showing Second IMO GHG Study 2009 estimates and uncertainty ranges
Figure 5: Summary graph of annual fuel use by international shipping, estimated using the top-down
and bottom-up methods, showing Second IMO GHG Study 2009 estimates and uncertainty ranges
Executive Summary 9
because, in conjunction with the quality (Section 1.4) and uncertainty (Section 1.5) analyses, it provides
evidence that increasing confidence can be placed in both analytical approaches.
Figure 4: Summary graph of annual fuel use by all ships, estimated using the top-down and bottom-up
methods, showing Second IMO GHG Study 2009 estimates and uncertainty ranges
Figure 5: Summary graph of annual fuel use by international shipping, estimated using the top-down
and bottom-up methods, showing Second IMO GHG Study 2009 estimates and uncertainty ranges
10 Third IMO GHG Study 2014
Table 2 – International, domestic and fishing CO
2
emissions 2007–2011, using top-down method
Marine sector Fuel type 2007 2008 2009 2010 2011
International shipping HFO 542.1 551.2 516.6 557.1 554.0
MDO 83.4 72.8 79.8 90.4 94.9
LNG 0.0 0.0 0.0 0.0 0.0
Top-down international total All 625.5 624.0 596.4 6 47. 5 648.9
Domestic navigation HFO 62.0 44.2 47. 6 44.5 39.5
MDO 72.8 76.6 75.7 82.4 87.8
LNG 0.1 0.1 0.1 0.1 0.2
Top-down domestic total All 134.9 121.0 123.4 127.1 127. 6
Fishing HFO 3.4 3.4 3.1 2.5 2.5
MDO 17. 3 15.7 16.0 16.7 16.4
LNG 0.1 0.1 0.1 0.1 0.1
Top-down fishing total All 20.8 19.2 19.3 19.2 19.0
Total CO
2
emissions 781.2 764.1 739.1 793.8 795.4
Table 3 – International, domestic and fishing CO
2
emissions 2007–2012, using bottom-up method
Marine sector Fuel type 2007 2008 2009 2010 2011 2012
International shipping HFO 773.8 802.7 736.6 650.6 716.9 6 67.9
MDO 97. 2 102.9 104.2 102.2 109.8 105.2
LNG 13.9 15.4 14.2 18.6 22.8 22.6
Bottom-up international total All 884.9 920.9 855.1 771.4 849.5 795.7
Domestic navigation HFO 53.8 57.4 32.5 45.1 61.7 39.9
MDO 142.7 138.8 80.1 88.2 98.1 91.6
LNG 0 0 0 0 0 0
Bottom-up domestic total All 196.5 196.2 112.6 133.3 159.7 131.4
Fishing HFO 1.6 1.5 0.9 0.8 1.4 1.1
MDO 17.0 16.4 9.3 9.2 10.9 9.9
LNG 0 0 0 0 0 0
Bottom-up fishing total All 18.6 18.0 10.2 10.0 12.3 11.0
Total CO
2
emissions 1,100.1 1,135.1 977.9 914.7 1,021.6 938.1
The fuel split between residual (HFO) and distillate (MDO) for the top-down approach is explicit in the
fuel sales statistics from IEA. However, the HFO/MDO allocation for the bottom-up inventory could not be
finalized without considering the top-down sales insights. This is because the engine-specific data available
through IHSF are too sparse, incomplete or ambiguous with respect to fuel type for large numbers of main
engines and nearly all auxiliary engines on vessels. QA/QC analysis with regard to fuel type assignment in the
bottom-up model was performed using top-down statistics as a guide, along with fuel allocation information
from the Second IMO GHG Study 2009. This iteration was important in order to finalize the QA/QC on fuel-
determined pollutant emissions (primarily SO
x
) and resulted in slight QA/QC adjustments for other emissions.
In addition to the uncertainties behind the total shipping emissions and fuel type allocations in each year, both
methods contain separate but important uncertainty about the allocation of fuel consumption and emissions
to international and domestic shipping. Where international shipping is defined as shipping between ports of
different countries, and one tank of fuel is used for multiple voyages, there is an intrinsic shortcoming in the
top-down method. More specifically, fuel can be sold to a ship engaged in both domestic and international
voyages but only one identifier (international or domestic) can be assigned to the report of fuel sold. Using the
bottom-up method, while location information is available, the AIS coverage is not consistently high enough
to be able to resolve voyage-by-voyage detail. Section 1.2 discusses possible alternative approaches to the
classification of international and domestic fuel consumption using the bottom up method and the selection
of definition according to ship type and size category.
Particular care must be taken when interpreting the domestic fuel consumption and emissions estimates
from both the top-down and the bottom-up methods. Depending on where the fuel for domestic shipping
and fishing is bought, it may or may not be adequately captured in the IEA marine bunkers. For example,
inland or leisure and fishing vessels may purchase fuel at locations where fuel is also sold to other sectors
of the economy and therefore it may be misallocated. In the bottom-up method, fuel consumption is only
included for ships that appear in the IHSF database (and have an IMO number). While this should cover
all international shipping, many domestic vessels (inland, fishing or cabotage) may not be included in this
database. An indication of the number of vessels excluded from the bottom-up method was obtained from
the count of MMSI numbers observed on the AIS for which no match with the IHSF database was obtained.
The implications of this count for both the bottom-up and top-down analyses are discussed in Section 1.4.
2007–2012 trends in CO
2
emissions and drivers of emissions
Figure 6, Figure 7 and Figure 8 present indexed time series of the total CO
2
emissions during the period studied
for three ship types: oil tankers, container ships and bulk carriers (all in-service ships). The figures also present
several key drivers of CO
2
emissions that can be used to decompose the fleet, activity and CO
2
emission
trends, estimated using the bottom-up method. All trends are indexed to their values in 2007. Despite rising
transport demand in all three fleets, each fleet’s total emissions are shown either to remain approximately
constant or to decrease slightly.
The contrast between the three plots in Figures 6–8 shows that these three sectors of the shipping industry
have experienced different changes over the period 2007–2012. The oil tanker sector has reduced its emissions
by a total of 20%. During the same period the dry bulk and container ship sectors also saw absolute emissions
reductions but by smaller amounts. All ship types experienced similar reductions in average annual fuel
consumption but differences in the number of ships in service, which explains the difference in fleet total CO
2
emissions trends. The reduction in average days at sea during the period studied is greatest in the dry bulk
fleet, while the container ship fleet has seen a slight increase. Consistent with the results presented in Table
4, container ships adopted slow steaming more than any other ship type. So, over the same period of time,
similar reductions in average fuel consumption per ship have come about through different combinations of
slow steaming and days at sea.
Figure 6: Time series for trends in emissions and drivers of emissions in the oil tanker fleet 2007–2012.
All trends are indexed to their values in 2007
Executive Summary 11
Particular care must be taken when interpreting the domestic fuel consumption and emissions estimates
from both the top-down and the bottom-up methods. Depending on where the fuel for domestic shipping
and fishing is bought, it may or may not be adequately captured in the IEA marine bunkers. For example,
inland or leisure and fishing vessels may purchase fuel at locations where fuel is also sold to other sectors
of the economy and therefore it may be misallocated. In the bottom-up method, fuel consumption is only
included for ships that appear in the IHSF database (and have an IMO number). While this should cover
all international shipping, many domestic vessels (inland, fishing or cabotage) may not be included in this
database. An indication of the number of vessels excluded from the bottom-up method was obtained from
the count of MMSI numbers observed on the AIS for which no match with the IHSF database was obtained.
The implications of this count for both the bottom-up and top-down analyses are discussed in Section 1.4.
2007–2012 trends in CO
2
emissions and drivers of emissions
Figure 6, Figure 7 and Figure 8 present indexed time series of the total CO
2
emissions during the period studied
for three ship types: oil tankers, container ships and bulk carriers (all in-service ships). The figures also present
several key drivers of CO
2
emissions that can be used to decompose the fleet, activity and CO
2
emission
trends, estimated using the bottom-up method. All trends are indexed to their values in 2007. Despite rising
transport demand in all three fleets, each fleet’s total emissions are shown either to remain approximately
constant or to decrease slightly.
The contrast between the three plots in Figures 6–8 shows that these three sectors of the shipping industry
have experienced different changes over the period 2007–2012. The oil tanker sector has reduced its emissions
by a total of 20%. During the same period the dry bulk and container ship sectors also saw absolute emissions
reductions but by smaller amounts. All ship types experienced similar reductions in average annual fuel
consumption but differences in the number of ships in service, which explains the difference in fleet total CO
2
emissions trends. The reduction in average days at sea during the period studied is greatest in the dry bulk
fleet, while the container ship fleet has seen a slight increase. Consistent with the results presented in Table
4, container ships adopted slow steaming more than any other ship type. So, over the same period of time,
similar reductions in average fuel consumption per ship have come about through different combinations of
slow steaming and days at sea.
Figure 6: Time series for trends in emissions and drivers of emissions in the oil tanker fleet 2007–2012.
All trends are indexed to their values in 2007
12 Third IMO GHG Study 2014
Figure 7: Time series for trends in emissions and drivers of emissions in the container ship fleet 2007–2012.
All trends are indexed to their values in 2007
Figure 8: Time series for trends in emissions and drivers of emissions in the bulk carrier fleet 2007–2012.
All trends are indexed to their values in 2007
Note: Further data on historical trends and relationship between transport supply and demand can be found in the Second
IMO GHG Study 2009.
The bottom-up method constructs the calculations of ship type and size totals from calculations for the fuel
consumption of each individual in-service ship in the fleet. The method allows quantification of both the
variability within a fleet and the influence of slow steaming. Across all ship types and sizes, the average ratio of
operating speed to design speed was 0.85 in 2007 and 0.75 in 2012. In relative terms, ships have slowed down
in line with the reported widespread adoption of slow steaming, which began after the financial crisis. The
consequence of this observed slow steaming is a reduction in daily fuel consumption of approximately 27%,
expressed as an average across all ship types and sizes. However, that average value belies the significant
operational changes that have occurred in certain ship type and size categories. Table 4 describes, for three of
the ship types studied, the ratio between slow steaming percentage (average at-sea operating speed expressed
as a percentage of design speed), the average at-sea main engine load factor (a percentage of the total installed
power produced by the main engine) and the average at-sea main engine daily fuel consumption. Many of the
Executive Summary 13
larger ship sizes in all three categories are estimated to have experienced reductions in daily fuel consumption
in excess of the average value for all shipping of 27%.
Table 4 also shows that the ships with the highest design speeds (container ships) have adopted the greatest
levels of slow steaming (in many cases operating at average speeds that are 60–70% of their design speeds),
relative to oil tankers and bulk carriers. Referring back to Figure 8, it can be seen that for bulk carriers, the
observed trend in slow steaming is not concurrent with the technical specifications of the ships remaining
constant. For example, the largest bulk carriers (200,000+ dwt capacity) saw increases in average size (dwt
capacity) as well as increased installed power (from an average of 18.9 MW to 22.2 MW), as a result of a large
number of new ships entering the fleet over the period studied. (The fleet grew from 102 ships in 2007 to 294
ships in 2012.)
The analysis of trends in speed and days at sea is consistent with the findings in Section 3 that the global
fleet is currently at or near the historic low in terms of productivity (transport work per unit of capacity).
The consequence is that these (and many other) sectors of the shipping industry represent latent emissions
increases, because the fundamentals (number of ships in service) have seen upward trends that have been
offset as economic pressures act to reduce productivity (which in turn reduces emissions intensity). Whether
and when the latent emissions may appear is uncertain, as it depends on the future market dynamics of the
industry. However, the risk is high that the fleet could encounter conditions favouring the conversion of latent
emissions to actual emissions; this could mean that shipping reverts to the trajectory estimated in the Second
IMO GHG Study 2009. This upward potential is quantified as part of sensitivity analysis in Section 3.
A reduction in speed and the associated reduction in fuel consumption do not relate to an equivalent
percentage increase in efficiency, because a greater number of ships (or more days at sea) are required to do
the same amount of transport work. This relationship is discussed in greater detail in Section 3.
14 Third IMO GHG Study 2014
Table 4 – Relationship between slow steaming, engine load factor (power output) and fuel consumption for 2007 and 2012
Ship type Size category Units
2007 2012
% change in
average at-sea
tonnes per day
(tpd) 2007–2012
Ratio of average
at-sea speed to
design speed
Average at-sea
main engine
load factor
(% MCR)
At-sea
consumption in
tonnes per day
(tpd)
Ratio of average
at-sea speed to
design speed
Average at-sea
main engine
load factor
(% MCR)
At-sea
consumption in
tonnes per day
(tpd)
Bulk carrier 0–9,999 dwt 0.92 92% 7. 0 0.84 70% 5.5
-24%
10,000–34,999 0.86 68% 22.2 0.82 59% 17.6
-23%
35,000–59,999 0.88 73% 29.0 0.82 58% 23.4
-21%
60,000–99,999 0.90 78% 37.7 0.83 60% 28.8
-27%
100,000–199,999 0.89 77% 55.5 0.81 57% 42.3
-27%
200,000–+
0.82 66% 51.2 0.84 62% 56.3 10%
Container 0–999 TEU 0.82 62% 17. 5 0.77 52% 14.4
-19%
1,000 –1,999 0.80 58% 33.8 0.73 45% 26.0
-26%
2,000–2,999 0.80 58% 55.9 0.70 39% 38.5
-37%
3,000–4,999 0.80 59% 90.4 0.68 36% 58.7
-42%
5,000–7,999 0.82 63% 151.7 0.65 32% 79.3
-63%
8,000–11,999 0.85 69% 200.0 0.65 32% 95.6
-71%
12,000–14,500 0.84 67% 231.7 0.66 34% 107. 8
-73%
14,500 –+
– – – 0.60 28% 100.0 –
Oil tanker 0–4,999 dwt 0.89 85% 5.1 0.80 67% 4.3
-18%
5,000–9,999 0.83 64% 9.2 0.75 49% 7.1
-26%
10,000–19,999 0.81 61% 15.3 0.76 49% 10.8
-34%
20,000–59,999 0.87 72% 28.8 0.80 55% 22.2
-26%
60,000–79,999 0.91 83% 45.0 0.81 57% 31.4
-35%
80,000–119,999 0.91 81% 49.2 0.78 51% 31.5
-44%
120,000–199,999 0.92 83% 65.4 0.77 49% 39.4
-50%
200,000–+
0.95 90% 103.2 0.80 54% 65.2
-45%
Summary of Section 2: Inventories of emissions of GHGs and other relevant
substances from international shipping 2007–2012
All data are calculated using the bottom-up method and the results of this study are compared with the Second
IMO GHG Study 2009 results in Figure 9 (all shipping). Figure 10 (international, domestic and fishing) presents
the time series of GHGs and other relevant substance emissions over the period of this study (2007–2012).
Calculations performed using the top-down method are presented in Section 2.3.
The trends are generally well correlated with the time series trend of CO
2
emissions totals, which is in turn
well correlated to fuel consumption. A notable exception is the trend in CH
4
emissions, which is dominated
by the increase in LNG fuel consumption in the LNG tanker fleet (related to increases in fleet size and activity)
during the years 2007–2012.
Agreements with the Second IMO GHG Study 2009 estimates are generally good, although there are some
differences, predominantly related to the emissions factors used in the respective studies and how they have
been applied. The Second IMO GHG Study 2009 estimated CH
4
emissions from engine combustion to be
approximately 100,000 tonnes in the year 2007.
Executive Summary 15
Summary of Section 2: Inventories of emissions of GHGs and other relevant
substances from international shipping 2007–2012
All data are calculated using the bottom-up method and the results of this study are compared with the Second
IMO GHG Study 2009 results in Figure 9 (all shipping). Figure 10 (international, domestic and fishing) presents
the time series of GHGs and other relevant substance emissions over the period of this study (2007–2012).
Calculations performed using the top-down method are presented in Section 2.3.
The trends are generally well correlated with the time series trend of CO
2
emissions totals, which is in turn
well correlated to fuel consumption. A notable exception is the trend in CH
4
emissions, which is dominated
by the increase in LNG fuel consumption in the LNG tanker fleet (related to increases in fleet size and activity)
during the years 2007–2012.
Agreements with the Second IMO GHG Study 2009 estimates are generally good, although there are some
differences, predominantly related to the emissions factors used in the respective studies and how they have
been applied. The Second IMO GHG Study 2009 estimated CH
4
emissions from engine combustion to be
approximately 100,000 tonnes in the year 2007.
16 Third IMO GHG Study 2014
a. CO
2
b. CH
4
c. N
2
O d. SO
x
e. NO
x
f. PM
g. CO h. NMVOC
Figure 9: Time series of bottom-up results for GHGs and other substances (all shipping). The green bar
represents the Second IMO GHG Study 2009 estimate
a. CO
2
b. CH
4
c. N
2
O d. SO
x
e. NO
x
f. PM
g. CO h. NMVOC
Figure 10: Time series of bottom-up results for GHGs and other substances (international shipping, domestic
navigation and fishing)
Executive Summary 17
a. CO
2
b. CH
4
c. N
2
O d. SO
x
e. NO
x
f. PM
g. CO h. NMVOC
Figure 10: Time series of bottom-up results for GHGs and other substances (international shipping, domestic
navigation and fishing)
18 Third IMO GHG Study 2014
Summary of Section 3: Scenarios for shipping emissions 2012–2050
Shipping projection scenarios are based on the Representative Concentration Pathways (RCPs) for future
demand of coal and oil transport and Shared Socioeconomic Pathways (SSPs) for future economic growth. SSPs
have been combined with RCPs to develop four internally consistent scenarios of maritime transport demand.
These are BAU scenarios, in the sense that they assume that the current policies on the energy efficiency and
emissions of ships remain in force, and that no increased stringencies or additional policies will be introduced.
In line with common practice in climate research and assessment, there are multiple BAU scenarios to reflect
the inherent uncertainty in projecting economic growth, demographics and the development of technology.
In addition, for each of the BAU scenarios, this study developed three policy scenarios that have increased
action on either energy efficiency or emissions or both. Hence, there are two fuel-mix/ECA scenarios: one
keeps the share of fuel used in ECAs constant over time and has a slow penetration of LNG in the fuel mix; the
other projects a doubling of the amount of fuel used in ECAs and has a higher share of LNG in the fuel mix.
Moreover, two efficiency trajectories are modelled: the first assumes an ongoing effort to increase the fuel
efficiency of new and existing ships, resulting in a 60% improvement over the 2012 fleet average by 2050; the
second assumes a 40% improvement by 2050. In total, emissions are projected for 16 scenarios.
Maritime transport demand projections
The projections of demand for international maritime transport show a rapid increase in demand for unitized
cargo transport, as it is strongly coupled to GDP and statistical analyses show no sign of demand saturation.
The increase is largest in the SSP that projects the largest increase of global GDP (SSP5) and relatively more
modest in the SSP with the lowest increase (SSP3). Non-coal dry bulk is a more mature market where an
increase in GDP results in a modest increase in transport demand.
Figure 11: Historical data to 2012 on global transport work for non-coal combined bulk dry cargoes
and other dry cargoes (billion tonne-miles) coupled to projections driven by GDPs from SSP1
through to SSP5 by 2050
Demand for coal and oil transport has historically been strongly linked to GDP. However, because of climate
policies resulting in a global energy transition, the correlation may break down. Energy transport demand
projections are based on projections of energy demand in the RCPs. The demand for transport of fossil fuels
is projected to decrease in RCPs that result in modest global average temperature increases (e.g. RCP2.6) and
to continue to increase in RCPs that result in significant global warming (e.g. RCP8.5).
Figure 12: Historical data to 2012 on global transport work for ship-transported coal and
liquid fossil fuels (billion tonne-miles) coupled to projections of coal and energy demand
driven by RCPs 2.6, 4.5, 6.0 and 8.5 by 2050
Executive Summary 19
Demand for coal and oil transport has historically been strongly linked to GDP. However, because of climate
policies resulting in a global energy transition, the correlation may break down. Energy transport demand
projections are based on projections of energy demand in the RCPs. The demand for transport of fossil fuels
is projected to decrease in RCPs that result in modest global average temperature increases (e.g. RCP2.6) and
to continue to increase in RCPs that result in significant global warming (e.g. RCP8.5).
Figure 12: Historical data to 2012 on global transport work for ship-transported coal and
liquid fossil fuels (billion tonne-miles) coupled to projections of coal and energy demand
driven by RCPs 2.6, 4.5, 6.0 and 8.5 by 2050
20 Third IMO GHG Study 2014
Maritime emissions projections
Maritime CO
2
emissions are projected to increase significantly. Depending on future economic and energy
developments, our four BAU scenarios project an increase of between 50% and 250% in the period up to
2050 (see Figure 13). Further action on efficiency and emissions could mitigate emissions growth, although all
but one scenarios project emissions in 2050 to be higher than in 2012, as shown in Figure 14.
Figure 13: BAU projections of CO
2
emissions from international maritime transport 2012–2050
Figure 14: Projections of CO
2
emissions from international maritime transport. Bold lines are BAU scenarios.
Thin lines represent either greater efficiency improvement than BAU or
additional emissions controls or both
Figure 15 shows the impact of market-driven or regulatory improvements in efficiency contrasted with
scenarios that have a larger share of LNG in the fuel mix. These four emissions projections are based on the
same transport demand projections. The two lower projections assume an efficiency improvement of 60%
instead of 40% over 2012 fleet average levels in 2050. The first and third projections have a 25% share of LNG
in the fuel mix in 2050 instead of 8%. Under these assumptions, improvements in efficiency have a larger
impact on emissions trajectories than changes in the fuel mix.
Figure 15: Projections of CO
2
emissions from international maritime transport under the
same demand projections. Larger improvements in efficiency have a higher impact on
CO
2
emissions than a larger share of LNG in the fuel mix
Table 5 shows the projection of the emissions of other substances. For each year, the median (minimum–
maximum) emissions are expressed as a share of their 2012 emissions. Most emissions increase in parallel with
CO
2
and fuel, with some notable exceptions. Methane emissions are projected to increase rapidly (albeit from
a very low base) as the share of LNG in the fuel mix increases. Emissions of sulphurous oxides, nitrogen oxides
and particulate matter increase at a lower rate than CO
2
emissions. This is driven by MARPOL Annex VI
requirements on the sulphur content of fuels (which also impact PM emissions) and the NO
x
technical code. In
scenarios that assume an increase in the share of fuel used in ECAs, the impact of these regulations is stronger.
Executive Summary 21
Figure 15 shows the impact of market-driven or regulatory improvements in efficiency contrasted with
scenarios that have a larger share of LNG in the fuel mix. These four emissions projections are based on the
same transport demand projections. The two lower projections assume an efficiency improvement of 60%
instead of 40% over 2012 fleet average levels in 2050. The first and third projections have a 25% share of LNG
in the fuel mix in 2050 instead of 8%. Under these assumptions, improvements in efficiency have a larger
impact on emissions trajectories than changes in the fuel mix.
Figure 15: Projections of CO
2
emissions from international maritime transport under the
same demand projections. Larger improvements in efficiency have a higher impact on
CO
2
emissions than a larger share of LNG in the fuel mix
Table 5 shows the projection of the emissions of other substances. For each year, the median (minimum–
maximum) emissions are expressed as a share of their 2012 emissions. Most emissions increase in parallel with
CO
2
and fuel, with some notable exceptions. Methane emissions are projected to increase rapidly (albeit from
a very low base) as the share of LNG in the fuel mix increases. Emissions of sulphurous oxides, nitrogen oxides
and particulate matter increase at a lower rate than CO
2
emissions. This is driven by MARPOL Annex VI
requirements on the sulphur content of fuels (which also impact PM emissions) and the NO
x
technical code. In
scenarios that assume an increase in the share of fuel used in ECAs, the impact of these regulations is stronger.
22 Third IMO GHG Study 2014
Table 5 – Summary of the scenarios for future emissions from international shipping, GHGs and
other relevant substances
Scenario
2012 2020 2050
index (2012 = 100) index (2012 = 100) index (2012 = 100)
Greenhouse
gases
CO
2
low LNG 100
108 (107 - 112) 183 (105 - 347)
high LNG 100
106 (105 - 109) 173 (99 - 328)
CH
4
low LNG 100
1.600 (1.600 - 1.700) 10.500 (6.000 - 20.000)
high LNG 100
7. 55 0 ( 7. 5 0 0 - 7.9 0 0) 32.000 (19.000 - 61.000)
N
2
O low LNG 100
108 (107 - 112) 181 (104 - 345)
high LNG 100
105 (104 - 109) 168 (97 - 319)
HFC 100
106 (105 - 108) 173 (109 - 302)
PFC
- - -
SF
6
- - -
Other
relevant
substances
NO
x
constant ECA 100
107 (106 - 110) 161 (93 - 306)
more ECAs 100
99 (98 - 103) 130 (75 - 247)
SO
x
constant ECA 100
64 (63 - 66) 30 (17 - 56)
more ECAs 100
55 (54 - 57) 19 (11 - 37)
PM constant ECA 100
77 (76 - 79) 84 (48 - 159)
more ECAs 100
65 (64 - 67) 56 (32 - 107)
NMVOC constant ECA 100
108 (107 - 112) 183 (105 - 348)
more ECAs 100
106 (105 - 110) 175 (101 - 333)
CO constant ECA 100
112 (111 - 115) 206 (119 - 392)
more ECAs 100
123 (122 - 127) 246 (142 - 468)
Note: Emissions of PFC and SF
6
from international shipping are insignificant.
Summary of the data and methods used (Sections 1, 2 and 3)
Key assumptions and method details
Assumptions are made in Sections 1, 2 and 3 for the best-estimate international shipping inventories and
scenarios. The assumptions are chosen on the basis of their transparency and connection to high-quality, peer-
reviewed sources. Further justification for each of these assumptions is presented and discussed in greater detail
in Sections 1.4 and 2.4. The testing of key assumptions consistently demonstrates that they are of high quality.
The uncertainty analysis in Section 1.5 examines variations in the key assumptions, in order to quantify the
consequences for the inventories. For future scenarios, assumptions are also tested through the deployment
of multiple scenarios to illustrate the sensitivities of trajectories of emissions to different assumptions. Key
assumptions made are that:
• the IEA data on marine fuel sales are representative of shipping’s fuel consumption;
• in 2007 and 2008, the number of days that a ship spends at sea per year can be approximated by the
associated ship-type- and size-specific days at sea given in the Second IMO GHG Study 2009 (for
the year 2007);
• in 2009, the number of days that a ship spends at sea per year can be approximated by a representative
sample of LRIT data (approximately 10% of the global fleet);
• in 2010–2012, the annual days at sea can be derived from a combined satellite and shore-based AIS
database;
• in all years, the time spent at different speeds can be estimated from AIS observations of ship activity,
even when only shore-based AIS data are available (2007–2009);
• in all years, the total number of active ships is represented by any ship defined as in service in the
IHSF database;
• ships observed in the AIS data that cannot be matched or identified in the IHSF data must be involved
in domestic shipping only;
• combinations of RCPs and SSPs can be used to derive scenarios for future transport demand of
shipping; and
• technologies that could conceivably reduce ship combustion emissions to zero (for GHGs and other
substances) will either not be available or not be deployed cost-effectively in the next 40 years on
both new and existing ships.
Inventory estimation methods overview (Sections 1 and 2)
Top-down and bottom-up methods provide two different and independent analysis tools for estimating
shipping emissions. Both methods are used in this study.
The top-down estimate mainly used data on marine bunker sales (divided into international, domestic and
fishing sales) from IEA. Data availability for 2007–2011 enabled top-down analysis of annual emissions for
these years. In addition to the marine bunker fuel sales data, historical IEA statistics were used to understand
and quantify the potential for misallocation in the statistics resulting in either under- or overestimations of
marine energy use and emissions.
The bottom-up estimate combined the global fleet technical data (from IHSF) with fleet activity data derived
from AIS observations. Estimates for individual ships in the IHSF database were aggregated by vessel category
to provide statistics describing activity, energy use and emissions for all ships for each of the years 2007–2012.
For each ship and each hour of that ship’s operation in a year, the bottom-up model relates speed and draught
to fuel consumption using equations similar to those deployed in the Second IMO GHG Study 2009 and
the wider naval architecture and marine engineering literature. Until the Third IMO GHG Study 2014, vessel
activity information was obtained from shore-based AIS receivers with limited temporal and geographical
coverage (typically a range of approximately 50nmi) and this information informed general fleet category
activity assumptions and average values. With low coverage comes high uncertainty about estimated activity
and, therefore, uncertainty in estimated emissions. To address these methodological shortcomings and
maximize the quality of the bottom-up method, the Third IMO GHG Study 2014 has accessed the most
globally representative set of vessel activity observations by combining AIS data from a variety of providers
(both shore-based and satellite-received data), shown in Figure 16.
The AIS data used in this study provide information for the bottom-up model describing a ship’s identity and
its hourly variations in speed, draught and location over the course of a year.
This work advances the activity-based modelling of global shipping by improving geographical and temporal
observation of ship activity, especially for recent years.
Table 6 – AIS observation statistics of the fleet identified in the IHSF database
as in service in 2007 and 2012
Total in-service ships Average % of in-service ships observed
on AIS (all ship types)
Average % of the hours in the year
that each ship is observed on AIS
(all ship types)
2007 51,818 76% 42%
2012 56,317 83% 71%
In terms of both space and time, the AIS data coverage is not consistent year-on-year during the period studied
(2007–2012). For the first three years (2007–2009), no satellite AIS data were available, only AIS data from
shore-based stations. This difference can be seen by contrasting the first (2007) and last (2012) years’ AIS
data sets, as depicted for their geographical coverage in Figure 16. Table 6 describes the observation statistics
(averages) for the different ship types. These data cannot reveal the related high variability in observation
depending on ship type and size. Larger oceangoing ships are observed very poorly in 2007 (10–15% of the
hours of the year) and these observations are biased towards the coastal region when the ships are either
moving slowly as they approach or leave ports, at anchor or at berth. Further details and implications of this
coverage for the estimate of shipping activity are discussed in greater detail in Sections 1.2, 1.4 and 1.5.
Executive Summary 23
• ships observed in the AIS data that cannot be matched or identified in the IHSF data must be involved
in domestic shipping only;
• combinations of RCPs and SSPs can be used to derive scenarios for future transport demand of
shipping; and
• technologies that could conceivably reduce ship combustion emissions to zero (for GHGs and other
substances) will either not be available or not be deployed cost-effectively in the next 40 years on
both new and existing ships.
Inventory estimation methods overview (Sections 1 and 2)
Top-down and bottom-up methods provide two different and independent analysis tools for estimating
shipping emissions. Both methods are used in this study.
The top-down estimate mainly used data on marine bunker sales (divided into international, domestic and
fishing sales) from IEA. Data availability for 2007–2011 enabled top-down analysis of annual emissions for
these years. In addition to the marine bunker fuel sales data, historical IEA statistics were used to understand
and quantify the potential for misallocation in the statistics resulting in either under- or overestimations of
marine energy use and emissions.
The bottom-up estimate combined the global fleet technical data (from IHSF) with fleet activity data derived
from AIS observations. Estimates for individual ships in the IHSF database were aggregated by vessel category
to provide statistics describing activity, energy use and emissions for all ships for each of the years 2007–2012.
For each ship and each hour of that ship’s operation in a year, the bottom-up model relates speed and draught
to fuel consumption using equations similar to those deployed in the Second IMO GHG Study 2009 and
the wider naval architecture and marine engineering literature. Until the Third IMO GHG Study 2014, vessel
activity information was obtained from shore-based AIS receivers with limited temporal and geographical
coverage (typically a range of approximately 50nmi) and this information informed general fleet category
activity assumptions and average values. With low coverage comes high uncertainty about estimated activity
and, therefore, uncertainty in estimated emissions. To address these methodological shortcomings and
maximize the quality of the bottom-up method, the Third IMO GHG Study 2014 has accessed the most
globally representative set of vessel activity observations by combining AIS data from a variety of providers
(both shore-based and satellite-received data), shown in Figure 16.
The AIS data used in this study provide information for the bottom-up model describing a ship’s identity and
its hourly variations in speed, draught and location over the course of a year.
This work advances the activity-based modelling of global shipping by improving geographical and temporal
observation of ship activity, especially for recent years.
Table 6 – AIS observation statistics of the fleet identified in the IHSF database
as in service in 2007 and 2012
Total in-service ships Average % of in-service ships observed
on AIS (all ship types)
Average % of the hours in the year
that each ship is observed on AIS
(all ship types)
2007 51,818 76% 42%
2012 56,317 83% 71%
In terms of both space and time, the AIS data coverage is not consistent year-on-year during the period studied
(2007–2012). For the first three years (2007–2009), no satellite AIS data were available, only AIS data from
shore-based stations. This difference can be seen by contrasting the first (2007) and last (2012) years’ AIS
data sets, as depicted for their geographical coverage in Figure 16. Table 6 describes the observation statistics
(averages) for the different ship types. These data cannot reveal the related high variability in observation
depending on ship type and size. Larger oceangoing ships are observed very poorly in 2007 (10–15% of the
hours of the year) and these observations are biased towards the coastal region when the ships are either
moving slowly as they approach or leave ports, at anchor or at berth. Further details and implications of this
coverage for the estimate of shipping activity are discussed in greater detail in Sections 1.2, 1.4 and 1.5.
24 Third IMO GHG Study 2014
Figure 16: Geographical coverage in 2007 (top) and 2012 (bottom), coloured according to the
intensity of messages received per unit area. This is a composite of both vessel activity and
geographical coverage; intensity is not solely indicative of vessel activity
AIS coverage, even in the best year, cannot obtain readings of vessel activity 100% of the time. This can be
due to disruption to satellite or shore-based reception of AIS messages, the nature of the satellite orbits and
interruption of a ship’s AIS transponder’s operation. For the time periods when a ship is not observed on
AIS, algorithms are deployed to estimate the unobserved activity. For 2010, 2011 and 2012, those algorithms
deploy heuristics developed from the observed fleet. However, with the low level of coverage in 2007, 2008
and 2009, the consortium had to use methods similar to previous studies that combined sparse AIS-derived
speed and vessel activity characteristics with days-at-sea assumptions. These assumptions were based on the
Second IMO GHG Study 2009 expert judgements. Conservatively, the number of total days at sea is held
constant for all three years (2007–2009) as no alternative, more reliable, source of data exists for these years.
Given the best available data, and by minimizing the amount of unobserved activity, uncertainties in both the
top-down and the bottom-up estimates of fuel consumption can be more directly quantified than previous
global ship inventories. For the bottom-up method, this study investigates these uncertainties in two ways:
1 The modelled activity and fuel consumption are validated against two independent data sources
(Section 1.4):
a LRIT data were obtained for approximately 8,000 ships and four years (2009–2012) and used to
validate both the observed and unobserved estimates of the time that a ship spends in different
modes (at sea, in port), as well as its speeds.
b Noon report data were collected for 470 ships for the period 2007–2012 (data for all ships were
available in 2012, with fewer ships’ data available in earlier years). The data were used to validate
both the observed and unobserved activity estimates and the associated fuel consumption.
2 The comparison between the modelled data and the validation data samples enabled the uncertainty
in the model to be broken down and discussed in detail. An analysis was undertaken to quantify the
different uncertainties and their influence on the accuracy of the estimation of a ship’s emissions in a
given hour and a given year, and the emissions of a fleet of similar ships in a given year.
Figure 17 presents the comparison of bottom-up and noon report data used in the validation process of 2012
analysis (further plots and years of data are included in Section 1.4). For each comparison, a ship is identified
by its IMO number in the two data sets so that the corresponding quarterly noon report and bottom-up
model output can be matched. The red line represents an ideal match (equal values) between the bottom-up
and noon-report outputs, the solid black line the best fit through the data and the dotted black lines the 95%
confidence bounds on the best fit. The “x” symbols represent individual ships, coloured according to the ship
type category as listed in the legend.
The comparative analysis demonstrates that there is a consistent and robust agreement between the bottom-up
method and the noon report data at three important stages of the modelling:
1 The average at-sea speed plot demonstrates that, in combination with high coverage AIS data, the
extrapolation algorithm estimates key activity parameters (e.g. speed) with high reliability.
2 The average daily fuel consumption plot demonstrates the reliability of the marine engineering and
naval architecture relationships and assumptions used in the model to convert activity into power and
fuel consumption.
3 The total quarterly fuel consumption plot demonstrates that the activity data (including days at
sea) and the engineering assumptions combine to produce generally reliable estimates of total fuel
consumption. The underestimate in the daily fuel consumption of the largest container ships can also
be seen in this total quarterly fuel consumption.
Figure 17: Total noon-reported quarterly fuel consumption of the main engine, compared with the
bottom-up estimate over each quarter of 2012, with a filter to select only days with high reliability
observations of the ship for 75% of the time or more
Executive Summary 25
b Noon report data were collected for 470 ships for the period 2007–2012 (data for all ships were
available in 2012, with fewer ships’ data available in earlier years). The data were used to validate
both the observed and unobserved activity estimates and the associated fuel consumption.
2 The comparison between the modelled data and the validation data samples enabled the uncertainty
in the model to be broken down and discussed in detail. An analysis was undertaken to quantify the
different uncertainties and their influence on the accuracy of the estimation of a ship’s emissions in a
given hour and a given year, and the emissions of a fleet of similar ships in a given year.
Figure 17 presents the comparison of bottom-up and noon report data used in the validation process of 2012
analysis (further plots and years of data are included in Section 1.4). For each comparison, a ship is identified
by its IMO number in the two data sets so that the corresponding quarterly noon report and bottom-up
model output can be matched. The red line represents an ideal match (equal values) between the bottom-up
and noon-report outputs, the solid black line the best fit through the data and the dotted black lines the 95%
confidence bounds on the best fit. The “x” symbols represent individual ships, coloured according to the ship
type category as listed in the legend.
The comparative analysis demonstrates that there is a consistent and robust agreement between the bottom-up
method and the noon report data at three important stages of the modelling:
1 The average at-sea speed plot demonstrates that, in combination with high coverage AIS data, the
extrapolation algorithm estimates key activity parameters (e.g. speed) with high reliability.
2 The average daily fuel consumption plot demonstrates the reliability of the marine engineering and
naval architecture relationships and assumptions used in the model to convert activity into power and
fuel consumption.
3 The total quarterly fuel consumption plot demonstrates that the activity data (including days at
sea) and the engineering assumptions combine to produce generally reliable estimates of total fuel
consumption. The underestimate in the daily fuel consumption of the largest container ships can also
be seen in this total quarterly fuel consumption.
Figure 17: Total noon-reported quarterly fuel consumption of the main engine, compared with the
bottom-up estimate over each quarter of 2012, with a filter to select only days with high reliability
observations of the ship for 75% of the time or more
26 Third IMO GHG Study 2014
Scenario estimation method overview (Section 3)
The consortium developed emissions projections by modelling the international maritime transport demand
and allocating it to ships, projecting regulation- and market-driven energy efficiency changes for each ship.
These are combined with fuel-mix scenarios and projections for the amount of fuel used by international
maritime transport. For most emissions, the energy demand is then multiplied by an emissions factor to arrive
at an emissions projection.
The basis for the transport demand projections is a combination of RCPs and SSPs that have been developed
for IPCC. The RCPs contain detailed projections about energy sources, which is relevant for fossil-fuel transport
projections. The SSPs contain long-term projections of demographic and economic trends, which are relevant
for the projections of demand for transport of non-energy cargoes. RCPs and SSPs are widely used across the
climate community.
The long-term projections are combined with a statistical analysis of historical relationships between changes
in transport demand, economic growth and fossil-fuel consumption.
The energy efficiency improvement projections are part regulation-driven, part market-driven. The relevant
regulations are EEDI for new ships and SEEMP for all ships. Market driven efficiency improvements have been
calculated using MACCs.
1
Inventories of CO
2
emissions from
international shipping 2007–2012
1.1 Top-down CO
2
inventory calculation method
1.1.1 Introduction
Section 1.1 provides a top-down estimate of emissions from shipping for the period 2007–2012. This task
also provides a comparison of this update with the methods used in the Second IMO GHG Study 2009. The
top-down approach is based on statistical data derived from fuel delivery reports to internationally registered
vessels. The top-down approach also considers allocation to domestic and international shipping, as reported
in national statistics.
Calculations of emissions using top-down fuel consumption estimates are presented. For CO
2
, these estimates
use CO
2
emissions factors consistent with those used in the bottom-up calculations in Section 1.2. Specifically,
the top-down inventory uses the CO
2
emissions factors reported in Section 2.2.7. For marine fuel oil (HFO),
this study uses 3.114 grams CO
2
per gram fuel; for marine gas oil (MDO), this study uses 3.206 grams CO
2
per
gram fuel; and for natural gas (LNG), this study uses 2.75 grams CO
2
per gram fuel.
1.1.2 Methods for review of IEA data
The World Energy Statistics published by IEA are used both in the Second IMO GHG Study 2009 and this
study. Both studies reviewed several years of IEA data, mainly as a quality assurance measure, but IEA statistics
provided the main top-down comparator with bottom-up results in that study.
The Second IMO GHG Study 2009 used IEA data for 1971–2005 (2007 edition). Two types of oil product
(fuel oil and gas/diesel) and three sectors (international marine bunkers, domestic navigation and fishing) were
reported, and the study subsequently projected those data for 2007 using tonne-miles transported.
For this study, the consortium reviewed data from IEA (2013) for all available years. Figure 18 shows the long-
run statistics for total marine consumption of energy products (international, domestic and fishing) over the
period 2007–2011. IEA statistics for international marine bunkers, domestic navigation and fishing data were
specifically examined for the fuels known to be most used by ships: fuel oil (residual), gas diesel oil, motor
gasoline, lubricants, non-specified fuel and natural gas fuel.
IEA statistics indicate that marine bunker consumption volumes of motor gasoline, lubricants, non-specified
fuel and natural gas are very small. Each of these features has less than 0.10% of fuel oil consumption as
international marine bunkers. Considering domestic and international marine fuels together, only motor
gasoline is reported at quantities equivalent to more than 1% of fuel oil used by ships. No natural gas is
reported as international marine bunkers consumption in IEA (2013), but a small quantity of natural gas is
reported for domestic navigation and fishing.
Other energy products are used in shipping, such as a small amount of primary solid biofuels (domestic and
fishing) and heat and electricity (exclusively in fishing). Given that the statistics identify none of these fuels as
used in international shipping, and given their very small volumes, these fuels were determined to be outside
the scope of this study. Comparison of top-down statistics is therefore limited to fuel oil (HFO), gas diesel oil
(MDO) and natural gas (LNG).
1
Inventories of CO
2
emissions from
international shipping 2007–2012
1.1 Top-down CO
2
inventory calculation method
1.1.1 Introduction
Section 1.1 provides a top-down estimate of emissions from shipping for the period 2007–2012. This task
also provides a comparison of this update with the methods used in the Second IMO GHG Study 2009. The
top-down approach is based on statistical data derived from fuel delivery reports to internationally registered
vessels. The top-down approach also considers allocation to domestic and international shipping, as reported
in national statistics.
Calculations of emissions using top-down fuel consumption estimates are presented. For CO
2
, these estimates
use CO
2
emissions factors consistent with those used in the bottom-up calculations in Section 1.2. Specifically,
the top-down inventory uses the CO
2
emissions factors reported in Section 2.2.7. For marine fuel oil (HFO),
this study uses 3.114 grams CO
2
per gram fuel; for marine gas oil (MDO), this study uses 3.206 grams CO
2
per
gram fuel; and for natural gas (LNG), this study uses 2.75 grams CO
2
per gram fuel.
1.1.2 Methods for review of IEA data
The World Energy Statistics published by IEA are used both in the Second IMO GHG Study 2009 and this
study. Both studies reviewed several years of IEA data, mainly as a quality assurance measure, but IEA statistics
provided the main top-down comparator with bottom-up results in that study.
The Second IMO GHG Study 2009 used IEA data for 1971–2005 (2007 edition). Two types of oil product
(fuel oil and gas/diesel) and three sectors (international marine bunkers, domestic navigation and fishing) were
reported, and the study subsequently projected those data for 2007 using tonne-miles transported.
For this study, the consortium reviewed data from IEA (2013) for all available years. Figure 18 shows the long-
run statistics for total marine consumption of energy products (international, domestic and fishing) over the
period 2007–2011. IEA statistics for international marine bunkers, domestic navigation and fishing data were
specifically examined for the fuels known to be most used by ships: fuel oil (residual), gas diesel oil, motor
gasoline, lubricants, non-specified fuel and natural gas fuel.
IEA statistics indicate that marine bunker consumption volumes of motor gasoline, lubricants, non-specified
fuel and natural gas are very small. Each of these features has less than 0.10% of fuel oil consumption as
international marine bunkers. Considering domestic and international marine fuels together, only motor
gasoline is reported at quantities equivalent to more than 1% of fuel oil used by ships. No natural gas is
reported as international marine bunkers consumption in IEA (2013), but a small quantity of natural gas is
reported for domestic navigation and fishing.
