Content uploaded by Johnny Reis
Author content
All content in this area was uploaded by Johnny Reis on Sep 08, 2023
Content may be subject to copyright.
Atmospheric Environment 312 (2023) 120026
Available online 15 August 2023
1352-2310/© 2023 The Authors. Published by Elsevier Ltd. This is an open access article under the CC BY-NC license (http://creativecommons.org/licenses/by-
nc/4.0/).
An exploratory approach to estimate point emission sources
D. Lopes
*
, D. Graça , S. Rafael , M. Rosa , H. Relvas , J. Ferreira , J. Reis , M. Lopes
CESAM & Department of Environment and Planning, University of Aveiro, 3810-193, Aveiro, Portugal
HIGHLIGHTS
•A point emission source inventory was developed.
•Manufactures and construction activities were the main emission sources.
•Country-specic proxies improved the temporal variation of the emissions.
•Improvement of the spatial allocation of emission sources.
•Scalability and replicability of the developed approach in other case studies.
ARTICLE INFO
Keywords:
Big data
High-resolution
Critical air pollutants
Air quality modelling
Emission inventory
ABSTRACT
The current atmospheric emission inventories do not fully meet the spatial and temporal resolution requirements
of air quality modelling applications. Considering Portugal as a case study and focusing on combustion point
emission sources (i.e., public power, reneries, manufacturers, and construction activities), this work proposes a
methodological approach and dataset to estimate anthropogenic emissions suitable for different spatial scales
(from regional to local). The obtained results were similar to the annual values reported by the Portuguese
Environment Agency with the maximum emissions being estimated for manufacturing and construction activ-
ities. No signicant differences were recorded between the temporal proles developed in this and previous
studies. However, the country-specic proxies from the developed database allowed us to better represent the
temporal and spatial patterns of the Portuguese atmospheric emissions. The combination of the BigAir database
with a comprehensive and standardized approach could help policymakers dene mitigation and/or plan
measures to reduce emissions from point sources, support countries worldwide (with a lack of data) to develop
high-resolution emission inventories, and improve the current global and European inventories.
1. Introduction
During the last decades, the importance of air quality modelling tools
has increased, and four main reasons can explain this (Baklanov and
Zhang, 2020): i) increased risks of air pollution for human health, with
the World Health Organization reporting that about 7 million people’s
deaths are anticipated due to exposure to outdoor and indoor air
pollution (WHO, 2021); ii) increased studies of climate change, namely
to understand the interactions between atmospheric pollution and
climate change; iii) better computational capabilities; and iv) a better
understanding of physical and chemical processes in the atmosphere.
By enabling the diagnosis of air quality states and the assessment of
mitigation and/or planning measures to avoid human exposure and its
adverse health impacts, air quality modelling can offer several societal
and economic benets (Relvas and Miranda, 2018). As a result, guar-
anteeing the accuracy of the results of such models is essential to
increasing policymakers’ condence in the robustness of the outcomes
and recommendations.
A set of studies has investigated the potential causes of air quality
model inaccuracies (Baklanov and Zhang, 2020; Holnicki and Nahorski,
2015; Maxim and van der Sluijs, 2011), highlighting several factors that
contribute to these. Among those factors, three categories pop out: i)
input data related to emission inventories, meteorological data, and
boundary conditions; ii) structure of the mathematical model, which
includes simplications and parameterizations of physical and chemical
processes; and iii) numerical approximation scheme.
According to Russell and Dennis (2000), major air quality modelling
uncertainties are due to input data rather than the model itself. Among
* Corresponding author.
E-mail address: diogojlopes@ua.pt (D. Lopes).
Contents lists available at ScienceDirect
Atmospheric Environment
journal homepage: www.elsevier.com/locate/atmosenv
https://doi.org/10.1016/j.atmosenv.2023.120026
Received 24 January 2023; Received in revised form 14 June 2023; Accepted 13 August 2023
Atmospheric Environment 312 (2023) 120026
2
the input data, atmospheric emissions are pointed out as the main source
of uncertainty (Borrego et al., 2016; Dias et al., 2018). As a result, un-
derstanding the spatial and temporal distribution of air pollutants and
greenhouse gases is vital for further advances in air quality modelling.
