Greenhouse gas emissions in coffee grown with differing input levels under conventional and organic management

Abstract

Coffee plays a key role in sustaining millions of livelihoods around the world. Understanding GHG emissions from coffee supply chains is important in evaluating options for climate change mitigation within the sector. We use data from two long-term coffee agroforestry experiments in Costa Rica and Nicaragua to calculate carbon footprints (CF) for coffee and identify emission hotspots within different management systems, levels of inputs and shade types. Management system and input level were the main cause of variation in CFs. Carbon footprints for 1 kg of fresh coffee cherries were between 0.26 and 0.67 kgCO2e for conventional and 0.12 and 0.52 kgCO2e for organic management systems. The main contributor to GHG emissions for all management systems was the inputs of organic and inorganic nitrogen. Nitrous oxide emissions from pruning inputs contributed between 7% and 42% of CFs. However, these estimates were strongly influenced by the choice of emission factor used in the calculations. Research is required to develop emission factors that account for different qualities and management of nitrogen inputs to enable effective calculation of the CF from different management strategies, and especially from the pruning and organic inputs managed in agroforestry systems. As such, effective climate change mitigation strategies can only be developed from site-specific studies which utilise accurate accounting and regional-specific emission factors.

Full text

Greenhouse gas emissions in coffee grown with differing input levels under conventional and
organic management
Martin R.A. Noponenab,1, Gareth Edwards-Jonesae, Jeremy P. Haggarbc, Gabriela Sotob, Nicola
Attarzadehad, John R. Healeya
a School of Environment, Natural Resource and Geography, Bangor University, Bangor, Gwynedd, LL57 2UW,
United Kingdom
b Centro Agronómico Tropical de Investigación y Enseñanza (CATIE), Turrialba, CATIE 7170, Costa Rica
c Natural Resource Institute (NRI), University of Greenwich at Medway, Chatham, ME4 4TB, United Kingdom
d CarbonRoots, York YO30 7DN, United Kingdom
e Deceased
Keywords: Agroforestry systems; Carbon footprinting; Climate change; Coffee; Nitrous oxide.
Abstract
Coffee plays a key role in sustaining millions of livelihoods around the world. Understanding GHG
emissions from coffee supply chains is important in evaluating options for climate change mitigation
within the sector. We use data from two long-term coffee agroforestry experiments in Costa Rica
and Nicaragua to calculate carbon footprints (CF) for coffee and identify emission hotspots within
different management systems, levels of inputs and shade types. Management system and input
level were the main cause of variation in CFs. Carbon footprints for 1 kg of fresh coffee cherries were
between 0.26 and 0.67 kgCO2e for conventional and 0.12 and 0.52 kgCO2e for organic management
systems. The main contributor to GHG emissions for all management systems was the inputs of
organic and inorganic nitrogen. Nitrous oxide emissions from pruning inputs contributed between
7% and 42 % of CFs. However, these estimates were strongly influenced by the choice of emission
1 Corresponding author. Current address: School of the Environment and Natural Resource and Geography,
Bangor University, Bangor, Gwynedd, LL57 2UW, United Kingdom. Tel.: +44 1904 399 860
E-mail address: m.noponen@bangor.ac.uk

factor used in the calculations. Research is required to develop emission factors that account for
different qualities and management of nitrogen inputs to enable effective calculation of the CF from
different management strategies, and especially from the pruning and organic inputs managed in
agroforestry systems. As such, effective climate change mitigation strategies can only be developed
from site-specific studies which utilise accurate accounting and regional-specific emission factors.
1. Introduction
The need for sustainable intensification of food production has recently been emphasised in the
development of global food policy (Foresight, 2011). Given the likely impacts of climate change and
rising human populations (UN 2009), a key challenge for achieving such sustainable intensification is
to develop farming systems which produce increased yields without associated increases in
greenhouse gas (GHG) emissions. In order to achieve this aim there is a need to fully understand the
types and amounts of GHGs that are emitted by different food production systems.
Product carbon footprinting (PCF) (often referred to as ‘carbon footprinting’ (CF)) is
commonly used to calculate the GHG emissions released from food supply chains. Developing a CF
has some similarities to developing a life cycle assessment (LCA), and many of the CF methods
currently in use are based upon the ISO method for Life Cycle Assessment, ISO 14040/44 (e.g., the
GHG Protocol’s draft Product Life Cycle Accounting and Reporting Standard (World Resources
Institute & World Business Council for Sustainable Development, 2009) and the British Standard
Institute’s Publically Available Specification 2050:2008 (hereafter referred to as PAS 2050) for
assessment of the life-cycle greenhouse gas emissions of goods and services (BSI, 2008)). Both the
draft GHG Protocol method and PAS 2050 have been developed in response to a call for
standardised and transparent CF methods, as ISO 14040 and 14044 have been criticised for being
flexible in their approach and therefore open to some interpretation in application (Plassmann et al.,
2010). By maintaining consistency in the calculation method it should be possible to compare the

CFs of different supply chains, thus enabling identification of systems with lower GHG emissions per
unit of production.
A number of problems, however, exist with the methods currently used for making CF calculations,
most notably the fact that despite the calls for consistency, different CF schemes do adopt different
analytical methods (Bolwig and Gibbon, 2009; Plassmann et al., 2010). For example, Plassmann et al.
(2010) found that the CF of a kilogram of sugar can vary by up to 1900% when calculated by different
CF methods. By far the greatest contributor to CF variation was the treatment of land use change
emissions (emissions released during the conversion of non-agricultural land to agriculture). This is
of concern for agricultural production in developing countries, where contemporary conversion of
land use from non-agricultural tree-dominated to agriculture is more likely than in developed
countries, and where few data currently exist to enable the accurate calculation of these emissions
(Brenton et al., 2009; Plassmann et al., 2010).
A second problem associated with CFs relates to the availability of relevant emission factors (EFs). In
essence, CFs arecalculatedbymultiplyingthequantitiesofallinputswhichcontributetoaproduct’s
life cycle (e.g. kg fertilisers, kWh electricity, litres diesel) by their relative EF, and summing these
emissions together to form the total CF. Emission factors represent the contribution of a product or
process to global warming, and are expressed in units of carbon dioxide equivalents (CO2e). Emission
factors are published in commercial databases and the scientific literature, but as the majority of CF
research and method-development to-date has taken place within industrialised countries there is a
lack of location-specific EFs for many production systems that occur in less industrialised countries,
e.g. coffee. This is a major challenge for understanding the levels of emissions from these regions.
As one of the most traded commodities in the world and with over 10 million hectares of land
devoted to its production (FAO, 2011), coffee continues to be one of the most widely grown cash
crops, sustaining the livelihoods of up to 25 million people globally (IIED, 1997). As a result, the
coffee supply chain is an important contributor to global GHG emissions. However, whilst a major
emission hotspot within the coffee supply chain has been found to lie within the production of