Other energy products are used in shipping, such as a small amount of primary solid biofuels (domestic and
fishing) and heat and electricity (exclusively in fishing). Given that the statistics identify none of these fuels as
used in international shipping, and given their very small volumes, these fuels were determined to be outside
the scope of this study. Comparison of top-down statistics is therefore limited to fuel oil (HFO), gas diesel oil
(MDO) and natural gas (LNG).
28 Third IMO GHG Study 2014
Figure 18: Oil products and products from other sources used in shipping (international,
domestic and fishing) 1971–2011
There are significant gaps in the IEA (2013) data for 2012, at the time of this analysis. For example, international
navigation fuel sales were available for only 29 countries, representing less than 20% of total sales in 2011
(see Table 7). IEA acknowledges that recent data are based on mini-questionnaires from OECD nations and
supply data for non-OECD nations; 2012 marine fuel statistics will be updated in future editions (IEA 2013).
The IMO Secretariat scope specifies that the Third IMO GHG Study 2014 should compute annual emissions
“as far as statistical data are available”. Given the incomplete data, this work therefore excludes year 2012
from this top-down analysis.
Table 7 – Comparison of 2011 and 2012 marine fuels reporting to IEA
2011 2012
Nations reporting
Fuel oil
(ktonnes)
Gas/diesel
(ktonnes)
Fuel oil
(ktonnes)
Gas/diesel
(ktonnes)
29 reporting nations in 2012
(Australia, Belgium, Canada, Chile, Denmark,
Estonia, Finland, France, Germany, Greece, Iceland,
Ireland, Israel, Italy, Japan, Korea [Republic of],
Mexico, Netherlands, New Zealand, Norway,
Poland, Portugal, Slovenia, Spain, Sweden,
Switzerland, Turkey, United Kingdom, United States)
74,833 16,479 70,359 17, 532
Other 98 nations reporting in 2011 102,658 12,655
Percentage of 2011 fuel reported by 29 nations
reporting in 2012
42% 57%
1.1.3 Top-down fuel consumption results
This section presents the Third IMO GHG Study 2014 top-down results for the period of 2007–2011.
Review of Second IMO GHG Study 2009 top-down estimates
The consortium reviewed the Second IMO GHG Study 2009 results, including updates based on current
versions of IEA statistics. Table 8 presents a summary of the information reported in the Second IMO GHG
Study 2009 (from appendix 1, Tables A1–17), with updated information from the IEA (2013) World Energy
Statistics.
It is important to note that top-down information reported in the Second IMO GHG Study 2009 is not
definitive. First, the estimated value for 2007 (derived from 2005 using a tonne-miles adjustment in the Second
IMO GHG Study 2009) can be compared with the IEA value reported in today’s World Energy Statistics.
The 2007 IEA value is approximately 9% greater than the estimated 2007 value in the Second IMO GHG
Study 2009. Second, IEA updated the 2005 reported value with an amended total for all marine fuels that is
approximately 5% greater than the published IEA data used in the Second IMO GHG Study 2009. Most of that
difference results from amended statistics for domestic navigation and fishing, with IEA statistics updates for
marine fuels that are less than 2% of the values reported in the 2009 study. Lastly, the IEA statistics explicitly
designate whether the fuel data aggregate was originally allocated to vessels identified as international shipping,
domestic shipping or fishing. These categories are defined by IEA and described in the Key Definitions section
of this report. IEA definitions are consistent with the IPCC 2006 Guidelines.
Table 8 – Comparison of Second IMO GHG Study 2009 top-down ship fuel consumption data
(million tonnes)
Second IMO GHG
Study 2009
Current IEA
Marine sector Fuel type 2005 2007 2004 2005 2006 2007
International marine bunkers HFO 150 159 144 153 164 174
MDO 26 27 28 26 27 26
International total 176 186 172 179 191 200
Domestic navigation HFO 13 14 17 17 18 20
MDO 20 21 20 21 22 23
Domestic total 33 35 37 38 40 43
Fishing HFO 0 1 1 1 1 1
MDO 5 6 6 7 6 5
Fishing total 5 7 7 8 7 6
Total 214 228 216 225 238 249
% difference from Second IMO GHG Study 2009
5% 9%
Top-down results for the period 2007–2011
Fuel statistics allocated to international shipping, domestic navigation and fishing are presented in Figures19–21
and Table 9. Figure 19 shows a generally flat trend in fuel oil consumption statistics since 2007 for each
shipping category (fishing, international navigation and domestic navigation). Similarly, Figure 20 shows a
generally increasing trend for gas/diesel while Figure 21 shows an increasing trend in natural gas sales in
domestic shipping and interannual variation in natural gas sales to fishing vessels.
Figure 19: IEA fuel oil sales in shipping 2007–2011
Inventories of CO2 emissions from international shipping 2007–2012 29
difference results from amended statistics for domestic navigation and fishing, with IEA statistics updates for
marine fuels that are less than 2% of the values reported in the 2009 study. Lastly, the IEA statistics explicitly
designate whether the fuel data aggregate was originally allocated to vessels identified as international shipping,
domestic shipping or fishing. These categories are defined by IEA and described in the Key Definitions section
of this report. IEA definitions are consistent with the IPCC 2006 Guidelines.
Table 8 – Comparison of Second IMO GHG Study 2009 top-down ship fuel consumption data
(million tonnes)
Second IMO GHG
Study 2009
Current IEA
Marine sector Fuel type 2005 2007 2004 2005 2006 2007
International marine bunkers HFO 150 159 144 153 164 174
MDO 26 27 28 26 27 26
International total 176 186 172 179 191 200
Domestic navigation HFO 13 14 17 17 18 20
MDO 20 21 20 21 22 23
Domestic total 33 35 37 38 40 43
Fishing HFO 0 1 1 1 1 1
MDO 5 6 6 7 6 5
Fishing total 5 7 7 8 7 6
Total 214 228 216 225 238 249
% difference from Second IMO GHG Study 2009
5% 9%
Top-down results for the period 2007–2011
Fuel statistics allocated to international shipping, domestic navigation and fishing are presented in Figures19–21
and Table 9. Figure 19 shows a generally flat trend in fuel oil consumption statistics since 2007 for each
shipping category (fishing, international navigation and domestic navigation). Similarly, Figure 20 shows a
generally increasing trend for gas/diesel while Figure 21 shows an increasing trend in natural gas sales in
domestic shipping and interannual variation in natural gas sales to fishing vessels.
Figure 19: IEA fuel oil sales in shipping 2007–2011
30 Third IMO GHG Study 2014
Figure 20: IEA gas/diesel sales in shipping 2007–2011
Figure 21: IEA natural gas sales in shipping 2007–2011
The IEA statistics explicitly designate fuel to ships as either international or domestic navigation, while fishing
vessel fuel statistics group international and domestic fishing activities together from 2007. The allocation
of total marine fuels provided to ships depends upon the data quality aggregated by IEA from national fuel
reports and ancillary statistical sources. (Issues of data quality and uncertainty in IEA statistics are addressed
in Sections 1.4 and 1.5.) For completeness, this section reports the top-down allocation provided in the IEA
statistics for all three marine fuel designations.
Table 9 reports a summary of IEA data of the fuels most used in shipping over the three different categories
in million tonnes, where natural gas data were converted to tonnes oil equivalent using IEA unit conversions
(1TJ = 0.0238845897 ktoe).
Table 9 – Summary of IEA fuel sales data in shipping (million tonnes)
Marine sector Fuel type 2007 2008 2009 2010 2011
International marine bunkers HFO 174.1 17 7. 0 165.9 178.9 17 7.9
MDO 26.0 22.7 24.9 28.2 29.6
LNG 0 0 0 0 0
International total 200.1 199.7 190.8 2 07.1 207. 5
Domestic navigation HFO 19.9 14.2 15.3 14.3 12.7
MDO 22.7 23.9 23.6 25.7 27. 4
LNG 0.04 0.05 0.05 0.05 0.07
Domestic total 42.64 38.15 38.95 40.05 40.17
Fishing HFO 1.1 1.1 1.0 0.8 0.8
MDO 5.4 4.9 5.0 5.2 5.1
LNG 0.04 0.02 0.04 0.02 0.05
Fishing total 6.54 6.02 6.04 6.02 5.95
Total 249.28 243.87 235.79 253.17 253.62
The time series for top-down fuel inventories reveals some correlation, which may be interpreted as a response
to the economic conditions (lower fuel consumption). The consortium evaluated the top-down consumption
data trends for international marine fuel oil and the world GDP trends as reported by the World Bank’s
World Development Indicators. “World Development Indicators (WDI) is the primary World Bank collection
of development indicators, compiled from officially recognized international sources. It presents the most
current and accurate global development data available, and includes national, regional and global estimates”
(World Bank, November 2013).
Figure 22 illustrates this correlation graphically and shows the correlation coefficient for 2000–2012 to be very
high (96.5%). This trend also shows correlation with the start of economic recovery in 2009. The divergence
between fuel oil consumption and GDP trends since 2010 could be a function of three factors:
1 energy efficiency measures adopted by shipping in response to price;
2 fuel-switching to gas diesel or natural gas fuels;
3 a lag in shipping activity change compared to world GDP change.
Further time series and additional analysis beyond the scope of this study would be required to evaluate
post-recession changes further.
Figure 22: Correlation between world GDP and international bunker fuel oil during the recession
Inventories of CO2 emissions from international shipping 2007–2012 31
Table 9 – Summary of IEA fuel sales data in shipping (million tonnes)
Marine sector Fuel type 2007 2008 2009 2010 2011
International marine bunkers HFO 174.1 17 7. 0 165.9 178.9 17 7.9
MDO 26.0 22.7 24.9 28.2 29.6
LNG 0 0 0 0 0
International total 200.1 199.7 190.8 2 07.1 207. 5
Domestic navigation HFO 19.9 14.2 15.3 14.3 12.7
MDO 22.7 23.9 23.6 25.7 27. 4
LNG 0.04 0.05 0.05 0.05 0.07
Domestic total 42.64 38.15 38.95 40.05 40.17
Fishing HFO 1.1 1.1 1.0 0.8 0.8
MDO 5.4 4.9 5.0 5.2 5.1
LNG 0.04 0.02 0.04 0.02 0.05
Fishing total 6.54 6.02 6.04 6.02 5.95
Total 249.28 243.87 235.79 253.17 253.62
The time series for top-down fuel inventories reveals some correlation, which may be interpreted as a response
to the economic conditions (lower fuel consumption). The consortium evaluated the top-down consumption
data trends for international marine fuel oil and the world GDP trends as reported by the World Bank’s
World Development Indicators. “World Development Indicators (WDI) is the primary World Bank collection
of development indicators, compiled from officially recognized international sources. It presents the most
current and accurate global development data available, and includes national, regional and global estimates”
(World Bank, November 2013).
Figure 22 illustrates this correlation graphically and shows the correlation coefficient for 2000–2012 to be very
high (96.5%). This trend also shows correlation with the start of economic recovery in 2009. The divergence
between fuel oil consumption and GDP trends since 2010 could be a function of three factors:
1 energy efficiency measures adopted by shipping in response to price;
2 fuel-switching to gas diesel or natural gas fuels;
3 a lag in shipping activity change compared to world GDP change.
Further time series and additional analysis beyond the scope of this study would be required to evaluate
post-recession changes further.
Figure 22: Correlation between world GDP and international bunker fuel oil during the recession
32 Third IMO GHG Study 2014
1.2 Bottom-up CO
2
inventory calculation method
The bottom-up method derives estimates of emissions from data sources describing shipping activity. The
primary source of vessel activity used is the AIS data, which describe, among other things, a ship’s identity,
position, speed and draught at a given time-stamp. The data are transmitted by the ship with a broadcast
frequency of one message every six seconds. The data are received by shore-based stations, satellites and
other ships and the consortium acquired access to a number of shore-based station and satellite receiver
archives. These were used to build time histories of shipping activity, which could be used, in conjunction
with ship specifications, to calculate the time histories of fuel consumption and emissions. Calculations were
carried out for every individual ship identified as in service in the IHSF database and for every hour of the year.
1.2.1 Overall bottom-up approach
The bottom-up method is split into two stages:
1 initial estimation of observed per-ship activity, energy consumption and emissions;
2 estimation of per-ship activity and associated energy consumption and emissions for ships not
observed in the AIS database.
The first stage is performed only on ships that appear coincidentally in both the IHSF and AIS databases. The
second stage is performed for all ships listed as “in service/commission” within the IHSF database and uses
estimated activity for similar ships in stage 1, in combination with IHSF technical specifications to estimate
power requirements, fuel consumption and emissions. The total energy consumption and emissions for a fleet
of similar ships is then found by summing the calculations for each ship, estimated either at stage 1 or stage 2.
The total shipping emissions are then found by summing across all ship type and size categories. International
shipping emissions are estimated by defining which ship type and size categories are involved in international
shipping.
Figure 23 is a diagram of the flow of data through the processes and calculation stages that make up the
bottom-up method.
Figure 23: Data assembly and method for Sections 1.2 and 2.2
1.2.2 Summary of data and method input revisions
Data
Access to increasingly detailed data on ships’ activity was enabled by the advent of S-AIS, which began
providing significant coverage in 2010. These data enable the specifics of any ship’s operation to be identified
on an hourly basis, or even more frequently if required. S-AIS allows greater fidelity in the calculation of the
fleet’s aggregate operational characteristics. For the first time in global inventory calculations, the activity of
specific individual ships (e.g. actual vessel speed over ground) and consequent engine load and emissions can
be considered as a component of an overall inventory calculation. In the Second IMO GHG Study 2009, a
limited sample of terrestrial AIS data was used to calculate ship activity parameters (speeds, days at sea, etc.).
In that study, ship activity could only be observed for a subset of the fleet and only within approximately 50
nmi of available shore-based receivers (only partial coverage of coastal regions), which left the activity of
vessels in the open ocean unobserved. In this study, the consortium brings together a number of data sets
from both terrestrial and satellite receiver operators and merges the data to provide extensive spatial and
temporal coverage of shipping activity observations. A visualization of the merged AIS data for 2012 is shown
in Figure24.
Figure 24: Chart showing the coverage of one of the merged AIS data sets used in this study (2012,
all sources, but no LRIT)
Observations in the merged AIS data set of ship activity (speeds, time spent in modes) are compared to similar
data derived from samples of the global fleet from LRIT. In all, data concerning approximately 8,000 ships
were put together (see Section 1.4 for details). LRIT data were not used in the Second IMO GHG Study 2009.
A visualization of the LRIT data for 2012 is shown in Figure 25. LRIT data are of lower temporal resolution than
AIS data but provide higher reliability and therefore enable important quality checks for the AIS data set and
the bottom-up calculations of average speeds and days at sea.
Inventories of CO2 emissions from international shipping 2007–2012 33
1.2.2 Summary of data and method input revisions
Data
Access to increasingly detailed data on ships’ activity was enabled by the advent of S-AIS, which began
providing significant coverage in 2010. These data enable the specifics of any ship’s operation to be identified
on an hourly basis, or even more frequently if required. S-AIS allows greater fidelity in the calculation of the
fleet’s aggregate operational characteristics. For the first time in global inventory calculations, the activity of
specific individual ships (e.g. actual vessel speed over ground) and consequent engine load and emissions can
be considered as a component of an overall inventory calculation. In the Second IMO GHG Study 2009, a
limited sample of terrestrial AIS data was used to calculate ship activity parameters (speeds, days at sea, etc.).
In that study, ship activity could only be observed for a subset of the fleet and only within approximately 50
nmi of available shore-based receivers (only partial coverage of coastal regions), which left the activity of
vessels in the open ocean unobserved. In this study, the consortium brings together a number of data sets
from both terrestrial and satellite receiver operators and merges the data to provide extensive spatial and
temporal coverage of shipping activity observations. A visualization of the merged AIS data for 2012 is shown
in Figure24.
Figure 24: Chart showing the coverage of one of the merged AIS data sets used in this study (2012,
all sources, but no LRIT)
Observations in the merged AIS data set of ship activity (speeds, time spent in modes) are compared to similar
data derived from samples of the global fleet from LRIT. In all, data concerning approximately 8,000 ships
were put together (see Section 1.4 for details). LRIT data were not used in the Second IMO GHG Study 2009.
A visualization of the LRIT data for 2012 is shown in Figure 25. LRIT data are of lower temporal resolution than
AIS data but provide higher reliability and therefore enable important quality checks for the AIS data set and
the bottom-up calculations of average speeds and days at sea.
34 Third IMO GHG Study 2014
Figure 25: Chart showing the coverage of one of the LRIT data sets used in this study (2012)
The quality of the bottom-up model’s activity and fuel consumption calculations was also checked against
operators’ fuel consumption data, contained in noon reports and fuel audits (see Section 1.4). No equivalent
data were reportedly used in the Second IMO GHG Study 2009.
This study uses IHSF data to obtain the technical characteristics of individual ships. While IHSF data were
used in the Second IMO GHG Study 2009, this study includes data on the status of a ship (in service, etc.).
Ship status data are obtained on a quarterly basis, so that ships that are reportedly active for only part of the
year are considered appropriately.
Method
The method developed by the consortium to conduct this study uses a comparable structure to the
methodology of the Second IMO GHG Study 2009 for the collation of aggregate data on activity parameters,
engine load and emissions. However, it is underpinned by analysis carried out at each calculation stage
on a complete database of the global fleet (i.e. all calculations are performed at the level of the individual
ship with aggregation of results only used for presentation purposes). This approach avoids the potential for
asymmetry or data bias that might reduce fidelity and accuracy. This represents a substantial progression in
the technology and practice of activity-based inventory methods for international shipping.
1.2.3 Aggregation of ship types and sizes
The algorithms used for vessel aggregation, developed by the consortium, build on aggregation methodologies
for EEDI (taken from IMO MEPC.231(65), expanded for ship classes not included in EEDI) and divide vessels
further into bins based on cargo capacity or ship size. The aggregations use definitions aligned as closely as
possible with those used in the Second IMO GHG Study 2009. In some cases, however, this was not possible
because the taxonomy used in the earlier study was not reported explicitly and because it did not always align
with the EEDI categories. Aggregation uses IHSF Statcode3, Statcode5, and relevant capacity fields to group
similar ships. IHSF organizes vessels into four types of ship:
• cargo-carrying;
• non-merchant;
• non-seagoing merchant;
• work vessels.
Most international shipping is represented by cargo-carrying transport ships, which are the primary focus of this
study. However, the other classes are needed to compare the bottom-up estimate with the top-down estimate
where both international and domestic voyages by oceangoing ships may be represented. The consortium
subdivided cargo-carrying vessel types into 13 classes, the non-merchant ships and non-seagoing merchant
ship types into two and one classes respectively, and the work vessel type into three classes. As shown in
Table 10, a total of 19 classes are defined.
Table 10 – IHSF vessel types and related vessel classes
Vessel group Vessel class
Cargo-carrying transport ships 1. Bulk carrier
2. Chemical tanker
3. Container
4. General cargo
5. Liquified gas tanker
6. Oil tanker
7. Other liquids tanker
8. Ferry – passengers (pax) only
9. Cruise
10. Ferry – roll-on/passengers (ro-pax)
11. Refrigerated cargo
12. Roll-on/roll-off (ro-ro)
13. Vehicle
Non-merchant ships 14. Yacht
15. Miscellaneous – fishing
1
Non-seagoing merchant ships 16. Miscellaneous – other
2
Work vessels 17. Service – tug
18. Offshore
19. Service – other
For each vessel class a capacity bin system was developed to further aggregate vessels by either their physical
size or cargo-carrying capacity, based on the following metrics: deadweight tonnage (dwt); 20-foot equivalent
units (TEU); cubic metres (cbm); gross tonnage (gt); or vehicle capacity (see Table 12). The capacity bins are the
same for all vessels in a class. Wherever possible, bin sizes are aligned to the Second IMO GHG Study 2009,
although there are some discrepancies due to differences in the class definitions. It should be noted that the
Third IMO GHG Study 2014 provides higher resolution by class/subclass/capacity bin than the Second IMO
GHG Study 2009. Further details of the approach used and the definitions applied can be found in Annex 1.
1.2.4 Estimating activity using AIS data
The primary purpose of AIS is to report the current location of vessels in order to avoid collisions. Under IMO
regulations (SOLAS, chapter V), all vessels over 300 gt on international transport (IMO, 2002) are required to
carry transmitters. AIS information is reported in different message types depending on the reporting entity
(e.g. vessel, base station) and the nature of the message (i.e. dynamic or static). The messages of interest
for this study are static and dynamic vessel messages (see ITU (2010) for further details of message types).
Dynamic messages (types 1, 2 and 3) report more frequently and provide frequently changing information,
such as location and speed. Static messages (types 5 and 24) contain voyage information, such as draught,
destination and (importantly) the IMO number of the vessel. Static and dynamic messages are linked through
the MMSI number, which is reported in both message types. These messages are collected through receivers
on land (T-AIS) and through a satellite network (S-AIS). Due to temporal and spatial coverage issues, explained
elsewhere (Smith et al. 2012; Second IMO GHG Study 2009), quality can be improved using a combination
of these sources as they offer complementary spatial and temporal coverage.
The consortium used multiple data sources. Annex 1 describes the process adopted for the processing of
the raw data to obtain hourly estimates of speed, draught and region of operation, and their merger into a
single, combined data set for use in the bottom-up model. Information in message 18 transmitted from ClassB
transponders was not used to estimate activity and emissions.
1
misc. fishing vessels fall into the non-merchant ships and non-seagoing merchant ships categories.
2
misc. other vessels fall into the non-seagoing merchant ships and work vessels categories.
Inventories of CO2 emissions from international shipping 2007–2012 35
ship types into two and one classes respectively, and the work vessel type into three classes. As shown in
Table 10, a total of 19 classes are defined.
Table 10 – IHSF vessel types and related vessel classes
Vessel group Vessel class
Cargo-carrying transport ships 1. Bulk carrier
2. Chemical tanker
3. Container
4. General cargo
5. Liquified gas tanker
6. Oil tanker
7. Other liquids tanker
8. Ferry – passengers (pax) only
9. Cruise
10. Ferry – roll-on/passengers (ro-pax)
11. Refrigerated cargo
12. Roll-on/roll-off (ro-ro)
13. Vehicle
Non-merchant ships 14. Yacht
15. Miscellaneous – fishing
1
Non-seagoing merchant ships 16. Miscellaneous – other
2
Work vessels 17. Service – tug
18. Offshore
19. Service – other
For each vessel class a capacity bin system was developed to further aggregate vessels by either their physical
size or cargo-carrying capacity, based on the following metrics: deadweight tonnage (dwt); 20-foot equivalent
units (TEU); cubic metres (cbm); gross tonnage (gt); or vehicle capacity (see Table 12). The capacity bins are the
same for all vessels in a class. Wherever possible, bin sizes are aligned to the Second IMO GHG Study 2009,
although there are some discrepancies due to differences in the class definitions. It should be noted that the
Third IMO GHG Study 2014 provides higher resolution by class/subclass/capacity bin than the Second IMO
GHG Study 2009. Further details of the approach used and the definitions applied can be found in Annex 1.
1.2.4 Estimating activity using AIS data
The primary purpose of AIS is to report the current location of vessels in order to avoid collisions. Under IMO
regulations (SOLAS, chapter V), all vessels over 300 gt on international transport (IMO, 2002) are required to
carry transmitters. AIS information is reported in different message types depending on the reporting entity
(e.g. vessel, base station) and the nature of the message (i.e. dynamic or static). The messages of interest
for this study are static and dynamic vessel messages (see ITU (2010) for further details of message types).
Dynamic messages (types 1, 2 and 3) report more frequently and provide frequently changing information,
such as location and speed. Static messages (types 5 and 24) contain voyage information, such as draught,
destination and (importantly) the IMO number of the vessel. Static and dynamic messages are linked through
the MMSI number, which is reported in both message types. These messages are collected through receivers
on land (T-AIS) and through a satellite network (S-AIS). Due to temporal and spatial coverage issues, explained
elsewhere (Smith et al. 2012; Second IMO GHG Study 2009), quality can be improved using a combination
of these sources as they offer complementary spatial and temporal coverage.
The consortium used multiple data sources. Annex 1 describes the process adopted for the processing of
the raw data to obtain hourly estimates of speed, draught and region of operation, and their merger into a
single, combined data set for use in the bottom-up model. Information in message 18 transmitted from ClassB
transponders was not used to estimate activity and emissions.
1
misc. fishing vessels fall into the non-merchant ships and non-seagoing merchant ships categories.
2
misc. other vessels fall into the non-seagoing merchant ships and work vessels categories.
36 Third IMO GHG Study 2014
1.2.5 Ship technical data
Ship technical data are required to estimate ship emissions in the bottom-up model. The primary source of
technical data used for this study is the IHSF ship registry database. Ship technical data from the IHSF data
sets used in this study include Statcode3, Statcode5, gt, dwt, length, beam, maximum draught, vessel speed,
installed main engine power, engine revolutions per minute (RPM), various cargo capacity fields, date of build,
keel laid date, propulsion type, number of screws, and main engine fuel consumption and stroke type. In
addition to technical data, the IHSF data set includes a ship status field that indicates whether a ship is active,
laid up, being built, etc. The consortium had access to quarterly IHSF data sets from 2007 through to 2012.
Each year’s specific data were used for the individual annual estimates.
It should be noted that the data sets do not provide complete coverage for all ships and all fields needed. In
cases where data are missing, values are estimated either from interpolation or by referencing another publicly
available data source. The details of the approach taken for the missing data and the technical and operational
data themselves are discussed further in Section 1.4.3 and Annex 1.
For auxiliary engine operational profiles, neither IHSF nor the other vessel characteristic data services
provide auxiliary engine use data by vessel mode. In the Second IMO GHG Study 2009, auxiliary loads were
estimated by assuming the number and load of auxiliary engines operated by vessel type, and were based
on the rated auxiliary engine power gauged from the limited data provided in IHS. To improve this approach,
the consortium used Starcrest’s Vessel Boarding Program (VBP) data, which had been collected at the Port
of Los Angeles, the Port of Long Beach (Starcrest, 2013), the Port Authority of New York & New Jersey,
the Port of Houston Authority, the Port of Seattle and the Port of Tacoma. The VBP data set includes over
1,200 vessels of various classes. For over 15 years, Starcrest has collected data on-board vessels specifically
related to estimating emissions from ships and validating its models. Auxiliary loads (in kW) are recorded
for at-berth, at-anchorage, manoeuvring and at-sea vessel modes. The vessel types boarded as part of VBP
include bulk carriers, chemical tankers, cruise passenger ships, oil tankers, general cargo ships, container ships
and refrigerated cargo ships.
For container and refrigerated cargo ships, vessel auxiliary engine and boiler loads (kW), by mode, were
developed based on the VBP data set and averages by vessel type and bin size were used. This approach
assumes that the vessels boarded are representative of the world fleet for the same classes.
For bulk carriers, chemical tankers, cruise passenger ships, oil tankers and general cargo ships, a hybrid
approach was used combining VBP data, data collected from the Finnish Meteorological Institute (FMI) and
the Second IMO GHG Study 2009 approach. The earlier study’s approach was based on average auxiliary
engine rating (kW), assumption of number of engines running expressed in operational days per year (if greater
than 365, it was assumed that more than one engine was running), a single load factor for each vessel type,
and capacity bins. A hybrid method was used for vessels boarded as part of VBP but this was not considered
to be robust enough to use on its own. VBP data were used to inform and align the estimate of number of
engines used and the ratios between various modes and to review the results for reasonableness.
For vessel classes not previously boarded by VBP, data collected by FMI (from engine manufacturers,
classification societies and other sources) were used to determine the ratio between main engines and auxiliary
engines. The number of engines assumed to be installed and running was derived from either the Second IMO
GHG Study 2009 or professional judgement. This information was used for the various vessel types and bin
sizes to develop vessel-weighted average auxiliary loads in kW. Consistent with the approach of the Second
IMO GHG Study 2009, these loads are applied across all operational modes in this study.
LIke auxiliary engine loads, there is no commercial data source that provides information about auxiliary
boiler loads by operational mode. Auxiliary boiler loads were developed using VBP data and the professional
judgement of members of the consortium. Auxiliary boiler loads are typically reported in tons of fuel per day
but these rates have been converted to kW (Starcrest, 2013). Boilers are used for various purposes on ships
and their operational profile can change by mode.
Further details of the approach used to develop auxiliary engine and boiler loads by vessel type and mode
can be found in Annex 1.
1.2.6 Sources and alignment/coverage of data sources
For the bottom-up method, calculations are performed on each individual ship’s technical and activity data.
For this, the consortium mainly used the IHSF database and AIS data sources and the majority of ships can
be identified in each of these for a given year. However, during the method development, the consortium has
recognized several ships for which a corresponding IHSF and activity data match does not occur (e.g. an IMO
number is not reported or the MMSI number does not match). Treatment of ships in such categories can be
summarized by the diagram presented in Figure 26, and is discussed below so that their contribution to global
CO
2
emissions estimates can be better understood.
Figure 26: Venn diagram describing the sets of ships observed in the two main data types used in the
bottom-up method (IHSF and AIS)
Type 1: IMO number is missing but MMSI number appears in both IHSF and activity data set
The SOLAS convention (chapter V) requires that all ships of >300 gt should install a class-A AIS transponder.
Furthermore, ships of <300 gt are urged to install class-B AIS transponders voluntarily. The consortium
recognized the MEPC request to calculate CO
2
emissions from all ships of >100 gt, therefore the consortium
retrieved both class-A and class-B data for this purpose.
Each AIS transponder has an individual MMSI code. MMSI transponder data from non-ships (e.g. fixed
structures, SAR aircraft) were excluded using message ID and the first three digits of the MMSI. Of the
remaining ships, for which no IMO numbers are reported in the activity data, the match was carried out on
MMSI number alone. However, this is not fully reliable because the record of MMSI numbers in the IHSF data
set is imperfect.
Type 2: MMSI appears in the activity data set only
The consortium recognized that some ships appeared in the activity data set only and did not match any ships
registered in the IHSF database. Three reasons could explain this mismatch:
1 erroneous or incomplete records in the IHSF database (e.g. incomplete list of MMSI numbers);
2 ships are operated only for domestic navigation purposes (in which case, the ships will be controlled
under each individual administration and do not need to be registered in IHSF);
3 the AIS equipment has reverted back to default “factory settings” of IMO/MMSI numbers.
1
1
See the Maritime and Coastguard Agency note MIN 298 (M+F): “AIS (Automatic Identification Systems) Operational Notification –
Safety of Navigation. ACR/Nauticast AIS”.
Inventories of CO2 emissions from international shipping 2007–2012 37
1.2.6 Sources and alignment/coverage of data sources
For the bottom-up method, calculations are performed on each individual ship’s technical and activity data.
For this, the consortium mainly used the IHSF database and AIS data sources and the majority of ships can
be identified in each of these for a given year. However, during the method development, the consortium has
recognized several ships for which a corresponding IHSF and activity data match does not occur (e.g. an IMO
number is not reported or the MMSI number does not match). Treatment of ships in such categories can be
summarized by the diagram presented in Figure 26, and is discussed below so that their contribution to global
CO
2
emissions estimates can be better understood.
Figure 26: Venn diagram describing the sets of ships observed in the two main data types used in the
bottom-up method (IHSF and AIS)
Type 1: IMO number is missing but MMSI number appears in both IHSF and activity data set
The SOLAS convention (chapter V) requires that all ships of >300 gt should install a class-A AIS transponder.
Furthermore, ships of <300 gt are urged to install class-B AIS transponders voluntarily. The consortium
recognized the MEPC request to calculate CO
2
emissions from all ships of >100 gt, therefore the consortium
retrieved both class-A and class-B data for this purpose.
Each AIS transponder has an individual MMSI code. MMSI transponder data from non-ships (e.g. fixed
structures, SAR aircraft) were excluded using message ID and the first three digits of the MMSI. Of the
remaining ships, for which no IMO numbers are reported in the activity data, the match was carried out on
MMSI number alone. However, this is not fully reliable because the record of MMSI numbers in the IHSF data
set is imperfect.
Type 2: MMSI appears in the activity data set only
The consortium recognized that some ships appeared in the activity data set only and did not match any ships
registered in the IHSF database. Three reasons could explain this mismatch:
1 erroneous or incomplete records in the IHSF database (e.g. incomplete list of MMSI numbers);
2 ships are operated only for domestic navigation purposes (in which case, the ships will be controlled
under each individual administration and do not need to be registered in IHSF);
3 the AIS equipment has reverted back to default “factory settings” of IMO/MMSI numbers.
1
1
See the Maritime and Coastguard Agency note MIN 298 (M+F): “AIS (Automatic Identification Systems) Operational Notification –
Safety of Navigation. ACR/Nauticast AIS”.
38 Third IMO GHG Study 2014
In some countries with cabotage, such as the United States, Japan and China, some ships may be employed
in domestic navigation only and this could be consistent with explanation 2. As the bottom-up method will
include both international and domestic fuel consumption and emissions (in order to assist in separating out
international fuel consumption and emissions alone), this category of ships will have to be included in the
method, but with high uncertainty because they cannot be given technical characteristics.
Type 3: ship appears in IHSF but cannot be identified in the activity data set
After the matching process, a number of ships may be identified in the IHSF database with no corresponding
activity data. Explanations for this could be:
a the ships were not active or had their transponders turned off; e.g. FPSOs, barges, platforms and older
ships awaiting scrapping;
b the ships may be less than 300 gt without any AIS installation.
If the ship was >300 gt, it was assumed to be inactive and omitted from the model. If the ship was <300 gt,
it was assumed that its absence from the AIS data was because it did not have a transponder. In this case the
vessel was assigned a typical activity model from similar identifiable ships.
Classifications for each type are summarized in Table 11. Category 0 includes ships that have no identification/
matching issues. All of the other four categories require assumptions, which are studied in greater detail in
Sections 1.4 and 1.5.
Table 11 – Classification of ships in the bottom-up approach
Type Identified in activity data set Identified in IHSF database Reason for non-matching Target for estimation
0 Yes Yes Yes
1 Yes Yes on MMSI number Incompletion in data Yes
2 Yes No Ships are operated for
domestic navigation only,
therefore not registered
in IHSF
Yes
3a No Yes Ship is not active No
3b No Yes
Ships of <300 gt
and without any AIS
transponder
Yes
1.2.7 Bottom-up fuel and emissions estimation
The bottom-up method combines activity data (derived from AIS and LRIT raw data sources) and technical
data (derived from IHSF and a series of empirical data and assumptions derived from the literature).
The model is composed of a main programme that calls up a number of subroutines (as listed in Annex1). Each
ship has a total of 8,760 unique activity observations per year (8,784 in a leap year) and with approximately
100,000 ships included in a given year’s fleet, the run time of the model is significant on conventional hardware.
The model can only perform calculations for ships for which both activity and IHSF activity data are available.
Procedures for estimating the fuel demands and emissions of ships that are not matched are described in
greater detail in Annex 1.
1.2.8 Classification of international and domestic fuel
Estimation of bottom-up fuel totals is performed without pre-identifying international versus domestic
allocations, because bottom-up methods focus on characteristics of vessel activity, irrespective of ports of
departure and arrival. Therefore, top-down allocations according to IEA and IPCC definitions cannot be directly
extracted from bottom-up results without route identification. However, some approaches can produce
estimates of the fraction of fuels reported in bottom-up totals that may represent a delineation of international
shipping, domestic navigation and fishing. These approaches can be summarized as three allocation methods:
1 apply heuristic from T-D statistics as a ratio of international to total shipping;
Inventories of CO2 emissions from international shipping 2007–2012 39
2 assign fleet sectors to domestic service and subtract from fleet;
3 combine T-D heuristics and fleet sector information to match the vessel types most likely to serve
domestic shipping (bottom-up) with expectations of total fraction likely to use domestic bunkers
(top-down).
The Second IMO GHG Study 2009 used method 3, a combined application of the top-down heuristic
and removal of some vessel types. However, the study noted significant uncertainties with this approach.
Specifically, it assumed that ship activity was proportional to data on seaborne transport. The study noted
that, over the course of a year’s activity, a given vessel could be engaged in both international shipping
and domestic navigation. “Since the [Second IMO GHG Study 2009] activity-based model cannot separate
domestic shipping from international shipping, figures from bunker statistics for emissions from domestic
shipping [were] used in the calculation of emissions from international shipping” (Second IMO GHG Study
2009, paragraph 3.17). This study explicitly removed fleet sectors associated with fishing, fixed offshore
installations (production vessels) and domestic navigation relying on fuel totals reported in their top-down
analysis based on IEA statistics.
The Third IMO GHG Study 2014 consortium chose not to apply allocation methods 1 or 3 and selected
method 2, for several reasons. Method 1 requires a simplistic and arbitrary direct application of the top-down
fuel ratios to bottom-up totals. The main disadvantage of method 1 is that it can be applied to the inventory
total only; results cannot be tied to bottom-up insights within vessel categories. A related disadvantage is
that the assumption may be untestable, preventing direct quality assurance or control and disabling any
quantitative consideration of uncertainty.
Allocation method 3 requires subjective judgements to be imposed on the bottom-up data beyond a testable
set of assumptions applied to vessel types. For example, the 2009 study imposed additional definitions of
oceangoing and coastwise shipping, designating some fleet sectors like cruise passenger ships, service and
fishing vessels and smaller ro-pax vessels as coastwise. However, that study did not reconcile or discuss
whether the fuel totals allocated to coastwise vessels corresponded to an international versus domestic
determination within its activity-based method. Moreover, an attempt to determine which shipping was
coastwise, as opposed to transiting along a coastal route, was beyond scope of the study.
The Third IMO GHG Study 2014 applies allocation method 2 with information provided in AIS to support
the bottom-up methodology. Based on general voyage behaviour, some ship types are likely to engage in
international shipping more often than domestic navigation. These types include transport and larger ferry
vessels, as listed in Table 12. This allocation, therefore, also identifies ship types that can be expected to
engage mostly in domestic navigation, including non-transport vessels, such as offshore and service vessels,
yachts and smaller regional ferry vessels (see Table 13). Results using allocation method 2 allow comparison
between bottom-up and top-down allocation of international shipping and domestic navigation. As a caveat,
method 2 might overestimate international shipping and could increase uncertainty, which is discussed in
Sections 1.4 and 1.5.
40 Third IMO GHG Study 2014
Table 12 – Summary of vessel types and sizes that can be expected to engage in international shipping
Vessel type Capacity bin Capacity unit
Bulk carrier 0–9,999 dwt
10,000–34,999
35,000–59999
60,000–99,999
100,000–199,999
200,000–+
Chemical tanker 0–4,999 dwt
5,000–9,999
10,000–19,999
20,000–+
Container 0–999 TEU
1,000 –1,999
2,000–2,999
3,000–4,999
5,000–7,999
8,000–11,999
12,000–14,500
14,500 –+
Cruise 0–1,999 gt
2,000–9,999
10,000–59,999
60,000–99,999
100,000–+
Ferry – pax only
2,000–+
gt
Ferry – ro-pax
2,000–+
gt
General cargo 0–4,999 dwt
5,000–9,999
10,000–+
Liquefied gas tanker 0–49,999 cubic metres (cbm)
50,000–199,999
200,000–+
Oil tanker 0–4,999 dwt
5,000–9,999
10,000–19,999
20,000–59,999
60,000–79,999
80,000–119,999
120,000–199,999
200,000–+
Other liquids tankers
0–+
dwt
Refrigerated cargo 0–1,999 dwt
Ro-ro 0–4,999 gt
5,000–+
Vehicle 0–3,999 vehicles
4,000–+
Table 13 – Summary of vessel types and sizes that can be expected to engage in domestic shipping
Vessel type Capacity bin Capacity unit
Ferry: pax only 0–1,999 gt
Ferry: ro-pax 0–1,999 gt
Miscellaneous – fishing All sizes gt
Miscellaneous – other All sizes gt
Offshore All sizes gt
Service – other All sizes gt
Service – tug All sizes gt
Yacht All sizes gt
1.3 Inventories of CO
2
emissions calculated using both the top-down
and bottom-up methods
1.3.1 CO
2
emissions and fuel consumption by ship type
Figure 27 presents CO
2
emissions by ship type, calculated using the bottom-up method. Equivalent ship-type-
specific results cannot be presented for the top-down method because the reported marine fuel sales statistics
are only available in three categories: international, domestic and fishing.
Figure 27: Bottom-up CO
2
emissions from international shipping by ship type (2012)
Figure 28 shows the relative fuel consumption among vessel types in 2012 (both international and domestic
shipping), estimated using the bottom-up method. The figure also identifies relative fuel consumption between
the main engine (predominantly propulsion), auxiliary engine (electricity generation) and boilers (steam
generation). The total shipping fuel consumption is shown to be dominated by three ship types: oil tankers,
bulk carriers and container ships. In each of these ship types, the main engine consumes the majority of the
fuel. The same plots recreated for earlier years (2007–2011) are included in Annex 2.
Inventories of CO2 emissions from international shipping 2007–2012 41
Table 13 – Summary of vessel types and sizes that can be expected to engage in domestic shipping
Vessel type Capacity bin Capacity unit
Ferry: pax only 0–1,999 gt
Ferry: ro-pax 0–1,999 gt
Miscellaneous – fishing All sizes gt
Miscellaneous – other All sizes gt
Offshore All sizes gt
Service – other All sizes gt
Service – tug All sizes gt
Yacht All sizes gt
1.3 Inventories of CO
2
emissions calculated using both the top-down
and bottom-up methods
1.3.1 CO
2
emissions and fuel consumption by ship type
Figure 27 presents CO
2
emissions by ship type, calculated using the bottom-up method. Equivalent ship-type-
specific results cannot be presented for the top-down method because the reported marine fuel sales statistics
are only available in three categories: international, domestic and fishing.
Figure 27: Bottom-up CO
2
emissions from international shipping by ship type (2012)
Figure 28 shows the relative fuel consumption among vessel types in 2012 (both international and domestic
shipping), estimated using the bottom-up method. The figure also identifies relative fuel consumption between
the main engine (predominantly propulsion), auxiliary engine (electricity generation) and boilers (steam
generation). The total shipping fuel consumption is shown to be dominated by three ship types: oil tankers,
bulk carriers and container ships. In each of these ship types, the main engine consumes the majority of the
fuel. The same plots recreated for earlier years (2007–2011) are included in Annex 2.
42 Third IMO GHG Study 2014
Figure 28: Summary graph of annual fuel consumption (2012), broken down by ship type and
machinery component (main, auxiliary and boiler)
The detailed results for 2012, broken down by ship type and size category, are presented in Table 14. This
table displays the differences between ship types and sizes; for example, differences in installed power, speeds
(both design speed and operational speed) and as a result differences in fuel consumption. There are also
important differences between the amounts (number of ships) in each of the ship type and size categories.
When aggregated to a specific ship type, in sum, these explain the differences observed in Figure 27 and
Figure 28, and the differences presented in the last column (“Total CO
2
emissions”).
The table also displays information about the coverage of the fleet on AIS. The “IHSF” column under “Number
active” lists the number of ships reported as being in service in the IHSF database for that year. The “AIS”
column under “Number active” lists the number of ships that are observed in the AIS data at any point in time
during the year. In general, the coverage of the in-service fleet on AIS is consistently high (e.g. 95% and above)
for the larger ship sizes but less so for some smaller ship size categories (the smallest general cargo carriers in
particular). This could be indicative of a number of issues:
• low quality in certain size and type categories of the IHSF database for maintaining information on a
ship’s status (in-service indication);
• low-quality AIS coverage for the smallest ship types;
• low compliance with SOLAS, chapter V (that ships above a certain size must fit an AIS transponder).
The discussion of quality of coverage is extended in Section 1.4.
Further tables listing the same specifics for the earlier years of the analysis are included in Annex 2.