Several emission inventories, with different calculation approaches,
are available for Europe, namely the EMEP (European Monitoring and
Evaluation Programme) (EMEP, 2017), the E-PRTR (European Pollutant
Release and Transfer Register database) (EEA, 2022), the TNO
(Netherlands Organization for Applied Scientic Research) (Kuenen
et al., 2014), CAMS-REG (Copernicus Atmosphere Monitoring Service)
(Kuenen et al., 2021) and EDGAR (Emission Database for Global At-
mospheric Research) (Crippa et al., 2019). EMEP, TNO, CAMS-REG,and
EDGAR provide annual gridded anthropogenic emissions with a coarse
horizontal resolution (EMEP ≈0.1◦×0.1◦; TNO_MACC-III ≈0.125◦×
0.0625◦; CAMS-REG ≈0.05◦×0.1◦, EDGAR ≈0.1◦×0.1◦) while
E-PRTR only supplies data for the largest sources of emissions from the
industrial sector. These inventories are being widely used in air quality
studies across Europe, namely in Portugal.
However, the available emission data are not as accurate as desired
due to an incorrect magnitude of emissions values related to inadequate
emission factors and activities data (Dias et al., 2018), imprecise emis-
sion locations due to the coarse horizontal resolution, unsuitable tem-
poral (i.e., monthly, weekday, and hourly proles) and speciation
proles applied to the annual atmospheric emission values (Alves et al.,
2015; Li et al., 2019). The uncertainties of emission data are especially
noteworthy and challenging in the case of urban or industrial areas.
These areas are characterized by different emission source categories,
technological characteristics, fuel types (and related fuel parameters),
pollutant composition, and emission intensities (Holnicki and Nahorski,
2015). This implies that the accuracy of emissions data is sensitive to the
activity data and their underlying assumptions, as well as to the meth-
odology used to estimate the emissions. Despite many approaches hav-
ing been proposed to deal with several types of emission sources and
pollutants, the development of high spatiotemporal emissions in-
ventories remains an outstanding major challenge (Shami et al., 2022).
In recent years, there has been a notable increase in the use of big
data for the development of emission inventories, particularly in the
transport sector. Several studies have focused on leveraging big data to
improve the accuracy and resolution of emission estimations. For
instance, Sun et al. (2018) utilized big data analysis to develop an in-
ventory for exhaust emissions in the Qingdao port, enabling the pre-
diction of annual emissions and a better understanding of associated
maritime pollution. Similarly, Deng et al. (2020) and Ma et al. (2022)
employed big data approaches to update vehicle emission inventories,
specically for light-duty passenger vehicles in China. Studies have also
been conducted to develop high-resolution emission inventories by
incorporating big data and updating estimation methods, emission fac-
tors, activity data, and allocation proles (Huang et al., 2021; Lin et al.,
2022).
In this context, two questions need to be explored: How can big data
be used to improve the accuracy of available emission inventories? Can a
singular methodology be developed to estimate emission data suitable
for different spatial scales? Therefore, the main goal of this work is to
propose a method and dataset – the so-called BigAir approach - to es-
timate anthropogenic emissions suitable to different spatial scales (from
regional to local), focused on combustion point sources (i.e., public
power, reneries, manufacturers, and mobile machinery of construction
activities) and, using Portugal as a case study. These industrial activities
are relevant sources of European emissions (Degraeuwe et al., 2020;
Relvas et al., 2022) and previous studies have demonstrated in-
consistencies (Thunis et al., 2021) and overestimation of the developed
inventories (Ferreira et al., 2020; Lopes et al., 2022) mainly due to their
inaccurate spatial and temporal distribution. The main strength of the
developed approach is the integration and harmonization of a tradi-
tional method of emission quantication with updated open-source data
into a unied framework that has wide scalability (from regional to
urban areas) and replicability (application in other countries/regions)
potential.