coffee at the farm level (PCF Pilotprojekt Deutschland, 2008), its GHG emissions remains relatively
understudied (Hergoualc’h et al., 2008; Verchot et al., 2006). Against this background the present
study uses PAS 2050:2008 and IPCC CF methods to (i) estimate the relative GHG emissions from
different levels of management and material inputs (high versus moderate) and from different types
of input (organic versus conventional production systems), (ii) identify the greatest source of GHG
emissionsfromeachsystem(their“emissionhotspots”)and(iii)determine the effects of uncertainty
in EF on the overall CF. Results from studies such as this should make an important contribution to
quantifying global GHG emissions from agricultural production and designing sustainable and
efficient systems that can meet human needs with a reduced environmental impact.
2. Methods
2.1. Site description
The research was conducted at two 3-ha field sites, in Costa Rica and Nicaragua respectively, chosen
to represent low altitude coffee growing regions, and both managed by the ‘Centro Agronómico
Tropical de Investigación y Enseñanza’(CATIE). Both sites were established at the end of 2000. The
Costa Rica site is located in Turrialba (9°53’44”N,83°40’7”W) at 685 m above sea level. The climate is
humid tropical with no marked dry season: annual precipitation is 2600 mm yr-1 and annual mean
temperature is 22 °C (Haggar et al., 2011). Two soil types have been identified at the site and
classified as Typic Endoaquepts and Typic Endoaquults under the USDA Soil Taxonomy classification
system (Soil Survey Staff, 1999); both are poorly drained. The previous land-use was sugar cane
cultivation. For establishment of the current experiment, the site was prepared with extensive
drainage channels of up to 1.5 m in depth. The coffee cultivar Coffea arabica L. ‘Caturra’ was
planted. The Nicaragua site is locatedinMasatepe(11°53’54”N, 86°08’56”W) at 455 m above sea
level. The climate is semi-dry tropical with a distinct rainy season between May and November:
mean annual rainfall is 1386 mm and mean annual temperature is 24 °C (Haggar et al., 2011). Two
soil types have been identified at the site and classified as Andisols or Andosols (Humic Durustands

and Humic Haplustands) under the USDA Soil Taxonomy classification system (Soil Survey Staff,
1999). The previous land-use was long-established shaded coffee. In the experiment, the coffee
cultivar planted was Coffea arabica var. ‘Pacas’. A more detailed description of the experiments and
their productivity is reported elsewhere (Haggar et al., 2011).
2.2. Experimental design
The experiments were set up to study the ecological efficiencies of coffee production. A main aim is
to compare organic and conventional coffee production systems under various types of shade. The
five main-plot treatments at each site are full sun and four different individual species or
combinations of shade tree (Table 1). The four sub-plot treatments are systems combining the two
different types and levels of nutrient and pest management inputs (Table 3). The tree species used in
the experiment (Table 2) are selected from those most commonly grown in association with coffee
production in the two regions. The design is a randomized block with three blocks per site, each
containing one replicate of each treatment combination. An incomplete factorial design comprising
14 of the potential 20 main-plot/sub-plot treatment combinations at each site was chosen as some
combinations are not representative of real farming systems (e.g. full sun with organic management,
Table 1). The sub-plots range in size between 500 and 800 m2 including borders. Coffee bushes were
planted at a density of 4000 and 5000 plants per ha in Nicaragua and Costa Rica respectively which
did not differ amongst the main-plot or sub-plot treatments. Shade trees were planted in 2000 at a
density of 416 and 667 trees per ha-1 in Costa Rica and Nicaragua respectively but have since been
progressively thinned and pruned to achieve a uniform shade level (Table 1).
The tree management regime varied according to species; Erythrina poeppigiana in Costa Rica and
Inga laurina in Nicaragua (both Leguminosae) were pruned for the management of shade and to
provide organic matter (including N) input to soil. All E. poeppigiana trees were heavily pruned twice
per year and their prunings left on the ground. In the conventional intensive (CI) sub-plot treatments
of E. poeppigiana, trees were pruned at a height of 1.8-2.0 m with the removal of all branches above

this height (pollarding). This practice is frequently found in conventional high-intensity coffee
agroforestry systems in Costa Rica. In the other three sub-plot treatments, however, E. poeppigiana
trees were managed according to the recommendations of Muschler (2001) whereby trees were
pruned at a height of around 4 m and a minimum of three branches were left for partial shade cover.
In Nicaragua, I. laurina was managed to create a homogeneous canopy cover of approximately 40%,
through annual pruning of branches at any height, accounting for overall smaller pruning residue
inputs compared with E. poepiggiana in Costa Rica. In contrast, the timber tree species were
managed to promote the development of a straight trunk and thus maximise timber value but were
not subjected to a systematic pruning regime. Trunks and major branches of thinned and pruned
timber trees were removed from the plots whereas leaf and small branch material was left as an
organic amendment. All the material pruned from coffee bushes was also left in the plots (coffee
bushes were pruned according to standard coffee agronomic practice, to the same level across all
treatments).
2.3. Calculation of carbon footprints
PAS 2050 (BSI, 2008) is the only transparent and publically available product CF method published
to-date and has therefore been chosen here for all CF calculations. Within this method, all GHGs
(including CO2, N2O and CH4) are accounted for and converted into units of CO2-equivalents (CO2e)
according to their global warming potential (GWP) over 100 years. All GHG emissions associated with
the provision and use of raw materials and energy are included in the calculation. Capital goods,
human energy inputs such as manual labour, transport of employees to and from the workplace and
animals providing transport are excluded from PAS 2050.
Of specific relevance to agricultural CFs are non-CO2 emissions from livestock, their manure and
soils, which must be included, calculated according to IPCC Guidelines for National GHG Inventories
(De Klein et al., 2006). Nitrous oxide emissions from soils are accounted for by including direct and
indirect emissions resulting from N additions, deposition and leaching. As all land under study here

was in agricultural production prior to 1990, no direct emissions from land use change (LUC) have
been included. Changes in soil carbon, either as emissions, sequestration or in eroded material, are
excluded from PAS 2050 unless they are a direct result of LUC activities. Carbon stored in living
organisms such as trees or perennial crops is excluded from the PAS 2050 method; therefore if LUC
results in net carbon storage, no recognition is given by way of a reduced CF. Although this is of
particular relevance to agroforestry systems with perennial crops such as coffee, which have been
shown to provide long-term carbon stores in shade-tree and crop biomass (Segura et al., 2006;
Dossa et al., 2008), currently these gain no recognition for their net carbon storage benefit when
compared, for example, with coffee grown in full sun or with annual crops.
2.4. Data Collection
As the aim of this study is to compare emissions from different farming methods, the system
boundaries were drawn at the farm gate, including only those emissions directly associated with the
production and management of a particular system. Carbon footprint calculations for each system
were based on annualised averages of all inputs and yields since the second year of coffee
production, to best represent the different production systems. The functional unit (unit of
production) was set at 1 kg of non-processed fresh coffee cherries.
Data on coffee yields, management and material inputs were recorded for all sub-plot treatments.
For both conventional managements (Table 3), emissions from the production of inorganic fertilisers
and pesticides were extracted from the Ecoinvent database (Nemecek et al., 2007). For all four sub-
plot management treatments, only commercial fertiliser and pesticide products were assigned
production emissions; PAS2050 states that emissions should beassigned according to a product’s
economic value rather than its mass, thus the production emission from one industry (e.g. chicken
farming) should be partitioned between its products (e.g. chicken meat and manure) according to
their respective commercial values. In the case of these coffee production systems, however,
organic fertilisers such as chicken manure and coffee pulp were assumed to be waste products of