Inventories of CO2 emissions from international shipping 2007–2012 43
Table 14 – Tabular data for 2012 describing the fleet (international, domestic and fishing) analysed using the bottom-up method
Ship type Size category Units
Number
active
Decimal AIS
coverage of
in-service
ships
Avg.
dead-
weight
(tonnes)
Avg.
installed
power
(kW)
Avg.
design
speed
(knots)
Avg.
days at
sea
Avg.*
sea
speed
(knots)
Avg.* consumption
(‘000 tonnes)
Total CO
2
emissions
(‘000 tonnes)
IHSF AIS Main Auxiliary Boiler
Bulk carrier 0–9,999 dwt 1,216 670 0.55 3,341 1,640 11.6 167 9.4 0.9 0.5 0.1 5,550
10,000–34,999 dwt 2,317 2,131 0.92 27, 6 69 6,563 14.8 168 11.4 3.0 0.5 0.1 24,243
35,000–59,999 dwt 3,065 2,897 0.95 52,222 9,022 15.3 173 11.8 4.0 0.7 0.1 4 4,116
60,000–99,999 dwt 2,259 2,145 0.95 81,876 10,917 15.3 191 11.9 5.4 1.1 0.3 45,240
100,000–199,999 dwt 1,246 1,169 0.94 176,506 17, 330 15.3 202 11.7 8.5 1.1 0.2 36,340
200,000–+
dwt 294 274 0.93 271,391 22,170 15.7 202 12.2 11.0 1.1 0.2 10,815
Chemical tanker 0–4,999 dwt 1,502 893 0.59 2,158 1,387 11.9 159 9.8 0.8 0.5 0.6 5,479
5,000–9,999 dwt 922 863 0.94 7,497 3,292 13.4 169 10.6 1.6 0.6 0.4 7,19 9
10,000–19,999 dwt 1,039 1,004 0.97 15,278 5,260 14.1 181 11.7 3.0 0.6 0.4 12,318
20,000–+
dwt 1,472 1,419 0.96 42,605 9,297 15.0 183 12.3 5.0 1.4 0.4 30,027
Container 0–999 TEU 1,126 986 0.88 8,634 5,978 16.5 190 12.4 2.8 0.9 0.2 12,966
1,000 –1,999 TEU 1,306 1,275 0.98 20,436 12,578 19.5 200 13.9 5.2 2.2 0.4 31,015
2,000–2,999 TEU 715 689 0.96 36,735 22,253 22.2 208 15.0 8.0 3.1 0.5 25,084
3,000–4,999 TEU 968 923 0.95 54,160 36,549 24.1 236 16.1 13.9 3.9 0.6 53,737
5,000–7,999 TEU 575 552 0.96 75,036 54,838 25.1 246 16.3 19.5 4.1 0.6 42,960
8,000–11,999 TEU 331 325 0.98 108,650 67,676 25.5 256 16.3 24.4 4.5 0.7 30,052
12,000–14,500 TEU 103 98 0.95 176,783 83,609 28.9 241 16.1 23.7 4.9 0.8 8,775
14,500 –+
TEU 8 7 0.88 158,038 80,697 25.0 251 14.8 25.3 6.1 1.1 806
General cargo 0–4,999 dwt 11,620 5,163 0.44 1,925 1,119 11.6 161 8.7 0.5 0.1 0.0 23,606
5,000–9,999 dwt 2,894 2,491 0.86 7, 339 3,320 13.6 166 10.1 1.4 0.4 0.1 16,949
10,000–+
dwt 1,972 1,779 0.90 22,472 7,418 15.8 174 12.0 3.4 1.2 0.1 27, 6 01
Liquefied gas tanker 0–49,999 cbm 1,104 923 0.84 6,676 3,815 14.2 180 11.9 2.4 0.6 0.4 11,271
50,000–199,999 cbm 463 444 0.96 68,463 22,600 18.5 254 14.9 17.9 4.1 0.6 29,283
200,000–+
cbm 45 43 0.96 121,285 37,358 19.3 277 16.9 33.5 4.0 1.0 5,406
44 Third IMO GHG Study 2014
Table 14 – Tabular data for 2012 describing the fleet (international, domestic and fishing) analysed using the bottom-up method (continued)
Ship type Size category Units
Number
active
Decimal AIS
coverage of
in-service
ships
Avg.
dead-
weight
(tonnes)
Avg.
installed
power
(kW)
Avg.
design
speed
(knots)
Avg.
days at
sea
Avg.*
sea
speed
(knots)
Avg.* consumption
(‘000 tonnes)
Total CO
2
emissions
(‘000 tonnes)
IHSF AIS Main Auxiliary Boiler
Oil tanker 0–4,999 dwt 3,500 1,498 0.43 1,985 1,274 11. 5 144 8.7 0.6 0.6 0.2 14,991
5,000–9,999 dwt 664 577 0.87 6,777 2,846 12.6 147 9.1 1.1 1.0 0.3 4,630
10,000–19,999 dwt 190 171 0.90 15,129 4,631 13.4 149 9.6 1.6 1.7 0.4 2,121
20,000–59,999 dwt 659 624 0.95 43,763 8,625 14.8 164 11.7 3.7 2.0 0.6 12,627
60,000–79,999 dwt 391 381 0.97 72,901 12,102 15.1 183 12.2 5.8 1.9 0.6 9,950
80,000–119,999 dwt 917 890 0.97 109,259 13,813 15.3 186 11.6 5.9 2.6 0.8 25,769
120,000–199,999 dwt 473 447 0.95 162,348 18,796 16.0 206 11.7 8.0 3.1 1.0 17, 2 30
200,000–+
dwt 601 577 0.96 313,396 27,6 8 5 16.0 233 12.5 15.3 3.6 1.1 36,296
Other liquids
tankers
0–+
dwt 149 39 0.26 670 558 9.8 116 8.3 0.3 1.3 0.5 5,550
Ferry – pax only 0–1,999 gt 3,081 1,145 0.37 135 1,885 22.7 182 13.9 0.8 0.4 0.0 10,968
2,000–+
gt 71 52 0.73 1,681 6,594 16.6 215 12.8 3.9 1.0 0.0 1,074
Cruise 0–1,999 gt 198 75 0.38 137 914 12.4 102 8.8 0.3 1.0 0.5 1,105
2,000–9,999 gt 69 53 0.77 1,192 4,552 16.0 161 9.9 1.3 1.1 0.4 580
10,000–59,999 gt 115 108 0.94 4,408 19,657 19.9 217 13.8 9.1 9.2 1.4 6,929
60,000–99,999 gt 87 85 0.98 8,425 53,293 22.2 267 15.7 30.8 26.2 0.6 15,415
100,000–+
gt 51 51 1.00 11,711 76,117 22.7 261 16.4 47. 2 25.5 0.5 10,906
Ferry – ro-pax 0–1,999 gt 1,669 732 0.44 401 1,508 13.0 184 8.4 0.6 0.2 0.0 4,308
2,000–+
gt 1,198 1,046 0.87 3,221 15,491 21.6 198 13.9 6.0 1.4 0.0 26,753
Refrigerated bulk 0–1,999 dwt 1,090 763 0.70 5,695 5,029 16.8 173 13.4 3.0 2.3 0.4 17,9 45
Ro-ro 0–4,999 dwt 1,330 513 0.39 1,031 1,482 10.7 146 8.8 1.1 2.5 0.3 15,948
5,000–+
dwt 415 396 0.95 11,576 12,602 18.6 209 14.2 6.8 3.6 0.4 13,446
Vehicle 0–3,999 vehicle 279 261 0.94 9,052 9,084 18.3 222 14.2 5.4 1.6 0.3 6,200
4,000–+
vehicle 558 515 0.92 19,721 14,216 20.1 269 15.5 9.0 1.4 0.2 18,302
Yacht
0–+
gt 1,750 1,110 0.63 171 2,846 16.5
66 10.7 0.4 0.5 0.0 3,482
Service – tug
0–+
gt 14,641 5,043 0.34 119 2,313 11.8 100 6.7 0.4 0.1 0.0 21,301
Inventories of CO2 emissions from international shipping 2007–2012 45
Table 14 – Tabular data for 2012 describing the fleet (international, domestic and fishing) analysed using the bottom-up method (continued)
Ship type Size category Units
Number active
Decimal AIS
coverage of
in-service
ships
Avg.
dead-
weight
(tonnes)
Avg.
installed
power
(kW)
Avg.
design
speed
(knots)
Avg.
days at
sea
Avg.*
sea
speed
(knots)
Avg.* consumption
(‘000 tonnes)
Total CO
2
emissions
(‘000 tonnes)
IHSF AIS Main Auxiliary Boiler
Miscellaneous
– fishing
0–+
gt 22,130 4,510 0.20 181 956 11.5 164 7.4 0.4 0.4 0.0 50,959
Offshore
0–+
gt 6,480 5,082 0.78 1,716 4,711 13.8 106 8.0 0.7 0.6 0.0 27, 397
Service – other
0–+
gt 3,423 2,816 0.82 2,319 3,177 12.8 116 7.9 0.7 0.4 0.0 11,988
Miscellaneous
– other
0–+
gt 3,008 64 0.02 59 2,003 12.7 117 7. 3 0.4 0.4 0.0 7,425
* indicates the use of weighted averaging (weighted by days at sea for each individual ship).
Note: slight differences in Table 14 and Table 16 totals are due to rounding in values reported in the report. For 2012 the dfference is approximately 0.1%.
46 Third IMO GHG Study 2014
1.3.2 CO
2
and fuel consumption for multiple years 2007–2012
Figure 29 shows the year-on-year trends for the total CO
2
emissions of each ship type, as estimated using the
bottom-up method. Figure 30 and Figure 31 show the associated total fuel consumption estimates for all years
of the study, from both the top-down and bottom-up methods. The total CO
2
emissions aggregated to the
lowest level of detail in the top-down analysis (international, domestic and fishing) are presented in Table15
and Table 16.
Figure 29: CO
2
emissions by ship type (international shipping only),
calculated using the bottom-up method for all years 2007–2012
Table 15 – International, domestic and fishing CO
2
emissions 2007–2011 (million tonnes),
using the top-down method
Marine sector Fuel type 2007 2008 2009 2010 2011
International shipping HFO 542.1 551.2 516.6 557.1 554.0
MDO 83.4 72.8 79.8 90.4 94.9
LNG 0.0 0.0 0.0 0.0 0.0
Top-down international total All 625.5 624.0 596.4 6 47.5 648.9
Domestic navigation HFO 62.0 44.2 47.6 44.5 39.5
MDO 72.8 76.6 75.7 82.4 87.8
LNG 0.1 0.1 0.1 0.1 0.2
Top-down domestic total All 134.9 121.0 123.4 127.1 127.6
Fishing HFO 3.4 3.4 3.1 2.5 2.5
MDO 17. 3 15.7 16.0 16.7 16.4
LNG 0.1 0.1 0.1 0.1 0.1
Top-down fishing total All 20.8 19.2 19.3 19.2 19.0
Total CO
2
emissions 781.2 764.1 739.1 793.8 795.4
Table 16 – International, domestic and fishing CO
2
emissions 2007–2012 (million tonnes),
using the bottom-up method
Marine sector Fuel type 2007 2008 2009 2010 2011 2012
International shipping HFO 773.8 802.7 736.6 650.6 716.9 6 67.9
MDO 97. 2 102.9 104.2 102.2 109.8 105.2
LNG 13.9 15.4 14.2 18.6 22.8 22.6
Bottom-up international total All 884.9 920.9 855.1 771.4 849.5 795.7
Domestic navigation HFO 53.8 57.4 32.5 45.1 61.7 39.9
MDO 142.7 138.8 80.1 88.2 98.1 91.6
LNG 0 0 0 0 0 0
Bottom-up domestic total All 196.5 196.2 112.6 133.3 159.7 131.4
Fishing HFO 1.6 1.5 0.9 0.8 1.4 1.1
MDO 17.0 16.4 9.3 9.2 10.9 9.9
LNG 0 0 0 0 0 0
Bottom-up fishing total All 18.6 18.0 10.2 10.0 12.3 11.0
Total CO
2
emissions 1,100.1 1,135.1 977.9 914.7 1,021.6 938.1
Total fuel consumption estimates for 2007–2012 using the bottom-up method are presented in Figure 30
for all ships and in Figure 31 for international shipping. These results are presented alongside the multi-year
top-down fuel consumption results presented in Section 1.1.3. Section 1.4.4 discusses the differences between
fuel consumption and emissions estimates from these methods.
Figure 30: Summary graph of annual fuel use by all ships,
estimated using the top-down and bottom-up methods
Inventories of CO2 emissions from international shipping 2007–2012 47
Table 16 – International, domestic and fishing CO
2
emissions 2007–2012 (million tonnes),
using the bottom-up method
Marine sector Fuel type 2007 2008 2009 2010 2011 2012
International shipping HFO 773.8 802.7 736.6 650.6 716.9 667.9
MDO 97. 2 102.9 104.2 102.2 109.8 105.2
LNG 13.9 15.4 14.2 18.6 22.8 22.6
Bottom-up international total All 884.9 920.9 855.1 771.4 849.5 795.7
Domestic navigation HFO 53.8 57.4 32.5 45.1 61.7 39.9
MDO 142.7 138.8 80.1 88.2 98.1 91.6
LNG 0 0 0 0 0 0
Bottom-up domestic total All 196.5 196.2 112.6 133.3 159.7 131.4
Fishing HFO 1.6 1.5 0.9 0.8 1.4 1.1
MDO 17.0 16.4 9.3 9.2 10.9 9.9
LNG 0 0 0 0 0 0
Bottom-up fishing total All 18.6 18.0 10.2 10.0 12.3 11.0
Total CO
2
emissions 1,100.1 1,135.1 977.9 914.7 1,021.6 938.1
Total fuel consumption estimates for 2007–2012 using the bottom-up method are presented in Figure 30
for all ships and in Figure 31 for international shipping. These results are presented alongside the multi-year
top-down fuel consumption results presented in Section 1.1.3. Section 1.4.4 discusses the differences between
fuel consumption and emissions estimates from these methods.
Figure 30: Summary graph of annual fuel use by all ships,
estimated using the top-down and bottom-up methods
48 Third IMO GHG Study 2014
Figure 31: Summary graph of annual fuel use by international shipping,
estimated using the top-down and bottom-up methods
Particular care must be taken when interpreting the domestic fuel consumption and emissions estimates from
both the top-down and the bottom-up methods. Depending on where domestic shipping and fishing buys its
fuel, it may or may not be adequately captured in the IEA marine bunkers. For example, inland or leisure and
fishing vessels may purchase fuel at locations that also sell fuel to other sectors of the economy and therefore
be misallocated. In the bottom-up method, fuel consumption is included only for ships that appear in the
IHSF database (and have an IMO number). While this should cover all international shipping, many domestic
vessels (inland, fishing or cabotage) may not be included in this database. An indication of the number of
vessels excluded from the bottom-up method was obtained from the count of MMSI numbers observed on
AIS but for which no match to the IHSF database was obtained. The implications of this count for both the
bottom-up and top-down analysis are discussed in Section 1.4.
1.3.3 Trends in emissions and drivers of emissions 2007–2012
Figures 32–37 present indexed time series of the total CO
2
emissions for three ship types – oil tankers, container
ships and bulk carriers – during the period studied. The figures also present a number of key drivers of CO
2
emissions estimated in the bottom-up method that can be used to decompose CO
2
emissions trends:
• the total CO
2
emissions are a function of the total number of ships and average annual fuel consumption;
• the average annual fuel consumption is primarily a function of days at sea and the extent of adoption
of slow steaming;
• all trends are indexed to their values in 2007.
These drivers of average annual fuel consumption can also be influenced by changes in the average specification
of the fleet (average capacity, average installed power, etc.). These are of less significance than the key trends
of speed and days at sea.
The contrast between the three plots shows that these three sectors of the shipping industry have changed
in different ways over the period 2007–2012. The oil tanker sector reduced its emissions by a total of 20%.
During the same period the dry bulk and container ship sectors also saw absolute emissions reductions but by
smaller amounts. All ship types experienced similar reductions in average annual fuel consumption, but the
difference in fleet total CO
2
emissions is explained by the combination of these reductions with differences in
the number of ships in service. The reduction in average days at sea during the period studied is greatest in the
Inventories of CO2 emissions from international shipping 2007–2012 49
dry bulk fleet, whereas the container ship fleet has seen a slight increase. Consistent with the results presented
in Table 17, more container ships adopted slow steaming operations. In other words, similar reductions in
average fuel consumption per ship over the study period were achieved through different combinations of
speed and days at sea.
The analysis of trends in speed and days at sea are consistent with the findings from Section 3 that the global
fleet is currently at or near the historic low in terms of productivity (transport work per unit of capacity).
(See Section 3.2.4 and related text and Annex 7, Figures 38–40, for further details.) The consequence is that
these (and many other) sectors of the shipping industry represent latent emissions increases, because the
fundamentals (number of ships in service, fleet total installed power and demand tonne-miles) have seen
upward trends. These upward trends have been controlled because economic pressures (excess supply of
fleet as demonstrated by the relative supply and demand growth in each plot), together with high fuel prices,
have acted to reduce productivity (reducing both average operating speeds and days spent at sea in both the
oil tanker and bulk carrier fleets, and only operating speeds in the container fleets). These two components
of productivity are both liable to change if the supply and demand differential returns to historical long-run
trends. Therefore, whether and when the latent emissions may appear is uncertain, as this depends on the
future market dynamics of the industry. However, the risk is high that fleet “potential to emit” (e.g. fleet-
average installed power and design speeds) could encounter conditions favouring the conversion of latent
emissions to actual emissions; this could mean that shipping reverts to the trajectory estimated in the Second
IMO GHG Study 2009. The potential for latent emissions to be realized is quantified in the sensitivity analysis
in Section3.3.4 (see Figure 88 and related text).
50 Third IMO GHG Study 2014
Figure 32: Average trends in the tanker sector 2007–2012, indexed to 2007
Figure 33: Average trends in the bulk carrier sector 2007–2012, indexed to 2007
Figure 34: Average trends in the container ship sector 2007–2012, indexed to 2007
Figure 35: Fleet total trends in the oil tanker sector (2007–2012), indexed to 2007
Figure 36: Fleet total trends in the bulk carrier sector (2007¬–2012), indexed to 2007
Figure 37: Fleet total trends in the container ship sector (2007–2012), indexed to 2007
Inventories of CO2 emissions from international shipping 2007–2012 51
Figure 35: Fleet total trends in the oil tanker sector (2007–2012), indexed to 2007
Figure 36: Fleet total trends in the bulk carrier sector (2007¬–2012), indexed to 2007
Figure 37: Fleet total trends in the container ship sector (2007–2012), indexed to 2007
52 Third IMO GHG Study 2014
Figure 38: Variability within ship size categories in the bulk ship fleet (2012). Size category 1
is the smallest bulk carrier (0–9,999 dwt) and size category 6 is the largest (200,000+ dwt)
Figure 39: Variability within ship size categories in the container ship fleet (2012). Size category 1
is the smallest container ship (0–999 TEU) and size category 8 is the largest (14,500+ TEU)
Figure 40: Variability within ship size categories in the tanker fleet (2012). Size category 1
is the smallest oil tanker (0–9,999 dwt) and size category 8 is the largest (200,000+ dwt)
1.3.4 Variability between ships of a similar type and size and the impact of slow steaming
The bottom-up method calculates ship type totals by summing the calculations for each individual ship
identified as in service in the IHSF database. This study therefore supersedes the Second IMO GHG Study
2009 in providing insight into individual ships within fleets of similar ships. To illustrate this, Figures 38–40
display the statistics for the bulk carrier, container ship and tanker fleets. The plots represent each ship type’s
population by ship size category (on the x-axis). The box plots convey the average ship (red line in the middle
of the box), the interquartile range (between the 25th and 75th percentile of the population) and the 2nd
to 98th percentile range (the extremes of the “whiskers”). Tabular data characterizing each ship type and
size category studied are included in Annex 2.The average sailing speed in 2012 of container ships in size
categories 4–7 (3,000 TEU to 14,500 TEU capacity) is between 16 knots and 16.3 knots (Figure 39). The
interquartile range of sailing speed is approximately 1 knot to 2 knots, depending on the size. This shows little
variability in operating speed across the sector (nearly 2,000 ships). The average speed of ships in those four
size categories varies between 24 knots and 29 knots. Therefore the sailing speed plot also shows the extent
to which ships are slow steaming in 2012. The ratio of operating speed to design speed (here approximated as
the IHSF reference speed) can be seen in the bottom left-hand plot (Figure 39), showing that larger ships (bin
8 in Figure 39) are on average operating at between 55% and 65% of their design speed. Although they have
lower design speeds than the larger ships, in ratio terms the smaller container ships (sizes 1 and 2) are slow
steaming less than the larger ships.
The top left of the plots portrays the estimated total annual main engine fuel consumption. In this instance
there is a comparatively higher variability within the population than observed for sailing speed. Some of this
is due to the variability in ship technical specifications (hull form, installed power and design speed). There
is also variability in the total fuel consumption because of variability in the number of sailing days in a year
(bottom right-hand plot). Holding all else equal, an increase in days at sea will increase total annual main-
engine fuel consumption by the same percentage.
Inventories of CO2 emissions from international shipping 2007–2012 53
Figure 40: Variability within ship size categories in the tanker fleet (2012). Size category 1
is the smallest oil tanker (0–9,999 dwt) and size category 8 is the largest (200,000+ dwt)
1.3.4 Variability between ships of a similar type and size and the impact of slow steaming
The bottom-up method calculates ship type totals by summing the calculations for each individual ship
identified as in service in the IHSF database. This study therefore supersedes the Second IMO GHG Study
2009 in providing insight into individual ships within fleets of similar ships. To illustrate this, Figures 38–40
display the statistics for the bulk carrier, container ship and tanker fleets. The plots represent each ship type’s
population by ship size category (on the x-axis). The box plots convey the average ship (red line in the middle
of the box), the interquartile range (between the 25th and 75th percentile of the population) and the 2nd
to 98th percentile range (the extremes of the “whiskers”). Tabular data characterizing each ship type and
size category studied are included in Annex 2.The average sailing speed in 2012 of container ships in size
categories 4–7 (3,000 TEU to 14,500 TEU capacity) is between 16 knots and 16.3 knots (Figure 39). The
interquartile range of sailing speed is approximately 1 knot to 2 knots, depending on the size. This shows little
variability in operating speed across the sector (nearly 2,000 ships). The average speed of ships in those four
size categories varies between 24 knots and 29 knots. Therefore the sailing speed plot also shows the extent
to which ships are slow steaming in 2012. The ratio of operating speed to design speed (here approximated as
the IHSF reference speed) can be seen in the bottom left-hand plot (Figure 39), showing that larger ships (bin
8 in Figure 39) are on average operating at between 55% and 65% of their design speed. Although they have
lower design speeds than the larger ships, in ratio terms the smaller container ships (sizes 1 and 2) are slow
steaming less than the larger ships.
The top left of the plots portrays the estimated total annual main engine fuel consumption. In this instance
there is a comparatively higher variability within the population than observed for sailing speed. Some of this
is due to the variability in ship technical specifications (hull form, installed power and design speed). There
is also variability in the total fuel consumption because of variability in the number of sailing days in a year
(bottom right-hand plot). Holding all else equal, an increase in days at sea will increase total annual main-
engine fuel consumption by the same percentage.
54 Third IMO GHG Study 2014
The results for oil tankers show a similar level of variability within a given ship size group, a significant
(although not as significant as container ships) uptake of slow steaming and similarities between the larger ship
types in terms of sailing speeds and days spent at sea.
The bottom-up method also allows the influence of slow steaming to be quantified. Across all ship types and
sizes, the average ratio of operating speed to design speed was 0.85 in 2007 and 0.75 in 2012. This shows
that, in relative terms, ships have slowed down: the widely reported phenomenon of slow steaming that has
occurred since the financial crisis. The consequence of this observed slow steaming is a reduction in daily
fuel of approximately 27% expressed as an average across all ship types and sizes. However, that average
value belies the significant operational changes that have occurred in certain ship type and size categories.
Table 17 describes, for three of the ship types studied, the ratio between slow steaming percentage (average
at-sea operating speed expressed as a percentage of design speed), the average at-sea main engine load factor
(a percentage of the total installed power produced by the main engine) and average at-sea main engine
daily fuel consumption. Many of the larger ship sizes in all three ship type categories are estimated to have
experienced reductions in daily fuel consumption well in excess of the average value of 25%.
The ships with the highest design speeds have adopted the greatest levels of slow steaming (e.g. container
ships are operating at average speeds much lower than their design speeds); there is also widespread adoption
of significant levels of slow steaming in many of the oil tanker size categories. Concurrent with the observed
trend, technical specifications changed for ships. The largest bulk carriers (200,000+ dwt capacity) saw
increases in average size (dwt capacity), as well as increased installed power (from an average of 18.9MW to
22.2MW), as a result of a large number of new ships entering the fleet over the time period (the fleet grew
from 102 ships in 2007 to 294 ships in 2012).
A reduction in speed and the associated reduction in fuel consumption do not relate to an equivalent
percentage increase in efficiency, because a greater number of ships (or more days at sea) are required to do
the same amount of transport work. This relationship is discussed in greater detail in Section 3.
Inventories of CO2 emissions from international shipping 2007–2012 55
Table 17 – Relationship between slow steaming, engine load factor (power output) and fuel consumption for 2007 and 2012
Ship type Size category Unit 2007 2012 % change
in average
at-sea tonnes
per day (tpd)
2007–2012
Ratio of average
at-sea speed to
design speed
Average at-sea
main engine load
factor
(% MCR)
At-sea
consumption in
tonnes per day
(tpd)
Ratio of average
at-sea speed to
design speed
Average at-sea
main engine load
factor
(% MCR)
At-sea
consumption in
tonnes per day
(tpd)
Bulk carrier 0–9,999 dwt 0.92 92% 7.0 0.84 70% 5.5
-24%
10,000–34,999 0.86 68% 22.2 0.82 59% 17. 6
-23%
35,000–59,999 0.88 73% 29.0 0.82 58% 23.4
-21%
60,000–99,999 0.90 78% 37. 7 0.83 60% 28.8
-27%
100,000–199,999 0.89 77% 55.5 0.81 57% 42.3
-27%
200,000–+
0.82 66% 51.2 0.84 62% 56.3 10%
Container 0–999 TEU 0.82 62% 17. 5 0.77 52% 14.4
-19%
1,000 –1,999 0.80 58% 33.8 0.73 45% 26.0
-26%
2,000–2,999 0.80 58% 55.9 0.70 39% 38.5
-37%
3,000–4,999 0.80 59% 90.4 0.68 36% 58.7
-42%
5,000–7,999 0.82 63% 151.7 0.65 32% 79.3
-63%
8,000–11,999 0.85 69% 200.0 0.65 32% 95.6
-71%
12,000–14,500 0.84 67% 231.7 0.66 34% 10 7. 8
-73%
14,500 –+
– – – 0.60 28% 100.0 –
Oil tanker 0–4,999 dwt 0.89 85% 5.1 0.80 67% 4.3
-18%
5,000–9,999 0.83 64% 9.2 0.75 49% 7.1
-26%
10,000–19,999 0.81 61% 15.3 0.76 49% 10.8
-34%
20,000–59,999 0.87 72% 28.8 0.80 55% 22.2
-26%
60,000–79,999 0.91 83% 45.0 0.81 57% 31.4
-35%
80,000–119,999 0.91 81% 49.2 0.78 51% 31.5
-44%
120,000–199,999 0.92 83% 65.4 0.77 49% 39.4
-50%
200,000–+
0.95 90% 103.2 0.80 54% 65.2
-45%
56 Third IMO GHG Study 2014
1.3.5 Shipping’s CO
2
e emissions
Carbon dioxide equivalency (CO
2
e) is a quantity that describes, for a given amount of GHG, the amount of
CO
2
that would have the same global warming potential (GWP) as another long-lived emitted substance,
when measured over a specified timescale (generally, 100 years). A total CO
2
e estimate is produced by
combining CO
2
emissions totals estimated in Section 1 with other GHG substances estimated in Section 2
and their associated GWP.
The Fifth IPCC Assessment Report (AR5) has changed the 100-year global warming potentials (GWP100)
from previous assessments because of new estimates of lifetimes, impulse response functions and radiative
efficiencies. IPCC (2013) acknowledges that the inclusion of indirect effects and feedbacks in metric values
has been inconsistent in IPCC reports, and therefore the GWPs presented in previous assessments may
underestimate the relative impacts of non-CO
2
gases.
The GWPs reported in IPCC (2013) include climate-carbon feedbacks for the reference gas CO
2
, and for the
non-CO
2
gases, GWPs are presented both with and without climate-carbon feedbacks. In accord with IPCC
(2013), such feedbacks may have significant impacts on metrics and should be treated consistently.
Using GWP100 with climate-carbon feedbacks, primary GHGs (CO
2
, N
2
O and CH
4
) from shipping account
for approximately 961 million tonnes of CO
2
e in 2012. International shipping is estimated to account for
816million tonnes of CO
2
e for primary GHGs in 2012.
Time series of bottom-up CO
2
e emissions estimates with climate-carbon feedbacks can be found in Table 18
and Table 19 and are presented in Figure 41.
Table 18 – Bottom-up CO
2
e emissions estimates with climate-carbon feedbacks
from total shipping (thousand tonnes)
2007 2008 2009 2010 2011 2012
CH
4
6,018 6,657 6,369 8,030 9,807 9,802
N
2
O 14,879 15,404 13,318 12,453 13,428 12,707
CO
2
1,100,100 1,135,100 977,9 0 0 914,700 1,021,600 938,100
Total 1,120,997 1,157,160 997, 5 87 935,183 1,044,835 960,608
Table 19 – Bottom-up CO
2
e emissions estimates with climate-carbon feedbacks
from international shipping (thousand tonnes)
2007 2008 2009 2010 2011 2012
CH
4
5,929 6,568 6,323 7,9 69 9,740 9,742
N
2
O 12,152 12,689 11,860 10,615 11,437 10,931
CO
2
884,900 920,900 855,100 771,400 849,500 795,700
Total 902,981 940,157 873,284 789,983 870,678 816,372
a) Total shipping b) International shipping
Figure 41: Time series of bottom-up CO
2
e emissions estimates for a) total shipping and
b) international shipping
1.3.6 Shipping as a share of global emissions
Inventories of ship emissions can be compared with global anthropogenic totals to quantify the contribution
of shipping to GHG totals from all human activity. The consortium evaluated AR5, a comprehensive technical
document that has assembled global emissions estimates (IPCC 2013). AR5 provides global emissions totals
for the year 2010 for a number of GHG substances, including CO
2
, CH
4
and N
2
O. It also refers to two sources
that provide annual CO
2
emissions for the years 2007–2012 (Boden et al., 2013; Peters et al., 2013). Totals were
converted from elemental C to CO
2
for comparison with the current study.
Comparisons of major GHGs from shipping are presented in Tables 20–23, using global totals identified in the
recent AR5 (IPCC 2013). For the period 2007-2012, on average, shipping accounted for approximately 3.1% of
annual global CO
2
and approximately 2.8% of annual GHGs on a CO
2
e basis. International shipping accounts,
on average, for approximately 2.6% and 2.4% of CO
2
and GHGs on a CO
2
e basis, respectively. These CO
2
and CO
2
e comparisons are similar to, but slightly smaller than, the 3.3% and 2.7% of global CO
2
emissions
reported by Second IMO GHG Study 2009 for total shipping and international shipping respectively.
Table 20 – Shipping CO
2
emissions compared with global CO
2
(values in million tonnes CO
2
)
Third IMO GHG Study 2014
Year Global CO
2
1
Total shipping CO
2
Percentage
of global
International shipping CO
2
Percentage
of global
2007 31,409 1,100 3.5% 885 2.8%
2008 32,204 1,135 3.5% 921 2.9%
2009 32,047 978 3.1% 855 2.7%
2010 33,612 915 2.7% 771 2.3%
2011 34,723 1,022 2.9% 850 2.4%
2012 35,640 938 2.6% 796 2.2%
Average 33,273 1,015 3.1% 846 2.6%
1
Global comparator represents CO
2
from fossil fuel consumption and cement production, converted from Tg C y
–1
to million tonnes CO
2
.
Sources: Boden et al., 2013, for years 2007–2010; Peters et al., 2013, for years 2011–2012, as referenced in IPCC (2013).
Inventories of CO2 emissions from international shipping 2007–2012 57
a) Total shipping b) International shipping
Figure 41: Time series of bottom-up CO
2
e emissions estimates for a) total shipping and
b) international shipping
1.3.6 Shipping as a share of global emissions
Inventories of ship emissions can be compared with global anthropogenic totals to quantify the contribution
of shipping to GHG totals from all human activity. The consortium evaluated AR5, a comprehensive technical
document that has assembled global emissions estimates (IPCC 2013). AR5 provides global emissions totals
for the year 2010 for a number of GHG substances, including CO
2
, CH
4
and N
2
O. It also refers to two sources
that provide annual CO
2
emissions for the years 2007–2012 (Boden et al., 2013; Peters et al., 2013). Totals were
converted from elemental C to CO
2
for comparison with the current study.
Comparisons of major GHGs from shipping are presented in Tables 20–23, using global totals identified in the
recent AR5 (IPCC 2013). For the period 2007-2012, on average, shipping accounted for approximately 3.1% of
annual global CO
2
and approximately 2.8% of annual GHGs on a CO
2
e basis. International shipping accounts,
on average, for approximately 2.6% and 2.4% of CO
2
and GHGs on a CO
2
e basis, respectively. These CO
2
and CO
2
e comparisons are similar to, but slightly smaller than, the 3.3% and 2.7% of global CO
2
emissions
reported by Second IMO GHG Study 2009 for total shipping and international shipping respectively.
Table 20 – Shipping CO
2
emissions compared with global CO
2
(values in million tonnes CO
2
)
Third IMO GHG Study 2014
Year Global CO
2
1
Total shipping CO
2
Percentage
of global
International shipping CO
2
Percentage
of global
2007 31,409 1,100 3.5% 885 2.8%
2008 32,204 1,135 3.5% 921 2.9%
2009 32,047 978 3.1% 855 2.7%
2010 33,612 915 2.7% 771 2.3%
2011 34,723 1,022 2.9% 850 2.4%
2012 35,640 938 2.6% 796 2.2%
Average 33,273 1,015 3.1% 846 2.6%
1
Global comparator represents CO
2
from fossil fuel consumption and cement production, converted from Tg C y
–1
to million tonnes CO
2
.
Sources: Boden et al., 2013, for years 2007–2010; Peters et al., 2013, for years 2011–2012, as referenced in IPCC (2013).
58 Third IMO GHG Study 2014
Table 21 – Shipping CH
4
emissions compared with global CH
4
(values in thousand tonnes CH
4
)
Third IMO GHG Study 2014
Year Global CH
4
1
Total shipping CH
4
Percentage
of global
International shipping CH
4
Percentage
of global
Average
annual
CH
4
for
decade
2000–09
96,000 177 0.18% 174 0.18%
196 0.20% 193 0.20%
187 0.20% 186 0.19%
236 0.25% 234 0.24%
288 0.30% 286 0.30%
288 0.30% 287 0.30%
Average 229 0.24% 227 0.24%
1
Global comparator represents CH
4
from fossil fuel consumption and cement production. Source: IPCC (2013, Table 6.8).
Table 22 – Shipping N
2
O emissions compared with global N
2
O (values in thousand tonnes N
2
O)
Third IMO GHG Study 2014
Year Global N
2
O
1
Total shipping N
2
O
Percentage
of global
International shipping N
2
O
Percentage
of global
Average
annual
N
2
O for
decade
2000–09
700 50 7.1% 41 5.8%
52 7.4% 43 6.1%
45 6.4% 40 5.7%
42 6.0% 36 5.1%
45 6.4% 38 5.5%
43 6.1% 37 5.2%
Average 46 6.6% 39 5.6%
1
Global comparator represents N
2
O from fossil fuel consumption and cement production. Source: IPCC (2013, Table 6.9).
Table 23 – Shipping GHGs (in CO
2
e) compared with global GHGs (values in million tonnes CO
2
e)
Third IMO GHG Study 2014
Year Global CO
2
e
1
Total shipping CO
2
e
Percentage
of global
International shipping CO
2
e
Percentage
of global
2007 34,881 1,121 3.2% 903 2.6%
2008 35,677 1,157 3.2% 940 2.6%
2009 35,519 998 2.8% 873 2.5%
2010 37,0 85 935 2.5% 790 2.1%
2011 38,196 1,045 2.7% 871 2.3%
2012 39,113 961 2.5% 816 2.1%
Average 36,745 1,036 2.8% 866 2.4%
1
Global comparator represents N
2
O from fossil fuel consumption and cement production. Source: IPCC (2013, Table 6.9).
For the year 2012, total shipping emissions were approximately 938 million tonnes CO
2
and 961 million tonnes
CO
2
e for GHGs combining CO
2
, CH
4
and N
2
O. International shipping emissions for 2012 are estimated to be
796 million tonnes CO
2
and 816 million tonnes CO
2
e for GHGs combining CO
2
, CH
4
and N
2
O. International
shipping accounts for approximately 2.2% and 2.1% of CO
2
and GHGs on a CO
2
e basis, respectively.
Table 20 and Table 23 are also illustrated graphically in Figure 42 a) and b) respectively. The bar graphs may
show more intuitively that global CO
2
and CO
2
e are increasing at different rates than recently observed in
the bottom-up results for shipping presented here. In other words, ship fuel use, CO
2
emissions and GHG
emissions (on a CO
2
e basis) have trended nearly flat while estimated global totals of these emissions have
increased; this results in a recent-year decline in the percentage of shipping emissions as a fraction of global
totals.
Inventories of CO2 emissions from international shipping 2007–2012 59
Figure 42: Comparison of shipping with global totals: a) CO
2
emissions compared, where the
percentage indicates international shipping emissions of CO
2
as a percentage of global CO
2
from fossil fuels;
b) CO
2
e emissions compared, where the percentage indicates international shipping emissions of CO
2
e as a
percentage of global CO
2
e from fossil fuels
60 Third IMO GHG Study 2014
1.4 Quality assurance and control of top-down and bottom-up inventories
The quality analysis is presented in three sections. The first section discusses QA/QC for the top-down
emissions inventory. The second section summarizes the QA/QC elements of the bottom-up fuel and emissions
inventory. The third section contains a comparison of the top-down and bottom-up emissions inventories.
Sections 1.1 and 1.2 contain many detailed processes that constitute QA/QC effort; these sections therefore
discuss QA/QC mainly in summary and provide context for the quantitative bottom-up uncertainty analysis
in Section 1.5.
1.4.1 Top-down QA/QC
Top-down statistics were evaluated for transparency and any significant discrepancies that might reflect
confidence in inventories based on fuel statistics.
This section begins with a review of the Second IMO GHG Study 2009 and a brief discussion of data quality,
confidence and uncertainty. It reviews relevant data quality information provided by IEA, including information
about likely causes of potential under- or overestimation of marine fuel use (both domestic and international).
Top-down method QA/QC efforts undertaken specifically for this study are described. Lastly, this section
gives a QA/QC summary of the study.
Second IMO GHG Study 2009: review of top-down data quality
The Second IMO GHG Study 2009 performed qualitative analyses of errors and inconsistencies of IEA statistics
to help explore how the top-down and bottom-up discrepancy may be explained by uncertainty in reported
fuel statistics. That study identified the following potential issues with top-down data:
• different data quality between OECD and non-OECD countries (fishing);
• identical numbers from year to year for some countries;
• big swings from year to year for other countries;
• differences in EIA bunkers statistics.
Although a number of challenges were recognized, mainly arising from the use of different data sources, the
sources of uncertainty remained unexplored and potential corrections were not attempted.
The Second IMO GHG Study 2009 explicitly quoted provisions in the IEA Agreement on an International
Energy Program (IEP) that determined which fuels would be considered in national oil stocks and which were
considered to be counted as international data. In particular, international marine bunkers were “treated as
exports under a 1976 Governing Board decision incorporated into the Emergency Management Manual”
(Scott, 1994). This information and subsequent discussion in the Second IMO GHG Study 2009 suggested that
some degree of allocation error among international bunkers, exports and/or imports could be a factor in the
accuracy of top-down fuel statistics for shipping.
IEA statistics: review of top-down data quality
IEA collects data from OECD countries that have agreed to report mandatory data through monthly and joint
annual IEA/Eurostat/UNECE questionnaires. For non-OECD countries, IEA collects data through voluntary
submissions (using no standard format) or through estimates made by IEA or its contractors. Figure 43 presents
a map of OECD and non-OECD countries that provide energy data to IEA; not all of these countries have
marine fuel sales to report (Morel, 2013).
Figure 43: OECD versus non-OECD data collection system
IEA acknowledges that challenges remain in collecting international marine bunkers data worldwide; however,
compared to other sources, the IEA database seems consistent across the years and is regulary updated.
According to Morel (2013), the revisions in the IEA international marine bunkers database have improved its
quality. The database published in 2012 covers 139 individual countries compared to the 137 of the 2007
database. Of these 139 countries, the 54 countries that represent 80% of the total sale have used official
energy statistics. Another six countries, representing 14% of the total sale, have used other sources, such as
port authorities, oil companies and data provided by FACTS Global Energy (http://www.fgenergy.com). Lastly,
in 2012 edition, data have been estimated for 33 countries that represent only 6% of the total sale, considering,
for example, residual GDP growth and marine traffic growth (Morel, 2013).
In addition to directly reported IEA marine fuel statistics, the consortium reviewed the energy balances of each
fuel to inform the uncertainty analysis for top-down marine fuel consumption in Section 1.5. This provides
QA/QC and enables an estimate of potential uncertainty around reported fuel sales for the marine sector
(domestic and international).
For example, corroborating information about the potential for under- or overreporting international marine
bunkers includes:
1 From Energy Statistics for Non-OECD Countries, IEA, 2009 edition: “For a given product, imports and
exports may not sum up to zero at the world level for a number of reasons. Fuels may be classified
differently (i.e. residual fuel oil exports may be reported as refinery feedstocks by the importing
country; NGL exports may be reported as LPG by the importing country, etc.). Other possible reasons
include discrepancies in conversion factors, inclusion of international marine bunkers in exports,
timing differences, data reported on a fiscal year basis instead of calendar year for certain countries,
and underreporting of imports and exports for fiscal reasons.”
2 From the OECD Factbook 2013: Economic, Environmental and Social Statistics (“Energy supply”, page
108) and the Factbook website: “Data quality is not homogeneous for all countries and regions. In
some countries, data are based on secondary sources, and where incomplete or unavailable, the IEA
has made estimates. In general, data are likely to be more accurate for production and trade than for
international bunkers or stock changes. Moreover, statistics for biofuels and waste are less accurate
than those for traditional commercial energy data.”
In summary, IEA and OECD identify specific types of error in energy data that involve marine bunkers. The
first is allocation or classification error involving imports, exports and marine bunker statistics. The second is
country-to-country differences in data quality, specifically related to poor accuracy for international bunkers
or stock changes. These insights helped inform the consortium’s direct QA/QC and uncertainty efforts.
Inventories of CO2 emissions from international shipping 2007–2012 61
Figure 43: OECD versus non-OECD data collection system
IEA acknowledges that challenges remain in collecting international marine bunkers data worldwide; however,
compared to other sources, the IEA database seems consistent across the years and is regulary updated.
According to Morel (2013), the revisions in the IEA international marine bunkers database have improved its
quality. The database published in 2012 covers 139 individual countries compared to the 137 of the 2007
database. Of these 139 countries, the 54 countries that represent 80% of the total sale have used official
energy statistics. Another six countries, representing 14% of the total sale, have used other sources, such as
port authorities, oil companies and data provided by FACTS Global Energy (http://www.fgenergy.com). Lastly,
in 2012 edition, data have been estimated for 33 countries that represent only 6% of the total sale, considering,
for example, residual GDP growth and marine traffic growth (Morel, 2013).
In addition to directly reported IEA marine fuel statistics, the consortium reviewed the energy balances of each
fuel to inform the uncertainty analysis for top-down marine fuel consumption in Section 1.5. This provides
QA/QC and enables an estimate of potential uncertainty around reported fuel sales for the marine sector
(domestic and international).
For example, corroborating information about the potential for under- or overreporting international marine
bunkers includes:
1 From Energy Statistics for Non-OECD Countries, IEA, 2009 edition: “For a given product, imports and
exports may not sum up to zero at the world level for a number of reasons. Fuels may be classified
differently (i.e. residual fuel oil exports may be reported as refinery feedstocks by the importing
country; NGL exports may be reported as LPG by the importing country, etc.). Other possible reasons
include discrepancies in conversion factors, inclusion of international marine bunkers in exports,
timing differences, data reported on a fiscal year basis instead of calendar year for certain countries,
and underreporting of imports and exports for fiscal reasons.”
2 From the OECD Factbook 2013: Economic, Environmental and Social Statistics (“Energy supply”, page
108) and the Factbook website: “Data quality is not homogeneous for all countries and regions. In
some countries, data are based on secondary sources, and where incomplete or unavailable, the IEA
has made estimates. In general, data are likely to be more accurate for production and trade than for
international bunkers or stock changes. Moreover, statistics for biofuels and waste are less accurate
than those for traditional commercial energy data.”
In summary, IEA and OECD identify specific types of error in energy data that involve marine bunkers. The
first is allocation or classification error involving imports, exports and marine bunker statistics. The second is
country-to-country differences in data quality, specifically related to poor accuracy for international bunkers
or stock changes. These insights helped inform the consortium’s direct QA/QC and uncertainty efforts.
62 Third IMO GHG Study 2014
1.4.2 Top-down QA/QC efforts specific to this study
This study independently confirmed the statistical balances of IEA energy statistics on both global and large
regional scales. Specifically, the calculation of statistical difference at the national and regional levels was
verified and discrepancy between imports and exports reported by IEA was confirmed.
Second, as in the Second IMO GHG Study 2009, the consortium researched other international energy data
providers to understand whether international marine bunker records were considered to be similar to or
different from IEA statistics. This included research into data quality studies for non-IEA energy statistics.
Comparisons with EIA top-down statistics and other resources
The following resources were evaluated for a) their similarity to IEA statistics and b) complementary data
quality investigations.