2. Overview of european emission inventories
Currently, there are ve inventories available that provide atmo-
spheric inventories for public power, reneries, manufacturing, and
construction activities in the European region (Crippa et al., 2019; EEA,
2022; EMEP, 2017; Kuenen et al., 2021, 2014). These inventories have
been developed over the years by several research teams and present
their weaknesses and strengths. A brief description of these inventories,
focused on their main advantages/disadvantages, is presented in this
section.
The EMEP dataset, which has a resolution of 0.1⁰ x 0.1⁰ and yearly
activity data from 1990 to (N-2) years (EMEP, 2017), is one of the oldest
and most widely used atmospheric emission inventories (Russo et al.,
2019). The pollutant emissions and projections are collected by the
Center on Emission Inventories and Projections (CEIP) and include
acidifying air pollutants, heavy metals, particulate matter, and photo-
chemical oxidants (EMEP, 2017). Every four years, each European
country is responsible for reporting its inventory (i.e., projection data,
gridded data, and information on point sources) according to the EMEP
guidelines (EMEP, 2018). However, some European countries do not
report accurate gridded emissions and/or submit incomplete or no data.
Therefore, gap-lling is a key task for both total emissions by activity
(and per country) and their spatial distribution (Kuenen et al., 2022). To
ll these gaps, assumptions are considered based on approaches that
have uncertainties, such as activity data or inaccurate emission factors
(Russo et al., 2019).
In Portugal, the Portuguese Environment Agency (APA) provides the
total (year N-2) (APA, 2022) and by municipality (APA, 2021a) (year
N-3) annual emissions. Public power and renery emissions are esti-
mated based on annual activity from the European Union Emission
Trading System (EU-ETS) (APA, 2022) and then spatially distributed
considering the geographic coordinates of each unit (APA, 2021a). The
dataset used to quantify the total emissions from the industrial sector
(manufactures and construction) is the annual activity from EU-ETS and
the Portuguese energy balance (DGEG, 2020). The industrial emissions
are spatially allocated based on population distribution, fuel sales by the
municipality, and the E-PRTR dataset (APA, 2021a).
E-PRTR provides easily accessible key environmental data for the
main industrial facilities (which do not include all the industrial units) in
European Union Member States (EEA, 2022). Since 2017, the dataset has
been updated every year with annual data (year N-2) reported by about
34000 industrial units covering 65 economic activity sectors (including
public power, reneries, and manufactures) across Europe. Each in-
dustrial facility reports information to its national authority (e.g., in
Portugal, it is the APA) on the amounts of pollutants released into the
atmosphere. The E-PRTR dataset covers 91 key pollutants, including
heavy metals, pesticides, greenhouse gases, and dioxins (EEA, 2022).
CAMS-REG is one of the most recent, detailed, and updated European
inventories. It provides annual emissions with a spatial horizontal res-
olution of 0.05◦×0.1◦between the years 2000 and 2020 (year N-2) for
the critical air pollutants and greenhouse gases. The CAMS-REG in-
ventory is an update of the TNO_MACC-III inventory (Kuenen et al.,
2014). CAMS-REG uses the annual total emissions by sector reported by
the European countries to CEIP (same as EMEP) (Kuenen et al., 2021).