another industry with no economic value, and thus were assigned no production emissions.
Furthermore, although data on GHG emissions from poultry manure can be found within the
Ecoinvent database, we considered these values excessive for this study as the database values
include processing emissions from drying, granulation and packaging (Nemecek et al., 2007) which
are not part of the manure production process in Costa Rica or Nicaragua.
Emissions were calculated for the transportation of materials and fertilisers from their place of
purchase to the on-farm experimental sites; to allow for comparability a default transport distance
of 10 km was chosen for both sites. Emissions arising from the production and use of fuels such as
gasoline and lubricants, used mostly for weed control, and materials and sundries used in the farm
management of the experimental sites were also included in the calculations. Emission factors for
the production and manufacturing processes of individual inputs were obtained from the publically
available database of the Renewable Fuels Agency (RFA) and Ecoinvent (Althaus et al., 2007; Classen
et al., 2009). Costa Rica-specific EFs for diesel and gasoline were sourced from a report used in the
Costa Rican national GHG inventory (Ministerio de Ambiente y Energía de Costa Rica, 2007) and used
forbothcountries’footprints.Noelectricitywasconsumedintheon-farm operations.
For calculating N2O emissions from soil we followed IPCC Good Practice Guidelines for calculating
GHG emissions (De Klein et al., 2006) and chose a regional-specific EF (Table 1) from Costa Rica for N
fertiliser application of 1% for timber-tree and full-sun coffee production systems established by
Hergoualc’h et al. (2008), 1.2% for leguminous-shade systems and a value from the same study of
0.3% for N applications from pruning inputs (Hergoualc’h pers. comm.). To assess the effects of using
differentEF’s on the overall CF we compared the results of (i) using the IPCC tier 1 default value of
1% for all N inputs (scenario 1) (De Klein et al., 2006), (ii) using a region-specific EF (scenario 2) and
(iii) excluding emissions from pruning inputs (scenario 3) (Hergoualc’h et al., 2008). N contents of
pruning residues, needed to calculate soil N2O emissions, were obtained from analyses carried out at
the Laboratorio de Análisis de Suelos, Tejido Vegetal y Aguas at CATIE, in Costa Rica.

Table 1 Main-plot (shade tree combinations) and sub-plot (management inputs) treatments at the experimental sites in a) Costa Rica and b) Nicaragua. Sub-plot treatment
abbreviations are given in table 3.
a) Costa Rica
b)
Nicaragua
Main-plot treatments
Full sun
Erythrina
poeppigiana
Terminalia
amazonia
Chloroleucon
eurycyclum
Erythrina
poeppigiana/
Terminalia
amazonia
Full
sun
Simarouba
glauca/
Tabebuia
rosea
Samanea
saman/
Tabebuia
rosea
Inga
laurina/
Simarouba
glauca
Inga
laurina/
Samanea
saman
Abbreviation
FS
E
T
C
ET
FS
SGTR
SSTR
ILSG
ILSS
Sub-plot treatments
CM1, CI
OM, OI,
CM, CI
OM, OI,
CM, CI
OI, CM
OI, CM
CM, CI
OM, OI,
CM, CI
OI, CM
OI, CM
OM, OI,
CM, CI
Shade tree density (ha-1)
Emission factor for N
inputs (excluding
pruning)
-2
1%
2693/5834
1.2%
216
1%
257
1.2%
231
1.2%
-2
1%
286
1%
331
1.2%
336
1.2%
376
1.2%
1 Subplot treatments are shown in full in Table 3; 2 no shade trees are present in full sun treatments; 3 densities for OM, OI and CM sub-plot treatments; 4 densities for CI
sub-plot treatment
Table 2 Shade tree species used in the main-plot experimental treatments in the sites in Costa Rica and Nicaragua
a) Costa Rica
b) Nicaragua
Species
Terminalia
amazonia
(J.F. Gmel.) Exell
Chloroleucon
eurycyclum
Barneby & J.W.
Grimes
Erythrina
poeppigiana
(Walp.) O.F. Cook
Inga laurina
(Sw.) Willd.
Samanea saman
(Jacq.) Merr.
Simarouba glauca
DC.
Tabebuia rosea
(Bertol.) DC.
Phenology
evergreen
evergreen
evergreen
evergreen
evergreen
evergreen
deciduous
N-fixer
no
yes
yes
yes
yes
no
no
Dominant use
timber1
timber1
service2
service2
timber1
timber1
timber1
1 ’timber’=shadetreesthataremanagedfortheirtimber;2‘service’=shadetreesthataremanagedfortheir‘services’tocoffeeproduction,e.g.N-fixation, organic matter
inputs

Table 3 Experimental sub-plot coffee management treatments in the sites in Costa Rica and Nicaragua.
1 Quantities of soil amendments and yields are shown as mean values of known amounts applied annually over seven years.
Name of sub-treatment
Organic Moderate
Organic Intensive
Conventional Moderate
Conventional Intensive
Abbreviation
OM
OI
CM
CI
Soil amendments1
Organic-coffee pulp
(kg-1 ha-1 yr-1) Costa
Rica: N 66, P 2, K 44;
Nicaragua: N 140, P 8,
K 88
Organic-coffee pulp, chicken
manure, lime, rock
phosphate, potassium
sulphate, (mean kg-1 ha-1 yr-
1) Costa Rica: N 248, P 205, K
326; Nicaragua: N 372, P
179, K 145)
Inorganic fertiliser
(kg-1 ha-1 yr-1) Costa Rica: N
150, P 10, K 75; Nicaragua:
N 78, P 46, K 21)
Inorganic fertiliser
(kg-1 ha-1 yr-1)Costa Rica: N
287, P 20, K 150;
Nicaragua: N 153, P 98, K
57)
Disease management
None
Use of organic and plant
derived substances
dependent on disease
incidence
Use of up to 4 inorganic
fungicide applications
dependent on disease
incidence
Regular use of 3 - 4
inorganic fungicides
applications
Insect pest management
Reducing of “gleaning”
(fallen cherries) after
harvest
Manual removal and use of
organic and plant derived
substances dependent on
disease incidence
Manual removal and use
of up to 4 inorganic
insecticides dependent on
disease incidence
Manual practices and
regular use of 3 - 4
inorganic insecticide
applications
Weed management
2-4 routine machete
weedings per year
Manual selective weed
management between rows
and cleaning within the row
area
Selective weed
management between
row and cleaning within
the row area manually and
with herbicide
Soil maintained clear of
weeds with herbicides
Average yield (±SE) of coffee
cherries across treatments
(tha-1yr-1) in Costa Rica /
Nicaragua
4.8 (±1.1) / 4.7 (±0.7)
6.6 (±0.36) / 6.4 (±0.4)
7.0 (±0.4) / 5.5 (±0.3)
9.9 (±0.7) / 7.1 (±0.8)