The consortium evaluated EIA international marine bunker fuel oil data for 2007–2010 (IEA did not provide
more recent data than 2010 during the period in which this study was conducted). Moreover, the EIA statistics
available on the United States Department of Energy website did not provide data for gas diesel international
marine bunkers, nor break down domestic marine fuel consumption, nor identify fishing vessel consumption.
These data may be available from the EIA; however, given that additional EIA data provide limited opportunities
to improve QA/QC in top-down estimates, these data were not pursued.
Table 24 and Figure 44 illustrate continued discrepancies in statistical reporting between IEA and EIA, similar
to those documented in the Second IMO GHG Study 2009. Namely, the IEA data report consistently greater
fuel oil consumption than the EIA data for international marine bunkers. This is indicated in Figure 45 by the
scatter plot for the period 2000–2010, the regression line and the confidence interval of the best-fit line.
Table 24 – Comparison of fuel sales data between IEA and EIA in international shipping (million tonnes)
Fuel oil statistics Source 2007 2008 2009 2010 2011
International marine bunkers IEA 174.1 17 7.0 165.9 178.9 17 7.9
EIA 155.3 158.8 160.9 171.2
Percentage difference 11% 10% 3% 4%
Figure 44: Comparison of IEA and EIA international marine bunker fuel oil statistics
Figure 45: Confidence bands showing statistical difference between IEA and EIA data, 2000–2010
Results of top-down QA/QC
The top-down QA/QC provides a thorough understanding of the quality and limitations of the top-down
inventory. This review shows that IEA revisions to statistics can change the total fuel sales estimate by as much
as 10% owing to documented quality controls in place at IEA. A rigorous review of IEA QA/QC practices
indicates that the energy balances continue to represent high-quality representation of OECD and non-OECD
energy statistics.
Our IEA data comparison with EIA fuel oil statistics for international marine bunkers indicate that year-on-year
fuel sales data can differ by more than 10% and that IEA tends to report more international marine bunkers
over the period 2000–2010.
Lastly, the IEA presentation to the IMO Expert Workshop in 2013 indicated that significant uncertainties are
not fully documented and require further analysis (see Section 1.5). For example, under- or overestimates of
international marine bunkers could result from allocation or classification errors – imports, exports, marine
bunker statistics, fuel transfers between sectors (as is typical for blending marine bunkers with other fuels to
meet ship/engine fuel quality specifications) – and poor data quality among reporting countries could restrict
the accuracy of international bunkers estimates.
1.4.3 Bottom-up QA/QC
The key findings of the bottom-up quality assurance and quality control analysis include:
• Quality in fuel consumption totals is extensively analysed by a number of independent sources (both
independent of the data used in the model and independent of each other).
• This assurance effort represents significant progress relative to all prior global ship inventories (including
the Second IMO GHG Study 2009). These QA/QC efforts demonstrate that a reliable inventory of fuel
consumption broken down by fleets of ships and their associated activity statistics has been achieved
in this study.
• There is a step change improvement in quality in the bottom-up inventory between the earlier years
(2007–2009 inclusive) and the later years (2010–2012 inclusive), which can be attributed to the
increased coverage (both temporal and spatial) of AIS data and therefore the accuracy of the activity
estimate. This also underpins better confidence in bottom-up emissions totals, based on the same
methods, using consensus emissions factors derived from reviewing published emissions factors.
• The key data sources that have enabled the high quality of this study, particularly S-AIS data, continue
to increase in quality. This is owing to continuous improvement of the algorithms on the receivers,
Inventories of CO2 emissions from international shipping 2007–2012 63
Figure 45: Confidence bands showing statistical difference between IEA and EIA data, 2000–2010
Results of top-down QA/QC
The top-down QA/QC provides a thorough understanding of the quality and limitations of the top-down
inventory. This review shows that IEA revisions to statistics can change the total fuel sales estimate by as much
as 10% owing to documented quality controls in place at IEA. A rigorous review of IEA QA/QC practices
indicates that the energy balances continue to represent high-quality representation of OECD and non-OECD
energy statistics.
Our IEA data comparison with EIA fuel oil statistics for international marine bunkers indicate that year-on-year
fuel sales data can differ by more than 10% and that IEA tends to report more international marine bunkers
over the period 2000–2010.
Lastly, the IEA presentation to the IMO Expert Workshop in 2013 indicated that significant uncertainties are
not fully documented and require further analysis (see Section 1.5). For example, under- or overestimates of
international marine bunkers could result from allocation or classification errors – imports, exports, marine
bunker statistics, fuel transfers between sectors (as is typical for blending marine bunkers with other fuels to
meet ship/engine fuel quality specifications) – and poor data quality among reporting countries could restrict
the accuracy of international bunkers estimates.
1.4.3 Bottom-up QA/QC
The key findings of the bottom-up quality assurance and quality control analysis include:
• Quality in fuel consumption totals is extensively analysed by a number of independent sources (both
independent of the data used in the model and independent of each other).
• This assurance effort represents significant progress relative to all prior global ship inventories (including
the Second IMO GHG Study 2009). These QA/QC efforts demonstrate that a reliable inventory of fuel
consumption broken down by fleets of ships and their associated activity statistics has been achieved
in this study.
• There is a step change improvement in quality in the bottom-up inventory between the earlier years
(2007–2009 inclusive) and the later years (2010–2012 inclusive), which can be attributed to the
increased coverage (both temporal and spatial) of AIS data and therefore the accuracy of the activity
estimate. This also underpins better confidence in bottom-up emissions totals, based on the same
methods, using consensus emissions factors derived from reviewing published emissions factors.
• The key data sources that have enabled the high quality of this study, particularly S-AIS data, continue
to increase in quality. This is owing to continuous improvement of the algorithms on the receivers,
64 Third IMO GHG Study 2014
increased numbers of satellites providing greater spatial and temporal coverage, and increased
experience in filtering and processing the raw data for use in modelling.
• A quality advantage in this work is that our approach for the bottom-up activity-based inventory
uses calculations for individual vessels. By maximizing vessel-specific activity characterization using
AIS data sources, this work quantifies the variability among vessels within a type and size category.
This eliminates the dominant uncertainties reported by the Second IMO GHG Study 2009 and most
published inventories.
• The AIS-informed bottom-up methodologies cannot directly distinguish between fuel type and voyage
type, which requires additional analyses and some expert judgement. Our QA/QC on allocation
of residual/distillate fuels (HFO/MDO) and international/domestic shipping provides transparent and
reproducible methodologies, with the opportunity to adjust these if and when better information
becomes available in the future.
At the time that this report was written, there were too few data sets of on-board measurements of CO
2
emissions for any statistically representative quality assurance investigation of the modelled CO
2
emission to
be carried out. The closest that the quality assurance can therefore get to the end product of this study is the
fuel consumption comparison (modelled estimate compared with operator data), carried out using noon report
data. This is done for a sample of approximately 500 ships (approximately 1% of all vessels) representing over
60,000 days of at-sea operation. This sample is described in detail in Annex 3. It should be noted that noon
report data are not infallible; their reliability and the implications for the comparative analysis undertaken here
are discussed in greater detail in Annex 3.
To provide further assurance of the inputs and assumptions of the bottom-up method, specifically the activity
estimate, the consortium also performed analysis with LRIT data (approximately 8,000 ships and 10% of the
global fleet) and third-party literature study.
Noon reports, LRIT data and the literature were used for the following components of quality assurance work:
• The activity estimation quality was assured using:
– spatial coverage analysis with information on the number of messages received in different geographical
locations and contrasting the AIS coverage with coverage maps obtained from alternative sources (e.g.
LRIT);
– temporal coverage analysis to test whether the derived profiles of time spent in different modes of
operation (e.g. in port, at sea) and at different speeds are representative;
– comparison of the AIS-derived activity parameters speed and draught against noon report data;
– description of coverage statistics for each year and each fleet (to evaluate AIS completeness and
facilitate imputed algorithms to estimate CO
2
emissions from periods when observations are missing).
• Fleet specifications and model assumption quality were assured using:
– investigations into the robustness of the IHSF database;
– comparative evaluation of prior work, independently produced and published by consortium
members, including peer-reviewed reports and scientific articles;
– consultation of third-party inventory and shipping literature (including the work of consortium partners)
providing substantial fleet data.
• Fuel consumption estimate quality was assured using:
– comparison of calculated fuel consumption to operators’ data recorded in noon reports pooled from
data independently collected by several consortium partners.
It should be noted that noon report data are not infallible; their reliability and the implications for the
comparative analysis undertaken here are discussed in greater detail in Annex 3, along with detailed QA/QC
for the source data and other analyses.
Spatial coverage of activity estimates QA/QC
The AIS data coverage, in terms of both space and time, is not consistent year-on-year during the period
studied (2007–2012). For the first three years (2007–2009), no satellite AIS data were available, only data from
shore-based stations. This difference can be seen by contrasting the first (2007) and last (2012) years’ AIS data
sets, depicted by geographical coverage in Figure 46.
Figure 46: Geographical coverage in 2007 (top) and 2012 (bottom), coloured according to the intensity
of messages received per unit area. This is a composite of both vessel activity and geographical coverage;
intensity is not solely indicative of vessel activity
The consequence of the change in coverage over time and the quality of the regional coverage can be
inferred from an analysis of the number of messages received in different sea regions. Two investigations were
carried out, on large oil tankers and large bulk carriers, both ship types that were anticipated to be engaged
in activity on routes that encompassed most of the world’s sea areas. Figure 47 displays the trend over time
in the number of messages received in different sea regions for a random sample of 300 large oil tankers.
The number of messages received is a composite of the number of ships in an area, the duration of time
they spend in an area and the geographical coverage of an area. This analysis cannot isolate the change in
geographical coverage alone. However, the marked contrast in open ocean regions (e.g. Indian Ocean, South
Atlantic Ocean and North Atlantic Ocean) over time shows increased quality of coverage on a regional level.
Importantly, by 2012, there are no sea areas for which no activity is observed, which implies that by the latter
years coverage quality has minimal regional bias. Greater detail and maps of both AIS and LRIT data for further
years is provided in Annex 3 (details for Section 1.4).
Inventories of CO2 emissions from international shipping 2007–2012 65
shore-based stations. This difference can be seen by contrasting the first (2007) and last (2012) years’ AIS data
sets, depicted by geographical coverage in Figure 46.
Figure 46: Geographical coverage in 2007 (top) and 2012 (bottom), coloured according to the intensity
of messages received per unit area. This is a composite of both vessel activity and geographical coverage;
intensity is not solely indicative of vessel activity
The consequence of the change in coverage over time and the quality of the regional coverage can be
inferred from an analysis of the number of messages received in different sea regions. Two investigations were
carried out, on large oil tankers and large bulk carriers, both ship types that were anticipated to be engaged
in activity on routes that encompassed most of the world’s sea areas. Figure 47 displays the trend over time
in the number of messages received in different sea regions for a random sample of 300 large oil tankers.
The number of messages received is a composite of the number of ships in an area, the duration of time
they spend in an area and the geographical coverage of an area. This analysis cannot isolate the change in
geographical coverage alone. However, the marked contrast in open ocean regions (e.g. Indian Ocean, South
Atlantic Ocean and North Atlantic Ocean) over time shows increased quality of coverage on a regional level.
Importantly, by 2012, there are no sea areas for which no activity is observed, which implies that by the latter
years coverage quality has minimal regional bias. Greater detail and maps of both AIS and LRIT data for further
years is provided in Annex 3 (details for Section 1.4).
66 Third IMO GHG Study 2014
Figure 47: The average volume of AIS activity reports for a region reported by a vessel for up to 300
randomly selected VLCCs (2007–2012)
Temporal coverage of activity estimates QA/QC
LRIT data were used for the quality assurance of the AIS-derived activity estimates. The total time spent at sea
and in port for individual ships over the course of a year was analysed using both the LRIT data (which have
consistently high reliability) and the AIS data (for varying levels of coverage and reliability). This analysis was
carried out for each of the ships observed in both the LRIT and the AIS data sets (approximately 8,000, for
2009–2012). Figure 48 shows the evaluation of the difference between the LRIT- and AIS-derived estimates of
days at sea. In this comparison, the LRIT-derived estimate is assumed to be the benchmark; deviations from a
mean difference of zero therefore imply deterioration in quality of the AIS-derived estimate.
Figure 48 shows that in 2012, for reliable observation of a ship above 50% of the time during the year, the
mean difference between the AIS and LRIT converges to approximately zero. However, as the percentage of
time for which reliable observations reduces, a significant bias occurs with the AIS-derived activity estimate,
which appears to underestimate time spent at sea. Figures in Annex 3 demonstrate that a similar trend (good
quality of reliable observations for 50% of the year or more) can be observed in 2010 and 2011.
Figure 48: Activity estimate quality assurance (2012)
Greater detail of the derivation of parameters from the LRIT data sets and their application in this comparative
analysis is given in Annex 3, along with analysis for 2010 and 2011.
Activity estimates and derived parameters (speed and draught)
In addition to the analysis carried out using LRIT data, a further quality analysis of the bottom-up method’s
estimate of activity (time in mode, speed estimation, draught estimation and distance covered) can be obtained
using noon report data. Noon report data record information daily, including average speed during the period
of the report and distance travelled. Noon reports also record the date and time a voyage begins and ends.
This information was aggregated over quarters, compared with the same data calculated using the bottom-up
model, and aggregated to the same quarter of each year.
The results for 2012 are presented in Figure 49 and Figure 50. The red line represents an ideal match (equal
values) between the bottom-up and noon report outputs, the solid black line the best fit through the data and
the dotted black lines the 95% confidence bounds on the best fit. The “x” symbols represent individual ships,
coloured according to the ship type category listed in the legend. The plots include all results, with no outliers
removed.
The activity estimation of days at sea and at port can be seen to have some scatter. This scatter is related
to the fact that for some of the time the ship is not observed and an extrapolation algorithm is used to
estimate activity. For any one ship, the reliability of that extrapolation is low. However, overall, the distribution
is approximately even and does not represent a significant degree of bias, as the best-fit line shows. The
reliability of the estimate of at-port and at-sea days appears consistent regardless of ship type.
The quality of the estimation of ship speed when at sea is higher than the quality of the port- and sea-time
estimation. The best-fit line shows close alignment with the red equilibrium line, albeit with a trend towards
underestimating the speeds of the larger container ships. The confidence bounds are closely aligned to the
best-fit line.
The draught observation shows the lowest quality of fit. The observed scatter implies a bias for the bottom-up
method to slightly overestimate draught. The agreement for ship types with low draught variability (e.g.
container ships) is good. This implies that the overall poor reliability is likely to be due to infrequent updating
of the draught data reported to the AIS receiver.
In earlier years (see Annex 3 for the data), similar relative quality assurance between the variables plotted can
be obtained; however, the absolute quality reduces for the earlier years, particularly 2007, 2008 and 2009.
This can be seen by comparing the 2012 results with Figure 51, even accounting for the fact that in 2009 there
are fewer ships in the noon reports data set. Days at sea and at-sea speed have significantly more scatter and
therefore wider confidence bounds than the equivalent plots in 2012. With the exception of some outlier data
in 2009, the speed agreement is moderate. However, the days-at-sea agreement implies that there is some
Inventories of CO2 emissions from international shipping 2007–2012 67
Figure 48: Activity estimate quality assurance (2012)
Greater detail of the derivation of parameters from the LRIT data sets and their application in this comparative
analysis is given in Annex 3, along with analysis for 2010 and 2011.
Activity estimates and derived parameters (speed and draught)
In addition to the analysis carried out using LRIT data, a further quality analysis of the bottom-up method’s
estimate of activity (time in mode, speed estimation, draught estimation and distance covered) can be obtained
using noon report data. Noon report data record information daily, including average speed during the period
of the report and distance travelled. Noon reports also record the date and time a voyage begins and ends.
This information was aggregated over quarters, compared with the same data calculated using the bottom-up
model, and aggregated to the same quarter of each year.
The results for 2012 are presented in Figure 49 and Figure 50. The red line represents an ideal match (equal
values) between the bottom-up and noon report outputs, the solid black line the best fit through the data and
the dotted black lines the 95% confidence bounds on the best fit. The “x” symbols represent individual ships,
coloured according to the ship type category listed in the legend. The plots include all results, with no outliers
removed.
The activity estimation of days at sea and at port can be seen to have some scatter. This scatter is related
to the fact that for some of the time the ship is not observed and an extrapolation algorithm is used to
estimate activity. For any one ship, the reliability of that extrapolation is low. However, overall, the distribution
is approximately even and does not represent a significant degree of bias, as the best-fit line shows. The
reliability of the estimate of at-port and at-sea days appears consistent regardless of ship type.
The quality of the estimation of ship speed when at sea is higher than the quality of the port- and sea-time
estimation. The best-fit line shows close alignment with the red equilibrium line, albeit with a trend towards
underestimating the speeds of the larger container ships. The confidence bounds are closely aligned to the
best-fit line.
The draught observation shows the lowest quality of fit. The observed scatter implies a bias for the bottom-up
method to slightly overestimate draught. The agreement for ship types with low draught variability (e.g.
container ships) is good. This implies that the overall poor reliability is likely to be due to infrequent updating
of the draught data reported to the AIS receiver.
In earlier years (see Annex 3 for the data), similar relative quality assurance between the variables plotted can
be obtained; however, the absolute quality reduces for the earlier years, particularly 2007, 2008 and 2009.
This can be seen by comparing the 2012 results with Figure 51, even accounting for the fact that in 2009 there
are fewer ships in the noon reports data set. Days at sea and at-sea speed have significantly more scatter and
therefore wider confidence bounds than the equivalent plots in 2012. With the exception of some outlier data
in 2009, the speed agreement is moderate. However, the days-at-sea agreement implies that there is some
68 Third IMO GHG Study 2014
bias, with the bottom-up method consistently overestimating the time that the ship is at sea. This supports the
findings of the activity estimate quality assurance work undertaken using LRIT data.
A more detailed description of the noon report data sources, the method for assembling the data for comparison
purposes and further years’ analysis results can be found in Annex 3.
Figure 49: Comparison of at-sea and at-port days, calculated using both the bottom-up model output
(y-axis) and noon report data (x-axis) (2012)
Figure 50: Comparison of average ship speed and average ship draught calculated
using both the bottom-up model output (y-axis) and noon report data (x-axis) (2012)
Inventories of CO2 emissions from international shipping 2007–2012 69
Figure 50: Comparison of average ship speed and average ship draught calculated
using both the bottom-up model output (y-axis) and noon report data (x-axis) (2012)
70 Third IMO GHG Study 2014
Figure 51: Comparison of at-sea days and average ship speed, calculated using both the bottom-up model
output (y-axis) and noon report data (x-axis) (2009)
Fleet specifications and model assumptions quality assurance
Fleet specifications were based on the IHS vessel characteristics database, which was used in the following
ways:
• identifying the various vessel types using Statcode3 and Statcode5;
• counts of vessels within the various vessel types making up the world fleet;
Inventories of CO2 emissions from international shipping 2007–2012 71
• subdividing common vessel types into bin sizes based on deadweight tonnage or various capacity
parameters;
• providing vessel technical details, such as installed main engine power, maximum sea trial speed and
other parameters used in estimating vessel emissions;
• determining each vessel’s operational status by quarter for each year inventoried.
The IHS data were treated as accurate; however, this accuracy assumption introduces uncertainties if the
data fields used are inaccurate or unrepresentative. Potential uncertainties with the vessel characteristics data
include:
• data quality – does the field consistently represent the actual ship’s parameter?
• data source accuracy – is the field measured/recorded/verified on board the ship directly and is the
field accurate?
• update frequency – is the field updated at least quarterly (when a change has occurred)?
Data fields that have been independently spot-checked by consortium members indicate that the vessel class
fields (Statcode3 and Statcode5), main engine installed power, maximum sea trial speed and deadweight
tonnage appear to be generally representative of actual vessel conditions. The ship status field, which is
used to identify whether the ship is in service, is shown consistently to include more ships than are observed
in AIS (see Section 1.4 for details), for all ship size and type categories. There are two explanations for this
observation: either that the AIS coverage is not capturing all in-service ships, or that the IHSF database is
incomplete in its coverage of the number of active ships.
Another uncertainty associated with the vessel characteristics database concerns blanks and zeros in fields
that should not be blank or contain zero (i.e. length, deadweight, speed, etc.). To fill blanks or zeros, valid
entries were averaged on a field-by-field basis for each vessel type and bin size. These averages were used
to fill blanks and zeros (as appropriate) within the same vessel type and bin size to allow emission estimates
to be completed. The fields in which gap filling was used included main engine installed power, deadweight
tonnage, length, draught maximum, maximum sea trial speed, RPM and gross tonnage. This assumes that the
average of each vessel type and bin size is representative of vessels with a blank or zero and that the blanks
and zeros are evenly distributed across the bin.
In addition to the uncertainties listed here, there is uncertainty about the auxiliary engine and boiler loads by
vessel class and mode. As stated previously in Section 1.2.5 and Annex 1, there are no definitive data sets that
include loads by vessel class and operational mode for auxiliary engines and boilers. This study incorporates
observed vessel data collected by Starcrest as part of VBP programmes in North America (Starcrest, 2013) and
vessel auxiliary engine data collected by the Finnish Meteorological Institute for use in its modelling to build
upon the Second IMO GHG Study 2009 findings in this topic area. This improvement injects real observed
data and additional technical details but still relies on significant assumptions. Owing to the nature of the
sources profiled, the wide array of vessel configurations and operational characteristics, this area of the global
vessel emissions inventory will remain an area of significant assumption for the foreseeable future.
Relating to auxiliary engine and boiler loads, by mode, the following uncertainties that are inherent in AIS and
satellite data have a direct impact on the emissions estimated. For example:
• Vessels moving at less than 1 knot, for a certain period of time, are assumed to be at berth. This
assumption has implications for the oil tanker vessel class in which tankers at berth and not moving
faster than 1 knot will have auxiliary loads associated with discharging cargoes, which are significantly
higher than a vessel at anchorage.
• Vessels moving at less than 3 knots are assumed to be at anchorage. This assumption will cover vessels
that are manoeuvring and that will typically have a higher auxiliary load than those at anchorage.
However, tankers at offshore discharge buoys would not be assigned at-berth discharging loads for
the auxiliary boilers.
Finally, EF and SFOC remain areas of uncertainty. Emissions testing is typically limited for vessels and when
the various engine types, vessel propulsion and auxiliary engine system configurations and diverse operational
conditions are considered, emissions tests do not cover all the combinations. Testing that has been conducted
to date relies on previously agreed duty cycles, like the E3 duty cycle for direct-drive propulsion engines.
With the advent of slow steaming, is the E3 duty cycle still relevant? There are very few tests that evaluate
72 Third IMO GHG Study 2014
engine loads below 25%, which is the lowest load in the E3 cycle. Further, when looking at emissions beyond
NO
x
, which is required to be tested during engine certification, the number of valid tests available for review
significantly drops off. Similar to EF testing, published SFOC data are limited, particularly over wide engine
load factor ranges (% MCR). There is uncertainty around the effects that engine deterioration has on an
engine’s emissions profile and SFOC.
Boiler usage
Hot steam on board ships is used to provide cargo and fuel oil heating as well as to run cargo operations
with steam-driven pumps. The energy required to run these operations is usually taken from auxiliary boilers
running on fossil fuels, mainly HFO. During voyages, waste heat from the main engine is used to provide the
energy needed for steam generation. However, at low engine loads, the heat provided by the exhaust boiler is
not enough to meet all the heating demand on board. At low engine loads, both the auxiliary boiler and waste
heat recovery provide the heat needed by the vessels. The shift from exhaust to auxiliary boilers happens at
20%–50% engine load range (Mys´ków & Borkowski, 2012), as illustrated in Figure 52.
Figure 52: General boiler operation profile (Mys´ków & Borkowski, 2012)
With lower engine loads, the auxiliary boiler is the main source of heat on board a vessel. With sufficiently
high engine loads, waste heat recovery can produce enough steam for the vessel and the auxiliary boiler may
be switched off. The operational profile of the auxiliary boiler of a container carrier is presented in Figure 53.
Figure 53: Operational profile of an auxiliary boiler of a container vessel during six months of operations
(Mys´ków & Borkowski, 2012)
For a container vessel, less than half of the auxiliary boiler capacity was reported in use most of the time. Over
six months of operation, 40%–60% of the boiler steam capacity was used for nearly 100 days. Of the total
182 days, 125 were spent at port or in low load conditions, where auxiliary boilers were needed (Mys´ków &
Borkowski, 2012).
Sources of boiler data
Determination of installed boiler capacity on board vessels cannot be done based on IHSF data, because this
information is excluded. Class societies report boiler installations and capacity for their vessels only rarely
and scant details about boilers are available from publications like Significant Ships. Because of this lack of
data, boiler usage profiles have been estimated from vessel boarding programmes and crew interviews. This
method is similar to the data collection procedure used to obtain information about auxiliary engine power
profiles.
Waste heat recovery (exhaust economizers) is assumed to be in use during cruising. Vessel operational profiles
for low load manoeuvring, berthing and anchoring have auxiliary boiler use. For further details, see Annex 3.
Fuel consumption estimate quality assurance
Following the same method used to produce the activity estimate comparison between the bottom-up model
and noon report data, Figure 54 and Figure 55 show the results for average daily fuel consumption at sea (main
engine and auxiliary engine), and the total main and auxiliary fuel consumption at sea (excluding port fuel
consumption) in 2012. (Comparative analysis results for all other years of the study can be found in Annex 3.)
No data were available in the noon report data set for the fuel consumption in boilers and so the quality of
boiler information from noon reports could not be independently verified for quality.
The average daily fuel consumption plot for the main engine demonstrates the reliability of the marine
engineering and naval architecture relationships and assumptions used in the model to convert activity into
power and fuel consumption. An exception to this is the largest container ships, whose daily fuel consumption
appears to be consistently underestimated in the bottom-up method.
The total quarterly fuel consumption for the main engine plot demonstrates that the activity data (including
days at sea) and the engineering assumptions combine to produce generally reliable estimates of total fuel
consumption, at least in recent years when AIS observations are more complete. The underestimation of the
daily fuel consumption of the container ships can also be seen in this total quarterly fuel consumption.
Inventories of CO2 emissions from international shipping 2007–2012 73
Figure 53: Operational profile of an auxiliary boiler of a container vessel during six months of operations
(Mys´ków & Borkowski, 2012)
For a container vessel, less than half of the auxiliary boiler capacity was reported in use most of the time. Over
six months of operation, 40%–60% of the boiler steam capacity was used for nearly 100 days. Of the total
182 days, 125 were spent at port or in low load conditions, where auxiliary boilers were needed (Mys´ków &
Borkowski, 2012).
Sources of boiler data
Determination of installed boiler capacity on board vessels cannot be done based on IHSF data, because this
information is excluded. Class societies report boiler installations and capacity for their vessels only rarely
and scant details about boilers are available from publications like Significant Ships. Because of this lack of
data, boiler usage profiles have been estimated from vessel boarding programmes and crew interviews. This
method is similar to the data collection procedure used to obtain information about auxiliary engine power
profiles.
Waste heat recovery (exhaust economizers) is assumed to be in use during cruising. Vessel operational profiles
for low load manoeuvring, berthing and anchoring have auxiliary boiler use. For further details, see Annex 3.
Fuel consumption estimate quality assurance
Following the same method used to produce the activity estimate comparison between the bottom-up model
and noon report data, Figure 54 and Figure 55 show the results for average daily fuel consumption at sea (main
engine and auxiliary engine), and the total main and auxiliary fuel consumption at sea (excluding port fuel
consumption) in 2012. (Comparative analysis results for all other years of the study can be found in Annex 3.)
No data were available in the noon report data set for the fuel consumption in boilers and so the quality of
boiler information from noon reports could not be independently verified for quality.
The average daily fuel consumption plot for the main engine demonstrates the reliability of the marine
engineering and naval architecture relationships and assumptions used in the model to convert activity into
power and fuel consumption. An exception to this is the largest container ships, whose daily fuel consumption
appears to be consistently underestimated in the bottom-up method.
The total quarterly fuel consumption for the main engine plot demonstrates that the activity data (including
days at sea) and the engineering assumptions combine to produce generally reliable estimates of total fuel
consumption, at least in recent years when AIS observations are more complete. The underestimation of the
daily fuel consumption of the container ships can also be seen in this total quarterly fuel consumption.
74 Third IMO GHG Study 2014
Both auxiliary engine comparisons (daily and total quarterly) imply that the bottom-up estimates of auxiliary
fuel consumption are of lower quality than those of the main engine. There are two possible explanations for
this: the low quality of noon report data for auxiliary fuel consumption, or the low quality of bottom-up method
estimates. Both are likely. Auxiliary fuel consumption in the noon report data set is commonly reported as
zero. This could be because:
1 a shaft generator is used;
2 the main and auxiliary power is derived from the same engine (in the case of LNG carriers);
3 the auxiliary fuel consumption is not monitored or reported.
Discussion with the operators from whom the data originated suggested that the second and third explanations
are the most likely.
As described in Section 1.2, the method for auxiliary engine fuel consumption estimation is derived from
samples taken from vessel boardings and averaged for ship type- and size-specific modes (at berth, at anchor
and at sea). This method is used because of the scarcity of data about the installed auxiliary engine in the
IHSF database and the shortage of other information in the public domain describing operational profiles of
auxiliary engines.
Figure 56 presents the comparison between the noon report and the bottom-up method in 2012, but with a
filter applied to include only data for which the AIS-derived activity was deemed reliable for more than 75%
of the time in the quarter. Otherwise, the data source is the same. The marked improvement of the agreement
is demonstration of the reliability of the bottom-up method in converting activity into fuel consumption and
shows that the largest source of uncertainty in the total fuel consumption is the estimate of activity, particularly
the estimate of days at sea.
Figure 57 and Figure 58 present the comparison between the noon report and the bottom-up method for 2007
and 2009 respectively. These quality assurance plots show that, consistent with the comparison of the activity
estimate data to noon report data, quality deteriorates between the earlier years (2007, 2008 and 2009) and
later years (2010, 2011 and 2012). The availability of noon report data in the earlier years is also limited, which
makes rigorous quality assurance difficult. However, even with the sample sizes available, the confidence
bounds clearly indicate that the quality deteriorates.
Table 25 summarizes the findings from the quality assurance analysis of the fuel consumption. Further data
from earlier years can be found in Annex 3.
Table 25 – Summary of the findings on the QA of the bottom-up method estimated fuel consumption
using noon report data
Consumer Quality, as assessed using noon report data Importance to the inventory of fuel
consumption and emissions
Main engine Good: consistent agreement and close
confidence bounds to the best fit
High (71% of total fuel in 2012)
Auxiliary engine Poor: moderate, with some ships showing good
agreement but many anomalies (very low values
in noon reports)
Low (25% of total fuel in 2012)
Boilers Unassessed Very low (3.7% of total fuel in 2012)
Figure 54: Average noon-reported daily fuel consumption of the main and auxiliary engines
compared with the bottom-up estimate over each quarter of 2012
Inventories of CO2 emissions from international shipping 2007–2012 75
Figure 54: Average noon-reported daily fuel consumption of the main and auxiliary engines
compared with the bottom-up estimate over each quarter of 2012
76 Third IMO GHG Study 2014
Figure 55: Total noon-reported quarterly fuel consumption of the main and auxiliary engines
compared with the bottom-up estimate over each quarter of 2012
Figure 56: Total noon-reported quarterly fuel consumption of the main engine
compared with the bottom-up estimate over each quarter of 2012,
with a filter to select only days with high reliability observations of the ship
for 75% of the time or more
Figure 57: Total noon-reported quarterly fuel consumption of the main engine
compared with the bottom-up estimate over each quarter of 2007
Inventories of CO2 emissions from international shipping 2007–2012 77
Figure 56: Total noon-reported quarterly fuel consumption of the main engine
compared with the bottom-up estimate over each quarter of 2012,
with a filter to select only days with high reliability observations of the ship
for 75% of the time or more
Figure 57: Total noon-reported quarterly fuel consumption of the main engine
compared with the bottom-up estimate over each quarter of 2007
78 Third IMO GHG Study 2014
Figure 58: Total noon-reported quarterly fuel consumption of the main engine
compared with the bottom-up estimate over each quarter of 2009
Coverage statistics and fleet size quality assurance
The total emissions for each fleet (and the sum of emissions of all fleets) are found from:
• the emissions of any ships observed on AIS, during the period of observation;
• extrapolation to cover periods of time when the observed ships are not currently under observation
by AIS;
• estimation for ships that are deemed “active” in the IHSF database but are not observed on AIS at all.
The maximum reliability of the inventory is achieved if all the ships are observed all the time, as demonstrated
by the main engine comparison in Figure 56. However, the reality is that AIS coverage is not perfect. The
statistics of coverage by AIS therefore provide important insight into the quality of the estimate and the
quantity of emissions calculated directly versus the quantity of data calculated from imputed and extrapolated
estimates of activity. This section examines the quality of the AIS data coverage of the fleets of international,
domestic and fishing ships by answering two questions:
• How many of the in-service ships are observed in the AIS data set?
• Of the ships that are observed, what is the duration of the observation period?
The number of in-service ships observed in the AIS data set
Table 26 describes the size of the fleet in the IHSF database in each year along with the percentage of the total
fleet classified as in service and, of those ships, the percentage that also appears in the AIS database.
Transport ships are ships that carry goods and people (all merchant shipping, ferries and cruise passenger
ships); non-transport ships include service vessels, workboats, yachts and fishing vessels.
Table 26: Observed, unobserved and active ship counts (2007–2012)
Year
Transport ships Non-transport ships
Total % in service
% of in-service
ships observed
on AIS
Total % in service
% of in-service
ships observed
on AIS
2007 58,074 89% 62% 49,396 99% 19%
2008 59,541 89% 66% 50,704 98% 24%
2009 61,065 90% 69% 50,872 100% 29%
2010 69,431 83% 68% 52,941 98% 31%
2011 72,462 75% 69% 51,961 96% 32%
2012 60,670 93% 76% 54,077 96% 42%
There is a large discrepancy between the number of AIS-observed and in-service ships, with fewer in-service
ships appearing on AIS than would be expected. This discrepancy is far greater for non-transport ships but still
significant for transport ships. Explanations for this discrepancy include:
• a large number of ships classified as in service were not actually so;
• the AIS transponders of in-service ships were not turned on during the year or were faulty/sending
spurious signals;
• ships were out of range of any AIS receiving equipment (shore-based or satellite).
The maps of AIS coverage shown in Annex 3 demonstrate that the third explanation (out of range) is plausible
for the shortfall in the earlier years (2007, 2008 and 2009). However, the consistency in the shortfall between
the number of observed ships and in-service ships across the years (particularly from 2010 onwards, when
satellite AIS data are available) does not support this as the only explanation.
Table 27 lists the statistics for four ship types: bulk carriers, container ships, general cargo ships and oil tankers.
For these fleets, which account for the majority of shipping emissions, the percentage of in-service ships that
also appear in AIS is generally excellent (90%–100%), although there are some notable exceptions. Only 50%
or less of the smallest size category of oil tankers, bulk carriers and general cargo ships are observed on AIS,
regardless of the year and the quality of AIS coverage.
This implies that the quality of AIS coverage for the ships most important to the inventory is good, but that
there are shortcomings in the quality of either the AIS coverage or the IHSF database for the smallest ship
size categories. Even as the geographical coverage of the AIS database increases over time, there are many
ship types and sizes for which the percentage of in-service ships observed in AIS reduces over time (this is
particularly true of the larger container ships and bulk carriers). This trend is indicative of deterioration in the
quality of the IHSF status indicator since 2007, 2008 and 2009.
The average duration period for ships that are observed
Table 27 also describes the percentage of the year for which there is a reliable estimate of activity for ships
observed on AIS. (The method and judgement of reliability are described in detail in Annex 3.) Consistent with
the switch from solely shore-based AIS in 2007, 2008 and 2009 to shore-based and satellite AIS in the later
years, there is a substantial increase over the period of this study in the percentage of the year for which a ship
can be reliably observed from its AIS transmissions. Many of the smaller ship categories are well observed
even in the early years of this study, which is indicative of the ships being operated in coastal areas of land
masses where there was good shore-based AIS reception (e.g. particularly Europe and North America).
A composite of the number of ships observed and the duration for which they are observed can be found by
taking the product of the two statistics in Table 27:
% total in-service coverage = % in-service ships on AIS × % of the year for which they are observed
Figure 59 displays the trend over time of the percentage of total in-service coverage for four of the fleets
sampled in Table 27. As expected from the increased geographical coverage of AIS data with time, the total
in-service coverage increases. In 2012, the average in-service large container or bulk carrier can be reliabily
observed in the AIS data set assembled by this consortium for nearly 70% of the time. Coverage of the largest
bulk carriers nearly tripled between 2010 and 2012, showing that rapid improvements have been observed
Inventories of CO2 emissions from international shipping 2007–2012 79
Table 26: Observed, unobserved and active ship counts (2007–2012)
Year
Transport ships Non-transport ships
Total % in service
% of in-service
ships observed
on AIS
Total % in service
% of in-service
ships observed
on AIS
2007 58,074 89% 62% 49,396 99% 19%
2008 59,541 89% 66% 50,704 98% 24%
2009 61,065 90% 69% 50,872 100% 29%
2010 69,431 83% 68% 52,941 98% 31%
2011 72,462 75% 69% 51,961 96% 32%
2012 60,670 93% 76% 54,077 96% 42%
There is a large discrepancy between the number of AIS-observed and in-service ships, with fewer in-service
ships appearing on AIS than would be expected. This discrepancy is far greater for non-transport ships but still
significant for transport ships. Explanations for this discrepancy include:
• a large number of ships classified as in service were not actually so;
• the AIS transponders of in-service ships were not turned on during the year or were faulty/sending
spurious signals;
• ships were out of range of any AIS receiving equipment (shore-based or satellite).
The maps of AIS coverage shown in Annex 3 demonstrate that the third explanation (out of range) is plausible
for the shortfall in the earlier years (2007, 2008 and 2009). However, the consistency in the shortfall between
the number of observed ships and in-service ships across the years (particularly from 2010 onwards, when
satellite AIS data are available) does not support this as the only explanation.
Table 27 lists the statistics for four ship types: bulk carriers, container ships, general cargo ships and oil tankers.
For these fleets, which account for the majority of shipping emissions, the percentage of in-service ships that
also appear in AIS is generally excellent (90%–100%), although there are some notable exceptions. Only 50%
or less of the smallest size category of oil tankers, bulk carriers and general cargo ships are observed on AIS,
regardless of the year and the quality of AIS coverage.
This implies that the quality of AIS coverage for the ships most important to the inventory is good, but that
there are shortcomings in the quality of either the AIS coverage or the IHSF database for the smallest ship
size categories. Even as the geographical coverage of the AIS database increases over time, there are many
ship types and sizes for which the percentage of in-service ships observed in AIS reduces over time (this is
particularly true of the larger container ships and bulk carriers). This trend is indicative of deterioration in the
quality of the IHSF status indicator since 2007, 2008 and 2009.
The average duration period for ships that are observed
Table 27 also describes the percentage of the year for which there is a reliable estimate of activity for ships
observed on AIS. (The method and judgement of reliability are described in detail in Annex 3.) Consistent with
the switch from solely shore-based AIS in 2007, 2008 and 2009 to shore-based and satellite AIS in the later
years, there is a substantial increase over the period of this study in the percentage of the year for which a ship
can be reliably observed from its AIS transmissions. Many of the smaller ship categories are well observed
even in the early years of this study, which is indicative of the ships being operated in coastal areas of land
masses where there was good shore-based AIS reception (e.g. particularly Europe and North America).
A composite of the number of ships observed and the duration for which they are observed can be found by
taking the product of the two statistics in Table 27:
% total in-service coverage = % in-service ships on AIS × % of the year for which they are observed
Figure 59 displays the trend over time of the percentage of total in-service coverage for four of the fleets
sampled in Table 27. As expected from the increased geographical coverage of AIS data with time, the total
in-service coverage increases. In 2012, the average in-service large container or bulk carrier can be reliabily
observed in the AIS data set assembled by this consortium for nearly 70% of the time. Coverage of the largest
bulk carriers nearly tripled between 2010 and 2012, showing that rapid improvements have been observed
80 Third IMO GHG Study 2014
during the period of this study. The trend for the smaller ship types is for increased coverage but the average
total in-service coverage remains 40% and lower for the smallest general cargo carriers and bulk carriers.
Figure 59: Total percentage of in-service time for which high-reliability activity estimates
are available from AIS
However, for the purposes of a high-quality inventory, it is more important for the quality of the AIS coverage
for the ship types and sizes with the greatest share of emissions to be high. Since the coverage statistics of the
highest contributing CO
2
emitters (i.e. the largest ship types and sizes) are also the highest, this is generally the
case. Figure 60 displays the CO
2
emissions weighted average of the percentage of total in-service coverage.
This is decomposed into two categories: i) ships classified as international shipping (see Section 1.2) and ii)
ships classified as domestic and fishing. The subject of the inventories in Section 1.3, international shipping,
has significantly higher coverage quality than the domestic and fishing fleet.
In Section 1.4, where the days-at-sea estimate from LRIT is compared with the estimate obtained from AIS,
there is a significant improvement in quality for the AIS-derived activity estimates when reliable coverage
exceeds 50% of the year. When this finding is placed in the context of the coverage statistics described in this
section, it can be seen that in 2011 and 2012 the coverage statistics lead to high-quality activity and therefore
inventory estimates. However, in the earlier years of this study, the comparatively lower coverage statistics,
relative to the later years, increase the uncertainty of the estimated inventories.
Figure 60: Emissions weighted average of the total percentage of in-service time
for which high-reliability activity estimates are available from AIS
Inventories of CO2 emissions from international shipping 2007–2012 81
Table 27 – Statistics of the number of in-service ships observed on AIS and of the average amount of time during the year for which a ship is observed
% of in-service ships observed on AIS % of year for which high-reliability activity estimates are available
Ship type Size category 2007 2008 2009 2010 2011 2012 2007 2008 2009 2010 2011 2012
Bulk carrier 0–9,999 34% 41% 45% 50% 47% 55% 75% 71% 72% 68% 71% 74%
10,000–34,999 91% 92% 91% 89% 86% 92% 23% 26% 31% 38% 56% 65%
35,000–59,999 97% 96% 97% 96% 91% 95% 17% 20% 22% 28% 52% 65%
60,000–99,999 99% 99% 98% 98% 93% 95% 14% 16% 19% 24% 51% 65%
100,000–199,999 99% 99% 99% 99% 93% 94% 10% 14% 18% 28% 51% 63%
200,000–+
98% 99% 99% 100% 95% 93% 11% 12% 17% 23% 52% 68%
Container 0–999 89% 90% 90% 84% 82% 88% 42% 45% 49% 49% 62% 70%
1,000 –1,999 99% 98% 99% 96% 92% 98% 21% 30% 36% 36% 56% 65%
2,000–2,999 99% 98% 99% 96% 92% 96% 20% 26% 29% 36% 63% 66%
3,000–4,999 99% 98% 99% 97% 92% 95% 23% 28% 28% 35% 63% 70%
5,000–7,999 99% 98% 99% 100% 95% 96% 27% 33% 30% 32% 59% 70%
8,000–11,999 100% 100% 99% 100% 91% 98% 31% 39% 37% 36% 60% 68%
12,000–14,500 100% 100% 100% 97% 94% 95% 56% 34% 38% 68% 65% 81%
14,500 –+
0% 0% 0% 0% 0% 88% 0% 0% 0% 0% 0% 71%
General cargo 0–4,999 38% 41% 42% 40% 39% 44% 72% 72% 73% 79% 78% 83%
5,000–9,999 81% 83% 83% 79% 77% 86% 34% 37% 41% 49% 56% 64%
10,000–+
87% 89% 89% 84% 84% 90% 28% 29% 34% 43% 55% 66%
Oil tanker 0–4,999 72% 30% 34% 37% 38% 43% 46% 43% 48% 48% 58% 66%
5,000–9,999 98% 70% 73% 78% 78% 87% 15% 17% 22% 26% 49% 59%
10,000–19,999 100% 78% 81% 77% 80% 90% 69% 67% 50% 23% 53% 59%
20,000–59,999 24% 93% 93% 91% 91% 95% 84% 81% 80% 77% 78% 81%
60,000–79,999 62% 95% 95% 96% 90% 97% 56% 53% 55% 46% 55% 59%
80,000–119,999 74% 98% 98% 97% 91% 97% 44% 45% 47% 43% 52% 58%
120,000–199,999 91% 98% 98% 97% 91% 95% 29% 30% 33% 38% 52% 61%
200,000–+
95% 99% 99% 95% 95% 96% 27% 29% 33% 34% 55% 64%
82 Third IMO GHG Study 2014
1.4.4 Comparison of top-down and bottom-up inventories
Four main comparators are essential to understanding the similarities, differences and joint insights that derive
from the top-down and bottom-up inventories:
1 estimates of fuel totals (in million tonnes);
2 allocation of fuel totals by fuel type (residual, distillate and natural gas, or HFO, MDO and LNG as
termed in this study);
3 estimates of CO
2
totals (in million tonnes), which depend in part upon the allocation of different fuel
types with somewhat different carbon contents;
4 allocation of fuel totals as international and not international (e.g. domestic and fishing).