For public power, reneries, and industrial sectors (manufacturing and
construction), the annual emissions are spatially distributed using the
E-PRTR and CORINE land cover datasets (Kuenen et al., 2022). The main
advantages of this inventory compared to EMEP are its completeness,
consistency between years, and the consistent spatial distribution pat-
terns that have been applied (ECCAD, 2022). The CAMS-REG inventory
also provides monthly, weekly, and daily emission proles per pollutant
and country with a horizontal spatial resolution of 0.05◦×0.1◦across
Europe. The monthly, weekly, and daily proles of energy activities
D. Lopes et al.
Atmospheric Environment 312 (2023) 120026
3
were estimated from electricity production statistics, while the monthly
emissions variation of the industrial sectors was based on the Industrial
Production Index, which measures the monthly evolution of the pro-
ductive activity of different industrial branches, including
manufacturing activities (Guevara et al., 2020). Despite the
TNO_MACC-III inventory being outdated, its weekly and daily temporal
proles for the industrial sector are still being used by the CAMS-REG
dataset (Denier van der Gon et al., 2011). It should be noted that the
TNO_MACC-III temporal proles only vary by SNAP (Selected Nomen-
clature for Air Pollution). For the year 2020, Guevara et al. (2022)
developed a European dataset of daily sector-, pollutant-, and
country-dependent emission adjustment factors associated with the
COVID-19 mobility restrictions using energy statistics datasets (herein-
after referred to as CAMS-COVID).
The EDGAR inventory is a global database that provides historical
emission time series and grid maps (with a spatial horizontal resolution
of 0.1◦×0.1◦) for all countries from 1970 to 2018, for both critical air
pollutants and greenhouse gases. A uniform approach is applied
worldwide to calculate the emissions from public power, reneries, and
industries using international statistics and emission factors (Crippa
et al., 2018). The industrial emissions are spatially distributed using
several datasets, which include E-PRTR and in-house EDGAR proxies (e.
g., population) (Janssens-Maenhout et al., 2013). This inventory also
provides monthly, weekly, and hourly temporal proles per country and
IPCC source category between 1980 and 2017 (Crippa et al., 2020).
Table 1 summarizes the temporal coverage and spatial resolution of the
European atmospheric emission inventories as well as indicates the
availability of total emissions and temporal proles by activity.
3. Methodology
This section describes the proposed methodology to estimate atmo-
spheric emissions with high spatial and temporal resolution (Section
3.1), focused on point sources, and presents the approach considered to
perform statistical analysis of the obtained results (Section 3.2).
3.1. BigAir emission approach
In this study, the BigAir approach is proposed to provide atmospheric
emissions with high spatial and temporal resolutions, and it is charac-
terized by three main features: i) scalability, since the methodology can
be adapted to any spatial scale (country, region, city); ii) adaptability,
since the spatial and temporal proxies can be adapted according to the
available information; and iii) evolution, since it is possible to have
constant updates and improvements.
The BigAir approach is composed of two stages and combines a
traditional method of emissions calculation with an innovative way of
gathering and processing data. The rst stage includes the construction
of the BigAir database, which contains information about source loca-
tions (facilities borders and residential areas), activity data (energy
production by fuel and fuel consumption), emission factors by sector
(public power, reneries, manufactures, and construction), fuel type (e.
g., natural gas), spatial proxies (installed power, company share capital,
business volume, and facility area), and temporal proxies (industrial
production index, European proles, holiday periods, and noise levels).
In the second stage, the mathematical equations provided by the Euro-
pean air pollutant emission inventory guidebook (EEA, 2019) were used
within the BigAir datasets to estimate atmospheric emissions with a high
spatial and temporal resolution, focused on the following main air pol-
lutants: PM
10
(particles with an aerodynamic equivalent diameter less
than or equal to 10
μ
m), PM
2.5
(particles with an aerodynamic equiva-
lent diameter less than or equal to 2.5
μ
m), NOx (nitrogen oxides), CO
(carbon monoxide), SOx (sulfur oxides), NH
3
(ammonia), NMVOC
(non-methane volatile organic compounds), and CH
4
(methane). The
hourly atmospheric emissions are presented as lower, mean, and upper
estimates (95% condence interval – CI). The range of the emission
estimates was based on the range of emission factors. For example, the
low estimates were calculated as the product of Ef
f,p
(lower) ×A
u,f,t
. The
BigAir approach is illustrated in Fig. 1. In addition, owcharts of the
methodology considered to quantify the public power, renery,
manufacturing, and construction emissions are presented in Figs. S1–S3
of the supplement material.