2.5. Statistical analysis
To investigate the relationship between main-plot and sub-plot treatment effects on individual CFs in
the experiment we fitted linear mixed effects models in R (R Development Core Team, 2010) using the
lme4 package (Bates et al., 2011). Main-plot/sub-plot treatment combinations were fitted as a factor
with 15 levels for each country (model 1: fixed effects = main-plot + sub-subplot; model 2: fixed effect =
main-plot; model 3: fixed effect = sub-plot). Results were assessed using the Akaike Information
Criterion (AIC, Burnham and Anderson, 1998), and the model presenting the smallest AIC selected.
3. Results
3.1. Carbon footprinting of different management systems
The best fitting model for predicting carbon footprints from the main-plot and sub-plot treatments of
the experiments is model 3 based on only sub-plot treatments as a fixed effect with random slope
effects of replicate blocks and main-plot treatments nested within replicate blocks (the AIC values for
this model for Costa Rica and Nicaragua were respectively -32.4 and -66.0; in comparison those for
model 1 were -13.6 and -49.5, and for model 2 were -4.0 and -35.6 respectively). This shows that
management system and input level (the sub-plot treatment) accounts for most variation in CFs
amongst the treatments in the experiment with little remaining variation explained by shade type (the
main-plot treatment). This is reflected in their relative coefficients of variation between treatment
mean values (CV is 0.17 and 0.18 amongst the sub-plot treatments and only 0.03 and 0.08 amongst the
main-plot treatments for Costa Rica and Nicaragua respectively). Interactions between main-plot and
sub-plot treatments cannot be tested separately for each of the experiments as a whole because of the
incomplete factorial design. Therefore, results presented here are largely aggregated at the sub-plot
level (Figure 1 and Table 4). Nonetheless, for both countries, based on non-overlap of 84% confidence
intervals of sub-plot intercepts for the best-fit model (Payton et al., 2003), there were no significant
differences at the p < 0.05 level amongst the sub-plot treatments (Figure 1). However, in Costa Rica
there was a notable trend in the association of CF with management type (conventional versus organic)

followed by input level, with the conventional intensive (CI) treatment showing the highest mean CF,
followed by conventional moderate (CM), then organic intensity (OI) and finally organic moderate
(OM). In Nicaragua, the positive association with level of inputs was dominant over management type:
the highest mean CF was again shown by the CI sub-plot treatment, but it was followed by OI, and then
CM and, again last, OM.
a) Costa Rica b) Nicaragua
Figure 1 Mean coffee product carbon footprints based on model predictions for four sub-plot
treatments across five main-plot shade treatments and three replicate blocks in a) Costa Rica and b)
Nicaragua. Conventional intensive (CI); Conventional moderate (CM); Organic intensive (OI); Organic
moderate (OM). The bars represent the mean CF per kg of fresh coffee cherries (kgCO2e); whiskers
indicate the upper and lower boundaries of the 84% confidence interval values (appropriate for judging
significance of differences at p < 0.05).
Direct and indirect soil N2O emissions account for a high proportion of the total product CF (average of
67% across treatments) and are therefore highly correlated with total CF for both conventional and
organic management systems (Figure 2). These emissions result from inorganic and organic fertilisers
and from pruned material from coffee bushes and shade trees (Appendix Tables 1 and 2). Nitrogen
inputs vary considerably across the main-plot/sub-plot treatment combinations due to variation in
CI CM OI OM
Subplot treatment
Mean carbon footprint per kg fresh coffee cherries ( kgCO2e)
0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7
CI CM OI OM
Subplot treatment
Mean carbon footprint per kg fresh coffee cherries ( kgCO2e)
0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7

pruning inputs from shade trees and in coffee bush management (Tables 2 and 3). The significantly
steeper (CF/soil N2O emissions) slope of the conventional than organic treatments is due to the fact
that soil N2O emissions form a greater proportion of the CF for the organic treatments.
Figure 2 Relationship between soil N2O emissions (direct and indirect) resulting from applications of
organic and inorganic N in fertiliser and prunings and the overall carbon footprint of conventional ( )
and organic ( ) coffee management treatments in Costa Rica and Nicaragua. Fitted lines: yconventional(CF)
= 0.031 + 2.03x (kgCO2e); yorganic(CF) = 0.007 + 1.11x (kgCO2e). There was no significant difference
between the intercepts but there were significant differences in the slopes between conventional and
organic management systems (as judged by the non-overlap of 84% confidence intervals of sub-plot
intercepts for best-fit model predictions), highlighting a significant difference between the two groups.
3.2. Carbon footprintemission‘hotspots’
The main CF emission hotspots for the conventional management treatments in both countries were
from fertiliser production and direct and indirect soil N2O emissions from fertiliser N inputs (Table 4).
Emissions from fertiliser production accounted for 50% and 45% of the CI and CM footprints
respectively, averaged across both countries. The main CF emission hotspots for the organic
0.1 0.2 0.3 0.4 0.5
0.1 0.2 0.3 0.4 0.5 0.6 0.7
Soil N2O emissions per kg of fresh coffee cherries ( kgCO2e)
Carbon footprint per kg fresh coffee cherries ( kgCO2e)

management treatments were direct and indirect soil N2O emissions, resulting from applications of
organic fertiliser (such as chicken manure or coffee pulp) and prunings. Soil N 2O emissions accounted
for 92% and 82% of the CF for OI and OM treatments respectively, averaged across both countries, in
contrast to only 45% and 47% for CI and CM treatments respectively. The contribution of N2O emissions
specifically from pruning inputs varied greatly amongst the four treatments ranging from 7% in CI to 42
% in OM. The lower yields of coffee cherries with moderate (OM) compared to high input (OI) organic
management (Table 3) resulted in soil N2O emissions from pruning residues accounting for 1.6 and 1.4
times higher CF per kg of coffee cherry yield in the OM than the OI treatments in Costa Rica and
Nicaragua respectively. Similarly, between the different shade types, in Costa Rica emissions from
pruning residues were highest with E. poeppigiana (Appendix Table 1) at 0.14 kg CO2e per kg of coffee
cherries (averaged across all sub-plot treatments), followed by mixed legume and timber trees (ET) with
0.08 kg CO2e, timber trees (C, T) with 0.01-0.02 kg CO2e and lowest was full sun (FS) at 0.01 kg CO2e
(Appendix Table 2). In Nicaragua a similar trend was detected with the highest emissions from pruning
residues arising in the mixed legume and timber tree types (ILSG, ILSS) with 0.04-0.05 kg CO2e per kg of
coffee cherries followed by timber trees (SGTR, SSTR) with 0.02-0.03 kg CO2e and full sun was again the
lowest with 0.02 kg CO2e per functional unit (Appendix Table 2). The main difference in emissions from
pruning residues between the countries however is due to the fact that these are smaller in quantity in
Nicaragua compared with Costa Rica. A more detailed description of pruning residue inputs within the
experiments can be found in Haggar et al. (2011).
3.3. Impact of different emission factors and the importance of pruning residues
Nitrous oxide emissions released from soils following the addition of fertilisers are commonly estimated
using global, rather than location-specific, EFs. However, soil N2O emissions from pruning inputs are
often overlooked completely in CF analyses. To explore their impact on system CFs we calculated the
mean CF of the four sub-plot coffee management treatments in each country using three different EFs
for the soil N2O emissions resulting from fertiliser and pruning inputs. Using each of the three different