Given the results presented in Sections 1.1.3 and 1.3.2, there is a clear difference between the best estimates of
the top-down and bottom-up methods. This difference has been documented in the scientific peer-reviewed
literature and in previous IMO reports. This study finds that the best estimates of fuel consumption differ by
varying quantities across the years studied. Smaller differences between top-down and bottom-up total fuel
consumption are observed after the availability of better AIS coverage in 2010. However, in all cases, the
activity-based bottom-up results for all fuels are generally greater than the top-down statistics.
a) All marine fuels b) International shipping
Figure 61: Top-down and bottom-up comparison for a) all marine fuels and b) international shipping
Allocation of fuel inventories by fuel type is important and comparison of top-down allocations with initial
bottom-up fuel type results provided important QA/QC that helped reconcile bottom-up fuel type allocation.
The fuel split between residual (HFO) and distillate (MDO) for the top-down approach is explicit in the
fuel sales statistics from IEA. However, the HFO/MDO allocation for the bottom-up inventory could not be
finalized without consideration of top-down sales insights. This is because the engine-specific data available
through IHSF are too sparse, incomplete or ambiguous with respect to fuel type for large numbers of main
engines and nearly all auxiliary engines on vessels. QA/QC analysis with regard to fuel type assignment
in the bottom-up model was performed using top-down statistics as a guide together with fuel allocation
information from the Second IMO GHG Study 2009. This iteration was important in order to finalize QA/QC
on fuel-determined pollutant emissions (primarily SO
x
and PM), and results in slight QA/QC adjustments for
other emissions. Figure 62 presents a side-by-side comparison of top-down, initial and updated bottom-up
approaches to fuel type allocations.
a) Top-down fuel type allocation
b) Initial bottom-up allocation c) Updated bottom-up allocation
Figure 62: Comparison of top-down fuel allocation with initial and updated bottom-up fuel allocation
(2007–2012)
Figure 62 a) and c) show that relative volumes of residual to distillate marine fuel (HFO to MDO) are similar.
This is because the updated allocation in the bottom-up inventory is constrained to replicate the reported IEA
fuel sales ratios. The year-on-year allocations are also constrained by bottom-up analysis that identifies vessel
categories with engines likely to use distillate fuel. A further constraint is that an MDO assignment applied to
a vessel category in any year requires that MDO be assigned to that category in every year.
The CO
2
comparison corresponds closely to the total fuel values, with the exception of the LNG consumption
identified in the bottom-up inventory. The IEA statistics report zero international marine bunkers of natural
gas (LNG), as shown in Table 9 in Section 1.1.3. Trends in CO
2
emissions are nearly identical to total fuel
estimates, with negligible modification by the fuel type allocation. Trends in the top-down inventory suggest
a low-growth trend in energy use by ships during the period 2007–2012. This is consistent with known
adaptations and innovations in the international shipping fleet to conserve fuel during a period of increasing
energy prices and global recession.
Inventories of CO2 emissions from international shipping 2007–2012 83
a) Top-down fuel type allocation
b) Initial bottom-up allocation c) Updated bottom-up allocation
Figure 62: Comparison of top-down fuel allocation with initial and updated bottom-up fuel allocation
(2007–2012)
Figure 62 a) and c) show that relative volumes of residual to distillate marine fuel (HFO to MDO) are similar.
This is because the updated allocation in the bottom-up inventory is constrained to replicate the reported IEA
fuel sales ratios. The year-on-year allocations are also constrained by bottom-up analysis that identifies vessel
categories with engines likely to use distillate fuel. A further constraint is that an MDO assignment applied to
a vessel category in any year requires that MDO be assigned to that category in every year.
The CO
2
comparison corresponds closely to the total fuel values, with the exception of the LNG consumption
identified in the bottom-up inventory. The IEA statistics report zero international marine bunkers of natural
gas (LNG), as shown in Table 9 in Section 1.1.3. Trends in CO
2
emissions are nearly identical to total fuel
estimates, with negligible modification by the fuel type allocation. Trends in the top-down inventory suggest
a low-growth trend in energy use by ships during the period 2007–2012. This is consistent with known
adaptations and innovations in the international shipping fleet to conserve fuel during a period of increasing
energy prices and global recession.
84 Third IMO GHG Study 2014
Table 28 – International, domestic and fishing CO
2
emissions 2007–2011 (million tonnes),
using top-down method
Marine sector Fuel type 2007 2008 2009 2010 2011
International shipping HFO 542.1 551.2 516.6 557.1 554.0
MDO 83.4 72.8 79.8 90.4 94.9
LNG 0.0 0.0 0.0 0.0 0.0
Top-down international total All 625.5 624.0 596.4 6 47.5 648.9
Domestic navigation HFO 62.0 44.2 47.6 44.5 39.5
MDO 72.8 76.6 75.7 82.4 8 7. 8
LNG 0.1 0.1 0.1 0.1 0.2
Top-down domestic total All 134.9 121.0 123.4 127.1 127.6
Fishing HFO 3.4 3.4 3.1 2.5 2.5
MDO 17. 3 15.7 16.0 16.7 16.4
LNG 0.1 0.1 0.1 0.1 0.1
Top-down fishing total All 20.8 19.2 19.3 19.2 19.0
All fuels top-down 781.2 764.1 739.1 793.8 795.4
Table 29 – International, domestic and fishing CO
2
emissions 2007–2012 (million tonnes),
using bottom-up method
Marine sector Fuel type 2007 2008 2009 2010 2011 2012
International shipping HFO 773.8 802.7 736.6 650.6 716.9 667.9
MDO 97. 2 102.9 104.2 102.2 109.8 105.2
LNG 13.9 15.4 14.2 18.6 22.8 22.6
Bottom-up international total All 884.9 920.9 855.1 771.4 849.5 795.7
Domestic navigation HFO 53.8 57.4 32.5 45.1 61.7 39.9
MDO 142.7 138.8 80.1 88.2 98.1 91.6
LNG 0 0 0 0 0 0
Bottom-up domestic total All 196.5 196.2 112.6 133.3 159.7 131.4
Fishing HFO 1.6 1.5 0.9 0.8 1.4 1.1
MDO 17.0 16.4 9.3 9.2 10.9 9.9
LNG 0 0 0 0 0 0
Bottom-up fishing total All 18.6 18.0 10.2 10.0 12.3 11.0
All fuels bottom-up 1,100.1 1,135.1 977.9 914.7 1,021.6 938.1
Across the set of years 2007–2012, CO
2
emissions from international shipping range between approximately
739 million and 795 million tonnes, according to top-down methods, and between approximately 915 million
and 1,135 million tonnes, according to bottom-up methods. The trend in top-down totals has been generally
flat or slightly increasing since the low point of the recession in 2009; the trend in bottom-up totals can be
interpreted as generally flat (since 2010 at least, when AIS data coverage became consistently global).
Domestic navigation and fishing
The top-down results are explicit in distinguishing between fuel delivered to international shipping, domestic
navigation or fishing. (Potential uncertainty in this explicit classification is discussed in Section 1.6.) Bottom-up
methods do not immediately identify international shipping, so the consortium considered ways to deduct
domestic navigation or fishing fuel from the total fuel estimates. For example, bottom-up results allow for
categorical identification of fishing fuel by virtue of ship type.
For domestic navigation and fishing, some categories of vessel presumably would be devoted mainly to
domestic navigation service, according to allocation method 2 in Section 1.2.8. To evaluate the quality of
this method, the consortium visually inspected AIS plots of service vessels, passenger ferries, ro-pax ferries
and other vessel types without respect to vessel size. The intensity of AIS reporting revealed generally local
operations for service vessels, as expected. Service vessels were observed operating in international waters,
but their patterns strongly conformed to EEZ boundaries as a rule. These were interpreted as non-transport
services that would result in a domestic-port-to-domestic-port voyage with offshore service to domestic
platforms for energy exploration, extraction, scientific missions, etc. Similar behaviour was observed for
offshore vessels and miscellaneous vessel categories (other than fishing). Cruise passenger ships exhibited
much more international voyage behaviour than passenger ferries (with some exceptions attributed to larger
ferries); similar observations were made after visualizing ro-pax vessel patterns. Moreover, no dominant
patterns of local operations for bulk cargo ships, container ships or tankers were identified.
The consortium mapped the set of AIS-observed but unidentified vessels and observed that these vessels
generally (but not exclusively) operated in local areas. This led to an investigation of the available message
data in these AIS observations. It was possible to evaluate the MMSI numbers that were unmatched with IHSF
vessel information, at least according to the MMSI code convention. A count of unique MMSI numbers was
made for each year and associated with its region code; only vessel identifiers were included.
Europe, Asia and North America were the top regions with unknown vessels, accounting for more than 85% of
the umatched MMSI numbers on average across 2007–2012 (approximately 36%, 30% and 21% respectively).
Oceania, Africa and South America each accounted for approximately 6%, 5% and 3% respectively. To
evaluate whether these vessel operations might qualify as domestic navigation, the top-down domestic fuel
sales statistics from IEA were classified according to these regions and the pattern of MMSI counts was
confirmed as mostly correlated with domestic marine bunker sales. This is illustrated in Table 30, which shows
that correlations in all but one year were greater than 50%. This evidence allows for a designation of these
vessels as mostly in domestic service, although it is not conclusive.
Table 30 – Summary of average domestic tonnes of fuel consumption per year (2007–2012),
MMSI counts and correlations between domestic fuel use statistics
Correlations: 0.87 0.56 0.66 0.13 0.66 0.87
Domestic fuel
consumption
(tonnes per year)
2007 MMSI 2008 MMSI 2009 MMSI 2010 MMSI 2011 MMSI 2012 MMSI
Africa 430 4,457 7, 39 9 2,501 3,336 10,801 13,419
Asia 9,900 18,226 23,588 15,950 12,530 82,198 112,858
Europe 3,000 13,856 23,368 20,972 75,331 94,379 88,286
North and
Central America
and Caribbean
4,800 14,100 48,261 16,104 22,590 26,878 55835
Oceania 430 3,903 7,18 8 4,135 5,200 13,889 21,320
South America 1,300 1,023 2,583 1,939 1,842 6,808 9,532
Grand total 19,900 55,565 112,387 8,301 120,829 234,953 301,250
1.5 Analysis of the uncertainty of the top-down and bottom-up CO
2
inventories
Section 1.5 requires an analysis of the uncertainties in the emissions estimates to provide IMO with reliable
and up-to-date information on which to base its decisions. Uncertainties are associated with the accuracy
of top-down fuel statistics and with the emissions calculations derived from marine fuel sales statistics.
Uncertainties also exist in the bottom-up calculations of energy use and emissions from the world fleet
of ships. These uncertainties can affect the totals, distributions among vessel categories and allocation of
emissions between international and domestic shipping.
1.5.1 Top-down inventory uncertainty analysis
An overview of the twofold approach applied to top-down statistics and emissions estimates is provided. A full
description of this approach is given in Annex 4. First, this work builds upon the QA/QC findings that suggest
that sources of uncertainty in fuel statistics relate to data quality and work to quantify the bounding impacts of
these. Second, this analysis quantifies uncertainties associated with emissions factors used to estimate GHGs
using top-down statistics.
Inventories of CO2 emissions from international shipping 2007–2012 85
operations for service vessels, as expected. Service vessels were observed operating in international waters,
but their patterns strongly conformed to EEZ boundaries as a rule. These were interpreted as non-transport
services that would result in a domestic-port-to-domestic-port voyage with offshore service to domestic
platforms for energy exploration, extraction, scientific missions, etc. Similar behaviour was observed for
offshore vessels and miscellaneous vessel categories (other than fishing). Cruise passenger ships exhibited
much more international voyage behaviour than passenger ferries (with some exceptions attributed to larger
ferries); similar observations were made after visualizing ro-pax vessel patterns. Moreover, no dominant
patterns of local operations for bulk cargo ships, container ships or tankers were identified.
The consortium mapped the set of AIS-observed but unidentified vessels and observed that these vessels
generally (but not exclusively) operated in local areas. This led to an investigation of the available message
data in these AIS observations. It was possible to evaluate the MMSI numbers that were unmatched with IHSF
vessel information, at least according to the MMSI code convention. A count of unique MMSI numbers was
made for each year and associated with its region code; only vessel identifiers were included.
Europe, Asia and North America were the top regions with unknown vessels, accounting for more than 85% of
the umatched MMSI numbers on average across 2007–2012 (approximately 36%, 30% and 21% respectively).
Oceania, Africa and South America each accounted for approximately 6%, 5% and 3% respectively. To
evaluate whether these vessel operations might qualify as domestic navigation, the top-down domestic fuel
sales statistics from IEA were classified according to these regions and the pattern of MMSI counts was
confirmed as mostly correlated with domestic marine bunker sales. This is illustrated in Table 30, which shows
that correlations in all but one year were greater than 50%. This evidence allows for a designation of these
vessels as mostly in domestic service, although it is not conclusive.
Table 30 – Summary of average domestic tonnes of fuel consumption per year (2007–2012),
MMSI counts and correlations between domestic fuel use statistics
Correlations: 0.87 0.56 0.66 0.13 0.66 0.87
Domestic fuel
consumption
(tonnes per year)
2007 MMSI 2008 MMSI 2009 MMSI 2010 MMSI 2011 MMSI 2012 MMSI
Africa 430 4,457 7, 39 9 2,501 3,336 10,801 13,419
Asia 9,900 18,226 23,588 15,950 12,530 82,198 112,858
Europe 3,000 13,856 23,368 20,972 75,331 94,379 88,286
North and
Central America
and Caribbean
4,800 14,100 48,261 16,104 22,590 26,878 55835
Oceania 430 3,903 7,18 8 4,135 5,200 13,889 21,320
South America 1,300 1,023 2,583 1,939 1,842 6,808 9,532
Grand total 19,900 55,565 112,387 8,301 120,829 234,953 301,250
1.5 Analysis of the uncertainty of the top-down and bottom-up CO
2
inventories
Section 1.5 requires an analysis of the uncertainties in the emissions estimates to provide IMO with reliable
and up-to-date information on which to base its decisions. Uncertainties are associated with the accuracy
of top-down fuel statistics and with the emissions calculations derived from marine fuel sales statistics.
Uncertainties also exist in the bottom-up calculations of energy use and emissions from the world fleet
of ships. These uncertainties can affect the totals, distributions among vessel categories and allocation of
emissions between international and domestic shipping.
1.5.1 Top-down inventory uncertainty analysis
An overview of the twofold approach applied to top-down statistics and emissions estimates is provided. A full
description of this approach is given in Annex 4. First, this work builds upon the QA/QC findings that suggest
that sources of uncertainty in fuel statistics relate to data quality and work to quantify the bounding impacts of
these. Second, this analysis quantifies uncertainties associated with emissions factors used to estimate GHGs
using top-down statistics.
86 Third IMO GHG Study 2014
Table 31 – Upper range of top-down fuel consumption by vessel type (million tonnes)
Fuel type 2007 2008 2009 2010 2011
MDO 71 73 77 64 73
HFO 258 258 245 256 244
All fuels 329 331 321 319 318
Fuel type 2007 2008 2009 2010 2011
MDO 22% 22% 24% 20% 23%
HFO 78% 78% 76% 80% 77%
All fuels 100% 100% 100% 100% 100%
The Third IMO GHG Study 2014 acknowledges that additional uncertainty about marine fuel sales to
consumers is not identified in the IEA data and cannot be quantified. For example, some ships that purchase
fuel (probably domestic and almost certainly MDO) are identified by IEA as “transport sector”. This includes
fuel purchased in places that might not be counted as “marine bunkers” (e.g. leisure ports and marinas). The
quantities of fuel sold to boats in a global context appear to be small compared to the volumes reported as
bunker sales but this cannot be evaluated quantitatively. Given that these sales are all domestic, the additional
uncertainty does not affect estimates of international shipping fuel use. However, uncertainty in the HFO/
MDO allocation may be slightly affected but remains unquantified; again, this analysis suggests such fuel
allocation uncertainty appears to be small.
Export-import discrepancy represents the primary source of uncertainty, as measured by the quantity of
adjustment that is supported by our analysis. This discrepancy exists because the total fuel volumes reported
as exports exceeds the total fuel volumes reported as imports. Evidence associating the export-import
discrepancy with marine fuels includes the known but unquantified potential to misallocate bunker fuel sales
as exports, as documented above. The magnitude of this error increased during the period of globalization,
particularly since the 1980s. In fact, the percentage adjustment due to export-import allocation uncertainty
has never been lower than 22% since 1982, as discussed in Annex 4. Table 32 and Figure 63 illustrate the
top-down adjustment for the years 2007–2011. During these years, the average adjustment due to export-
import allocation uncertainty averaged 28%.
Table 32 – Results of quantitative uncertainty analysis on top-down statistics (million tonnes)
Marine sector 2007 2008 2009 2010 2011
Total marine fuel consumption (reported) 249.2 243.7 235.9 253.0 253.5
Adjustment for export-import discrepancy 71.5 79.4 78.0 59.0 56.0
Adjustment for fuel transfers balance 8.1 8.1 7. 5 7. 5 8.2
Adjusted top-down marine fuel estimate 329.8 331.2 321.4 319.5 317. 7
Figure 63: Adjusted marine fuel sales based on quantitative uncertainty results (2007–2011)
1.5.2 Bottom-up inventory uncertainty analysis
Bottom-up uncertainty in this study is conditioned on the quality control of information for specific vessels,
application of known variability in vessel activity to observed vessels within similar ship type and size fleets
and the way in which activity assumptions are applied to unobserved vessels within similar ship type and size
fleets. In other words, the quantification of uncertainty is linked to the quality control section of this report.
One of the most important contributions of this study in reducing uncertainty is the explicit quality control
to calculate fuel use and emissions using specific vessel technical details; this directly accounts for variability
within a fleet bin, and replaces the uncertainty with the average technical parameters in the Second IMO
GHG Study 2009 calculations with the average technical parameters. Another important contribution to
reducing uncertainty is the direct observation of activity data for individual vessels, i.e. speed and draught
aggregated hourly, then annually.
Figure 64 presents the uncertainty ranges around the top-down and bottom-up fuel totals for the years studied.
The vertical bars attached to the total fuel consumption estimate for each year and each method represent
uncertainty. This study estimates higher uncertainty in the bottom-up method in the earlier years (2007, 2008
and 2009), with the difference between these uncertainty estimates being predominantly attributable to the
change in AIS coverage over the period of the study. The uncertainty in the earlier years is dominated by
uncertainty in the activity data, due to the lack of satellite AIS data. In later years (2010, 2011 and 2012), this
uncertainty reduces, but the discrepancy between the number of ships identified as in service in IHSF and the
ships observed on AIS increases (relative to the earlier years). The result is that the total bottom-up uncertainty
only reduces slightly in the later years when improved AIS data is available.
The top-down estimates are also uncertain, and include observed discrepancies between global imports and
exports of fuel oil and distillate oil, observed transfer discrepancies among fuel products that can be blended
into marine fuels and the potential for misallocation of fuels between sectors of shipping (international,
domestic and fishing).
Inventories of CO2 emissions from international shipping 2007–2012 87
Figure 63: Adjusted marine fuel sales based on quantitative uncertainty results (2007–2011)
1.5.2 Bottom-up inventory uncertainty analysis
Bottom-up uncertainty in this study is conditioned on the quality control of information for specific vessels,
application of known variability in vessel activity to observed vessels within similar ship type and size fleets
and the way in which activity assumptions are applied to unobserved vessels within similar ship type and size
fleets. In other words, the quantification of uncertainty is linked to the quality control section of this report.
One of the most important contributions of this study in reducing uncertainty is the explicit quality control
to calculate fuel use and emissions using specific vessel technical details; this directly accounts for variability
within a fleet bin, and replaces the uncertainty with the average technical parameters in the Second IMO
GHG Study 2009 calculations with the average technical parameters. Another important contribution to
reducing uncertainty is the direct observation of activity data for individual vessels, i.e. speed and draught
aggregated hourly, then annually.
Figure 64 presents the uncertainty ranges around the top-down and bottom-up fuel totals for the years studied.
The vertical bars attached to the total fuel consumption estimate for each year and each method represent
uncertainty. This study estimates higher uncertainty in the bottom-up method in the earlier years (2007, 2008
and 2009), with the difference between these uncertainty estimates being predominantly attributable to the
change in AIS coverage over the period of the study. The uncertainty in the earlier years is dominated by
uncertainty in the activity data, due to the lack of satellite AIS data. In later years (2010, 2011 and 2012), this
uncertainty reduces, but the discrepancy between the number of ships identified as in service in IHSF and the
ships observed on AIS increases (relative to the earlier years). The result is that the total bottom-up uncertainty
only reduces slightly in the later years when improved AIS data is available.
The top-down estimates are also uncertain, and include observed discrepancies between global imports and
exports of fuel oil and distillate oil, observed transfer discrepancies among fuel products that can be blended
into marine fuels and the potential for misallocation of fuels between sectors of shipping (international,
domestic and fishing).
88 Third IMO GHG Study 2014
Figure 64: Summary of uncertainty on top-down and bottom-up fuel inventories for a) all ships and
b) international shipping
1.6 Comparison of the CO
2
inventories in this study
to the Second IMO GHG Study 2009 inventories
The Third IMO GHG Study 2014 produces multi-year inventories including 2007, which is the year that
the Second IMO GHG Study 2009 selected for its most detailed inventory. The two top-down inventories
compare very closely, at 249 million versus 234 million metric tonnes of fuel for the 2014 and 2009 studies
respectively. Top-down comparisons differ by less than 10% and can be explained by the extrapolation of
2005 IEA data used by the Second IMO GHG Study 2009 to estimate 2007 top-down totals. Similarly, the best
estimates for bottom-up global fuel inventories for 2007 in both studies differ by just over 5%, at 352 million
versus 333 million metric tonnes fuel respectively. Bottom-up fuel inventories for international shipping differ
by less than 3%.
Figure 65 and Figure 66 present results from this study (all years) and also from the Second IMO GHG Study
2009 (2007 only), including the uncertainty ranges for this work as presented in Section 1.5. The comparison
of the estimates in 2007 shows that for both the top-down and bottom-up analysis methods, for both the total
fuel inventory and international shipping, the results of the Third IMO GHG Study 2014 are in close agreement
with findings from the Second IMO GHG Study 2009. Similarly, the CO
2
estimate of 1,054 million metric
tonnes reported by the Second IMO GHG Study 2009 falls within the multi-year range of CO
2
estimates
reported in the bottom-up method for this study.
Figure 65: Top-down and bottom-up inventories for all ship fuels, from the Third IMO GHG Study 2014 and
the Second IMO GHG Study 2009
Inventories of CO2 emissions from international shipping 2007–2012 89
1.6 Comparison of the CO
2
inventories in this study
to the Second IMO GHG Study 2009 inventories
The Third IMO GHG Study 2014 produces multi-year inventories including 2007, which is the year that
the Second IMO GHG Study 2009 selected for its most detailed inventory. The two top-down inventories
compare very closely, at 249 million versus 234 million metric tonnes of fuel for the 2014 and 2009 studies
respectively. Top-down comparisons differ by less than 10% and can be explained by the extrapolation of
2005 IEA data used by the Second IMO GHG Study 2009 to estimate 2007 top-down totals. Similarly, the best
estimates for bottom-up global fuel inventories for 2007 in both studies differ by just over 5%, at 352 million
versus 333 million metric tonnes fuel respectively. Bottom-up fuel inventories for international shipping differ
by less than 3%.
Figure 65 and Figure 66 present results from this study (all years) and also from the Second IMO GHG Study
2009 (2007 only), including the uncertainty ranges for this work as presented in Section 1.5. The comparison
of the estimates in 2007 shows that for both the top-down and bottom-up analysis methods, for both the total
fuel inventory and international shipping, the results of the Third IMO GHG Study 2014 are in close agreement
with findings from the Second IMO GHG Study 2009. Similarly, the CO
2
estimate of 1,054 million metric
tonnes reported by the Second IMO GHG Study 2009 falls within the multi-year range of CO
2
estimates
reported in the bottom-up method for this study.
Figure 65: Top-down and bottom-up inventories for all ship fuels, from the Third IMO GHG Study 2014 and
the Second IMO GHG Study 2009
90 Third IMO GHG Study 2014
Figure 66: Top-down and bottom-up inventories for international shipping fuels,
from the Third IMO GHG Study 2014 and the Second IMO GHG Study 2009
Differences between the bottom-up and top-down estimated values are consistent with the Second IMO
GHG Study 2009. This convergence is important because, in conjunction with the quality (Section 1.4) and
uncertainty (Section 1.5) analyses, it provides evidence that increasing confidence can be placed in both
analytic approaches.
There are some important explanatory reasons for the detailed activity method reported here to have
fundamental similarity with other activity-based methods, even if they are less detailed. Crossplot comparisons
in Figure 67 indicate that the fundamental input data to the bottom-up inventory in the Second IMO GHG
Study 2009 appear valid, compared to the best available data used in the Third IMO GHG Study 2014.
a) Deadweight tonnes b) Gross registered tonnes
c) Main engine power installed
Figure 67: Crossplots of deadweight tonnes, gross tonnes and average installed main engine power
for the year 2007, as reported by the Second IMO GHG Study 2009 (x-axis)
and the Third IMO GHG Study 2014 (y-axis)
There are differences in parameters between the studies. The most important uncertainty identified by the
Second IMO GHG Study 2009 was engine operating days, especially for main engines. The 2009 study
considered confidence to be “moderate, but dominat[ing] uncertainty”, and explained that the coverage
accuracy of the AIS data would affect uncertainty in several ways. Uncertainty in main engine load was
reported as the second most important parameter affecting confidence in the 2009 bottom-up calculations.
Generally, uncertainty in auxiliary engine inputs was assessed as moderate to low in the Second IMO GHG
Study 2009 (i.e. the study reported confidence in these to be moderate to high). The 2009 study identified
several ways in which auxiliary engine information was uncertain, including engine size, auxiliary engine
operating days, auxiliary engine load and auxiliary engine specific fuel oil consumption. The IHSF data on
auxiliary engines used in the Third IMO GHG Study 2014 remained sparse, although the consortium was able
to access auxiliary data for more than 1,000 ships from noon reports, previous vessel boardings, etc. These
are shown in Figure 68.
Inventories of CO2 emissions from international shipping 2007–2012 91
a) Deadweight tonnes b) Gross registered tonnes
c) Main engine power installed
Figure 67: Crossplots of deadweight tonnes, gross tonnes and average installed main engine power
for the year 2007, as reported by the Second IMO GHG Study 2009 (x-axis)
and the Third IMO GHG Study 2014 (y-axis)
There are differences in parameters between the studies. The most important uncertainty identified by the
Second IMO GHG Study 2009 was engine operating days, especially for main engines. The 2009 study
considered confidence to be “moderate, but dominat[ing] uncertainty”, and explained that the coverage
accuracy of the AIS data would affect uncertainty in several ways. Uncertainty in main engine load was
reported as the second most important parameter affecting confidence in the 2009 bottom-up calculations.
Generally, uncertainty in auxiliary engine inputs was assessed as moderate to low in the Second IMO GHG
Study 2009 (i.e. the study reported confidence in these to be moderate to high). The 2009 study identified
several ways in which auxiliary engine information was uncertain, including engine size, auxiliary engine
operating days, auxiliary engine load and auxiliary engine specific fuel oil consumption. The IHSF data on
auxiliary engines used in the Third IMO GHG Study 2014 remained sparse, although the consortium was able
to access auxiliary data for more than 1,000 ships from noon reports, previous vessel boardings, etc. These
are shown in Figure 68.
92 Third IMO GHG Study 2014
a) Days at sea b) Engine load (average % MCR)
c) Auxiliary engine fuel consumption
Figure 68: Crossplots for days at sea, average engine load (% MCR) and auxiliary engine fuel use
for the year 2007, as reported by the Second IMO GHG Study 2009 (x-axis)
and the Third IMO GHG Study 2014 (y-axis)
As a result, activity-based calculations of fuel consumption are generally similar. Figure 69 presents crossplots
showing that average main engine fuel consumption and average total vessel fuel consumption patterns are
consistent between the Second IMO GHG Study 2009 and the Third IMO GHG Study 2014.
a) Main engine fuel consumption b) Total vessel fuel consumption
Figure 69: Crossplots for average main engine daily fuel consumption and total vessel
daily fuel consumption for 2007, as reported by the Second IMO GHG Study 2009 (x-axis)
and the Third IMO GHG Study 2014 (y-axis)
Figure 70 demonstrates good agreement between the various components of the calculation of fuel
consumption. This provides evidence that observed good agreement in total fuel consumption is underpinned
by good agreement in model design. These crossplots are most directly related to the international shipping
totals reported in Figure 66. This is because the crossplots are limited to vessel categories that are known to
be engaged in international shipping and where the Third IMO GHG Study 2014 categories can be directly
matched to categories reported in 2009 study.
a) Main engine fuel consumption b) Total vessel fuel consumption
c) Vessel type annual fuel c) Vessel type annual CO
2
Figure 70: Crossplots for main engine annual fuel consumption, total vessel annual fuel consumption,
aggregated vessel type annual fuel consumption and CO
2
for the year 2007,
as reported by the Second IMO GHG Study 2009 (x-axis) and the Third IMO GHG Study 2014 (y-axis)
Table 33 summarizes this discussion by making explicit the key differences between the 2009 study and
the current study. Given these observations, the general conclusion is that better AIS data on activity are
determinants of the precision of individual vessel calculations for activity-based emissions inventories. The
variation between vessel voyage days, vessels in a vessel category and other important variations can only be
evaluated with access to very detailed activity data. However, if a more general approach uses representative
input parameters that reflect the best composite activity data, the results will generally be similar.
Table 33 – Summary of major differences between the Second IMO GHG Study 2009
and Third IMO GHG Study 2014
Key variable Differences 2009 study 2014 study Overall effect
Days at sea Data and method Annual IHSF status
indicator only
Uses quarterly IHSF status indicator to
indicate if laid up for part of the year
Minor decrease
in emissions
At-sea main
engine MCR
Data and method AIS-informed expert
judgement
Uses AIS data extrapolation, quality-checked
using LRIT and noon reports
Minor increase
in emissions
Auxiliary engine Data and method Expert judgement
annual aggregates
Auxiliary power outputs derived from vessel
boarding data and applied according to
mode of operation
Minor increase
in emissions
Inventories of CO2 emissions from international shipping 2007–2012 93
Figure 70 demonstrates good agreement between the various components of the calculation of fuel
consumption. This provides evidence that observed good agreement in total fuel consumption is underpinned
by good agreement in model design. These crossplots are most directly related to the international shipping
totals reported in Figure 66. This is because the crossplots are limited to vessel categories that are known to
be engaged in international shipping and where the Third IMO GHG Study 2014 categories can be directly
matched to categories reported in 2009 study.
a) Main engine fuel consumption b) Total vessel fuel consumption
c) Vessel type annual fuel c) Vessel type annual CO
2
Figure 70: Crossplots for main engine annual fuel consumption, total vessel annual fuel consumption,
aggregated vessel type annual fuel consumption and CO
2
for the year 2007,
as reported by the Second IMO GHG Study 2009 (x-axis) and the Third IMO GHG Study 2014 (y-axis)
Table 33 summarizes this discussion by making explicit the key differences between the 2009 study and
the current study. Given these observations, the general conclusion is that better AIS data on activity are
determinants of the precision of individual vessel calculations for activity-based emissions inventories. The
variation between vessel voyage days, vessels in a vessel category and other important variations can only be
evaluated with access to very detailed activity data. However, if a more general approach uses representative
input parameters that reflect the best composite activity data, the results will generally be similar.
Table 33 – Summary of major differences between the Second IMO GHG Study 2009
and Third IMO GHG Study 2014
Key variable Differences 2009 study 2014 study Overall effect
Days at sea Data and method Annual IHSF status
indicator only
Uses quarterly IHSF status indicator to
indicate if laid up for part of the year
Minor decrease
in emissions
At-sea main
engine MCR
Data and method AIS-informed expert
judgement
Uses AIS data extrapolation, quality-checked
using LRIT and noon reports
Minor increase
in emissions
Auxiliary engine Data and method Expert judgement
annual aggregates
Auxiliary power outputs derived from vessel
boarding data and applied according to
mode of operation
Minor increase
in emissions
2
Inventories of emissions of GHGs
and other relevant substances from
international shipping 2007–2012
2.1 Top-down other relevant substances inventory calculation method
2.1.1 Method for combustion emissions
The top-down calculation of non-CO
2
GHGs and other relevant substances is divided into two components:
• emissions resulting from the combustion of fuels;
• other emissions (HFCs, PFCs and SF
6
) from ships.
The emissions from the combustion of fuels are found in the fuel sales statistics (see Section 1.1) and emissions
factors (EFs) data. The method for other emissions replicates the methods used in the Second IMO GHG Study
2009.
The data for the fuel sales statistics were obtained and compiled for all available years (2007–2011) and
are described in greater detail in Section 1.1. These fuel statistics, and their uncertainty, form the basis for
top-down emissions estimates.
Estimation of emissions factors
EFs are obtained from Section 2.2, as weighted averages for a given fuel type, taking into account the variation
in engine type and operation. These values are more general in some cases than EFs used in bottom-up
methods, because the limited detail for top-down methods does not allow the application of specific EFs
to auxiliaries, varying engine load or other activity-based conditions. Generally, EFs corresponding to Tier 0
(pre-2000) engines and load factors of 70% are listed in Table 34.
Where it is known that varying fuel sulphur levels can affect the SO
x
and PM emissions factors, that information
can be used to produce yearly EFs for these top-down calculations, as shown in Table 34 and Table 35. The
fuel statistics used are aggregated for fuel use in all engine types (main engine, boiler and auxiliary). These
emissions factors are therefore not machinery-type-specific but an aggregate for fuel use in all engine types
with the preliminary working assumption that representative EFs can be derived from main engines only.
2
Inventories of emissions of GHGs
and other relevant substances from
international shipping 2007–2012
2.1 Top-down other relevant substances inventory calculation method
2.1.1 Method for combustion emissions
The top-down calculation of non-CO
2
GHGs and other relevant substances is divided into two components:
• emissions resulting from the combustion of fuels;
• other emissions (HFCs, PFCs and SF
6
) from ships.
The emissions from the combustion of fuels are found in the fuel sales statistics (see Section 1.1) and emissions
factors (EFs) data. The method for other emissions replicates the methods used in the Second IMO GHG Study
2009.
The data for the fuel sales statistics were obtained and compiled for all available years (2007–2011) and
are described in greater detail in Section 1.1. These fuel statistics, and their uncertainty, form the basis for
top-down emissions estimates.
Estimation of emissions factors
EFs are obtained from Section 2.2, as weighted averages for a given fuel type, taking into account the variation
in engine type and operation. These values are more general in some cases than EFs used in bottom-up
methods, because the limited detail for top-down methods does not allow the application of specific EFs
to auxiliaries, varying engine load or other activity-based conditions. Generally, EFs corresponding to Tier 0
(pre-2000) engines and load factors of 70% are listed in Table 34.
Where it is known that varying fuel sulphur levels can affect the SO
x
and PM emissions factors, that information
can be used to produce yearly EFs for these top-down calculations, as shown in Table 34 and Table 35. The
fuel statistics used are aggregated for fuel use in all engine types (main engine, boiler and auxiliary). These
emissions factors are therefore not machinery-type-specific but an aggregate for fuel use in all engine types
with the preliminary working assumption that representative EFs can be derived from main engines only.
96 Third IMO GHG Study 2014
Table 34 – Emissions factors for top-down emissions from combustion of fuels
Emissions
substance
Marine HFO emissions factor
(g/g fuel)
Marine MDO emissions factor
(g/g fuel)
Marine LNG emissions factor
(g/g fuel)
CO
2
3.114 0 0 3.20600 2.75000
CH
4
0.00006 0.00006 0.05120
N
2
O 0.00016 0.00015 0.00011
NO
x
0.09300 0.08725 0.00783
CO 0.00277 0.00277 0.00783
NMVOC 0.00308 0.00308 0.00301
Table 35 – Year-specific emissions factors for sulphur-dependent emissions (SO
x
and PM)
% Sulphur content averages – wt IMO
1
Fuel type 2007 2008 2009 2010 2011 2012
Average non-ECA HFO S% 2.42 2.37 2.6 2.61 2.65 2.51
SO
x
EF (g/g fuel)
Marine fuel oil (HFO) 0.0 4749 0.04644 0.05066 0.05119 0.05171 0.04908
Marine gas oil (MDO) 0.00264 0.00264 0.00264 0.00264 0.00264 0.00264
Natural gas (LNG) 0.00002 0.00002 0.00002 0.00002 0.00002 0.00002
PM EF (g/g fuel)
Marine fuel oil (HFO) 0.00684 0.00677 0.00713 0.00713 0.00721 0.00699
Marine gas oil (MDO) 0.00102 0.00102 0.00102 0.00102 0.00102 0.00102
Natural gas (LNG) 0.00018 0.00018 0.00018 0.00018 0.00018 0.00018
1
Source: MEPC annual reports on Sulphur Monitoring Programme.
All emissions factors are in mass of emissions per unit mass of fuel and the data compiled in Section 1.1
are in units of mass of fuel, so for oil-based fuels the production of the total emissions is a straightforward
multiplication. Further work is needed to compile the gas fuel emissions factors and the method for emissions
calculation (the units for gas fuel use are mass of oil equivalent).
2.1.2 Methane slip
Some of the fuel used in gas engines is emitted unburned to the atmosphere. This feature is specific to LNG
marine engines running on LNG with low engine loads. A new generation of gas engines, based on the Otto
cycle (spark-ignited, lean-burn engines), is reported to reduce methane slip significantly with improvements
made to cylinder, cylinder head and valve systems. In this study, methane slip is included in the combustion EF
for CH
4
in LNG-fuelled engines. However, for the top-down analysis it was not feasible to estimate the energy
usage (kWh) for the global LNG fleet.
2.1.3 Method for estimation for non-combustion emissions
Refrigerants, halogenated hydrocarbons
Refrigerants are used on board vessels for air conditioning and provisional and cargo cooling purposes.
The ozone-depleting substances (HCFCs and CFCs) have been replaced with other refrigerants, like HFCs
1,1,1,2-tetrafluoroethane (R134a) and a mixture of pentafluoroethane, trifluoroethane and tetrafluoroethane
(R404a). All these refrigerants, including the replacements for ozone-depleting substances, have significant
GWP. The GWP is reported as CO
2
equivalent (CO
2
e): this describes the equivalent amount of CO
2
that would
be needed to achieve the same warming effect. The numerical values of GWP for different substances used
in this study were taken from the IPCC Fourth Assessment Report and are based on the latest IPCC estimate
of CO
2
concentration in the atmosphere.
This part of the report builds on the findings of two others: the United Nations Environmental Programme
(UNEP) 2010 Report of the Refrigeration, Air Conditioning and Heat Pumps Technical Options Committee
and the European Commission (DG Environment) 2007 report The analysis of the emissions of fluorinated
greenhouse gases from refrigeration and air conditioning equipment used in the transport sector other than
road transport and options for reducing these emissions – Maritime, Rail, and Aircraft Sector.
Inventories of emissions of GHGs and other relevant substances 97
Other refrigerants, SF
6
Sulphur hexafluoride (SF
6
) is a colourless, odourless, non-toxic, non-flammable gas that has a high dielectric
strength. It has been used as a dielectric in microwave frequencies, as an insulating medium for the power
supplies of high-voltage machines and in some military applications, for example as a torpedo propellant.
Sulphur hexafluoride is also gaining use in non-electrical applications, including blanketing of molten
magnesium (molten magnesium will oxidize violently in air), leak detection and plasma etching in the semi-
conductor industry. Sulphur hexafluoride also has some limited medical applications. SF
6
is extensively used
as a gaseous dielectric in various kinds of electrical power equipment, such as switchgear, transformers,
condensers and medium- to high-voltage (>1 kV) circuit breakers (Compressed Gas Association, 1990). In
circuit breakers, SF
6
is typically used in a sealed pressurized chamber to prevent electrical arcing between
conductors.
According to World Bank data (2010), global SF
6
emissions were 22,800 thousand tonnes CO
2
e, which
corresponds to 463 tons (i.e. short tons, as per key definition of “ton”) of SF
6
emitted from all sectors. (According
to UNFCCC, SF
6
has a GWP of 23,900.) The use of SF
6
in electrical switchgear in general (all land, air and sea
installations) is primarily (90%) concentrated on the high-voltage segment (>36 kV) and the remaining 10% for
the medium (1 kV–36 kV) voltage segment (Schneider 2003). Ships rarely use electrical systems over 11 kV and
typical nominal voltages are in the 1 kV–11 kV range (Ackermann and Planitz, 2009). The leaks from sealed
systems are small: EPA (2006) estimates a range of 0.2%–2.5% per year. However, the mass of SF
6
on board
the global fleet is unknown, which prohibits detailed analysis of SF
6
emissions from shipping.
If this 90%/10% division is assumed, which represents SF
6
use in high/medium voltage systems, and also
applies to emissions, medium-voltage systems would be responsible for 46.3 tons of SF
6
emitted annually. If all
medium-voltage systems were installed in ships (i.e. no medium-voltage installations on land), the maximum
contribution to total GHG emissions from shipping would be 1.1 million tons (46.3 tons × 22,800 CO
2
e/ton)
of CO
2
e (IPCC, 2007), which is less than 0.1% of the total CO
2
emissions from shipping in 2010. The actual
emissions of SF
6
are likely to be less than this, because alternative solutions (vacuum, CO
2
) are also available in
arc quenching. Because SF
6
emissions from ships are negligible, they are not considered further in this report.
Other refrigerants, PFCs
Several binary and ternary blends of various HFC, HCFC, PFC and hydrocarbon refrigerants have been
developed to address continuing service demand for CFC-12. These blends are tailored to have physical and
thermodynamic properties comparable to the requirements of the original CFC-12 refrigerant charge.
HFCs were used to replace halon-based systems in the mid 1990s. A small quantity of PFC (mainly C
4
F
10
) was
imported by a US company into the EU to be used as an alternative fluid in firefighting fixed systems. The
main application of these PFC-based fixed systems is for fire protection by flooding closed rooms (e.g. control
rooms) with halon to replace oxygen. Imports for new systems stopped in 1999, as this application of PFCs
was not regarded as an essential use (AEA, 2010). The electronics and metal industry is a large consumer of
PFC compounds, which are used as etching agents during manufacturing (IPCC/TEAP, 2005). The main PFC
used as a refrigerant is octafluoropropane (C
3
F
8
), which is a component of the R-413a refrigerant (Danish EPA,
2003). The composition of R-413a is 88% R134a, 9% C
3
F
8
and 3% isobutane and it is used in automotive air
conditioning (Danish EPA, 2003). Another refrigerant with C
3
F
8
is Isceon 89, a mixture of 86% HFC-125, 9%
C
3
F
8
and 5% propane. Isceon 89 is used for deep-freezing purposes (-40°C to -70°C), like freeze-dryers,
medical freezers and environmental chambers (DuPont, 2005).
The annual leakage of all refrigerants from cooling equipment of reefer and fishing vessels is estimated at
2,200 tons. The extreme worst-case estimate assumes that all this is Isceon 89, which contains 9% of C
3
F
8
.
This would total 201 tons of C
3
F
8
and correspond to (8,830 CO
2
e/ton × 201 tons) 1.8 million tons of CO
2
e,
which is about 0.2% of the total CO
2
emitted from ships in 2010. The emissions of C
3
F
8
from ships are likely
to be smaller than this value because the need for extreme cooling is limited; only some reefer cargo ships and
fishing vessels may need this temperature range. Because PFC emissions from ships are likely to be negligible,
they are not considered further in this report.
98 Third IMO GHG Study 2014
Method used in this study
In this study the use of ozone-depleting R-22 has been restricted to vessels built before 2000. The amounts of
refrigerant used in various types of ship for air conditioning of passenger areas and provision of refrigeration
(galley, cargo) are described in Table 36.
Table 36 – Amounts of refrigerants carried by various types of ships (from DG ENV report)
Ship type kg/AC kg/refr % vessels built after 1999
Bulk carrier 150 10 59%
Chemical tanker 150 10 63%
Container 150 10 59%
Cruise 6,000 400 37%
Ferry – pax only 500 20 23%
Ferry – ro-pax 500 20 27%
General cargo 150 10 27%
Liquefied gas tanker 150 10 53%
Miscellaneous –fishing 150 210
**
15%
Miscellaneous – other 150 10 32%
Offshore 150 10 56%
Oil tanker 150 10 45%
Other liquids tankers 150 10 45%
Refrigerated bulk 150 2,500
*
7%
Ro-ro 500 20 26%
Service – tug 150 10 45%
Service – other 150 10 32%
Vehicle 150 10 57%
Yacht 150 10 66%
Total, tons in global fleet 21,917 tons 8,569 tons
*
Vessels using cargo cooling are assigned 2,500kg refrigerant charge, which is an average of the range (1,000kg–5,000kg) indicated in the
DG ENV report.