The BigAir approach was tested in the Portuguese territory (for the
year 2020) to estimate the combustion atmospheric emissions for public
power, reneries, manufacturing, and mobile machinery of construction
(not including the emissions from the industrial process and fugitive
sources) activities since there are relevant sources of emissions in
Portugal (APA, 2022) and previous air quality modelling applications
demonstrate the need to improve the emission estimations from these
activity sectors (Ferreira et al., 2020; Lopes et al., 2022). It should be
noted that the year 2020 was chosen to test the BigAir approach because
it is the most recent year with all input datasets available. In addition, its
unusual emission pattern due to the COVID pandemic could demon-
strate the accuracy of the developed methodology to be applied in any
study period.
Fig. 2 shows the geographical location of the public power plants (30
facilities), reneries (2 facilities), and manufacturers (719 facilities) in
Portugal’s mainland, Madeira Islands, and the Azores islands. Portugal is
a country whose mainland is located in southwestern Europe, and whose
territory also includes the Atlantic archipelagos of the Madeira (three
islands: Madeira, Funchal, and Desertas) and Azores (nine islands: Flores
and Corvo, to the West; Graciosa, Terceira, S˜
ao Jorge, Pico, and Faial, in
the center; and S˜
ao Miguel, Santa Maria and the Formigas islet, to the
east).
Since the geographical location and consumption of the industrial
units were obtained from several sources of information, the method-
ologies and data sources for each activity sector are fully described in the
next section.
Table 1
Main features of the atmospheric emission inventories available across the European region.
Features EMEP APA E-PRTR CAMS-REG EDGAR TNO_MACC-III
Total ✓ ✓ ✓ ✓ ✓ ✓
Temporal proles Energy – – – Monthly
Weekly
Daily
Monthly
Weekly
Hourly
Monthly
Weekly
Daily
Industries – – – Monthly Weekly
Hourly
Monthly
Weekly
Daily
Temporal coverage 1980–2020 1990–2020 2007–2020 2000–2020 1970–2018 2000–2011
Spatial resolution 0.1◦×0.1◦Municipality Exact geographical location 0.05◦×0.1◦0.1◦×0.1◦0.125◦×0.0625◦
Spatial coverage Europe Portugal Europe Europe Global Europe
✓: data available.
-: no data.
D. Lopes et al.
Atmospheric Environment 312 (2023) 120026
4
3.1.1. Public power
The electric energy production in mainland Portugal is based on
eighteen public power units using fuel oil (power =855 MW), diesel oil
(power =2017 MW), natural gas (power =2317 MW), liquid petroleum
gas (LPG; power =510 MW), biomass (power =133 MW), hard coal
(power =785 MW), and biogas (power =135 MW). Electricity pro-
duction in the Azores and Madeira depends mostly on small and
medium-scale power plant units using imported fuels. Azores include
nine public power units (one facility per island) working on fuel oil
(power =241 MW) and diesel oil (power =260 MW), while Madeira has
three small power units using fuel oil (power =183 MW), diesel oil
(power =183 MW), natural gas (power =109 MW), and LPG (power =
48 MW) (APA, 2022) (Fig. 2 – blue triangles). The atmospheric
emissions from this activity were quantied using equations (1)–(3),
depending on the input data available.
Portugalu,f,t,pg
h=EFf,pg
Gj×Af,tMW
h×SPPP( − ) × 3.6GJ
MW1
Madeirau,f,t,pg
h=EFf,pg
Gj×Af,tMW
h×SPpp( − ) × 3.6GJ
MW2
Azoresu,f,t,pg
h=EFf,pg
Gj×Af,tMW
h×SPPP( − ) × TP ( − )
×3.6GJ
MW3
Fig. 1. The general methodology used to estimate the atmospheric emissions using the BigAir approach.
Fig. 2. The geographical location of the public power plants (blue triangles), reneries (orange squares), and manufacturers (grey circles) in Portugal’s mainland,
Madeira Islands, and Azores islands.