Table 4 Mean greenhouse gas emission contributions (kgCO2et-1 fresh cherries, ±SE) of each emission category to the total product carbon footprint, for
the four sub-plot treatments (Conventional intensive (CI); Conventional moderate (CM); Organic intensive (OI); Organic moderate (OM)) for a) Costa Rica
and b) Nicaragua. The emissions are shown on a per land area per time basis in Appendix Table 1.
Country
Sub-plot
management
treatments
Fertiliser
production
Pesticide
production
Fuels (used
for non-
transport
purposes)
Materials and
sundries
Transport
Direct/indirec
t soil N2O
emissions
from fertiliser
application
Direct/indirec
t soil N2O
emissions
from pruning
inputs
Total
a) Costa Rica
CI
305 (±25)
30 (±3)
-1
-2
-2
196 (±14)
35 (±12)
567 (±41)
CM
227 (±13)
13 (±1)
19 (±1)
-2
-2
152 (±9)
50 (±14)
463 (±29)
OI
7 (±0)
2
20 (±1)
-2
-2
244 (±15)
69 (±18)
345 (±27)
OM
10 (±4)
15 (±7)
51 (±23)
4 (±2)
3 (±1)
65 (±28)
108 (±30)
256 (±68)
b) Nicaragua
CI
162 (±16)
12 (±1)
-2
-2
-2
147 (±15)
25 (±4)
347 (±36)
CM
103 (±6)
13 (±1)
-2
-2
8 (±8)
93 (±6)
36 (±4)
255 (±18)
OI
18 (±1)
-2
-2
-2
5 (±0)
303 (±21)
32 (±4)
359 (±26)
OM
-2
-2
-2
3 (±1)
4 (±1)
94 (±22)
45 (±10)
145 (±34)
1 In Costa Rica CI weed control was managed with chemical herbicides applied manually. 2 Emissions are considered here to be negligible if < 1% of total CF. Sub-plot treatments in a) Costa
Rica (CI, n = 9; CM, n = 15; OI, n = 12; OM, n = 6) and b) Nicaragua (CI, n = 9; CM, n = 15; OI, n = 12; OM, n = 6).

EFs produced a similar trend in CF amongst all four sub-plot treatments in both countries. The greater
variation between the EFs for organic (24-244%) than for conventional (14-40%) management (Figure 3)
was mainly due to the effect of inputs of pruned material.
a) Costa Rica b) Nicaragua
Figure 3 Mean carbon footprint (kgCO2e) for all main-plot treatment x sub-plot treatment combinations
over the three replicate blocks for a) Costa Rica and b) Nicaragua using three different emission factor
scenarios (as described in the Methods section); scenario 1 ( ), scenario 2 ( ), scenario 3 ( ). Bars
represent the mean CF per kg of fresh coffee cherries (kgCO2e); whiskers indicate the upper and lower
boundaries of the 84% confidence interval values.
Scenario 1, which is based on IPCC tier 1 global default values for calculating direct and indirect soil N2O
emissions, does not distinguish between organic, inorganic or pruning/crop residue inputs; it assumes
that 1% of applied N in all the residues is lost as emissions. Scenario 1 produces a greater mean CF than
that from scenario 2, which uses the region-specific lower value of 0.3% for the proportion of N applied
to the soil in pruned material that is emitted as N2O. Scenario 3, which uses the same N fertiliser EFs as
scenario 2 but omits soil N2O emissions from pruning inputs, produced the lowest CF across all
treatments with the greatest reduction in OM sub-plots. Overall, the choice of EF did not change the
rank order of CFs across the four management treatments in either country. However, the effects of EF

choice are more marked in the organic management treatments because N2O inputs from pruning
inputs form a comparatively large proportion of their CF. Further, high variability in CF between main-
plot shade treatments is observed for both of the organic sub-plot treatments in Costa Rica in the
scenario 1 calculations due to the comparatively large contribution to the CF of pruning inputs from the
fast-growing leguminous shade tree E. poeppigiana with a 1% EF.
4. Discussion
4.1. Effect of coffee system management on carbon footprint
In Costa Rica and Nicaragua together, coffee cultivation covered over 212,000 ha of land in 2010 (FAO
2011), making it a significant contributor to bothcountries’agriculturalGHGbalance. Results from this
study found that the carbon footprint per kg of coffee production increased with higher levels of
management input in both conventional and organic systems in both the Costa Rican and Nicaraguan
experiments. The type of farm management was found by the mixed effects models to account for most
variation in CFs. By intensifying coffee farming systems within the experiments, GHG emissions per unit
output are increased for conventional and organic treatments.
However, no general conclusion can be made about the comparative CF of organic and conventional
systems because the results differed between the two countries. While the organic moderate intensity
(OM) treatment had the lowest CF in both countries, the organic intensive (OI) treatment in Nicaragua
had a slightly higher mean CF than the conventional moderate (CM) treatment, whereas in Costa Rica it
was lower. This difference between the countries is associated with the variation in local
implementation of‘conventional’and‘organic’systems.InNicaragua,the OI management had higher N
inputs than the three other management systems, whereas in Costa Rica, total inputs of organic and
inorganic N reduced from the CI to CM to OI to OM management (Table 3). To determine the effects of
organic compared to conventional systems on the carbon footprint it would be necessary to evaluate
the N-use efficiency of the two management strategies at the same level of inputs.