**
Refrigerant carried by fishing vessels has been calculated as a weighted average of 7,970 fishing vessels described in DG ENV report.
In addition to the vessels, there are 1.7 million refrigerated containers, each of which carries approximately
6kg of refrigerant (80% R134a, 20% R-22) (DG ENV, 2007).
Refrigerants used in the calculation are assumed as R-22 for both air conditioning and cooling for vessels
built before 2000. For newer vessels, R134a is assumed for air conditioning and R404a for provisional cooling
purposes. Refrigerant loss of 40% is assumed for all ships, except for passenger vessels for which 20% annual
loss of refrigerants is assumed.
Inventories of emissions of GHGs and other relevant substances 99
Fishing vessels and reefer ships
In Table 36, two distinctions between the existing reports (UNEP, DG ENV) are made. First, the refrigerant
charge carried by the world fishing fleet (“Miscellaneous – fishing”) was based on the DG ENV report, which
describes the use of refrigerants on board the European fishing fleet. In this study, the weighted average
(number of vessels, refrigerant charge carried) of the European fishing fleet (approximately 8,000 vessels) was
used to estimate the air conditioning and cooling needs of the global fishing fleet. The composition of the
EU fishing fleet is likely to be different from the global fleet, and this will be reflected in the estimates of the
refrigerant emissions of the global fishing fleet. The second difference concerns reefer ships. According to both
existing reports (UNEP, DG ENV), the reefer fleet carries 1 ton–5 tons of refrigerants per ship for cargo cooling.
This study takes the average (2.5 tons of refrigerants) and assumes R-22 to be used in vessels built before 2000
(DG ENV, 2007).
Reefer containers
Refrigerants can also be found in the cooling systems of reefer containers, which are used to provide a controlled
environment for perishable goods, like fruit, during cargo transport. The fleet of dedicated refrigerated cargo-
carrying vessels has decreased over the years and is slowly being replaced by container ships carrying reefer
containers. According to the DG ENV report (2007), each reefer container carries 6kg refrigerant charge, of
which 15% is lost annually. The number of refrigerated containers has been estimated in the DG ENV report
(2006 figure) as 1.6 million TEU. In this study the number of refrigerated containers for 2012 was based on
the projected number of reefer plugs of the world container fleet (1.7 million TEU). The reefer container count
was based on the IHS Fairplay data for 5,400 container ships (1.7 million TEU). The projection has some
inherent uncertainty, because reefer plug installations (rather than reefer TEU counts) have been used. Also,
the completeness of the container ship fleet in the data set used to determine the reefer plug count is likely to
have some impact on the reefer TEU numbers, because this data set consists of some 85,000 vessels and so
does not cover the complete global fleet.
Estimated emissions of refrigerants from ships
Both the UNEP and DG ENV reports use the 100 gt limit to indicate a vessel that has refrigerants on board.
This assumption was based on expert judgements on vessels that operate in a variety of climate conditions
and need air conditioning.
In this study, the fleet-wide assessment is made according to the vessel construction year (before 2000,
constructed that year or later) and refrigerant type is assigned on the basis of the vessels’ age. For old vessels,
HCFCs (R-22) were assumed, while new vessels use HFCs (R134a/R404a).
The estimated annual total of refrigerant loss in the global fleet in 2012 is described in Table 37.
100 Third IMO GHG Study 2014
Table 37 – Annual loss of refrigerants from the global fleet during 2012. Annual release of 40% total
refrigerant carried is assumed except for passenger-class vessels, where 20% refrigerant loss is assumed.
Ro-ro, pax, ro-pax and cruise vessels are calculated as passenger ships
Ship type Annual loss, air
conditioning, tons
Annual loss,
cooling, tons
R-22, tons R134a, tons R404, tons
Bulk carrier 466.9 31.1 195.7 275.4 14.6
Chemical tanker 221.7 14.8 83.6 140.0 6.7
Container 230.5 15.4 96.4 136.2 7. 2
Cruise 622.8 41.5 4 0 7.9 228.1 20.8
Ferry – pax only 313.9 12.6 245.9 72.8 6.3
Ferry – ro-pax 285.6 11.4 211.8 78.0 5.7
General cargo 740.0 49.3 555.2 196.9 28.5
Liquefied gas tanker 72.4 4.8 35.1 38.1 2.4
Miscellaneous – fishing 1,000.3 1,421.1 1,259.8 145.4 878.6
Miscellaneous – other 261.0 17.4 180.9 84.0 9.7
Offshore 309.2 20.6 138.2 174.0 9.9
Oil tanker 332.1 22.1 186.0 150.1 11.4
Other liquids tankers 6.7 0.4 3.8 3.0 0.2
Refrigerated bulk 48.7 812.3 297. 5 3.4 522.9
Ro-ro 173.9 7.0 130.7 45.7 3.5
Service – tug 657. 5 43.8 372.9 292.7 22.7
Service – other 26.1 1.7 18.1 8.4 1.0
Vehicle 37. 5 2.5 16.6 21.3 1.2
Yacht 70.2 4.7 24.2 46.6 2.1
Total, tons 5, 877.1 2,534.6 4,460.1 2,140.2 1,555.4
The estimated reefer TEU count globally is 1.7 million TEU, which would result in 10,070 tons of refrigerant
charge and 1,510 tons of refrigerant release in 2012. This means an additional 1,208 tons of R134a and 302
tons of R404 on top of the values in Table 37, if the 80:20 ratio of the DG ENV (2007) report is used.
There is large uncertainty about the leakage rate of refrigerants from ships. A range of 20%–40% is reported
by both UNEP and DG ENV, attributed to the permanent exposure of refrigerated systems to continuous
motion (waves), which can cause damage and leakage to piping (DG ENV). The average estimate, using a
30% leakage rate, is described in Table 37 and amounts to 8,412 tons. The corresponding values for low- and
high-bound estimates are 5,967 tons and 10,726 tons respectively. In the 2010 UNEP report, the annual loss
of refrigerants is reported as 7,850 tons, which is close to the estimate of this study. If the refrigerant emissions
from reefer containers are included, then an additional 1,510 tons (80% R134a, 20% R404a) should be added
to these numbers.
Global warming potential of refrigerant emission from ships
According to the results of this study, the share of R-22 is 70%, R134a 26% and R404a 4%. The balance of
refrigerant shares will shift towards R134a when old vessels using R-22 as a cooling agent are replaced with
new ships using HFCs (R134a). The use of R-22 in industrial refrigeration in developed countries is on the
decline because it is banned in new refrigerating units. However, the Montreal Protocol has determined that
it can be used until 2040 in developing countries.
Table 38 – Global warming potential of refrigerants commonly used in ships. The GWP100 is described
relative to CO
2
warming potential (IPCC Fourth Assessment Report: Climate Change 2007)
Refrigerant Warming potential
(relative to CO
2
)
R-22 1,810
R134a 1,430
R404a 3,260
The release of refrigerants from global shipping is estimated at 8,412 tons, which corresponds to 15 million
tons (range 10.8 million tons–19.1 million tons) in CO
2
e emissions. Inclusion of reefer container refrigerant
emissions yields 13.5 million tons (low) and 21.8 million tons (high) of CO
2
e emissions. If these numbers
are compared to CO
2
emissions of shipping during 2011 (top-down estimate of 794 million tons of CO
2
),
refrigerant emissions constitute about 1.9% of the GHG emissions of shipping. Inclusion of the reefer TEU
increases this to 2.2% of the total GHG emissions from shipping.
Refrigerant emissions from ships 2007–2012
The emissions of refrigerants from ships are mainly affected by changes in the size and composition of the
global fleet. The methodology used to assess refrigerant emissions is driven by the age structure of each ship
type rather than the activity patterns of vessels. This assumption makes the annual emission changes small
(Figure 71) but nevertheless consistent with the UNEP report (2010). Also, the dominant substance is R-22
(70% share), which is in line with previous studies (UNEP 2010; DG ENV 2007).
Figure 71: Estimated refrigerant emissions of the global fleet 2007–2012
The slow decrease of R-22 share in ship systems (Table 39) means that R-22 will be present for a long time,
possibily decades, before it is replaced by other substances.
Inventories of emissions of GHGs and other relevant substances 101
Table 38 – Global warming potential of refrigerants commonly used in ships. The GWP100 is described
relative to CO
2
warming potential (IPCC Fourth Assessment Report: Climate Change 2007)
Refrigerant Warming potential
(relative to CO
2
)
R-22 1,810
R134a 1,430
R404a 3,260
The release of refrigerants from global shipping is estimated at 8,412 tons, which corresponds to 15 million
tons (range 10.8 million tons–19.1 million tons) in CO
2
e emissions. Inclusion of reefer container refrigerant
emissions yields 13.5 million tons (low) and 21.8 million tons (high) of CO
2
e emissions. If these numbers
are compared to CO
2
emissions of shipping during 2011 (top-down estimate of 794 million tons of CO
2
),
refrigerant emissions constitute about 1.9% of the GHG emissions of shipping. Inclusion of the reefer TEU
increases this to 2.2% of the total GHG emissions from shipping.
Refrigerant emissions from ships 2007–2012
The emissions of refrigerants from ships are mainly affected by changes in the size and composition of the
global fleet. The methodology used to assess refrigerant emissions is driven by the age structure of each ship
type rather than the activity patterns of vessels. This assumption makes the annual emission changes small
(Figure 71) but nevertheless consistent with the UNEP report (2010). Also, the dominant substance is R-22
(70% share), which is in line with previous studies (UNEP 2010; DG ENV 2007).
Figure 71: Estimated refrigerant emissions of the global fleet 2007–2012
The slow decrease of R-22 share in ship systems (Table 39) means that R-22 will be present for a long time,
possibily decades, before it is replaced by other substances.
102 Third IMO GHG Study 2014
Table 39 – Annual emissions of refrigerants from the global fleet
and estimated shares of different refrigerants
Year
Refrigerant emissions,
tons, reefer TEU excluded
Low bound, tons High bound, tons %, R-22 %, R134a %, R404
2007 8,185 5,926 10,444 80% 17% 4%
2008 8,349 6,045 10,654 77% 19% 4%
2009 8,484 6,144 10,825 75% 21% 4%
2010 8,709 6,307 11,110 73% 23% 4%
2011 8,235 5,967 10,503 71% 24% 4%
2012 8,412 5,967 10,726 70% 26% 4%
UNEP 2010 7,850
Non-exhaust emissions of NMVOCs from ships
The reported global crude oil transport in 2012 was 1,929 million tons (UNCTAD Review of Maritime Transport
2013). This study applies the same methodology as the Second IMO GHG Study 2009 and uses the net
standard volume (= NSV at bill of lading - NSV at out-turn) loss of 0.177%. This corresponds to 0.124% mass
loss and results in VOC emissions of 2.4 million tons, which is very close to the value of the 2009 study figures
for 2006 (crude oil transport 1,941 million tons, VOC emissions 2.4 million tons).
2.2 Bottom-up other relevant substances emissions calculation method
2.2.1 Method
Three primary emission sources are found on ships: main engine(s), auxiliary engines and boilers. The
consortium studied emissions from main and auxiliary engines as well as boilers in this report. Emissions from
other energy-consuming sources were omitted because of their small overall contribution. Emissions from
non-combustion sources, such as HFCs, are estimated consistent with the Second IMO GHG Study 2009
methods.
2.2.2 Main engine(s)
Emissions from the main engine(s) or propulsion engine(s) (both in terms of magnitude and emissions factor)
vary as a function of main engine rated power output, load factor and the engine build year. The main engine
power output and load factor vary over time as a result of a ship’s operation and activity specifics: operational
mode (e.g. at berth, anchoring, manoeuvring), speed, loading condition, weather, etc. Emissions are also
specific to a ship, as individual ships have varying machinery and activity specifications. The bottom-up
model described in Section 1.2 calculates these specifics (main engine power output and load factor) for each
individual ship in the global fleet and for activity over the year disaggregated to an hourly basis. This same
model is therefore used for the calculations of the other main engine emissions substances.
2.2.3 Auxiliary engines
Emissions from auxiliary engines (both in terms of magnitude and emissions factor) vary as a function of
auxiliary power demand (typically changing by vessel operation mode), auxiliary engine rated power output,
load factor and the engine build year. Technical and operational data about auxiliary engines are often missing
from commercial databases, especially for older ships (constructed before 2000). Technical data (power rating,
stroke, model number, etc.) for auxiliary engines of new vessels can be found much more frequently than for
those of old vessels; however, these form a very small percentage of the entire fleet. There are typically two or
more auxiliary engines on a ship and the number and power rating (not necessarily the same for all engines on
a ship) of each engine is determined by the ship owner’s design criteria. This means that the actual operation
of the specific auxiliary engines, by vessel type and operational mode, can vary significantly from ship to
ship. There are no commercial databases that provide these operational profiles on the basis of operational
mode or vessel class. This lack of data will hinder the determination of auxiliary engine power estimation
using predetermined auxiliary engine load levels. For this reason, the approach taken in this study is based
on the vessel surveys conducted by Starcrest for various ports in North America. These surveys allow the
determination of auxiliary engine power requirements or total auxiliary loads in various operating modes of
vessels. Further information relating to the approaches used to estimate auxiliary engine loads are provided
in Section 1.2.5 and Annex 1. A detailed explanation of auxiliary engine power prediction can be found in
Starcrest (2013).
2.2.4 Boilers
Emissions from auxiliary boilers vary based on vessel class and operational mode. For example, tankers
typically have large steam plants powered by large boilers that supply steam to the cargo pumps and in some
cases heat cargoes. For most non-tanker class vessels, boilers are used to supply hot water to keep the main
engine(s) warm (during at-berth or anchorage calls) and for crew and other ancillary needs. These boilers are
typically smaller and are not used during open-ocean operations because of the waste heat recovery systems
(i.e. economizers) that take the waste head from the main engine(s). Unlike main and auxiliary engines, the
emissions factors do not change, as there are no regulatory frameworks associated with boilers. Of the three
emission source types, boilers typically have significantly fewer emissions than main and auxiliary engines.
Further details about auxiliary boilers are provided in Section 1.2.5 and Annex 1.
2.2.5 Operating modes
The auxiliary engine use profiles have been specifically defined for each ship type and size class. Furthermore,
auxiliary engine use varies according to vessel operating modes, which are defined by vessel speed ranges.
The modes used in this study are defined in Table 40. Auxiliary engine use during harbour visits is divided into
two modes: “at berth” describes the auxiliary engine use during cargo loading or unloading operations and
“anchoring” involves extended waiting periods when cargo operations do not take place.
Table 40 – Vessel operating modes used in this study
Speed Mode
Less than 1 knot At berth
1 knot–3 knots Anchored
Greater than 3 knots and less than 20% MCR Manoeuvring
Between 20% MCR and 65% MCR Slow-steaming
Above 65% MCR Normal cruising
Further details on auxiliary engine and boiler loads, by vessel class and mode, are given in Section 1.2.5 and
Annex 1.
2.2.6 Non-combustion emissions
Emissions from non-combustion sources (refrigerants and NMVOCs from oil transport) on board vessels were
evaluated with the top-down approach using the fleet-wide methodology described in Section 2.1.3 to maintain
consistency with the Second IMO GHG Study 2009. The emissions factors of non-combustion sources have
wide variations and the significance to overall GHG emissions is small (less than 3%). It is very unlikely that
the bottom-up approach to the modelling of non-combustion sources would change this conclusion.
Methane emissions
Emissions of CH
4
to the atmosphere are associated with LNG-powered vessels and include venting, leakage
and methane slip. Venting and leakage related to maritime LNG operations are not included in this report.
Methane slip during the combustion process is accounted for in the combustion emissions factors detailed in
Section 2.2.7.
NMVOC emissions from non-combustion sources
The NMVOC emissions from crude oil cargo operations and transport have not been included in the bottom-up
analysis. An estimate of global NMVOC emissions has been presented in the top-down analysis (see Section
2.1.3).
Inventories of emissions of GHGs and other relevant substances 103
vessels. Further information relating to the approaches used to estimate auxiliary engine loads are provided
in Section 1.2.5 and Annex 1. A detailed explanation of auxiliary engine power prediction can be found in
Starcrest (2013).
2.2.4 Boilers
Emissions from auxiliary boilers vary based on vessel class and operational mode. For example, tankers
typically have large steam plants powered by large boilers that supply steam to the cargo pumps and in some
cases heat cargoes. For most non-tanker class vessels, boilers are used to supply hot water to keep the main
engine(s) warm (during at-berth or anchorage calls) and for crew and other ancillary needs. These boilers are
typically smaller and are not used during open-ocean operations because of the waste heat recovery systems
(i.e. economizers) that take the waste head from the main engine(s). Unlike main and auxiliary engines, the
emissions factors do not change, as there are no regulatory frameworks associated with boilers. Of the three
emission source types, boilers typically have significantly fewer emissions than main and auxiliary engines.
Further details about auxiliary boilers are provided in Section 1.2.5 and Annex 1.
2.2.5 Operating modes
The auxiliary engine use profiles have been specifically defined for each ship type and size class. Furthermore,
auxiliary engine use varies according to vessel operating modes, which are defined by vessel speed ranges.
The modes used in this study are defined in Table 40. Auxiliary engine use during harbour visits is divided into
two modes: “at berth” describes the auxiliary engine use during cargo loading or unloading operations and
“anchoring” involves extended waiting periods when cargo operations do not take place.
Table 40 – Vessel operating modes used in this study
Speed Mode
Less than 1 knot At berth
1 knot–3 knots Anchored
Greater than 3 knots and less than 20% MCR Manoeuvring
Between 20% MCR and 65% MCR Slow-steaming
Above 65% MCR Normal cruising
Further details on auxiliary engine and boiler loads, by vessel class and mode, are given in Section 1.2.5 and
Annex 1.
2.2.6 Non-combustion emissions
Emissions from non-combustion sources (refrigerants and NMVOCs from oil transport) on board vessels were
evaluated with the top-down approach using the fleet-wide methodology described in Section 2.1.3 to maintain
consistency with the Second IMO GHG Study 2009. The emissions factors of non-combustion sources have
wide variations and the significance to overall GHG emissions is small (less than 3%). It is very unlikely that
the bottom-up approach to the modelling of non-combustion sources would change this conclusion.
Methane emissions
Emissions of CH
4
to the atmosphere are associated with LNG-powered vessels and include venting, leakage
and methane slip. Venting and leakage related to maritime LNG operations are not included in this report.
Methane slip during the combustion process is accounted for in the combustion emissions factors detailed in
Section 2.2.7.
NMVOC emissions from non-combustion sources
The NMVOC emissions from crude oil cargo operations and transport have not been included in the bottom-up
analysis. An estimate of global NMVOC emissions has been presented in the top-down analysis (see Section
2.1.3).
104 Third IMO GHG Study 2014
2.2.7 Combustion emissions factors
Emissions factors are used in conjunction with energy or fuel consumption to estimate emissions and can
vary by pollutant, engine type, duty cycle and fuel. Emissions tests are used to develop emissions factors in
g/kWh and are converted to fuel-based emissions factors (grams of pollutant per grams of fuel consumed) by
dividing by the brake-specific fuel consumption (BSFC) or specific fuel oil consumption (SFOC) corresponding
to the test associated with the emissions factors. Pollutant-specific information relating to emissions factors is
provided later in this section. Emissions factors vary by: engine type (main, auxiliary, auxiliary boilers); engine
rating (SSD, MSD, HSD); whether engines are pre-IMO Tier I or meet IMO Tier I or II requirements; and type
of service (duty cycle) in which they operate (propulsion or auxiliary). Emissions factors are adjusted further for
fuel type (HFO, MDO, MGO and LNG) and the sulphur content of the fuel being burned. Finally, engine load
variability is incorporated into the factors used for estimating emissions. All these variables were taken into
account when estimating the bottom-up emissions inventories (2007–2012) using the following methodology:
1 Identify baseline emissions factors with the following hierarchy: IMO emissions factors; if none
published, then consortium-recommended emissions factors from other studies that members are
using in their published work. Emissions factors come in two groups: energy-based in g pollutant/kWh
and fuel-based in g pollutant/g fuel consumed. The baseline fuel for the bottom-up emissions factors
is defined as HFO fuel with 2.7% sulphur content.
2 Convert energy-based baseline emissions factors in g pollutant/kWh to fuel-based emissions factors
in g pollutant/g fuel consumed, as applicable, using:
EF
baseline
(g pollutant)⁄(g fuel) =
EF
baseline
(g pollutant⁄kWh)
___________________
SFOC
baseline
(g fuel⁄kWh)
Eq. (1)
where
EF
baseline
= cited emissions factor
SFOC
baseline
= SFOC associated with the cited emissions factor
3 Use FCFs, as applicable, to adjust emissions factors for the specific fuel used by the engine:
EF
actual
(g pollutant)⁄(g fuel) = EF
baseline
(g pollutant)⁄(g fuel) × FCF Eq. (2)
Convert to kg pollutant/tonne fuel consumed (for presentation/comparison purposes consistent with
Second IMO GHG Study 2009).
4 Adjust EF
actual
based on variable engine loads using SFOC engine curves and low load adjustment
factors to adjust the SFOC.
Emissions factors were developed for the following GHGs and pollutants:
• carbon dioxide (CO
2
)
• nitrogen oxides (NO
x
)
• sulphur oxides (SO
x
)
• particulate matter (PM)
• carbon monoxide (CO)
• methane (CH
4
)
• nitrous oxide (N
2
O)
• non-methane volatile organic compounds (NMVOC)
An overview of baseline emissions factors, fuel correction factors and adjustments based on variable engine
loads and SFOC is provided in the following sections on GHGs and pollutants. For comparison purposes
with the Second IMO GHG Study 2009, emissions factors are provided in kg of pollutant per tonne of fuel.
Emissions factors in grams of pollutant per gram of fuel and grams of pollutant per kWh or g/kWh along with
associated references are provided in Table 22 in Annex 6.
CO
2
baseline
The carbon content of each fuel type is constant and is not affected by engine type, duty cycle or other
parameters when looking on the basis of kg CO
2
per tonne fuel. The fuel-based CO
2
emissions factors for
main and auxiliary engines at slow, medium and high speeds are based on MEPC 63/23, annex 8, and include:
HFO EF
baseline
CO
2
= 3,114 kg CO
2
/tonne fuel
MDO/MGO EF
baseline
CO
2
= 3,206 kg CO
2
/tonne fuel
LNG EF
baseline
CO
2
= 2,750 kg CO
2
/tonne fuel
It should be noted that CO
2
emissions are also unaffected by the sulphur content of the fuel burned. For further
information on specific emissions factors and references, see Annex 6.
NO
x
baseline
The NO
x
emissions factors for main and auxiliary engines rated at slow, medium and high speeds were
assigned according to the IMO NO
x
emission Tiers I and II standards as defined in MARPOL Annex VI,
regulation 13. Emissions for Tier 0 engines (constructed before 2000) were modelled in accordance with
Starcrest (2013). The SFOC corresponding to the energy-based emissions factors was used to convert to
fuel-based emissions factors. NO
x
EF
baseline
for boilers (denoted by STM in Table 41) remains the same, as
there are no IMO emissions standards that apply to boiler emissions. The emissions factors used in the study
are presented in Table 41.
Table 41 – NO
x
baseline emissions factors
IMO Tier Eng speed/type Fuel type SFOC ME/Aux ME EF
baseline
(kg/tonne fuel)
Aux eng EF
baseline
(kg/tonne fuel)
Reference
0 SSD
MSD
HSD
HFO
HFO
HFO
195/na
215/227
na/227
92.82
65.12
na
na
64.76
51.10
ENTEC, 2002
ENTEC, 2002
ENTEC, 2002
1 SSD
MSD
HSD
HFO
HFO
HFO
195/na
215/227
na/227
87.18
60.47
na
na
57. 27
45.81
IMO Tier I
IMO Tier I
IMO Tier I
2 SSD
MSD
HSD
HFO
HFO
MDO
195/na
215/227
na/227
78.46
52.09
na
na
49.34
36.12
IMO Tier II
IMO Tier II
IMO Tier II
all Otto LNG 166 7. 8 3 7. 8 3 Kristensen, 2012
na GT HFO 305 20.00 na IVL, 2004
na STM HFO 305 6.89 na IVL, 2004
Notes: GT – gas turbine; STM – steam boiler
Fuel consumption efficency improvements associated with Tier I and II engines is taken into account and
further explained in the SFOC variability with load section below.
It should be noted that NO
x
emissions are not affected by fuel sulphur content but do change slightly between
HFO and distillate fuels. For further information on specific emissions factors, FCFs and references, see Annex6.
SO
x
baseline
For all three ship emissions sources, SO
x
emissions are directly linked to the sulphur content of the fuel
consumed. For emission estimating purposes, the typical fuel types (based on ISO 8217 definitions) include:
• heavy fuel oil (HFO)/intermediate fuel oil (IFO);
• marine diesel oil (MDO)/marine gas oil (MGO);
• liquefied natural gas (LNG).
The SO
x
EF
baseline
factors are based on the percentage sulphur content of the fuel, with 97.54% of the fuel
sulphur fraction converted to SO
x
(IVL 2004), while the remaining fraction is emitted as a PM sulphate
component. Therefore, SO
x
and PM emissions are directly tied to the sulphur content of the fuel consumed.
Inventories of emissions of GHGs and other relevant substances 105
CO
2
baseline
The carbon content of each fuel type is constant and is not affected by engine type, duty cycle or other
parameters when looking on the basis of kg CO
2
per tonne fuel. The fuel-based CO
2
emissions factors for
main and auxiliary engines at slow, medium and high speeds are based on MEPC 63/23, annex 8, and include:
HFO EF
baseline
CO
2
= 3,114 kg CO
2
/tonne fuel
MDO/MGO EF
baseline
CO
2
= 3,206 kg CO
2
/tonne fuel
LNG EF
baseline
CO
2
= 2,750 kg CO
2
/tonne fuel
It should be noted that CO
2
emissions are also unaffected by the sulphur content of the fuel burned. For further
information on specific emissions factors and references, see Annex 6.
NO
x
baseline
The NO
x
emissions factors for main and auxiliary engines rated at slow, medium and high speeds were
assigned according to the IMO NO
x
emission Tiers I and II standards as defined in MARPOL Annex VI,
regulation 13. Emissions for Tier 0 engines (constructed before 2000) were modelled in accordance with
Starcrest (2013). The SFOC corresponding to the energy-based emissions factors was used to convert to
fuel-based emissions factors. NO
x
EF
baseline
for boilers (denoted by STM in Table 41) remains the same, as
there are no IMO emissions standards that apply to boiler emissions. The emissions factors used in the study
are presented in Table 41.
Table 41 – NO
x
baseline emissions factors
IMO Tier Eng speed/type Fuel type SFOC ME/Aux ME EF
baseline
(kg/tonne fuel)
Aux eng EF
baseline
(kg/tonne fuel)
Reference
0 SSD
MSD
HSD
HFO
HFO
HFO
195/na
215/227
na/227
92.82
65.12
na
na
64.76
51.10
ENTEC, 2002
ENTEC, 2002
ENTEC, 2002
1 SSD
MSD
HSD
HFO
HFO
HFO
195/na
215/227
na/227
87.18
60.47
na
na
57. 27
45.81
IMO Tier I
IMO Tier I
IMO Tier I
2 SSD
MSD
HSD
HFO
HFO
MDO
195/na
215/227
na/227
78.46
52.09
na
na
49.34
36.12
IMO Tier II
IMO Tier II
IMO Tier II
all Otto LNG 166 7. 8 3 7. 8 3 Kristensen, 2012
na GT HFO 305 20.00 na IVL, 2004
na STM HFO 305 6.89 na IVL, 2004
Notes: GT – gas turbine; STM – steam boiler
Fuel consumption efficency improvements associated with Tier I and II engines is taken into account and
further explained in the SFOC variability with load section below.
It should be noted that NO
x
emissions are not affected by fuel sulphur content but do change slightly between
HFO and distillate fuels. For further information on specific emissions factors, FCFs and references, see Annex6.
SO
x
baseline
For all three ship emissions sources, SO
x
emissions are directly linked to the sulphur content of the fuel
consumed. For emission estimating purposes, the typical fuel types (based on ISO 8217 definitions) include:
• heavy fuel oil (HFO)/intermediate fuel oil (IFO);
• marine diesel oil (MDO)/marine gas oil (MGO);
• liquefied natural gas (LNG).
The SO
x
EF
baseline
factors are based on the percentage sulphur content of the fuel, with 97.54% of the fuel
sulphur fraction converted to SO
x
(IVL 2004), while the remaining fraction is emitted as a PM sulphate
component. Therefore, SO
x
and PM emissions are directly tied to the sulphur content of the fuel consumed.
106 Third IMO GHG Study 2014
This study used the following SO
x
EF
baseline
factors, based on HFO with 2.7% sulphur content. The EF
baseline
factors for SO
x
are presented in Table 42. It should be noted that SO
x
and SO
2
are basically interchangeable
for marine-related engine emissions.
Table 42 – SO
x
baseline emissions factors
Eng speed/type Fuel
1
type ME EF
baseline
(kg/tonne fuel) Aux eng EF
baseline
(kg/tonne fuel) Reference
SSD
MSD
HSD
HFO
HFO
HFO
52.77
52.79
na
na
52.78
52.78
Mass balance
2
Mass balance
2
Mass balance
2
Otto LNG 0.02 0.02 Kunz & Gorse, 2013
GT HFO 52.79 na Mass balance
2
STM HFO 52.79 na Mass balance
2
Notes:
1
assumes HFO fuel with 2.7% sulphur content
2
assumes 97.54% of sulphur fraction is converted to SO
x
; remainder is converted to PM SO
4
These baseline emissions factors are adjusted using FCF to account for the changing annual fuel sulphur
content world averages (2007–2012) or as required regionally within an ECA. The global sulphur content
of marine fuel oils was modelled according to IMO global sulphur fuel oil monitoring reports, as presented
in Table 43. For regional variations driven by regulation (ECAs), the fuel sulphur content is assumed to be
equivalent to the minimum regulatory requirement (see the description in Section 1.2 on how the shipping
activity is attributed to different global regions). Further regional variations of fuel sulphur content were not
taken into account owing to the complexity associated with points of purchase of fuel and where and when it
is actually burned. It is assumed that the world average is representative across the world fleet for each year.
Table 43 – Annual fuel oil sulphur worldwide averages
Fuel type 2007 2008 2009 2010 2011 2012
HFO/IFO 2.42 2.37 2.6 2.61 2.65 2.51
MDO/MGO 0.15 0.15 0.15 0.15 0.14 0.14
For further information on specific emissions factors, FCFs and references, see Annex 6.
PM baseline
The current literature contains a rather large variation of PM emissions factors, which vary significantly
between studies because of differences in methodology, sampling and analysis techniques. The United States
Environmental Protection Agency (EPA) and the California Air Resources Board (CARB) evaluated the available
PM test data and determined that along with direct PM there is secondary PM associated with the sulphur
in fuel (2.46% fuel sulphur fraction is converted to secondary PM while the remainder is emitted as SO
x
,
as discussed previously). This study used the following PM EF
baseline
factors based on 2.7% sulphur content
HFO. The EF
baseline
factors for PM are presented in Table 44. It should be noted there is virtually no difference
between total PM and PM less than 10 microns or PM
10
for diesel-based fuels.
Table 44 – PM baseline emissions factors
Eng speed/type Fuel
1
type ME EF
baseline
(kg/tonne fuel) Aux eng EF
baseline
(kg/tonne fuel) Reference
SSD
MSD
HSD
HFO
HFO
HFO
7. 28
6.65
na
na
6.34
6.34
EPA, 2007
EPA, 2007
EPA, 2007
Otto LNG 0.18 0.18 Kristensen, 2012
GT HFO 0.20 na IVL, 2004
STM HFO 3.05 na IVL, 2004
Notes:
1
assumes HFO fuel with 2.7% sulphur content
The approach taken in this study is compatible with the Second IMO GHG Study 2009, which defined PM as
substances including sulphate, water associated with sulphate ash and organic carbons, measured by dilution
method. Therefore, the model can accommodate changes in fuel sulphur content. This reflects the changes in
PM emissions factors arising from ECAs as defined in MARPOL Annex VI.
For further information on specific emissions factors, FCFs and references, see Annex 6.
CO baseline
Emissions of carbon monoxide (CO) were determined by methods originally described in Sarvi et al. (2008),
Kristensen (2012) and IVL (2004). From these sources, the CO EF
baseline
factors presented in Table 45 were
used.
Table 45 – CO baseline emissions factors
Eng speed/type Fuel type ME EF
baseline
(kg/tonne fuel) Aux eng EF
baseline
(kg/tonne fuel) Reference
SSD
MSD
HSD
HFO
HFO
HFO
2.77
2.51
na
na
2.38
2.38
EPA, 2007
EPA, 2007
EPA, 2007
Otto LNG 7. 8 3 7. 8 3 Kristensen, 2012
GT HFO 0.33 na IVL, 2004
STM HFO 0.66 na IVL, 2004
It should be noted that CO emissions are also unaffected by the sulphur content of the fuel burned and are
the same for HFO and distillates. For further information on specific emissions factors and references, see
Annex6.
CH
4
baseline
Emissions of methane (CH
4
) were determined by analysis of test results reported in IVL (2004) and MARINTEK
(2010). Methane emissions factors for diesel-fuelled engines, steam boilers and gas turbines are taken from IVL
(2004), which states that CH
4
emissions are approximately 2% magnitude of VOC. Therefore, the EF
baseline
is
derived by multiplying the NMVOC EF
baseline
by 2%. The emissions factor for LNG Otto-cycle engines is 8.5g/
kWh, which is on a par with the data for LNG engines (MARINTEK, 2010 and 2014). However, this value may
be slightly low for older gas-fuelled engines, especially if run on low engine loads, and slightly high for the
latest generation of LNG engines (Wärtsilä, 2011). This emissions factor was used in the bottom-up approach
to determine the amount of methane released to the atmosphere from each of the vessels powered by LNG.
The majority of LNG-powered engines operating during the 2007–2012 time frame are assumed to be Otto-
cycle; all LNG engines have been modelled as low-pressure, spark injection Otto-cycle engines, which have
low NO
x
emissions. From these sources, the CH
4
EF
baseline
factors presented in Table 46 were used.
Table 46 – CH
4
baseline emissions factors
Eng speed/type Fuel type ME EF
baseline
(kg/tonne fuel) Aux eng EF
baseline
(kg/tonne fuel) Reference
SSD
MSD
HSD
HFO
HFO
HFO
0.06
0.05
na
na
0.04
0.04
IVL, 2004
IVL, 2004
IVL, 2004
Otto LNG 51.2 51.2 MARINTEK, 2010
GT HFO 0.01 na IVL, 2004
STM HFO 0.01 na IVL, 2004
It should be noted that CH
4
emissions are also unaffected by the sulphur content of the fuel burned and are
the same for HFO and distillates. For further information on specific emissions factors and references, see
Annex 6.
Inventories of emissions of GHGs and other relevant substances 107
The approach taken in this study is compatible with the Second IMO GHG Study 2009, which defined PM as
substances including sulphate, water associated with sulphate ash and organic carbons, measured by dilution
method. Therefore, the model can accommodate changes in fuel sulphur content. This reflects the changes in
PM emissions factors arising from ECAs as defined in MARPOL Annex VI.
For further information on specific emissions factors, FCFs and references, see Annex 6.
CO baseline
Emissions of carbon monoxide (CO) were determined by methods originally described in Sarvi et al. (2008),
Kristensen (2012) and IVL (2004). From these sources, the CO EF
baseline
factors presented in Table 45 were
used.
Table 45 – CO baseline emissions factors
Eng speed/type Fuel type ME EF
baseline
(kg/tonne fuel) Aux eng EF
baseline
(kg/tonne fuel) Reference
SSD
MSD
HSD
HFO
HFO
HFO
2.77
2.51
na
na
2.38
2.38
EPA, 2007
EPA, 2007
EPA, 2007
Otto LNG 7. 8 3 7. 8 3 Kristensen, 2012
GT HFO 0.33 na IVL, 2004
STM HFO 0.66 na IVL, 2004
It should be noted that CO emissions are also unaffected by the sulphur content of the fuel burned and are
the same for HFO and distillates. For further information on specific emissions factors and references, see
Annex6.
CH
4
baseline
Emissions of methane (CH
4
) were determined by analysis of test results reported in IVL (2004) and MARINTEK
(2010). Methane emissions factors for diesel-fuelled engines, steam boilers and gas turbines are taken from IVL
(2004), which states that CH
4
emissions are approximately 2% magnitude of VOC. Therefore, the EF
baseline
is
derived by multiplying the NMVOC EF
baseline
by 2%. The emissions factor for LNG Otto-cycle engines is 8.5g/
kWh, which is on a par with the data for LNG engines (MARINTEK, 2010 and 2014). However, this value may
be slightly low for older gas-fuelled engines, especially if run on low engine loads, and slightly high for the
latest generation of LNG engines (Wärtsilä, 2011). This emissions factor was used in the bottom-up approach
to determine the amount of methane released to the atmosphere from each of the vessels powered by LNG.
The majority of LNG-powered engines operating during the 2007–2012 time frame are assumed to be Otto-
cycle; all LNG engines have been modelled as low-pressure, spark injection Otto-cycle engines, which have
low NO
x
emissions. From these sources, the CH
4
EF
baseline
factors presented in Table 46 were used.
Table 46 – CH
4
baseline emissions factors
Eng speed/type Fuel type ME EF
baseline
(kg/tonne fuel) Aux eng EF
baseline
(kg/tonne fuel) Reference
SSD
MSD
HSD
HFO
HFO
HFO
0.06
0.05
na
na
0.04
0.04
IVL, 2004
IVL, 2004
IVL, 2004
Otto LNG 51.2 51.2 MARINTEK, 2010
GT HFO 0.01 na IVL, 2004
STM HFO 0.01 na IVL, 2004
It should be noted that CH
4
emissions are also unaffected by the sulphur content of the fuel burned and are
the same for HFO and distillates. For further information on specific emissions factors and references, see
Annex 6.
108 Third IMO GHG Study 2014
N
2
O baseline
Emissions factors for N
2
O and LNG were taken from the EPA 2014 report on GHGs and Kunz & Gorse
(2013), respectively. The LNG N
2
O EF
baseline
was converted from g/mmBTU to g/kWh assuming 38% engine
efficiency, and then converted to grams of N
2
O per gram of fuel using an SFOC of 166g fuel/kWh. From these
sources, the N
2
O EF
baseline
factors presented in Table 47 were used.
Table 47 – N
2
O baseline emissions factors
Eng speed/type Fuel type ME EF
baseline
(kg/tonne fuel) Aux eng EF
baseline
(kg/tonne fuel) Reference
SSD
MSD
HSD
HFO
HFO
HFO
0.16
0.16
na
na
0.16
0.16
EPA, 2014
EPA, 2014
EPA, 2014
Otto LNG 0.11 0.11 Kunz & Gorse, 2013
GT HFO 0.16 na EPA, 2014
STM HFO 0.16 na EPA, 2014
It should be noted that, similar to NO
x
, N
2
O emissions are unaffected by fuel sulphur content but do change
slightly between HFO and distillate fuels. For further information on specific emissions factors, FCFs and
references, see Annex 6.
NMVOC baseline
Emissions factors for non-methane volatile organic compounds (NMVOC) were taken from ENTEC (2002)
study and for LNG from Kristensen (2012) report. The LNG NMVOC emissions factor was conservatively
assumed to be the same as the hydrocarbon emissions factor. From these sources, the NMVOC EF
baseline
factors presented in Table 48 were used for this study. It should be noted that NMVOCs and non-methane
hydrocarbons have the same emissions factors.
Table 48 – NMVOC baseline emissions factors
Eng speed/type Fuel type ME EF
baseline
(kg/tonne fuel) Aux eng EF
baseline
(kg/tonne fuel) Reference
SSD
MSD
HSD
HFO
HFO
HFO
3.08
2.33
na
na
1.76
1.76
ENTEC, 2002
ENTEC, 2002
ENTEC, 2002
Otto LNG 3.01 3.01 Kristensen, 2012
GT HFO 0.33 na ENTEC, 2002
STM HFO 0.33 na ENTEC, 2002
NMVOC emissions are also unaffected by the sulphur content of the fuel burned and are the same for HFO
and distillates. For further information on specific emissions factors and references, see Annex 6.
SFOC variability with load
Marine diesel engines have been optimized to work within a designated load range, in which fuel economy
and engine emissions are balanced. Optimizing for fuel economy will lead to higher NO
x
emissions and
vice versa; MARPOL Annex VI NO
x
emission Tiers thus indirectly regulate the specific fuel oil consumption
(SFOC) range of the engine. Using an MDO outside the optimum load range (usually 85%–100% MCR) will
lead to higher specific fuel oil consumption per power unit (g/kWh) unless the electronic engine control unit
can adjust the engine accordingly (valve timing, fuel injection). This is possible to achieve with modern smart
engine control units by changing the engine control programming, but for older mechanical set-ups greater
effort may be required from the engine manufacturer. For slow steaming purposes, the optimum working load
range of a diesel engine can be adjusted to be lower than the default load range.
Figure 72: Impact of engine control tuning (ECT) to specific fuel oil consumption during low load operation
of MAN 6S80ME-C8.2. Standard tuning is shown by the solid black line, part load optimization by the solid
blue line and low load tuning by the broken line (from MAN, 2012)
The changes in specific fuel oil consumption (SFOC) for a specified maximum continuous rating (SMCR) of a
large two-stroke engine are illustrated in Figure 72. It is possible to achieve a lower optimum load range for the
purpose of slow steaming, but this will make the engine less efficient in the high load range.
SFOC assumptions used in this study for marine diesel engines
Engines are classified as SSD, MSD and HSD and assigned SFOC or BSFC in accordance with the Second
IMO GHG Study 2009.
Table 49 – Specific fuel oil consumption of marine diesel engines (ll values in g/kWh)
Engine age SSD MSD HSD
before 1983 205 215 225
1984–2000 185 195 205
post 2001 175 185 195
Table 49 gives the values used in this study. Main engines are typically SSD and MSD while auxiliary engines
are typically MSD and HSD. The SFOC data for turbine machinery, boilers and auxiliary engines are listed in
Table 50.
Inventories of emissions of GHGs and other relevant substances 109
Figure 72: Impact of engine control tuning (ECT) to specific fuel oil consumption during low load operation
of MAN 6S80ME-C8.2. Standard tuning is shown by the solid black line, part load optimization by the solid
blue line and low load tuning by the broken line (from MAN, 2012)
The changes in specific fuel oil consumption (SFOC) for a specified maximum continuous rating (SMCR) of a
large two-stroke engine are illustrated in Figure 72. It is possible to achieve a lower optimum load range for the
purpose of slow steaming, but this will make the engine less efficient in the high load range.
SFOC assumptions used in this study for marine diesel engines
Engines are classified as SSD, MSD and HSD and assigned SFOC or BSFC in accordance with the Second
IMO GHG Study 2009.
Table 49 – Specific fuel oil consumption of marine diesel engines (ll values in g/kWh)
Engine age SSD MSD HSD
before 1983 205 215 225
1984–2000 185 195 205
post 2001 175 185 195
Table 49 gives the values used in this study. Main engines are typically SSD and MSD while auxiliary engines
are typically MSD and HSD. The SFOC data for turbine machinery, boilers and auxiliary engines are listed in
Table 50.
110 Third IMO GHG Study 2014
Table 50 – Specific fuel oil consumption (SFOC
baseline
) of gas turbines, boiler and auxiliary engines used in
this study as the basis to estimate dependency of SFOC as a function of load. Unit is grams of fuel used per
power unit (g/kWh) (IVL, 2004)
Engine type HFO MDO/MGO HSD
Gas turbine 305 300 225
Steam boiler 305 300 205
Auxiliary
engine
225 225 195
The values in Table 49 and Table 50 represent the lowest point in the SFOC/load curve illustrated in Figure72.
In this study each MDO engine is assumed to maintain a parabolic dependency on engine load, which has
been applied to SSD/MSD/HSD engines. This approach is described further in Jalkanen et al. (2012). The
changes of SFOC as a function of engine load are computed using the base values in Table 49 and a parabolic
representation of changes over the whole engine load range.
SFOC(load) = SFOC
base
× (0.455 × load
2
- 0.71 × load + 1.28) Eq. (3)
In equation (3), engine load range (0–1) adjusts the base value of SFOC and describes the SFOC as a function
of the engine load. This provides a mechanism that will increase SFOC on low engine loads (see Table 49)
and allow the energy-based (grams of emissions per grams of fuel) and power-based (grams of emissions per
kWh used) emissions factors to be linked. Different curves are used for SSD, MSD and HSD, depending on the
values in Table 49, but all diesel engines use identical load dependency across the whole load range (0–100%)
in this study. The default engine tuning is assumed (SFOC lowest at 80% engine load) for all diesel engines
because it was not possible to determine the low load optimizations from the IHS Fairplay data.
Figure 73: Impact of engine load on brake-specific fuel consumption of various selected SSD,
MSD and HSD engines (emissions factors by engine type)
Figure 73 illustrates the change of SFOC as a function of engine load for a large two-stroke engine (31,620 kW,
MAN 6S90MC-C8), two medium-size four-stroke engines (6,000 kW, Wärtsilä 6L46; 6,000 kW, MaK M43C)
and a small four-stroke engine (1,700 kW, CAT 3512C HD). The methodology used in this study allows SFOC
changes of approximately 28% above the optimum engine load range.
Load dependency of SFOC in the case of a gas turbine
There is only a limited amount of information available about the load dependency and fuel economy of gas
turbines. In this study, gas turbine SFOC load dependency was not modelled.
SFOC of auxiliary boilers
In this study, a constant value of 305 g/kWh SFOC was used for auxiliary boilers.
SFOC of auxiliary engines
A constant value for auxiliary engine SFOC was used (indicated in Table 50). The load/SFOC dependency was
not used for auxiliary engines, because the engine load of operational auxiliary engines is usually adjusted by
switching multiple engines on or off. The optimum working range of auxiliary engines is thus maintained by
the crew and it is not expected to have large variability, in contrast to the main engine load.
CO
2
The power-based CO
2
emissions factors for main, auxiliary and boiler engines at slow, medium and high
speeds were taken from either ENTEC (2002) or IVL (2004) and were converted to mass-based factors using
the corresponding SFOC.
NO
x
The NO
x
emissions factors for main and auxiliary engines at slow, medium and high speeds were assigned
according to the three IMO NO
x
emission Tiers defined in MARPOL Annex VI. Emissions for Tier 0 engines
(constructed before 2000) were modelled in accordance with Starcrest (2013). This approach will give an
energy-based emissions factor as a function of engine RPM. The SFOC corresponding to the energy-based
emissions factor provided a link between the energy- and fuel-based emissions factors. NO
x
EF for boilers
remains the same, as there are no emissions standards that apply to boiler emissions.
SO
x
For all three emissions sources, SO
x
emissions are directly linked to the sulphur content of the fuel consumed.
For emissions estimating purposes, the typical fuel types (based on ISO 8217 definitions) include HFO, IFO,
MDO and MGO.
The emissions factor for SO
x
was determined directly from fuel sulphur content by assuming conversion of fuel
sulphur to gaseous SO
2
according to
EF(SO
x
) = SFOC × 2 × 0.97753 × fuel_sulphur_content Eq. (4)
Equation (4) includes a constant indicating that approximately 98% of the fuel sulphur will be converted to
gaseous SO
2
and that about 2% of the sulphur can be found in particulate matter (SO
4
) (IVL, 2004). In order
to obtain the mass-based emissions factors from the power-based factors given by equation (4), division with
SFOC was made. The SFOC was obtained from the SFOC
baseline
after adjusting with the load dependency
(Eq. (3)).
The global sulphur content of marine fuel oils was modelled according to IMO global sulphur fuel oil monitoring
reports, as shown in Table 51. For regional variations driven by regulation (ECAs), the fuel sulphur content is
assumed to be equivalent to the minimum regulatory requirement (see the description in Section 1.2 of how
shipping activity is attributed to different global regions).
Table 51 – Annual fuel oil sulphur worldwide averages
Fuel type 2007 2008 2009 2010 2011 2012
HFO/IFO 2.42 2.37 2.60 2.61 2.65 2.51
MDO/MGO 0.15 0.15 0.15 0.15 0.14 0.14
PM
The current literature contains a large range of PM emissions factors, which vary significantly between studies
because of differences in methodology, sampling and analysis techniques. Again, the approach taken in
the current study is compatible with the Second IMO GHG Study 2009, which defined PM as substances
including sulphate, water associated with sulphate ash and organic carbons, measured by dilution method.
Therefore, the model can accommodate changes in fuel sulphur content. This reflects the changes in PM
emissions factors arising from ECAs as defined in MARPOL Annex VI. For main engines, PM was adjusted for
low engine loads (<20%), as described in Starcrest (2013).
Inventories of emissions of GHGs and other relevant substances 111
SFOC of auxiliary engines
A constant value for auxiliary engine SFOC was used (indicated in Table 50). The load/SFOC dependency was
not used for auxiliary engines, because the engine load of operational auxiliary engines is usually adjusted by
switching multiple engines on or off. The optimum working range of auxiliary engines is thus maintained by
the crew and it is not expected to have large variability, in contrast to the main engine load.
CO
2
The power-based CO
2
emissions factors for main, auxiliary and boiler engines at slow, medium and high
speeds were taken from either ENTEC (2002) or IVL (2004) and were converted to mass-based factors using
the corresponding SFOC.
NO
x
The NO
x
emissions factors for main and auxiliary engines at slow, medium and high speeds were assigned
according to the three IMO NO
x
emission Tiers defined in MARPOL Annex VI. Emissions for Tier 0 engines
(constructed before 2000) were modelled in accordance with Starcrest (2013). This approach will give an
energy-based emissions factor as a function of engine RPM. The SFOC corresponding to the energy-based
emissions factor provided a link between the energy- and fuel-based emissions factors. NO
x
EF for boilers
remains the same, as there are no emissions standards that apply to boiler emissions.
SO
x
For all three emissions sources, SO
x
emissions are directly linked to the sulphur content of the fuel consumed.
For emissions estimating purposes, the typical fuel types (based on ISO 8217 definitions) include HFO, IFO,
MDO and MGO.
The emissions factor for SO
x
was determined directly from fuel sulphur content by assuming conversion of fuel
sulphur to gaseous SO
2
according to
EF(SO
x
) = SFOC × 2 × 0.97753 × fuel_sulphur_content Eq. (4)
Equation (4) includes a constant indicating that approximately 98% of the fuel sulphur will be converted to
gaseous SO
2
and that about 2% of the sulphur can be found in particulate matter (SO
4
) (IVL, 2004). In order
to obtain the mass-based emissions factors from the power-based factors given by equation (4), division with
SFOC was made. The SFOC was obtained from the SFOC
baseline
after adjusting with the load dependency
(Eq. (3)).
The global sulphur content of marine fuel oils was modelled according to IMO global sulphur fuel oil monitoring
reports, as shown in Table 51. For regional variations driven by regulation (ECAs), the fuel sulphur content is
assumed to be equivalent to the minimum regulatory requirement (see the description in Section 1.2 of how
shipping activity is attributed to different global regions).
Table 51 – Annual fuel oil sulphur worldwide averages
Fuel type 2007 2008 2009 2010 2011 2012
HFO/IFO 2.42 2.37 2.60 2.61 2.65 2.51
MDO/MGO 0.15 0.15 0.15 0.15 0.14 0.14
PM
The current literature contains a large range of PM emissions factors, which vary significantly between studies
because of differences in methodology, sampling and analysis techniques. Again, the approach taken in
the current study is compatible with the Second IMO GHG Study 2009, which defined PM as substances
including sulphate, water associated with sulphate ash and organic carbons, measured by dilution method.
Therefore, the model can accommodate changes in fuel sulphur content. This reflects the changes in PM
emissions factors arising from ECAs as defined in MARPOL Annex VI. For main engines, PM was adjusted for
low engine loads (<20%), as described in Starcrest (2013).
112 Third IMO GHG Study 2014
Figure 74: Comparison of PM emissions factors reported in Second IMO GHG Study 2009 [blue diamond]
(Figure 7.7, based on data from Germanischer Lloyd) with values from Jalkanen et al. (2012) [red square]
and Starcrest (2013) [green triangle]
CO
Emissions of CO were determined by the method originally described in Sarvi et al. (2008) and included in
Jalkanen et al. (2012). The methodology describing transient engine loads and their changes were not used and
all CO emissions factors represent steady-state operation and emissions. For main engines, PM was adjusted
for low engine loads (<20%), as described in Starcrest (2013).
CH
4
The power-based CH
4
emissions factors for main, auxiliary and boiler engines at slow, medium and high
speeds were taken from ENTEC (2002) and were converted to mass-based factors using the corresponding
SFOC. The main engine CH
4
emissions factors are further adjusted at low load (<20%) using engine load
adjustment, as reported in the Port of Los Angeles Inventory of Air Emissions – 2012 (Starcrest, 2013). The
mass-based factors are further adjusted for various loads dependent on SFOC, as described in Jalkanen et al.
(2012).
N
2
O
Emissions factors for N
2
O for main, auxiliary and boiler engines were taken from the ENTEC study (2002).
For main engines the factors were adjusted for low engine loads (<20%) as described in Starcrest (2013).
As for CH
4
, conversion from power-based to fuel-based emissions factors was carried out. In addition, the
mass-based factors are adjusted for various loads dependent on SFOC as described in Jalkanen et al. (2012)
(see Table 23 and Figure 35).
NMVOC
Emissions factors for NMVOC for main, auxiliary and boiler engines were taken from the ENTEC study (2002),
and for main engines were adjusted for low engine loads (<20%) as described in Starcrest (2013). As for CH
4
,
convesion from power-based to fuel-based emissions factors was carried out. In addition, the mass-based
factors are adjusted for various loads dependent on SFOC as described in Jalkanen et al. (2012) (see Figure 72).
2.3 Other relevant substances emissions inventories for 2007–2012
This section presents summary tables of top-down and bottom-up results for other substances besides CO
2
that are emitted from ships. Section 2.4.3 presents top-down and bottom-up inventory results graphically.
This section groups these tables (52–67) as follows:
• top-down fuel consumption (repeated from earlier sections);
• top-down GHG totals, including CH
4
and N
2
O;
• top-down air pollutant inventories, including SO
x
, NO
x
, PM, CO and NMVOC;
• bottom-up fuel consumption (repeated from earlier sections);
• bottom-up GHG totals, including CH
4
and N
2
O;
• bottom-up air pollutant inventories, including SO
x
, NO
x
, PM, CO and NMVOC.
2.3.1 Top-down fuel inventories
Table 52 – Top-down fuel consumption inventory (million tonnes)
Marine sector Fuel type 2007 2008 2009 2010 2011
International marine bunkers HFO 174.1 177 165.9 178.9 17 7.9
MDO 26 22.7 24.9 28.2 29.6
LNG 0 0 0 0 0
International total 200.1 199.7 190.8 2 07.1 207. 5
Domestic navigation HFO 19.9 14.2 15.3 14.3 12.7
MDO 22.7 23.9 23.6 25.7 27. 4
LNG 0.04 0.05 0.05 0.05 0.07
Domestic total 42.64 38.15 38.95 40.05 40.17
Fishing HFO 1.1 1.1 1 0.8 0.8
MDO 5.4 4.9 5 5.2 5.1
LNG 0.04 0.02 0.04 0.02 0.05
Fishing total 6.54 6.02 6.04 6.02 5.95
Total 249.28 243.87 235.79 253.17 253.62
2.3.2 Top-down GHG inventories
Table 53 – Top-down CH
4
emissions estimates (tonnes)
Marine sector Fuel type 2007 2008 2009 2010 2011
International marine bunkers HFO 10,446 10,620 9,954 10,734 10,674
MDO 1,560 1,362 1,494 1,692 1,776
LNG 0 0 0 0 0
International total 12,006 11,982 11,448 12,426 12,450
Domestic navigation HFO 1,194 852 918 858 762
MDO 1,362 1,434 1,416 1,542 1,644
LNG 2,048 2,560 2,560 2,560 3,584
Domestic total 4,604 4,846 4,894 4,960 5,990
Fishing HFO 66 66 60 48 48
MDO 324 294 300 312 306
LNG 2,048 1,024 2,048 1,024 2,560
Fishing total 2,438 1,384 2,408 1,384 2,914
Total 19,048 18,212 18,750 18,770 21,354
Inventories of emissions of GHGs and other relevant substances 113
This section groups these tables (52–67) as follows:
• top-down fuel consumption (repeated from earlier sections);
• top-down GHG totals, including CH
4
and N
2
O;
• top-down air pollutant inventories, including SO
x
, NO
x
, PM, CO and NMVOC;
• bottom-up fuel consumption (repeated from earlier sections);
• bottom-up GHG totals, including CH
4
and N
2
O;
• bottom-up air pollutant inventories, including SO
x
, NO
x
, PM, CO and NMVOC.
2.3.1 Top-down fuel inventories
Table 52 – Top-down fuel consumption inventory (million tonnes)
Marine sector Fuel type 2007 2008 2009 2010 2011
International marine bunkers HFO 174.1 177 165.9 178.9 17 7.9
MDO 26 22.7 24.9 28.2 29.6
LNG 0 0 0 0 0
International total 200.1 199.7 190.8 2 07.1 207. 5
Domestic navigation HFO 19.9 14.2 15.3 14.3 12.7
MDO 22.7 23.9 23.6 25.7 27. 4
LNG 0.04 0.05 0.05 0.05 0.07
Domestic total 42.64 38.15 38.95 40.05 40.17
Fishing HFO 1.1 1.1 1 0.8 0.8
MDO 5.4 4.9 5 5.2 5.1
LNG 0.04 0.02 0.04 0.02 0.05
Fishing total 6.54 6.02 6.04 6.02 5.95
Total 249.28 243.87 235.79 253.17 253.62
2.3.2 Top-down GHG inventories
Table 53 – Top-down CH
4
emissions estimates (tonnes)
Marine sector Fuel type 2007 2008 2009 2010 2011
International marine bunkers HFO 10,446 10,620 9,954 10,734 10,674
MDO 1,560 1,362 1,494 1,692 1,776
LNG 0 0 0 0 0
International total 12,006 11,982 11,448 12,426 12,450
Domestic navigation HFO 1,194 852 918 858 762
MDO 1,362 1,434 1,416 1,542 1,644
LNG 2,048 2,560 2,560 2,560 3,584
Domestic total 4,604 4,846 4,894 4,960 5,990
Fishing HFO 66 66 60 48 48
MDO 324 294 300 312 306
LNG 2,048 1,024 2,048 1,024 2,560
Fishing total 2,438 1,384 2,408 1,384 2,914
Total 19,048 18,212 18,750 18,770 21,354
114 Third IMO GHG Study 2014
Table 54 – Top-down N
2
O emissions estimates (tonnes)
Marine sector Fuel type 2007 2008 2009 2010 2011
International marine bunkers HFO 27, 856 28,320 26,544 28,624 28,464
MDO 3,900 3,405 3,735 4,230 4,440
LNG 0 0 0 0 0
International total 31,756 31,725 30,279 32,854 32,904
Domestic navigation HFO 3,184 2,272 2,448 2,288 2,032
MDO 3,405 3,585 3,540 3,855 4,110
LNG 4 6 6 6 8
Domestic total 6,593 5,863 5,994 6,149 6,150
Fishing HFO 176 176 160 128 128
MDO 810 735 750 780 765
LNG 4 2 4 2 6
Fishing total 990 913 914 910 899
Total 39,340 38,501 37,187 39,913 39,952
2.3.3 Top-down air pollutant inventories
Table 55 – Top-down SO
x
emissions estimates (thousand tonnes as SO
2
)
Marine sector Fuel type 2007 2008 2009 2010 2011
International marine bunkers HFO 8,268 8,220 8,404 9,158 9,199
MDO 69 60 66 74 78
LNG 0 0 0 0 0
International total 8,337 8,280 8,470 9,232 9,277
Domestic navigation HFO 945 659 775 732 657
MDO 60 63 62 68 72
LNG 0 0 0 0 0
Domestic total 1,005 723 837 800 729
Fishing HFO 52 51 51 41 41
MDO 14 13 13 14 13
LNG 0 0 0 0 0
Fishing total 66 64 64 55 55
Total 9,408 9,066 9,371 10,087 10,061
Table 56 – Top-down NO
x
emissions estimates (thousand tonnes as NO
2
)
Marine sector Fuel type 2007 2008 2009 2010 2011
International marine bunkers HFO 16,191 16,461 15,429 16,638 16,545
MDO 2,269 1,981 2,173 2,460 2,583
LNG 0 0 0 0 0
International total 18,460 18,442 17,6 01 19,098 19,127
Domestic navigation HFO 1,851 1,321 1,423 1,330 1,181
MDO 1,981 2,085 2,059 2,242 2,391
LNG 0 0 0 0 1
Domestic total 3,832 3,406 3,482 3,573 3,572
Fishing HFO 102 102 93 74 74
MDO 471 428 436 454 445
LNG 0 0 0 0 0
Fishing total 574 530 530 528 520
Total 22,865 22,378 21,613 23,199 23,219
Table 57 – Top-down PM emissions estimates (thousand tonnes)
Marine sector Fuel type 2007 2008 2009 2010 2011
International marine bunkers HFO 1,191 1,198 1,183 1,276 1,283
MDO 27 23 25 29 30
LNG 0 0 0 0 0
International total 1,217 1,221 1,208 1,304 1,313
Domestic navigation HFO 136 96 109 102 92
MDO 23 24 24 26 28
LNG 0 0 0 0 0
Domestic total 159 121 133 128 120
Fishing HFO 8 7 7 6 6
MDO 6 5 5 5 5
LNG 0 0 0 0 0
Fishing total 13 12 12 11 11
Total 1,390 1,354 1,354 1,444 1,443
Table 58 – Top-down CO emissions estimates (thousand tonnes)
Marine sector Fuel type 2007 2008 2009 2010 2011
International marine bunkers HFO 482.3 490.3 459.5 495.6 492.8
MDO 72.0 62.9 69.0 78.1 82.0
LNG 0.0 0.0 0.0 0.0 0.0
International total 554.3 553.2 528.5 573.7 574.8
Domestic navigation HFO 55.1 39.3 42.4 39.6 35.2
MDO 62.9 66.2 65.4 71.2 75.9
LNG 0.3 0.4 0.4 0.4 0.5
Domestic total 118.3 105.9 108.1 111.2 111.6
Fishing HFO 3.0 3.0 2.8 2.2 2.2
MDO 15.0 13.6 13.9 14.4 14.1
LNG 0.3 0.2 0.3 0.2 0.4
Fishing total 18.3 16.8 16.9 16.8 16.7
Total 690.9 675.9 653.6 701.6 703.1
Inventories of emissions of GHGs and other relevant substances 115
Table 56 – Top-down NO
x
emissions estimates (thousand tonnes as NO
2
)
Marine sector Fuel type 2007 2008 2009 2010 2011
International marine bunkers HFO 16,191 16,461 15,429 16,638 16,545
MDO 2,269 1,981 2,173 2,460 2,583
LNG 0 0 0 0 0
International total 18,460 18,442 17,6 01 19,098 19,127
Domestic navigation HFO 1,851 1,321 1,423 1,330 1,181
MDO 1,981 2,085 2,059 2,242 2,391
LNG 0 0 0 0 1
Domestic total 3,832 3,406 3,482 3,573 3,572
Fishing HFO 102 102 93 74 74
MDO 471 428 436 454 445
LNG 0 0 0 0 0
Fishing total 574 530 530 528 520
Total 22,865 22,378 21,613 23,199 23,219
Table 57 – Top-down PM emissions estimates (thousand tonnes)
Marine sector Fuel type 2007 2008 2009 2010 2011
International marine bunkers HFO 1,191 1,198 1,183 1,276 1,283
MDO 27 23 25 29 30
LNG 0 0 0 0 0
International total 1,217 1,221 1,208 1,304 1,313
Domestic navigation HFO 136 96 109 102 92
MDO 23 24 24 26 28
LNG 0 0 0 0 0
Domestic total 159 121 133 128 120
Fishing HFO 8 7 7 6 6
MDO 6 5 5 5 5
LNG 0 0 0 0 0
Fishing total 13 12 12 11 11
Total 1,390 1,354 1,354 1,444 1,443
Table 58 – Top-down CO emissions estimates (thousand tonnes)
Marine sector Fuel type 2007 2008 2009 2010 2011
International marine bunkers HFO 482.3 490.3 459.5 495.6 492.8
MDO 72.0 62.9 69.0 78.1 82.0
LNG 0.0 0.0 0.0 0.0 0.0
International total 554.3 553.2 528.5 573.7 574.8
Domestic navigation HFO 55.1 39.3 42.4 39.6 35.2
MDO 62.9 66.2 65.4 71.2 75.9
LNG 0.3 0.4 0.4 0.4 0.5
Domestic total 118.3 105.9 108.1 111.2 111.6
Fishing HFO 3.0 3.0 2.8 2.2 2.2
MDO 15.0 13.6 13.9 14.4 14.1
LNG 0.3 0.2 0.3 0.2 0.4
Fishing total 18.3 16.8 16.9 16.8 16.7
Total 690.9 675.9 653.6 701.6 703.1
116 Third IMO GHG Study 2014
Table 59 – Top-down NMVOC emissions estimates (thousand tonnes)
Marine sector Fuel type 2007 2008 2009 2010 2011
International marine bunkers HFO 536.2 545.2 511.0 551.0 5 47.9
MDO 80.1 69.9 76.7 86.9 91.2
LNG 0.0 0.0 0.0 0.0 0.0
International total 616.3 615.1 5 87.7 637.9 639.1
Domestic navigation HFO 61.3 43.7 47.1 44.0 39.1
MDO 69.9 73.6 72.7 79.2 84.4
LNG 0.1 0.2 0.2 0.2 0.2
Domestic total 131.3 117. 5 120.0 123.4 123.7
Fishing HFO 3.4 3.4 3.1 2.5 2.5
MDO 16.6 15.1 15.4 16.0 15.7
LNG 0.1 0.1 0.1 0.1 0.2
Fishing total 20.1 18.5 18.6 18.5 18.3
Total 767.8 751.1 726.2 779.8 781.1
2.3.4 Bottom-up fuel inventories
Table 60 – Bottom-up fuel consumption estimates (million tonnes)
Fleet sector 2007 2008 2009 2010 2011 2012
International shipping 283 294 275 248 274 257
Domestic navigation 42 43 24 26 35 27
Fishing 27 25 14 18 18 16
Total bottom-up estimate 352 363 313 293 327 300
2.3.5 Bottom-up GHG inventories
Table 61 – Bottom-up CH
4
emissions estimates (tonnes)
Fleet sector 2007 2008 2009 2010 2011 2012
International shipping 174,370 193,180 185,980 234,370 286,480 286,520
Domestic navigation 1,510 1,570 770 1,020 1,180 1,060
Fishing 1,110 1,040 570 780 780 700
Total bottom-up estimate 176,990 195,790 187, 320 236,170 288,440 288,280
Table 62 – Bottom-up N
2
O emissions estimates (tonnes)
Fleet sector 2007 2008 2009 2010 2011 2012
International shipping 40,780 42,580 39,800 35,620 38,380 36,680
Domestic navigation 5,220 5,380 2,790 3,440 3,950 3,560
Fishing 3,930 3,730 2,100 2,730 2,730 2,400
Total bottom-up estimate 49,930 51,690 44,690 41,790 45,060 42,640
2.3.6 Bottom-up air pollutant inventories
Table 63 – Bottom-up SO
x
emissions estimates (thousand tonnes as SO
2
)
Fleet sector 2007 2008 2009 2010 2011 2012
International shipping 10,771 11,041 11,16 4 9,895 10,851 9,712
Domestic navigation 278 331 202 251 358 268
Fishing 533 521 280 405 423 261
Total bottom-up estimate 11,581 11,892 11,646 10,550 11,632 10,240
Table 64 – Bottom-up NO
x
emissions estimates (thousand tonnes as NO
2
)
Fleet sector 2007 2008 2009 2010 2011 2012
International shipping 19,943 20,759 19,104 16,708 18,047 16,997
Domestic navigation 1,564 1,639 930 1,114 1,323 1,171
Fishing 1,294 1,242 722 935 940 834
Total bottom-up estimate 22,801 23,639 20,756 18,756 20,310 19,002
Table 65 – Bottom-up PM emissions estimates (thousand tonnes)
Fleet sector 2007 2008 2009 2010 2011 2012
International shipping 1,493 1,545 1,500 1,332 1,446 1,317
Domestic navigation 51 58 33 41 56 44
Fishing 78 76 41 59 61 41
Total bottom-up estimate 1,622 1,679 1,574 1,432 1,563 1,402
Table 66 – Bottom-up CO emissions estimates (thousand tonnes)
Fleet sector 2007 2008 2009 2010 2011 2012
International shipping 823 864 816 763 834 806
Domestic navigation 99 103 60 72 82 76
Fishing 76 72 46 59 58 53
Total bottom-up estimate 998 1,039 921 893 975 936
Table 67 – Bottom-up NMVOC emissions estimates (thousand tonnes)
Fleet sector 2007 2008 2009 2010 2011 2012
International shipping 696 727 672 593 643 609
Domestic navigation 76 78 38 51 59 53
Fishing 55 52 28 39 39 35
Total bottom-up estimate 827 858 739 683 741 696
While these global totals differ from primary GHGs in terms of regional distribution, typical substance lifetimes
and air quality impacts, NO
x
and SO
x
play indirect roles in tropospheric ozone formation and indirect aerosol
warming at regional scales; moreover, ship emissions of NO
x
and SO
x
have been compared with global
anthropogenic emissions.
These totals are slightly greater than reported in the Second IMO GHG Study 2009. The Third IMO GHG Study
2014 estimates multi-year (2007–2012) average annual totals of 11.3 million tonnes and 20.9 million tonnes for
SO
x
(as SO
2
) and NO
x
(as NO
2
) from all shipping respectively (corresponding to 5.6 million tonnes and 6.3
million tonnes converted to elemental weights for nitrogen and sulphur respectively). A multi-year average
of international shipping results in an annual average estimate of some 10.6 million tonnes and 18.6million
tonnes of SO
x
(as SO
2
) and NO
x
(as NO
2
) respectively; this converts to totals of 5.3 million tonnes and
5.6million tonnes of SO
x
and NO
x
respectively (as elemental sulphur and nitrogen respectively). These totals
can be compared with totals reported in AR5 (IPCC, 2013). Global NO
x
and SO
x
emissions from all shipping
represent about 15% and 13% of global NO
x
and SO
x
from anthropogenic sources respectively; international
Inventories of emissions of GHGs and other relevant substances 117
2.3.6 Bottom-up air pollutant inventories
Table 63 – Bottom-up SO
x
emissions estimates (thousand tonnes as SO
2
)
Fleet sector 2007 2008 2009 2010 2011 2012
International shipping 10,771 11,041 11,16 4 9,895 10,851 9,712
Domestic navigation 278 331 202 251 358 268
Fishing 533 521 280 405 423 261
Total bottom-up estimate 11,581 11,892 11,646 10,550 11,632 10,240
Table 64 – Bottom-up NO
x
emissions estimates (thousand tonnes as NO
2
)
Fleet sector 2007 2008 2009 2010 2011 2012
International shipping 19,943 20,759 19,104 16,708 18,047 16,997
Domestic navigation 1,564 1,639 930 1,114 1,323 1,171
Fishing 1,294 1,242 722 935 940 834
Total bottom-up estimate 22,801 23,639 20,756 18,756 20,310 19,002
Table 65 – Bottom-up PM emissions estimates (thousand tonnes)
Fleet sector 2007 2008 2009 2010 2011 2012
International shipping 1,493 1,545 1,500 1,332 1,446 1,317
Domestic navigation 51 58 33 41 56 44
Fishing 78 76 41 59 61 41
Total bottom-up estimate 1,622 1,679 1,574 1,432 1,563 1,402
Table 66 – Bottom-up CO emissions estimates (thousand tonnes)
Fleet sector 2007 2008 2009 2010 2011 2012
International shipping 823 864 816 763 834 806
Domestic navigation 99 103 60 72 82 76
Fishing 76 72 46 59 58 53
Total bottom-up estimate 998 1,039 921 893 975 936
Table 67 – Bottom-up NMVOC emissions estimates (thousand tonnes)
Fleet sector 2007 2008 2009 2010 2011 2012
International shipping 696 727 672 593 643 609
Domestic navigation 76 78 38 51 59 53
Fishing 55 52 28 39 39 35
Total bottom-up estimate 827 858 739 683 741 696
While these global totals differ from primary GHGs in terms of regional distribution, typical substance lifetimes
and air quality impacts, NO
x
and SO
x
play indirect roles in tropospheric ozone formation and indirect aerosol
warming at regional scales; moreover, ship emissions of NO
x
and SO
x
have been compared with global
anthropogenic emissions.
These totals are slightly greater than reported in the Second IMO GHG Study 2009. The Third IMO GHG Study
2014 estimates multi-year (2007–2012) average annual totals of 11.3 million tonnes and 20.9 million tonnes for
SO
x
(as SO
2
) and NO
x
(as NO
2
) from all shipping respectively (corresponding to 5.6 million tonnes and 6.3
million tonnes converted to elemental weights for nitrogen and sulphur respectively). A multi-year average
of international shipping results in an annual average estimate of some 10.6 million tonnes and 18.6million
tonnes of SO
x
(as SO
2
) and NO
x
(as NO
2
) respectively; this converts to totals of 5.3 million tonnes and
5.6million tonnes of SO
x
and NO
x
respectively (as elemental sulphur and nitrogen respectively). These totals
can be compared with totals reported in AR5 (IPCC, 2013). Global NO
x
and SO
x
emissions from all shipping
represent about 15% and 13% of global NO
x
and SO
x
from anthropogenic sources respectively; international
118 Third IMO GHG Study 2014
shipping NO
x
and SO
x
represent approximately 13% and 12% of global NO
x
and SO
x
totals respectively.
Comparisons with AR5 are also consistent with comparisons in peer-reviewed journal publications reporting
global SO
x
(Smith et al., 2011) and NO
x
(Miyazaki et al., 2012).
Multi-year averages for PM, CO and NMVOC are calculable but are rarely compared with global totals.
Moreover, AR5 only reports global values for CO and NMVOC, and IPCC reports substances of particulate
matter such as black carbon and organic carbon. Interested readers are referred to annex II of AR5 (IPCC,
2013) for tables with global totals for CO (AR5, Table All.2.16), NMVOC (AR5, Table All.2.17), organic carbon
(AR5, Table All.2.21) and black carbon (AR5, Table All.2.22).
2.4 Quality assurance and quality control of other relevant substances
emissions inventories
Because the input data and method for Section 2.1 and Section 2.2 have substantial similarity to the input
data and method for Sections 1.1 and Section 1.2, Section 2.4 is closely connected to Section 1.4. The two
areas where there is specific additional content are in the QA/QC of the emissions factors used and in the
comparison of emissions inventories obtained using the two approaches (bottom-up and top-down).
2.4.1 QA/QC of bottom-up emissions factors
As stated in Section 2.2.7, the emissions factors used in the Third IMO GHG Study 2014 were selected by
the consortium with first preference going to published IMO factors (e.g. NO
x
by fuel type). Other factors
were selected with the unanimous agreement of the emissions factor working group based on what various
members are currently using in their work. It should also be noted that emissions factors are typically derived
from emissions testing results and reported as energy-based (g pollutant/kWh) factors. Both the Second IMO
GHG Study 2009 and this study used fuel-based (g pollutant/g fuel) factors. The following observations can
be made about the comparison of the two sets of emissions factors:
• The Second IMO GHG Study 2009 emissions factors (presented in its Table 3.6) do not differentiate
for various engine types (SSD, MSD, HSD, auxiliary boilers, LNG Otto, steam, gas turbine), engine
tier (0, I, II) or duty cycle (propulsion, auxiliary). Exceptions to these are fuel type differentiation (CO
2
,
SO
2
, NO
x
, PM
10
) and auxiliary boilers (NO
x
). The Third IMO GHG Study 2014 includes each of these
differentiations and further adjusts the emissions factors based on engine load.
Since the emissions factors are significantly more detailed in the Third IMO GHG Study 2014, comparisons
are somewhat difficult; however, they are compared in Table 68.
Table 68 – Comparison of emissions factors, Second IMO GHG Study 2009
and Third IMO GHG Study 2014
Pollutant IMO Study Engine type Tier Fuel type EF
1
Correlation
2014/2009
EFs
Correlation
CO
2
2009 unk unk HFO 3,130
0.99 good 2014 all all HFO 3,114
2009 unk unk MDO 3,190
1.01 good 2014 all all MDO 3,206
NO
x
2009 SSD 0 ? 90
1.03 good 2014 SSD 0 HFO 92.82
2009 SSD 1 ? 78
1.12 good 2014 SSD 1 HFO 87.18
2009 MSD 0 ? 60
1.09 good 2014 MSD 0 HFO 65.12
2009 MSD 1 ? 51
1.19 moderate difference2014 MSD 1 HFO 60.47
2009 Boiler na ? 7
0.98 good 2014 Boiler na HFO 6.89
SO
x
2009 unk unk HFO 2.7% 54
0.98 good 2014 SSD 0 HFO 2.7% 52.77
2014 SSD 0 HFO 2.42% 47.49 0.88 as modelled for 2007
2009 unk unk MDO 0.5% 10
0.98 good 2014 SSD 0 MDO 0.5% 9.76
2014 SSD 0 MDO 0.15% 2.64 0.26 as modelled for 2007
PM 2009 unk unk HFO 2.7% 6.7
1.09 good 2014 SSD 0 HFO 2.7% 7. 28
2014 SSD 0 HFO 2.42% 6.84 1.02 as modelled for 2007
2009 unk unk MDO 0.5% 1.1
1.65 significant difference2014 SSD 0 MDO 0.5% 1.82
2014 SSD 0 MDO 0.1% 1.24 1.13 as modelled for 2007
CO 2009 unk unk unk 7. 4
0.37 significant difference2014 SSD 0 HFO 2.77
CH
4
2009 unk unk unk 0.3
0.20 significant difference2014 SSD 0 HFO 0.06
N
2
O 2009 unk unk unk 0.08
2.00 significant difference2014 SSD 0 HFO 0.16
NMVOC 2009 unk unk unk 2.4
1.28 significant difference2014 SSD 0 HFO 3.08
Notes:
1
kg pollutant/tonne of fuel; unk = unknown; moderate difference 10%–25%; significant difference >25%
In Table 68, some pollutant emissions factors do not correlate well (values in red) between the two studies
and are discussed further below:
• NO
x
– The Third IMO GHG Study 2014 MSD Tier I emissions factor is 19% higher than the Second
IMO GHG Study 2009, which could be because of the assumed SFOC rates.
• SO
x
– The modelled Third IMO GHG Study 2014 SSD Tier 0 HFO emissions factors are 12% lower
owing to use of the annual average IMO published fuel sulphur contents (2.42% for 2007) in the Third
IMO GHG Study 2014, compared to the 2.7% used in the Second IMO GHG Study 2009.
Inventories of emissions of GHGs and other relevant substances 119
Table 68 – Comparison of emissions factors, Second IMO GHG Study 2009
and Third IMO GHG Study 2014
Pollutant IMO Study Engine type Tier Fuel type EF
1
Correlation
2014/2009
EFs
Correlation
CO
2
2009 unk unk HFO 3,130
0.99 good 2014 all all HFO 3,114
2009 unk unk MDO 3,190
1.01 good 2014 all all MDO 3,206
NO
x
2009 SSD 0 ? 90
1.03 good 2014 SSD 0 HFO 92.82
2009 SSD 1 ? 78
1.12 good 2014 SSD 1 HFO 87.18
2009 MSD 0 ? 60
1.09 good 2014 MSD 0 HFO 65.12
2009 MSD 1 ? 51
1.19 moderate difference2014 MSD 1 HFO 60.47
2009 Boiler na ? 7
0.98 good 2014 Boiler na HFO 6.89
SO
x
2009 unk unk HFO 2.7% 54
0.98 good 2014 SSD 0 HFO 2.7% 52.77
2014 SSD 0 HFO 2.42% 47.49 0.88 as modelled for 2007
2009 unk unk MDO 0.5% 10
0.98 good 2014 SSD 0 MDO 0.5% 9.76
2014 SSD 0 MDO 0.15% 2.64 0.26 as modelled for 2007
PM 2009 unk unk HFO 2.7% 6.7
1.09 good 2014 SSD 0 HFO 2.7% 7. 28
2014 SSD 0 HFO 2.42% 6.84 1.02 as modelled for 2007
2009 unk unk MDO 0.5% 1.1
1.65 significant difference2014 SSD 0 MDO 0.5% 1.82
2014 SSD 0 MDO 0.1% 1.24 1.13 as modelled for 2007
CO 2009 unk unk unk 7. 4
0.37 significant difference2014 SSD 0 HFO 2.77
CH
4
2009 unk unk unk 0.3
0.20 significant difference2014 SSD 0 HFO 0.06
N
2
O 2009 unk unk unk 0.08
2.00 significant difference2014 SSD 0 HFO 0.16
NMVOC 2009 unk unk unk 2.4
1.28 significant difference2014 SSD 0 HFO 3.08
Notes:
1
kg pollutant/tonne of fuel; unk = unknown; moderate difference 10%–25%; significant difference >25%
In Table 68, some pollutant emissions factors do not correlate well (values in red) between the two studies
and are discussed further below:
• NO
x
– The Third IMO GHG Study 2014 MSD Tier I emissions factor is 19% higher than the Second
IMO GHG Study 2009, which could be because of the assumed SFOC rates.
• SO
x
– The modelled Third IMO GHG Study 2014 SSD Tier 0 HFO emissions factors are 12% lower
owing to use of the annual average IMO published fuel sulphur contents (2.42% for 2007) in the Third
IMO GHG Study 2014, compared to the 2.7% used in the Second IMO GHG Study 2009.
120 Third IMO GHG Study 2014
• SO
x
– The modelled Third IMO GHG Study 2014 SSD Tier 0 MDO emissions factors are 74% lower
owing to use of the annual average IMO published fuel sulphur contents (0.15% for 2007) in the Third
IMO GHG Study 2014 compared to the 0.5% used in the Second IMO GHG Study 2009.
• PM – the modelled Third IMO GHG Study 2014 SSD Tier 0 MDO emissions factors are 13% higher
owing to use of fuel correction factors as described in Section 2.2.7 and Annex 6 compared to the
value developed in the Second IMO GHG Study 2009.
• CO – The Third IMO GHG Study 2014 SSD Tier 0 HFO emissions factors are 63% lower than the
Second IMO GHG Study 2009. The 2009 study used CORINAIR emissions factors for CO, which can
be traced back to the Lloyd’s Register report Marine Exhaust Emissions Research Programme (1995).
The Third IMO GHG Study 2014 used an updated CO emissions factor that was recently supported
in the Kristensen (2012) report.
• CH
4
– The Third IMO GHG Study 2014 SSD Tier 0 HFO emissions factor is 80% lower than the
Second IMO GHG Study 2009. The 2009 study used the IPCC 2013 emissions factor for CH
4
, which
can be traced back to the Lloyd’s Register report Marine Exhaust Emissions Research Programme
(1995). The 2014 study used an updated CH
4
emissions factor. In addition to the CH
4
combustion
product, methane is also released into the atmosphere as an unburnt fuel from engines operating on
LNG Otto-cycle engines. In this report, the methane slip has been included in the methane emission
inventory and an additional non-combustion emissions factor has been assigned for CH
4
to account
for this feature. For further details, see Section 2.2.6.
• N
2
O – The Third IMO GHG Study 2014 SSD Tier 0 HFO emissions factor is two times higher than the
Second IMO GHG Study 2009. The 2009 study used CORINAIR emissions factors for N
2
O, which
can be traced back to the Lloyd’s Register report Marine Exhaust Emissions Research Programme
(1995). The IPCC guidelines state that the uncertainty of the emissions factor is as high as 140%. The
Third IMO GHG Study 2014 used an updated N
2
O emissions factor.
• NMVOC – The Third IMO GHG Study 2014 SSD Tier 0 HFO emissions factor is 28% higher than
the Second IMO GHG Study 2009. The 2009 study used CORINAIR emissions factors for NMVOC,
which can be traced back to Lloyd’s Register report Marine Exhaust Emissions Research Programme
(1995). The Third IMO GHG Study 2014 used an updated NMVOC emissions factor.
2.4.2 QA/QC of top-down emissions factors
The top-down emissions factors (Table 34, Section 2.1.1) are a subset of the bottom-up emissions factors and
were selected as described in Section 2.2.7. They have the same correlations to the Second IMO GHG Study
2009 as presented in Section 2.4.1.
2.4.3 Comparison of top-down and bottom-up inventories
Top-down and bottom-up time series for each pollutant inventory are presented in Figure 75 and Figure 76
respectively. These results are provided with the same units and similar (but not identical) scales for visual
comparison.
One clear difference is the trend pattern across years for some pollutants. For example, the top-down data
among all pollutants remains similar. Most top-down inventories reveal a decline after 2007, to 2009 or
so, and an increase in subsequent years. This can be explained because the top-down data do not include
technology detail and the inventories are therefore computed using a best-judgement fleet-average emissions
factor. Conversely, the bottom-up inventories can exhibit diverging patterns from one another and very
different patterns from the top-down inventory trends. Whereas CO
2
in the bottom-up results exhibits a trend
similar to SO
x
, NO
x
and PM, the pattern for CH
4
is increasing over the years. This is because the number of
larger vessels using LNG has increased, despite the fact that top-down statistics have not begun reporting any
LNG in international sales statistics.
a) CO
2
b) CH
4
c) N
2
O d) SO
X
e) NO
X
f) PM
g) CO h) NMVOC
Figure 75: Time series of top-down results for a) CO
2
, b) CH
4
, c) N
2
O, d) SO
x
, e) NO
x
, f) PM, g) CO,
and h) NMVOC, delineated by international shipping, domestic navigation and fishing
Inventories of emissions of GHGs and other relevant substances 121
a) CO
2
b) CH
4
c) N
2
O d) SO
X
e) NO
X
f) PM
g) CO h) NMVOC
Figure 75: Time series of top-down results for a) CO
2
, b) CH
4
, c) N
2
O, d) SO
x
, e) NO
x
, f) PM, g) CO,
and h) NMVOC, delineated by international shipping, domestic navigation and fishing
122 Third IMO GHG Study 2014
a) CO
2
b) CH
4
c) N
2
O d) SO
X
e) NO
X
f) PM
g) CO h) NMVOC
Figure 76: Time series of bottom-up results for a) CO
2
, b) CH
4
, c) N
2
O, d) SO
x
, e) NO
x
, f) PM, g) CO,
and h) NMVOC, delineated by international shipping, domestic navigation and fishing
2.5 Other relevant substances emissions inventory uncertainty analysis
The uncertainties involved with missing technical data for ships, incomplete geographical/temporal coverage
of activity data and resistance/powering prediction are described in Section 1. Other sources of uncertainty
include estimates of fuel consumption, allocation of fuel types consumed versus actual fuels consumed,
auxiliary engine and boiler loads by mode, assignment of modes based on AIS data, IMO sulphur survey
annual averages, and the factors used to estimate emissions. Uncertainty associated with these, with the
exception of the emissions factors, is discussed in Section 1.5. The uncertainties associated with emissions
factors include the vessels tested compared to the fleet modelled and robustness of the number of ships tested
in each subclass. While some emissions factors have remained within the same ranges since the Second
IMO GHG Study 2009, there were several pollutants that had moderate to significant changes, as detailed in
Section 2.4.1.
2.6 Other relevant substances emissions inventory comparison
against Second IMO GHG Study 2009
Figure 75 presents the time series results for non-CO
2
relevant substances estimated in this study using
bottom-up methods, with explicit comparison with the Second IMO GHG Study 2009 results. Section1.6
shows that for ship types that could be directly compared, fuel consumption and CO
2
emissions totals
estimated by the methods used in this study compare very well with methods used in the earlier study. As
reported in Section 1.6, the additional precision in observing vessel activity patterns in the Third IMO GHG
Study 2014 largely match general vessel activity assumptions in the 2009 study, at least for the inventory year
2007. (The updated methodology provides greatest value in the ability to observe year-on-year changes in
shipping patterns, which the Second IMO GHG Study 2009 methods were less able to do.)
Given that the Second IMO GHG Study 2009 “concluded that activity-based estimates provide a more correct
representation of the total emissions from shipping”, only bottom-up emissions for other relevant substances
can be compared. The Third IMO GHG Study 2014 estimates of non-CO
2
GHGs and some air pollutant
substances differ substantially from the Second IMO GHG Study 2009 results for the common year 2007. The
Third IMO GHG Study 2014 produces higher estimates of CH
4
and N
2
O than the Second IMO GHG Study
2009, higher by 43% and 40% respectively (approximate values). The Third IMO GHG Study 2014 estimates
lower emissions of SO
x
(approximately 30% lower) and approximately 40% of the CO emissions estimated in
the Second IMO GHG Study 2009. Estimates for NO
x
, PM and NMVOC in both studies are similar for 2007,
within 10%, 11% and 3% respectively (approximate values).
These underlying activity similarities essentially reduce the comparisons of other relevant substances
estimated in the Second IMO GHG Study 2009 to a description of differences in EFs, as illustrated in Table68.