Although the mixed effects model selection procedure showed that shade type had little overall
influence on total CF, there were notable differences in the calculated N2O emissions associated with
their prunings, which were highest from the heavily pruned legume shade trees (Appendix Table 2),
even when accounting for the differences in EF used between leguminous and non-leguminous shade
(the change in the overall CF for the highest pruning residue input system ECI due to a change in EF from
1.2% to 1% of N, is less than 5%). Hergoualc’hetal.(2008)concludedthatannualN2O emissions from a
legume-shaded tree system of coffee were 1.3 times higher compared with an un-shaded coffee
monoculture. With leguminous shade trees contributing 60-340 kg N ha-1 yr-1 through pruning residues
in coffee agroforestry systems (Beer, 1988), the resulting soil N2O emissions can account for a significant
part of the CF. This is reflected in the present study; relative N2O emissions per unit of coffee production
from pruning residues were 84% and 33% lower for timber shade tree only treatments compared with
those using the heavily pruned leguminous shade trees (E and IL) in Costa Rica and Nicaragua
respectively (Appendix Table 2). For FS treatments the emissions were 92% and 63% lower than those
with the heavily pruned leguminous trees respectively (Appendix Table 2). This underlines the
importance of quantifying the different factors that contribute to overall coffee production system
greenhouse gas emissions to provide a broader knowledge base to differentiate the emission factors
associated with different N2O sources.
4.2. Emission ‘hotspots’
In the intensive and moderate input organic coffee management treatments in Costa Rica, and all four
management treatments in Nicaragua, N2O emissions from soil were the greatest emission hotspot.
These emissions stem from the application of mineral and organic fertilisers to the soil, and from
decomposition of pruning residues where applicable. This is in-line with findings from studies of other
crops such as by Plassmann et al. (2010) and Röös et al. (2010) who found that N2O emissions form the
largest portion of the CF of a sugar cane farm in Mauritius and a potato farm in Sweden, respectively.
Despite coffee’s global significance economically and agro-ecologically, there appears to be only one

study published to-date which has analysed the GHG emissions from its cultivation; this pilot study of two
coffee estates in Tanzania found that the production and transport of agrochemicals formed over 79% of
the CF of coffee production and primary processing (PCF Pilotprojekt Deutschland, 2008). This is
comparable with findings from the intensive and moderate input conventional management systems in
Costa Rica in the present study, where fertiliser and pesticide production combined accounted for≥50%
of the CF (Table 4). However, the N2O emissions resulting from N fertiliser application were calculated to
be much higher in the present study than those in Tanzania. This may be because the present study
includes direct and indirect N2O emissions from soils, whereas the Tanzanian pilot study only included
direct emissions (PCF Pilotprojekt Deutschland, 2008). Furthermore, for the two organic management
systems of the present study, emission hotspots were dominated by release of N2O from soils, with
virtually no emissions included from fertiliser production because the fertilisers used are by-products or
wastes of other industries. Although the organic fertilisers used in these experiments contained relatively
small percentages of N, they were applied in large quantities – up to 10 tonnes of chicken manure and 7.5
tonnes of coffee pulp per ha per year in the intensive organic management. As a result, while in the
intensive organic treatment soil N2O emissions were largely caused by application of these organic
fertilisers, in the moderate input organic management, over half the N2O emissions resulted from pruning
inputs from the shade trees and coffee bushes (Table 4).
There is significant scope for managing farm-level GHG emissions through improved planning of N
application, and this should be seen as a priority by farm extension workers when making
recommendations for climate-friendly farming systems. Examples of such GHG-mitigating actions include
switching from urea to use of fertilisers with lower rates of nitrification such as ammonium nitrate;
improved timing of N application, taking into account crop requirements, weather patterns and
availability of mineral N in the rooting zone, so that N is applied at times of greatest demand by the plant;
and subsurface application of fertilisers to reduce losses of NO (Matson et al., 1996; Skiba et al., 1997;
Smith et al., 1997). However, currently the methods used to calculate the CF would not differentiate
between these management practices, and research is needed to quantify their impacts on N2O emissions

and develop appropriate emission factors associated with these practices. Any recommendations
requiring capital investment or a change in farming practice will need wider support in order to
encourage farmer uptake, and indeed further research on improving the efficiency of both organic and
mineral fertiliser use should be seen as a priority in order to determine optimal fertiliser management
mechanisms (Tilman et al., 2002).
4.3. Choice of emission factors
It is clear from this study that, for coffee production CFs, the accuracy of EFs used to calculate direct
and indirect N2O emissions from soil is important; within the production systems analysed here, N2O
emissions formed between 45% and 92% of the total CF, making them the single largest source of
emissions in the organic management treatments and the second largest emissions source in the
conventional treatments. As a result, using different EFs for calculating N2O emissions had a large effect
on CFs, with CF varying by between 14 - 244% depending on the EF used for individual coffee
management treatments (Figure 3). Three categories of soil N2O emissions are commonly accounted
for: direct emissions from N-fertilisation of soils, ‘secondary’ emissions resulting from various
transformations of N compounds, and indirect emissions resulting from leaching and volatilisation of
deposited N (Smith et al., 2010). ‘Secondary’ emissions include those produced by application of crop
residues or pruning material, dung and urine from livestock to the soil, and N mineralisation from soil
organic matter and root residues (Smith et al., 2010). In the IPCC tier 1 methodology (scenario 1 in the
present study), however, no differentiation is made between direct emissions from N-fertilised soils
and secondary emissions from crop residues or pruned material, as both are given the same EF.
Further, its value of 1% for direct N2O emissions has a large uncertainty of 30-300% depending on
localised variables such as climate, soil properties and the quality of the incorporated material (De Klein
et al., 2006). Therefore, calculating a farm’s CF with this global IPCC Tier 1 N2O emissions factor can
introduce significant error, and indeed its use will not enable the estimation of emission reductions
resulting from actions such as improved N use efficiency, as outlined in section 4.2.

In tree-based agricultural systems, and in particular in coffee agroforestry systems in which shade-tree
prunings contribute a significant proportion of “crop residues”, the choice of EF can have a large
influence on the overall CF result as shown in Figure 3. Here, we found that the heavily pruned
leguminous tree species (E. poeppigiana and I. laurina) had much higher relative emissions from
pruning residues per kg of fresh coffee cherries than other shade types (Appendix Table 2). However,
the complexity and interaction of variables influencing soil N2O emissions is vast and, because of their
major importance for the specification of accurate EFs, they should be a priority for further research to
underpin improved carbon footprinting. Factors found to affect N2O release from pruning residues
include: the presence of N-fixing tree species (Hergoualc'h et al., 2008; Verchot et al., 2008), the quality
or chemical composition of plant residues (Seneviratne, 2000; Baggs et al., 2001; Millar and Baggs,
2005) including specifically its C:N ratio (Millar and Baggs, 2004), the interaction between residues and
inorganic fertilisers (Frimpong and Baggs, 2010), and the timing of pruning relative to plant nutrient
demand and supply (Mosier et al., 2004). However, there is a lack of published literature to enable the
accurate calculation of N2O emissions from tropical agricultural systems (Matson et al., 1996; Erickson
et al., 2001; Mosier et al., 2004) and indeed the IPCC default EF is based heavily on data from
temperate and subtropical zones rather than from tropical regions (Stehfest and Bouwman, 2006).
4.4 Implications for carbon footprinting methodology
So far, carbon footprinting in agricultural systems has neglected the role of shade trees (often used in
coffee cultivation) in sequestering significant amounts of C, even beyond the lifetime of the crop. Indeed,
carbon storage in living biomass is omitted from the UK carbon footprinting specification, PAS 2050:2008,
and its recent revision (October 2011, subsequent to the completion of this study) only gives credit for
carbon stored in biomass when that carbon is sequestered as a direct result of land use change occurring
in the past 20 years. Importantly,‘landusechange’isdefined here as a change from one land use type
(e.g. forestry) to another (e.g. agriculture), therefore the addition (or removal) of trees within a coffee
farm during its lifetime would not be recognised as a form of land use change, thus the resulting change

to farm GHG balance would not be included in the carbon footprint. In the case of shade-grown coffee,
however, trees tend to be planted as a result of coffee farming taking place, thus stored carbon in these
systems arises as a direct result of the agricultural production system and should be recognised within the
farm GHG balance calculation. To allow for more representative analyses of agricultural systems, a full
balance based on emitted and sequestered carbon should be calculated, using the carbon footprinting
method followed in this study but including C sequestration and emissions from biomass and soil.
Sequestration of C in some shade systems could outweigh their emission costs resulting in a net C-balance
benefit and potentially making the whole production system carbon neutral over its life span.
4. Conclusions
Carbon footprinting enables improved understanding of the most important GHG emission hotspots
within a food supply chain. This will help in developing systems which achieve higher agricultural
productivity without a proportionate increase in emissions (or lower emissions without a proportionate
reduction in productivity). The results of this study highlight the importance of determining which
impacts and variables are relevant in calculating the net environmental efficiency of agricultural
production systems. While the moderate intensity organic coffee management system had the lowest CF
per kilogramme of coffee produced it also had substantially the lowest yield of coffee per hectare.
Maintenance of the overall level of production from such systems with low GHG emissions but also low
yield per unit area would require conversion of more land to coffee production, locally or elsewhere, but
if this land was converted from forest or grassland this would result in additional emissions. This
emphasises the potential conflict between increasing food production and creating incentives for climate
change mitigation (Angelsen, 2010), which CF methodology needs to encompass.
Identifying emission hotspots through carbon footprinting enables the targeting of farm
management recommendations to reduce the impact of agricultural production on GHG emissions. For six
of the eight coffee management systems studied here, N2O emissions from soil were the greatest
contributor to coffee production CFs, for the other two systems fertiliser production made a larger

contribution. This indicates the value of improvements in fertiliser use efficiency for mitigation of
agricultural GHG emissions on coffee farms.
While methodologies such as those of the IPCC are important in standardising estimation of the
contribution of overall N2O emissions to the CF for gross system comparisons, in order to compare CFs of
different supply chains, accurate emission factors have to be used for each, as demonstrated by the large
variability in CF found when using different EFs for calculating N2O emissions. However, for products such
as coffee, originating in developing countries, despite their huge global impact, there is a shortage of
evidence to enable calculation of EF for different sources and management of nitrogen inputs which are
locally specific. Although much has been published on soil N2O emissions from agricultural systems, a
more detailed understanding of the underlying processes is needed, particularly in tropical regions. Our
research supports the conclusions of Smith et al. (2010) that the link between input parameters and
release processes is a research priority in order to recommend changes in agricultural management that
will reduce emissions. In particular, we recommend new research into the effects of practices aimed to
improve N use efficiency, not only on soil N2O emissions, but also nitrogen use efficiency of coffee
production.
Acknowledgements
We thank CATIE for providing and managing the study sites; Mirna Barrios, Elias de Melo, Luis Romero,
Elvin Navarette and Ledis Navarette for their great efforts in helping to collect the data for this study;
James Gibbons and Aidan Keane for their advice on the statistical analysis; Rodolfo Munguia of the
National Agricultural University of Nicaragua (UNA) for his support. This research was funded by a UK
Economic & Social Research Council/Natural Environment Research Council studentship award and a
partial fieldwork grant by CAFNET and the Coalbourn Trust to MRAN.

References
Althaus, H., Chudacoff, M., Hischier, R., Jungbluth, N., Osses, M., Primas, A., 2007. Life cycle inventories
of chemicals. Final report ecoinvent data v2.0 No. 8, Swiss Centre for Life Cycle Inventories, Dübendorf,
CH.
Angelsen, A., 2010. Policies for reduced deforestation and their impact on agricultural production. Proc.
Natl. Acad. Sci. USA, 107, 19639-19644.
Baggs, E.M., Millar, N., Ndufa, J.K., Cadisch, G., 2001. Effect of residue quality on N2O emissions from
tropical soils, in: Rees, R.M., Ball, B.C., Campbell, C.D., Watson, R.T. (Eds.), Sustainable management of
soil organic matter. CAB International, Oxford, pp. 120-125.
Bates, D., Maechler, M., Bolker, B., 2011. R project. URL http://lme4.r-forge.r-project.org/
Beer, J., 1988. Litter production and nutrient cycling in coffee (Coffea arabica) or cacao (Theobroma
cacao) plantations with shade trees. Agrofor. Syst., 7, 103-114.
Bolwig, S., Gibbon, P., 2010. Product carbon footprinting schemes and standards, in: Gibbon, P., Ponte,
S., Lazaro, E. (Eds.), Global Agro-food Trade and Standards: Challenges for Africa, Palgrave Macmillan,
London, pp. 21-42.
Brenton, P., Edwards-Jones, G., Jensen, M.F., 2009. Carbon labelling and low-income country exports: a
review of the development issues. Dev. Policy Rev., 27, 243-267.
BSI, 2008. PAS 2050: Specification for the assessment of the life cycle greenhouse gas emissions of
goods and services. British Standards Institute, London, UK.
Burney, J.A., Davis, S.J., Lobell, D.B., 2010. Greenhouse gas mitigation by agricultural intensification.
Proc. Natl. Acad. Sci. USA, 107, 12052-12057.

Burnham, K., Anderson, D., 1998. Model Selection and Inference: A Practical Information Theoretical
Approach. Springer-Verlag, New York.
Classen, M., Althaus, H., Blaser, S., Doka, G., Jungbluth, N., Tuchschmid, M., 2009. Life Cycle Inventories
of metals. Final report ecoinvent data v2.1 No.10, Swiss Centre for Life Cycle Inventories, Dübendorf,
CH.
De Klein, C., Novoa, R.S.A., Ogle, S., 2006. N2O emissions from managed soils and CO2 emissions from
lime and urea application. In: IPCC Guidelines for National Greenhouse Gas Inventories, Chapter 11,
IGES, Japan.
Dossa, E., Fernandes, E., Reid, W., Ezui, K., 2008. Above- and belowground biomass, nutrient and carbon
stocks contrasting an open-grown and a shaded coffee plantation. Agrofor. Syst., 72, 103-115.
Erickson, H., Keller, M., Davidson, E.A., 2001. Nitrogen oxide fluxes and nitrogen cycling during
postagricultural succession and forest fertilization in the humid tropics. Ecosystems, 4, 67-84.
FAO, 2011. FAOSTAT: Land-use statistics. Food and Agricultural Organization Rome, Italy.
Foresight, 2011. The future of food and farming. Final project report. The Government Office for
Science, London, UK.
Frimpong, K.A., Baggs, E.M., 2010. Do combined applications of crop residues and inorganic fertilizer
lower emission of N2O from soil? Soil Use. Manage., 26, 412-424.
Haggar, J.P., Barrios, M., Bolaños, M., Merlo, M., Moraga, P., Munguia, R., Ponce, A., Romero, S., Soto,
G., Staver, C., Virginio, E. de M. F., 2011. Coffee agroecosystem performance under full sun, shade,
conventional and organic management regimes in Central America. Agrofor. Syst., 82, 285-301.