Differences in EFs essentially relate to static values in the 2009 study, which assumed an average MCR and
EFs representative of the average engine’s actual duty cycle. The consortium made more detailed calculations
in the Third IMO GHG Study 2014, which computes hourly fuel consumption and engine load factors and
applies a load-factor specific EF. In theory, if the average EFs across a duty cycle in the earlier study were
computed for the same or similar activity, then the average EFs would mathematically represent the weighted
average of the hourly load-dependent calculations. The crossplots presented in Section 1.6 provide evidence
that the duty cycle assumptions in the Second IMO GHG Study 2009 were generally consistent with the more
detailed analyses presented in this study.
Another source of differences may be related to fuel quality or engine parameter data representing a different
understanding of the fleet technology. For example, the 2009 study assumed fuel sulphur content was 2.7%,
while the current study documents that the typical fuel sulphur content in 2007 was closer to 2.4%. Moreover,
the Third IMO GHG Study 2014 updates these sulphur contents for later years.
Another example is natural-gas-fuelled engines, which are observed in the Third IMO GHG Study 2014
fleet but were not addressed in the 2009 study. This enables better characterization of methane emissions
(sometimes called methane slip), which has been significantly reduced through engine innovations. The
Second IMO GHG Study 2009 characterized methane losses due to evaporation during transport of fuels
as cargo (Second IMO GHG Study 2009, non-exhaust emissions, paragraph 3.47), and used a top-down
methodology to evaluate methane emissions from engine exhaust (Second IMO GHG Study 2009, Tables 3.6
and 3.7). The 2009 study reported total methane emissions (combining exhaust and cargo transport estimates
in Table 3.11), but did not determine any value for international shipping (Second IMO GHG Study 2009,
Inventories of emissions of GHGs and other relevant substances 123
2.5 Other relevant substances emissions inventory uncertainty analysis
The uncertainties involved with missing technical data for ships, incomplete geographical/temporal coverage
of activity data and resistance/powering prediction are described in Section 1. Other sources of uncertainty
include estimates of fuel consumption, allocation of fuel types consumed versus actual fuels consumed,
auxiliary engine and boiler loads by mode, assignment of modes based on AIS data, IMO sulphur survey
annual averages, and the factors used to estimate emissions. Uncertainty associated with these, with the
exception of the emissions factors, is discussed in Section 1.5. The uncertainties associated with emissions
factors include the vessels tested compared to the fleet modelled and robustness of the number of ships tested
in each subclass. While some emissions factors have remained within the same ranges since the Second
IMO GHG Study 2009, there were several pollutants that had moderate to significant changes, as detailed in
Section 2.4.1.
2.6 Other relevant substances emissions inventory comparison
against Second IMO GHG Study 2009
Figure 75 presents the time series results for non-CO
2
relevant substances estimated in this study using
bottom-up methods, with explicit comparison with the Second IMO GHG Study 2009 results. Section1.6
shows that for ship types that could be directly compared, fuel consumption and CO
2
emissions totals
estimated by the methods used in this study compare very well with methods used in the earlier study. As
reported in Section 1.6, the additional precision in observing vessel activity patterns in the Third IMO GHG
Study 2014 largely match general vessel activity assumptions in the 2009 study, at least for the inventory year
2007. (The updated methodology provides greatest value in the ability to observe year-on-year changes in
shipping patterns, which the Second IMO GHG Study 2009 methods were less able to do.)
Given that the Second IMO GHG Study 2009 “concluded that activity-based estimates provide a more correct
representation of the total emissions from shipping”, only bottom-up emissions for other relevant substances
can be compared. The Third IMO GHG Study 2014 estimates of non-CO
2
GHGs and some air pollutant
substances differ substantially from the Second IMO GHG Study 2009 results for the common year 2007. The
Third IMO GHG Study 2014 produces higher estimates of CH
4
and N
2
O than the Second IMO GHG Study
2009, higher by 43% and 40% respectively (approximate values). The Third IMO GHG Study 2014 estimates
lower emissions of SO
x
(approximately 30% lower) and approximately 40% of the CO emissions estimated in
the Second IMO GHG Study 2009. Estimates for NO
x
, PM and NMVOC in both studies are similar for 2007,
within 10%, 11% and 3% respectively (approximate values).
These underlying activity similarities essentially reduce the comparisons of other relevant substances
estimated in the Second IMO GHG Study 2009 to a description of differences in EFs, as illustrated in Table68.
Differences in EFs essentially relate to static values in the 2009 study, which assumed an average MCR and
EFs representative of the average engine’s actual duty cycle. The consortium made more detailed calculations
in the Third IMO GHG Study 2014, which computes hourly fuel consumption and engine load factors and
applies a load-factor specific EF. In theory, if the average EFs across a duty cycle in the earlier study were
computed for the same or similar activity, then the average EFs would mathematically represent the weighted
average of the hourly load-dependent calculations. The crossplots presented in Section 1.6 provide evidence
that the duty cycle assumptions in the Second IMO GHG Study 2009 were generally consistent with the more
detailed analyses presented in this study.
Another source of differences may be related to fuel quality or engine parameter data representing a different
understanding of the fleet technology. For example, the 2009 study assumed fuel sulphur content was 2.7%,
while the current study documents that the typical fuel sulphur content in 2007 was closer to 2.4%. Moreover,
the Third IMO GHG Study 2014 updates these sulphur contents for later years.
Another example is natural-gas-fuelled engines, which are observed in the Third IMO GHG Study 2014
fleet but were not addressed in the 2009 study. This enables better characterization of methane emissions
(sometimes called methane slip), which has been significantly reduced through engine innovations. The
Second IMO GHG Study 2009 characterized methane losses due to evaporation during transport of fuels
as cargo (Second IMO GHG Study 2009, non-exhaust emissions, paragraph 3.47), and used a top-down
methodology to evaluate methane emissions from engine exhaust (Second IMO GHG Study 2009, Tables 3.6
and 3.7). The 2009 study reported total methane emissions (combining exhaust and cargo transport estimates
in Table 3.11), but did not determine any value for international shipping (Second IMO GHG Study 2009,
124 Third IMO GHG Study 2014
Table 1-1). The 2009 study allocated significant discussion in sections describing potential reductions in GHGs
to characterizing natural gas methane emissions, and identified efforts to achieve reductions in methane
emissions from marine engines. The Third IMO GHG Study 2014 explicitly applied current knowledge of
methane slip in marine engines to those vessels fuelled by natural gas in our bottom-up inventories, thereby
better characterizing CH
4
emissions.
However, more detailed characterization of fleet technology can result in different technology mixes. For
example, the Second IMO GHG Study 2009 documented auxiliary boilers for crude oil tankers only, whereas
the Third IMO GHG Study 2014 identified boiler technology on some bulk carriers, chemical tankers,
container ships, general cargo ships, cruise passenger ships, refrigerated bulk, ro-ro and vehicle carriers. The
Third IMO GHG Study 2014 assigned engine-specific EFs at the individual ship level where possible, including
differentiating between MSD and SSD engines, and residual versus distillate fuel types. These differences can
help explain inventory differences between the two studies.
For CO
2
, NO
x
and PM, the Third IMO GHG Study 2014 values for 2007 closely match the results reported in
the Second IMO GHG Study 2009. The differences in these EFs are 1% for CO
2
, 3%–9% for NO
x
and 2%–13%
for PM respectively (approximate values). (The two values for NO
x
represent SSD and MSD respectively;
similarly, the two values for PM represent HFO and MDO typical values respectively.) The match is best where
vessel activity comparisons are similar, where observed fleet technology matches and where the emissions
factors have changed little. This again confirms that the general impact of the updated methodology is greater
precision and ability to update year-on-year variation in technology or activity among individual vessels in
the fleet. Major differences in emissions results for other relevant substances, therefore, can be explained by
the different EFs used in the Second IMO GHG Study 2009 compared with the more detailed assignment of
EFs in the Third IMO GHG Study 2014. This mainly relates to the emissions of CH
4
, N
2
O, CO and NMVOC.
These EF differences are 80%, 100% and 63% lower in the current study for CH
4
, N
2
O and CO respectively,
and 30% higher for NMVOC (approximate values). These emissions represent combustion emissions of fuels
and do not include evaporative losses from the transport of cargos; the Second IMO GHG Study 2009
estimated the CH
4
losses from the transport of crude oil to be 140,000 tonnes. Table 1-1 of that study added
direct emissions from engine combustion with the estimated losses of CH
4
from the transport of crude oil; no
equivalent calculation is performed here.
Differences in sulphur (SO
x
) emissions are similarly attributed to different fuel sulphur contents, using updated
IMO sulphur reports. In this study, the bottom-up model allocation of fuel types for auxiliaries and some main
engine technologies enables more detailed delineation of heavy residual and distillate fuel use; this accounts
for most of the difference in sulphur emissions inventories between the studies. Moreover, the use of updated
fuel sulphur contents can account for about 12% difference in the heavy residual fuel sulphur contents in
20 0 7.
a) CO
2
b) CH
4
c) N
2
O d) SO
X
e) NO
X
f) PM
g) CO h) NMVOC
Figure 77: Time series of bottom-up results for a) CO
2
, b) CH
4
, c) N
2
O, d) SO
x
, e) NO
x
, f) PM, g) CO,
and h) NMVOC. The green bar represents the Second IMO GHG Study 2009 estimate for comparison
Inventories of emissions of GHGs and other relevant substances 125
a) CO
2
b) CH
4
c) N
2
O d) SO
X
e) NO
X
f) PM
g) CO h) NMVOC
Figure 77: Time series of bottom-up results for a) CO
2
, b) CH
4
, c) N
2
O, d) SO
x
, e) NO
x
, f) PM, g) CO,
and h) NMVOC. The green bar represents the Second IMO GHG Study 2009 estimate for comparison
3
Scenarios for shipping emissions
2012–2050
3.1 Introduction
This chapter presents emissions scenarios for all six GHGs (CO
2
, CH
4
, N
2
O, HFCs, PFCs, SF
6
) and for other
relevant substances as defined in this study (NO
x
, NMVOC, CO, PM, SO
x
).
Emissions scenarios present possible ways in which emissions could develop, building on plausible
socioeconomic, energy and policy scenarios. The emissions scenarios can inform policymakers, scientists
and other stakeholders about the development of the environmental impacts of shipping, its drivers and the
relevance of possible policy instruments to address emissions.
3.1.1 Similarities with and differences from Second IMO GHG Study 2009
The emissions scenarios have been developed using a similar approach to that of the Second IMO GHG
Study 2009, i.e. by modelling the most important drivers of maritime transport and efficiency trends in order
to project energy demand in the sector. For most emissions, the energy demand is then multiplied by an
emissions factor to arrive at an emissions projection. More detail about the methods and modelling can be
found in Section 3.2.
Even though the approach is similar, the methods have been improved in important ways, taking into account
advances in the literature and newly developed scenarios. Some of the most important improvements are
highlighted below.
Socioeconomic and energy scenarios
In the Second IMO GHG Study 2009, a range of transport and corresponding emissions projections to 2050
were presented. The underlying overall basis for these projections were the IPCC Special Report on Emissions
Scenarios (SRES) (based upon the IPCC 2000 SRES, which were widely in use at the time). There has been
increased recognition across the climate-scenario-modelling community that there is a need for an updated
set of scenarios, but also recognition of the need to circumvent the time and expense associated with another
IPCC-focused exercise. Thus, the relevant community itself developed the concept of RCPs. Since these are
now in use across the climate community, they have been adopted for this study (see Section 3.2.2). Outside
the climate research community, other long-term scenarios exist (e.g. IEA, 2013; OECD, 2012; IMF, 2014; RTI,
2013).
Previously, shipping emissions scenarios were based more loosely on a consortium consensus approach,
the so-called Delphi method. This study adopts a more disaggregated numerical approach with explicit
improvements to the projection methodology by splitting the projections by ship type, using a non-linear
regression model of a type widely adopted in the econometric literature (as opposed to simple linear models),
and decoupling the transport of fossil fuels from GDP. In the previous report, there was no such discrimination
by type, or consideration of future worlds where fossil fuel energy demand is decoupled from GDP. More
details are provided in Section 3.2.2 and Annex 7.
3
Scenarios for shipping emissions
2012–2050
3.1 Introduction
This chapter presents emissions scenarios for all six GHGs (CO
2
, CH
4
, N
2
O, HFCs, PFCs, SF
6
) and for other
relevant substances as defined in this study (NO
x
, NMVOC, CO, PM, SO
x
).
Emissions scenarios present possible ways in which emissions could develop, building on plausible
socioeconomic, energy and policy scenarios. The emissions scenarios can inform policymakers, scientists
and other stakeholders about the development of the environmental impacts of shipping, its drivers and the
relevance of possible policy instruments to address emissions.
3.1.1 Similarities with and differences from Second IMO GHG Study 2009
The emissions scenarios have been developed using a similar approach to that of the Second IMO GHG
Study 2009, i.e. by modelling the most important drivers of maritime transport and efficiency trends in order
to project energy demand in the sector. For most emissions, the energy demand is then multiplied by an
emissions factor to arrive at an emissions projection. More detail about the methods and modelling can be
found in Section 3.2.
Even though the approach is similar, the methods have been improved in important ways, taking into account
advances in the literature and newly developed scenarios. Some of the most important improvements are
highlighted below.
Socioeconomic and energy scenarios
In the Second IMO GHG Study 2009, a range of transport and corresponding emissions projections to 2050
were presented. The underlying overall basis for these projections were the IPCC Special Report on Emissions
Scenarios (SRES) (based upon the IPCC 2000 SRES, which were widely in use at the time). There has been
increased recognition across the climate-scenario-modelling community that there is a need for an updated
set of scenarios, but also recognition of the need to circumvent the time and expense associated with another
IPCC-focused exercise. Thus, the relevant community itself developed the concept of RCPs. Since these are
now in use across the climate community, they have been adopted for this study (see Section 3.2.2). Outside
the climate research community, other long-term scenarios exist (e.g. IEA, 2013; OECD, 2012; IMF, 2014; RTI,
2013).
Previously, shipping emissions scenarios were based more loosely on a consortium consensus approach,
the so-called Delphi method. This study adopts a more disaggregated numerical approach with explicit
improvements to the projection methodology by splitting the projections by ship type, using a non-linear
regression model of a type widely adopted in the econometric literature (as opposed to simple linear models),
and decoupling the transport of fossil fuels from GDP. In the previous report, there was no such discrimination
by type, or consideration of future worlds where fossil fuel energy demand is decoupled from GDP. More
details are provided in Section 3.2.2 and Annex 7.
128 Third IMO GHG Study 2014
Business as usual and policy scenarios
The Second IMO GHG Study 2009 presented a multitude of scenarios but did not consider any of them to be
BAU. All scenarios presented in this study are combinations of trade scenarios, ship efficiency scenarios and
emissions scenarios. The trade scenarios are based on combinations of RCPs and SSPs and, as discussed in
detail in Section 3.2.2, all four are equally likely to occur. Their differences reflect either inherent uncertainties
about the future (e.g. economic development, demographics and technological development), or uncertainties
related to policy choices outside the remit of IMO (e.g. climate, energy efficiency or trade policies). In many
cases, these uncertainties are interrelated and cannot be disentangled.
The ship efficiency and emissions scenarios can be classified in two groups. Each of the scenarios has an
option in which no policies are assumed beyond the policies that are currently in place, and one in which
IMO continues to adopt policies to address air emissions or the energy efficiency of ships. The first type is
labelled BAU, as it does not require policy interventions. In this way, each of the four trade scenarios has
one BAU variant and three policy intervention variants. As both policy interventions result in lower GHG
emissions, all policy intervention scenarios have emissions below the BAU scenario. These lower emission
scenarios require additional policies beyond those that are currently adopted.
Marginal abatement cost curves
This study employs MACCs containing 22 measures in 15 groups (measures within the same group are
mutually exclusive), taking into account the fact that measures may be applicable to certain ship types only.
The benefit of using MACCs over holistic efficiency improvement assumptions is that they allow for feedback
between fuel prices and improvements in efficiency.
MARPOL Annex VI revisions (EEDI, SEEMP)
After the publication of the Second IMO GHG Study 2009, State Parties adopted a new chapter for MARPOL
Annex VI on energy efficiency for ships, mandating EEDI for new ships and SEEMP for all ships. The impact of
these regulations on the energy efficiency of ships is analysed and included in the model.
Ship types
Since the Second IMO GHG Study 2009, there has been a remarkable increase in ship size, especially for
container ships. The earlier study assumes that all container ships over 8,000 TEU would have an average
size of 100,000 dwt, but in 2011 the size of the average new-build ship had increased to 125,000 dwt, while
ships of 165,000 dwt have entered the fleet and larger ones are being studied. Larger ships are more efficient,
i.e. they require less energy to move an amount of cargo over an amount of distance. In response, this study
analyses the development of ship types in the last year and includes new categories for the largest ships.
3.1.2 Outline
The remainder of this chapter is organized as follows. Section 3.2 provides a brief description of the methods
and data used to project emissions. It begins by presenting the emissions model, the factors taken into account
in our projections and the long-term scenarios used as a basis for our projections. All the relevant factors of the
projections are then discussed individually, showing which assumptions are made in each case and the basis
on which they are made. Section 3.3 presents the projections of international maritime transport demand and
associated emissions of CO
2
and of other relevant substances up to 2050.
3.2 Methods and data
3.2.1 The emissions projection model
The model used to project emissions starts with a projection of transport demand, building on long-term
socioeconomic scenarios developed for IPCC (see Section 3.2.2). Taking into account developments in fleet
productivity (see Section 3.2.4) and ship size (see Section 3.2.5), it projects the fleet composition in each year.
Subsequently, it projects energy demand, taking into account regulatory and autonomous improvements in
efficiency (see Section 3.2.6). Fuel consumption is calculated together with the fuel mix (see Section 3.2.7);
this, combined with emissions factors (see Section 3.2.8), yields the emissions. Emissions are presented both
in aggregate and per ship type and size category.
A schematic presentation of the emissions projection model is shown in Figure 78.
Figure 78: Schematic presentation of the emissions projection model
3.2.2 Base scenarios
Scenario construction is necessary to gain a view of what may happen in the future. In the Second IMO GHG
Study 2009, background scenarios (SRES – see Section 3.1.1) were chosen from IPCC activities, since the 2009
study was primarily about emissions and it made sense to make the emissions scenarios consistent with other
associated climate projections. Here, this study basically follows the same logic; while other visions of the
future are available, and arguably equally plausible, since the overall subject of the present study is emissions,
this study follows the earlier precedent and uses approaches and assumptions that will ultimately allow the
projections to be used in climate studies. Moreover, data from climate projections studies include the essential
socioeconomic and energy drivers that are essential for the emissions projections made here.
After its Fourth Assessment Report, published in 2007, IPCC decided to update the projections to be used in
its next assessment report (AR5). The scenarios are called representative concentration pathways (RCPs). Their
naming and use are best explained in the quote below:
“The name ‘representative concentration pathways’ was chosen to emphasize the rationale behind their
use. RCPs are referred to as pathways in order to emphasize that their primary purpose is to provide
time-dependent projections of atmospheric greenhouse gas (GHG) concentrations. In addition, the term
pathway is meant to emphasize that it is not only a specific long-term concentration or radiative forcing
outcome, such as a stabilization level, that is of interest, but also the trajectory that is taken over time to
reach that outcome. They are representative in that they are one of several different scenarios that have
similar radiative forcing and emissions characteristics.” (Towards New Scenarios for Analysis of Emissions,
Climate Change, Impacts, and Response Strategies – IPCC Expert Meeting Report, 2007).
Scenarios for shipping emissions 2012–2050 129
this, combined with emissions factors (see Section 3.2.8), yields the emissions. Emissions are presented both
in aggregate and per ship type and size category.
A schematic presentation of the emissions projection model is shown in Figure 78.
Figure 78: Schematic presentation of the emissions projection model
3.2.2 Base scenarios
Scenario construction is necessary to gain a view of what may happen in the future. In the Second IMO GHG
Study 2009, background scenarios (SRES – see Section 3.1.1) were chosen from IPCC activities, since the 2009
study was primarily about emissions and it made sense to make the emissions scenarios consistent with other
associated climate projections. Here, this study basically follows the same logic; while other visions of the
future are available, and arguably equally plausible, since the overall subject of the present study is emissions,
this study follows the earlier precedent and uses approaches and assumptions that will ultimately allow the
projections to be used in climate studies. Moreover, data from climate projections studies include the essential
socioeconomic and energy drivers that are essential for the emissions projections made here.
After its Fourth Assessment Report, published in 2007, IPCC decided to update the projections to be used in
its next assessment report (AR5). The scenarios are called representative concentration pathways (RCPs). Their
naming and use are best explained in the quote below:
“The name ‘representative concentration pathways’ was chosen to emphasize the rationale behind their
use. RCPs are referred to as pathways in order to emphasize that their primary purpose is to provide
time-dependent projections of atmospheric greenhouse gas (GHG) concentrations. In addition, the term
pathway is meant to emphasize that it is not only a specific long-term concentration or radiative forcing
outcome, such as a stabilization level, that is of interest, but also the trajectory that is taken over time to
reach that outcome. They are representative in that they are one of several different scenarios that have
similar radiative forcing and emissions characteristics.” (Towards New Scenarios for Analysis of Emissions,
Climate Change, Impacts, and Response Strategies – IPCC Expert Meeting Report, 2007).
130 Third IMO GHG Study 2014
A useful summary and guide to the origin and formulation of the RCP scenarios is provided by Wayne (2013).
The “concentration” refers to that of CO
2
and the “pathways” are “representative” of possible outcomes of
energy, population, policy and other drivers that will ultimately determine the concentration of CO
2
in the
atmosphere. There are four main RCPs in use, detailed in Table 69.
Table 69 – Descriptions and sources of representative concentration pathways
RCP Description Source references Model
RCP2.6 (or 3PD)
Peak in radiative forcing at ~3 W/m
2
before 2100
and decline
Van Vuuren et al., 2006, 2007 IMAGE
RCP4.5 Stabilization without overshoot pathway to
4.5 W/m
2
at stabilization after 2100
Clarke et al., 2007; Wise et al., 2009 GCAM
RCP6.0 Stabilization without overshoot pathway to
6 W/m
2
at stabilization after 2100
Hijoka et al., 2008 AIM
RCP8.5 Rising radiative forcing pathway leading to
8.5 W/m
2
in 2100.
Riahi et al., 2007 MESSAGE
The numbers associated with the RCPs (2.6–8.5) simply refer to resultant radiative forcing in W/m
2
by 2100.
Further technical details of the RCPs are given in Moss et al. (2010). The RCPs cover a range of ultimate
temperature projections by 2100 (i.e. global mean surface temperature increases over the pre-industrial period
from GHGs), from around 4.9 °C (RCP8.5) to 1.5 °C in the most optimistic scenario (RCP2.6 or RCP3PD, where
PD refers to peak and decline).
These RCPs are used to project shipping coal and liquid fossil fuel transport work, on the basis of a historical
correlation with global coal and oil consumption (see Section 3.2.3), using the IAM energy demand projections
of different fuel/energy types (EJ/yr). A set of GDP projections from the associated five SSP scenarios (see
Kriegler et al., 2012) was used for non-fossil-fuel transport projections (see Section 3.2.3).
The five SSPs each have different narratives (Ebi et al., 2013) and are summarized in Table 70.
Table 70 – Short narratives of shared socioeconomic pathways
SSP number and name Short narrative
SSP1: Sustainability A world making relatively good progress towards sustainability, with ongoing efforts to achieve
development goals while reducing resource intensity and fossil fuel dependency. It is an
environmentally aware world with rapid technology development and strong economic growth,
even in low-income countries.
SSP2: Middle of the road A world that sees the trends typical of recent decades continuing, with some progress towards
achieving development goals. Dependency on fossil fuels is slowly decreasing. Development of
low-income countries proceeds unevenly.
SSP3: Fragmentation A world that is separated into regions characterized by extreme poverty, pockets of moderate
wealth and a large number of countries struggling to maintain living standards for a rapidly
growing population.
SSP4: Inequality A highly unequal world in which a relatively small, rich global elite is responsible for most
GHG emissions, while a larger, poor group that is vulnerable to the impact of climate changes
contributes little to the harmful emissions. Mitigation efforts are low and adaptation is difficult
due to ineffective institutions and the low income of the large poor population.
SSP5: Conventional
development
A world in which development is oriented towards economic growth as the solution to
social and economic problems. Rapid conventional development leads to an energy system
dominated by fossil fuels, resulting in high GHG emissions and challenges to mitigation.
This presented the problem of how to combine the RCPs with the SSPs and guidance was taken from Kriegler
et al. (2012), as follows.
In principle, several SSPs can result in the same RCP, so in theory many BAU scenarios can be developed.
However, in order to limit the number of scenarios, while still showing the variety in possible outcomes, it
was decided to combine each SSP with one RCP, under the constraint that this combination is feasible. The
SSPs are thus aligned with the RCPs on the basis of their baseline warming. Increased mitigation effort would
potentially result in less fossil fuel transport, probably somewhat lower economic growth until 2050 and
therefore probably lower transport demand and maritime emissions.
This procedure has resulted in the following scenarios:
• RCP8.5 combined with SSP5;
• RCP6 combined with SSP1;
• RCP4.5 combined with SSP3;
• RCP2.6 combined with SSP4/2.
In all the work by IPCC on future scenarios of climate and its impacts, it has never assumed a BAU underlying
growth scenario. IPCC has always argued that it does not produce any one emissions scenario that is more
likely than another, ergo no overall BAU scenario exists. This is therefore reflected in this study and no one
basic RCP/SSP scenario that underlies the shipping emissions scenarios can be considered more likely than
another: they are all BAU scenarios.
3.2.3 Transport demand projections
Transport work data (in billion tonne-miles per year) were kindly provided for the years 1970–2012 by
UNCTAD (see Annex 7). The categories considered were crude oil and oil products (combined), coal bulk dry
cargo, non-coal bulk dry cargo (iron ore, coal, grain, bauxite and aluminia and phosphate, all combined) and
other dry cargo (essentially considered as container and other similar purpose shipping). The data were for
international shipping only. Transport work (i.e. tonne-miles), as opposed to the absolute amount transported
(tonnes), is considered to be a better variable to predict transport demand and emissions. However, this
assumes that average hauls remain constant: this is in fact borne out by the data and the two variables
correlate significantly with an R
2
value of >0.95.
Cargo types were treated separately, as it is evident from the data that they are growing at different rates and
subject to different market demands.
Thus, as a refinement to the approach taken in the Second IMO GHG Study 2009, the current study has
developed the methodology of CE Delft (2012), which considered different ship types and has gone a step
further by decoupling the transport of fossil fuel (oil and coal products) from GDP, as in the RCP/SSP scenarios
in which fossil fuel use is decoupled from economic development.
In order to predict ship transport work (by type, or total), the general principle is to look for a predictor variable
that has a meaningful physical relationship with it. In previous scenario studies, global GDP has been used as
a predictor for total ship transport work, in that it has a significant positive statistical correlation, and is also
meaningful in the sense that an increase in global GDP is likely to result in an increase in global trade and
therefore ship transport of goods.
If an independent assessment of the predictor variable (e.g. GDP) is available for future years, this allows
prediction of ship transport work. It assumes that such a physical relationship is as robust for the future as it has
been for the past. Previously, a linear assumption has been made, i.e. a linear regression model has been used
between the ratio of historical transport work to historical GDP against time. In this study, this assumption
has been improved by the use of a non-linear model, commonly used in economics, that assumes classic
emergence, growth and maturation phases.
However, the assumption of a historical relationship between coal and oil transport by shipping and GDP
inherently means that GDP growth and fossil fuel use will remain tightly coupled in the future, i.e. that with
increased economic growth, it is not possible to limit fossil fuel use. This clearly does not reflect certain
desired policy and environmental outcomes, where a decrease in fossil fuel dependence and an increase in
GDP can be achieved.
In order to overcome this, this study has investigated the relationship between historical ship-transported coal
and oil and historical global coal and oil consumption. This relationship has been found to be as robust as that
between historical coal and oil transport work and historical GDP (r2 >0.9) and is arguably a better physical
relationship than between fossil fuel transported by shipping and GDP. The RCP scenarios have provided
projections of fossil fuel consumption, split between coal and oil. This conveniently allows us to use these
predictor variables to determine potential future ship transport of coal and oil but decoupled from GDP. Other
ship transported goods and products remain predicted by independent future GDP assessments provided by
the RCPs.
Scenarios for shipping emissions 2012–2050 131
This procedure has resulted in the following scenarios:
• RCP8.5 combined with SSP5;
• RCP6 combined with SSP1;
• RCP4.5 combined with SSP3;
• RCP2.6 combined with SSP4/2.
In all the work by IPCC on future scenarios of climate and its impacts, it has never assumed a BAU underlying
growth scenario. IPCC has always argued that it does not produce any one emissions scenario that is more
likely than another, ergo no overall BAU scenario exists. This is therefore reflected in this study and no one
basic RCP/SSP scenario that underlies the shipping emissions scenarios can be considered more likely than
another: they are all BAU scenarios.
3.2.3 Transport demand projections
Transport work data (in billion tonne-miles per year) were kindly provided for the years 1970–2012 by
UNCTAD (see Annex 7). The categories considered were crude oil and oil products (combined), coal bulk dry
cargo, non-coal bulk dry cargo (iron ore, coal, grain, bauxite and aluminia and phosphate, all combined) and
other dry cargo (essentially considered as container and other similar purpose shipping). The data were for
international shipping only. Transport work (i.e. tonne-miles), as opposed to the absolute amount transported
(tonnes), is considered to be a better variable to predict transport demand and emissions. However, this
assumes that average hauls remain constant: this is in fact borne out by the data and the two variables
correlate significantly with an R
2
value of >0.95.
Cargo types were treated separately, as it is evident from the data that they are growing at different rates and
subject to different market demands.
Thus, as a refinement to the approach taken in the Second IMO GHG Study 2009, the current study has
developed the methodology of CE Delft (2012), which considered different ship types and has gone a step
further by decoupling the transport of fossil fuel (oil and coal products) from GDP, as in the RCP/SSP scenarios
in which fossil fuel use is decoupled from economic development.
In order to predict ship transport work (by type, or total), the general principle is to look for a predictor variable
that has a meaningful physical relationship with it. In previous scenario studies, global GDP has been used as
a predictor for total ship transport work, in that it has a significant positive statistical correlation, and is also
meaningful in the sense that an increase in global GDP is likely to result in an increase in global trade and
therefore ship transport of goods.
If an independent assessment of the predictor variable (e.g. GDP) is available for future years, this allows
prediction of ship transport work. It assumes that such a physical relationship is as robust for the future as it has
been for the past. Previously, a linear assumption has been made, i.e. a linear regression model has been used
between the ratio of historical transport work to historical GDP against time. In this study, this assumption
has been improved by the use of a non-linear model, commonly used in economics, that assumes classic
emergence, growth and maturation phases.
However, the assumption of a historical relationship between coal and oil transport by shipping and GDP
inherently means that GDP growth and fossil fuel use will remain tightly coupled in the future, i.e. that with
increased economic growth, it is not possible to limit fossil fuel use. This clearly does not reflect certain
desired policy and environmental outcomes, where a decrease in fossil fuel dependence and an increase in
GDP can be achieved.
In order to overcome this, this study has investigated the relationship between historical ship-transported coal
and oil and historical global coal and oil consumption. This relationship has been found to be as robust as that
between historical coal and oil transport work and historical GDP (r2 >0.9) and is arguably a better physical
relationship than between fossil fuel transported by shipping and GDP. The RCP scenarios have provided
projections of fossil fuel consumption, split between coal and oil. This conveniently allows us to use these
predictor variables to determine potential future ship transport of coal and oil but decoupled from GDP. Other
ship transported goods and products remain predicted by independent future GDP assessments provided by
the RCPs.
132 Third IMO GHG Study 2014
In all cases of ship-transported products, the non-linear Verhulst regression model (with S-shaped curve) is
used to reflect more realistic market behaviour rather than continued linear relationships. The historical data
on transport work (by type) and demand and GDP are shown in Figure 79.
Figure 79: Historical data on world coal and oil consumption, coal and oil transported (upper panel),
total (non-coal) bulk dry goods, other dry cargoes and global GDP (lower panel)
Predicted proxy data of (separate) coal and oil demand and GDP were provided by the RCP/SSP scenarios and
the associated underlying IAMs. In one case (RCP6.0), fossil energy demand data could not be obtained and
data from the IAM GCAM were used.
3.2.4 Fleet productivity
For the emissions projection, the development of the tonnage of the different ship types is determined by a
projection of the ships’ productivity, defined as transport work per deadweight tonne. More precisely, the fleet
is assumed to grow if, given the projected productivity, the expected transport demand cannot be met by the
fleet. On the other hand, if, given the projected productivity, the expected transport demand could be met by
a smaller fleet, the active fleet is not assumed to decrease. This means that ships are assumed to reduce their
cargo load factor – i.e. become less productive – rather than being scrapped or laid up or reducing their speed.
The projection of ship productivity is based on the historical productivity of the ship types. For all ship types,
the 2012 productivity of the ship types is lower than the long-term historical average (see Annex 7 for more
details). This is assumed to be caused by the business cycle, rather than by structural changes in the shipping
market; this study therefore applies a future productivity development that converges towards the ship type’s
average productivity, reverting back to the 25-year
1
mean value within 10 years, i.e. until 2022.
The ship productivity indices used in the emissions projection model, which can be specified per five-year
period, are given in Table 71.
Table 71 – Ship type productivity indices used in emissions projection model
2012 2017 2022–2050
Liquid bulk vessels 100 113 125
Dry bulk vessels 100 102 104
Container ships 100 109 118
General cargo vessels 100 109 118
Liquefied gas carriers 100 106 113
All other vessels 100 100 100
3.2.5 Ship size development
In the emissions projection model, ship types are divided into the same ship size categories as in the emissions
inventory model. For the emissions projection, the future number of ships per size category has to be
determined.
The distribution of ships over their size categories can be expected to change over time according to the
number of ships that are scrapped and enter the fleet, as well as their respective sizes.
In the emissions projection model, it is assumed that total capacity per ship type meets projected transport
demand, that all ships have a uniform lifetime of 25 years and that the average size of the ships per size
category will not change compared to the base year 2012, while the number of ships per bin size will.
The development of the distribution of vessels over the size categories until 2050 is determined based on
a literature review, taking into account historical developments in distribution, expected structural changes
in the markets and infrastructural constraints. In Table 72 and Table 73, 2012 distributions and expected
distributions for 2050 are presented.
1
Due to a lack of historical data, for container vessels and liquefied gas vessels we take the average of the 1999–2012 period, i.e. a
13-year period.
Scenarios for shipping emissions 2012–2050 133
is assumed to grow if, given the projected productivity, the expected transport demand cannot be met by the
fleet. On the other hand, if, given the projected productivity, the expected transport demand could be met by
a smaller fleet, the active fleet is not assumed to decrease. This means that ships are assumed to reduce their
cargo load factor – i.e. become less productive – rather than being scrapped or laid up or reducing their speed.
The projection of ship productivity is based on the historical productivity of the ship types. For all ship types,
the 2012 productivity of the ship types is lower than the long-term historical average (see Annex 7 for more
details). This is assumed to be caused by the business cycle, rather than by structural changes in the shipping
market; this study therefore applies a future productivity development that converges towards the ship type’s
average productivity, reverting back to the 25-year
1
mean value within 10 years, i.e. until 2022.
The ship productivity indices used in the emissions projection model, which can be specified per five-year
period, are given in Table 71.
Table 71 – Ship type productivity indices used in emissions projection model
2012 2017 2022–2050
Liquid bulk vessels 100 113 125
Dry bulk vessels 100 102 104
Container ships 100 109 118
General cargo vessels 100 109 118
Liquefied gas carriers 100 106 113
All other vessels 100 100 100
3.2.5 Ship size development
In the emissions projection model, ship types are divided into the same ship size categories as in the emissions
inventory model. For the emissions projection, the future number of ships per size category has to be
determined.
The distribution of ships over their size categories can be expected to change over time according to the
number of ships that are scrapped and enter the fleet, as well as their respective sizes.
In the emissions projection model, it is assumed that total capacity per ship type meets projected transport
demand, that all ships have a uniform lifetime of 25 years and that the average size of the ships per size
category will not change compared to the base year 2012, while the number of ships per bin size will.
The development of the distribution of vessels over the size categories until 2050 is determined based on
a literature review, taking into account historical developments in distribution, expected structural changes
in the markets and infrastructural constraints. In Table 72 and Table 73, 2012 distributions and expected
distributions for 2050 are presented.
1
Due to a lack of historical data, for container vessels and liquefied gas vessels we take the average of the 1999–2012 period, i.e. a
13-year period.
134 Third IMO GHG Study 2014
Table 72 – 2012 distribution and expected distribution 2050 of container and LG carriers over bin sizes
Ship type Bin sizes (dwt)
Distribution in terms of numbers
2012 2050
Container vessels 0–999 22% 22%
1,000–1,999 TEU 25% 20%
2,000–2,999 TEU 14% 18%
3,000–4,999 TEU 19% 5%
5,000–7,999 TEU 11% 11%
8,000–11,999 TEU 7% 10%
12,000–14,500 TEU 2% 9%
14,500 –+ TEU
0.2% 5%
Liquefied gas carriers 0–49,000 m
3
68% 32%
50,000–199,999 m
3
29% 66%
>200,000 m
3
3% 2%
Table 73 – 2012 distribution and expected distribution 2050 of oil/chemical tankers
and dry bulk carriers over bin sizes
Ship type Bin sizes (dwt)
Distribution in terms of numbers
2012 2050
Oil/chemical tankers 0–4,999 1% 1%
5,000–9,999 1% 1%
10,000–19,999 1% 1%
20,000–59,999 7% 7%
60,000–79,999 7% 7%
80,000–119,999 23% 23%
120,000–199,999 17% 17%
200,000–+
43% 43%
Dry bulk carriers 0–9,999 1% 1%
10,000–34,999 9% 6%
35,000–59,999 22% 20%
60,000–99,999 26% 23%
100,000–199,999 31% 40%
200,000–+
11% 10%
For the other ship types, the 2012 size distribution is presumed not to change until 2050.
3.2.6 EEDI, SEEMP and autonomous improvements in efficiency
The projection of the future emissions of maritime shipping requires the projection of future developments in
the fleet’s fuel efficiency. In the period up to 2030, this study distinguishes between market-driven efficiency
changes and changes required by regulation, i.e. EEDI and SEEMP. Market-driven efficiency changes are
modelled using a MACC, assuming that a certain share of the cost-effective abatement options is implemented.
In addition, regulatory requirements may result in the implementation of abatement options irrespective of
their cost-effectiveness. Between 2030 and 2050, there is little merit in using MACCs, as the uncertainty about
the costs of technology and its abatement potential increases rapidly for untested technologies. In addition,
regulatory improvements in efficiency for the post 2030 period have been discussed but not defined. This
study therefore takes a holistic approach towards ship efficiency after 2030.
Our MACC is based on data collected for IMarEST and submitted to the IMO in MEPC 62/INF.7. The cost
curve uses data on the investment and operational costs and fuel savings of 22 measures to improve the
energy efficiency of ships, grouped into 15 groups (measures within one group are mutually exclusive and
cannot be implemented simultaneously on a ship). The MACC takes into account that some measures can be
implemented on specific ship types only. It is also assumed that not all cost-effective measures are implemented
immediately but that there is a gradual increase in the uptake of cost-effective measures over time.
EEDI will result in more efficient ship designs and consequently in ships that have better operational efficiency.
In estimating the impact of EEDI on operational efficiency, this study takes two counteracting factors into
account. First, the current normal distribution of efficiency (i.e. there are as many ships below as above
the average efficiency, and the larger the deviation from the mean, the fewer ships there are) is assumed to
change to a skewed distribution (i.e. most ships have efficiencies at or just below the limit, and the average
efficiency will be a little below the limit value). As a result, the average efficiency improvement will exceed
the imposed stringency limit. Second, the fact that most new-build ships install engines with a better specific
fuel consumption than has been assumed in defining the EEDI reference lines is also taken into account.
The result of these two factors is that operational improvements in efficiency of new ships will exceed the
EEDI requirements in the first three phases but will lag behind in the third (see Annex 7 for a more detailed
explanation).
It is likely that improvements in efficiency will continue after 2030, although it is impossible to predict what
share of the improvements will be market-driven and what regulation-driven. Because of the high uncertainty
of technological development over such a timescale, two scenarios are adopted. One coincides with the
highest estimates in the literature (excluding speed and alternative fuels, which are accounted for elsewhere):
a 60% improvement over current efficiency levels. The second has more conservative estimates: a 40%
improvement over current levels.
3.2.7 Fuel mix: market- and regulation-driven changes
Two main factors will determine the future bunker fuel mix of international shipping:
1 the relative costs of using the alternative fuels;
2 the relative costs of the sector’s alternative options for compliance with environmental regulation.
The environmental regulations that can be expected to have the greatest impact on the future bunker fuel mix
are the SO
x
and NO
x
limits set by the IMO (regulations 13 and 14 of MARPOL Annex VI), which will become
more stringent in the future. This will also apply in any additional ECAs that may be established in the future.
In the emissions projection model, two fuel mix scenarios are considered, a low LNG/constant ECAs case and
a high LNG/extra ECAs case.
In the low LNG/constant ECAs case, the share of fuel used in ECAs will remain constant. In this case, it is
assumed that half of the fuel currently used in ECAs is used in ECAs that control SO
x
only, and the other half
in ECAs where both SO
x
and NO
x
emissions are controlled. In this scenario, NO
x
controls are introduced in
half of the ECAs from 2016 and in the other half from 2025. In this case, demand for LNG is limited.
The high LNG/extra ECAs case assumes that new ECAs will be established in 2030, doubling the share of fuel
used in ECAs. In this case, there is a strong incentive to use LNG to comply with ECAs. In Table 74, the fuel
mix is given per scenario.
Scenarios for shipping emissions 2012–2050 135
cannot be implemented simultaneously on a ship). The MACC takes into account that some measures can be
implemented on specific ship types only. It is also assumed that not all cost-effective measures are implemented
immediately but that there is a gradual increase in the uptake of cost-effective measures over time.
EEDI will result in more efficient ship designs and consequently in ships that have better operational efficiency.
In estimating the impact of EEDI on operational efficiency, this study takes two counteracting factors into
account. First, the current normal distribution of efficiency (i.e. there are as many ships below as above
the average efficiency, and the larger the deviation from the mean, the fewer ships there are) is assumed to
change to a skewed distribution (i.e. most ships have efficiencies at or just below the limit, and the average
efficiency will be a little below the limit value). As a result, the average efficiency improvement will exceed
the imposed stringency limit. Second, the fact that most new-build ships install engines with a better specific
fuel consumption than has been assumed in defining the EEDI reference lines is also taken into account.
The result of these two factors is that operational improvements in efficiency of new ships will exceed the
EEDI requirements in the first three phases but will lag behind in the third (see Annex 7 for a more detailed
explanation).
It is likely that improvements in efficiency will continue after 2030, although it is impossible to predict what
share of the improvements will be market-driven and what regulation-driven. Because of the high uncertainty
of technological development over such a timescale, two scenarios are adopted. One coincides with the
highest estimates in the literature (excluding speed and alternative fuels, which are accounted for elsewhere):
a 60% improvement over current efficiency levels. The second has more conservative estimates: a 40%
improvement over current levels.
3.2.7 Fuel mix: market- and regulation-driven changes
Two main factors will determine the future bunker fuel mix of international shipping:
1 the relative costs of using the alternative fuels;
2 the relative costs of the sector’s alternative options for compliance with environmental regulation.
The environmental regulations that can be expected to have the greatest impact on the future bunker fuel mix
are the SO
x
and NO
x
limits set by the IMO (regulations 13 and 14 of MARPOL Annex VI), which will become
more stringent in the future. This will also apply in any additional ECAs that may be established in the future.
In the emissions projection model, two fuel mix scenarios are considered, a low LNG/constant ECAs case and
a high LNG/extra ECAs case.
In the low LNG/constant ECAs case, the share of fuel used in ECAs will remain constant. In this case, it is
assumed that half of the fuel currently used in ECAs is used in ECAs that control SO
x
only, and the other half
in ECAs where both SO
x
and NO
x
emissions are controlled. In this scenario, NO
x
controls are introduced in
half of the ECAs from 2016 and in the other half from 2025. In this case, demand for LNG is limited.
The high LNG/extra ECAs case assumes that new ECAs will be established in 2030, doubling the share of fuel
used in ECAs. In this case, there is a strong incentive to use LNG to comply with ECAs. In Table 74, the fuel
mix is given per scenario.
136 Third IMO GHG Study 2014
Table 74 – Fuel mix scenarios used for emissions projection (mass %)
High LNG/extra ECAs case LNG share Distillates and LSHFO* HFO
2012 0% 15% 85%
2020 10% 30% 60%
2030 15% 35% 50%
2050 25% 35% 40%
Low LNG/constant ECAs case LNG share Distillates and LSHFO* HFO
2012 0% 15% 85%
2020 10% 30% 60%
2030 15% 35% 50%
2050 25% 35% 40%
*
Sulphur content of 1% in 2012 and 0.5% from 2020.
Both scenarios assume that the global 0.5% sulphur requ