Hergoualc’h,K.,Skiba, U., Harmand, J., Hénault, C., 2008. Fluxes of greenhouse gases from Andosols
under coffee in monoculture or shaded by Inga densiflora in Costa Rica. Biogeochemistry, 89, 329-345.
International Institute for Environment and Development (IIED), 1997. Unlocking Trade opportunities:
changing consumption and production patterns. Report for the U.N. Department of Policy Coordination
and Sustainable Development, pp. 36.
Matson, P.A., Billow, C., Hall, S., Zachariassen, J., 1996. Fertilisation practices and soil variations control
nitrogen oxide emissions from tropical sugar cane. J. Geophys. Res. 101, 18533-18545.
Millar, N., Baggs, E.M., 2004. Chemical composition, or quality, of agroforestry residues influences N2O
emissions after their addition to soil. Soil Biol. Biochem., 36, 935-943.
Millar, N., Baggs, E.M., 2005. Relationships between N2O emissions and water-soluble C and N contents
of agroforestry residues after their addition to soil. Soil Biol. Biochem., 37, 605-608.
Ministerio de Ambiente y Energia, Costa Rica, 2007. Inventario e informe de gases con efecto
invernadero (GEI). Estrategia Nacional de Cambio Climatico, Ministerio de Ambient y Energia, Costa
Rica.
Mosier, A., Wassmann, R., Verchot, L., King, J., Palm, C., 2004. Methane and nitrogen oxide fluxes in
tropical agricultural soils: sources, sinks and mechanisms. Environ. Dev. Sustain., 6, 11-49.
Muschler, R., 2001. Shade improves coffee quality in a sub-optimal coffee-zone of Costa Rica. Agrofor.
Syst., 51, 131-139.
Nemecek, T., Kägi, T., Blaser, S., 2007. Life cycle inventories of agricultural production systems. Final
report ecoinvent v2.0. 15. Swiss Centre for Life Cycle Inventories, Dübendorf, CH.

Payton, M.E., Greenstone, M.H., Schenker, N., 2003. Overlapping confidence intervals or standard error
intervals: What do they mean in terms of statistical significance? J. Insect Sci., 3:34, Available online:
insectscience.org/3.34.
PCF Pilotprojekt Deutschland, 2008. Case study Tchibo Privat Kaffee Rarity Machare by Tchibo GmbH. 1,
1-60.
Plassmann, K., Norton, A., Attarzadeh, N., Jensen, M.P., Brenton, P., Edwards-Jones, G., 2010.
Methodological complexities of product carbon footprinting: a sensitivity analysis of key variables in a
developing country context. Environ. Sci. Policy, 13, 393-404.
R Development Core Team, 2010. R: a language and environment for statistical computing. R
Foundation for Statistical Computing, Vienna, Austria. URL http://www.R-project.org/.
Röös, E., Sundberg, C., Hansson, P., 2010. Uncertainties in the carbon footprint of food products: a case
study on table potatoes. Int. J. Life Cycle Ass., 15, 478-488.
Segura, M., Kanninen, M., Suarez, D., 2006. Allometric models for estimating aboveground biomass of
shade trees and coffee bushes grown together. Agrofor. Syst., 68, 143-150.
Seneviratne, G., 2000. Litter quality and nitrogen release in tropical agriculture: a synthesis. Biol. Fert.
Soils, 31, 60-64.
Smith, K.A., McTaggart, I.P., Tsuruta, H., 1997. Emissions of N2O and NO associated with nitrogen
fertilization in intensive agriculture, and the potential for mitigation. Soil Use Manage., 13, 296-304.
Smith, K., Crutzen, P., Mosier, A., Winiwarter, W., 2010. The global nitrous oxide budget: a
reassessment, in: Smith, K. (Ed.), Nitrous oxide and climate change. pp. 63-84, Earthscan, London, pp.
63-84.

Smith, P., Martino, D., Cai, Z., Gwary, D., Janzen, H., Kumar, P., McCarl, B., Ogle,S.,O’Mara,F.,Rice,C.,
Scholes, B., Sirotenko, O., 2007. Agriculture, in: Metz, B., Davidson, O.R., Bosch, P.R., Dave, R., Meyer,
L.A. (Eds.) Climate Change 2007: Mitigation. Contribution of Working Group III to the Fourth
Assessment Report of the Intergovernmental Panel on Climate Change. Cambridge University Press,
Cambridge, United Kingdom and New York, pp. 498 - 540.
Smith, P., Gregory, P.J., van Vuuren, D., Obersteiner, M., Havlík, P., Rounsevell, M., Woods, J., Stehfest,
E., Bellarby, J., 2010. Competition for land. Philos. Trans. R. Soc. Lond. B. Biol. Sci., 365, 2941-2957.
Soil Survey Staff, 1999. Soil taxonomy: a basic system of soil classification for making and interpreting
soil surveys. US Department of Agriculture Soil Conservation Service, Washington, DC.
Stehfest, E., Bouwman, L., 2006. N2O and NO emission from agricultural fields and soils under natural
vegetation: summarizing available measurement data and modelling of global annual emissions. Nutr.
Cycl. Agroecosys., 74, 207-228.
Tilman, D., Cassman, K.G., Matson, P.A., Naylor, R., Polasky, S., 2002. Agricultural sustainability and
intensive production practices. Nature, 418, 671-677.
UN, Population Division of the Department of Economic and Social Affairs of the United Nations
Secretariat, 2009. World Population Prospects: The 2008 Revision Population Database, New York, USA.
Verchot, L.V., Brienza Júnior, S., de Oliveira, V.C., Mutegi, J.K., Cattânio, J.H., Davidson, E.A., 2008.
Fluxes of CH4, CO2, NO, and N2O in an improved fallow agroforestry system in eastern Amazonia. Agr.
Ecosyst. Environ., 126, 113-121.
Verchot, L.V., Hutabarat, L., Hairiah, K., van Noordwijk, M., 2006. Nitrogen availability and soil N2O
emissions following conversion of forests to coffee in southern Sumatra. Global Biogeochem. Cy., 20,
GB4008.

World Resources Institute & World Business Council for Sustainable Development, 2009. The
Greenhouse Gas Protocol Initiative: Product Life Cycle Accounting and Reporting Standard. Review
Draft for Stakeholder Advisory Group, Washington, DC