D. Lopes et al.
Atmospheric Environment 312 (2023) 120026
5
where, Portugal
u,f,t,p
, Madeira
u,f,t,p
and Azores
u,f,t,p
are the emitted
pollutant p (e.g., PM
10
) by the fuel f (e.g., biomass and diesel oil) at unit
u at time t. EF (emission factor) is the emitted pollutant p by the fuel f
and A
f,t
is the energy produced by the fuel f at time t. SP
PP
and TP are,
respectively, the considered spatial (i.e., installed power) and temporal
proxy (e.g., hourly energy production).
The energy produced by the Portuguese power plant units was ob-
tained from different data sources. The national energy network (REN)
provided the total hourly energy (MW) produced by the eighteen public
power units located in mainland Portugal using natural gas, biomass,
hard coal, and other thermal energy (fuel oil +diesel oil +LPG +
biogas) (REN, 2022). Due to the unavailability of the data, the total
hourly energy was distributed by public power units according to their
installed power for each fuel type. For example, it was assumed that the
power plant unit with the highest power production using natural gas
has the highest hourly energy production. In addition, hourly informa-
tion of electricity generation data per individual facility (2 in total) was
obtained from the ENTSO-E transparency platform (ENTSO-E, 2023).
For the three power plant units located in Madeira, the Madeira Electric
Lighting Company (EEM) supplied total hourly energy production by
fuel (i.e., thermal fuel and natural gas). The hourly energy production
for each facility was estimated based on their installed power (EEM,
2020). The Electricity Company of the Azores provided the monthly
energy production for each facility by fuel (i.e., diesel, fuel oil, and
biogas). In this sense, due to similar energy needs between Madeira and
the Azores (Pearson correlation =0.874; p-value <0.001; see section
3.2), the hourly energy production prole from EEM was applied to
obtain hourly production datasets for the nine power plant units in the
Azores (EDA, 2020).
3.1.2. Reneries
There were two oil rening plants in operation in mainland Portugal
(in Porto and Sines cities) (Fig. 2 – orange squares). The Porto renery
converts crude oil and other intermediate materials received from the
Sines renery by atmospheric and vacuum distillation, cracking, plat-
forming, and several treatment processes (desulphurization). Oils, lu-
bricants, and aromatics (e.g., benzene, hexane, toluene, and xylene) are
also produced in this facility. The Sines renery also has an extensive
transformation process for crude products after atmospheric and vac-
uum distillation, which are subjected to uid catalytic cracking (FCC),
platforming, hydrocracking, alkylation, and asphalt blowing (APA,
2022). The emitted pollutants from this activity were estimated using
equation (4):
Refineriesu,f,t,pg
h=EFf,pg
Gj×Au,f,t(GJ) × TP ( − ) 4
where, Reneries
u,f,t,p
is the emitted pollutant p (e.g., PM
10
) by the fuel f
(e.g., biomass, and diesel oil) at unit u at time t. EF is the pollutant
emitted p by the fuel f and A
u,f,t
is the consumption of fuel f at time t
(annual data). TP is the temporal proxy considered to obtain hourly
emissions (industrial production index and TNO_MACC-III dataset).
The annual energy consumption for each renery by fuel (i.e., other
gases and natural gas) was obtained from Industrial Reporting under the
Industrial Emissions Directive 2010/75/EU and European Pollutant
Release and Transfer Register (E-PRTR) Regulation (EC) No 166/2006
(EEA, 2022). To obtain the monthly energy consumption, the monthly
average of the Portuguese industrial production index was used (INE,
2020). Due to the lack of country-specic data, the weekly and hourly
temporal proles provided in the TNO_MACC-III dataset for the industry
were considered (Denier van der Gon et al., 2011). It should be noted
that these proles are also considered in the up-to-date CAMS-REG
temporal proles dataset.
3.1.3. Manufactures and construction
The Portuguese production includes iron and steel (Nomenclature for
Reporting – NFR – 1.A.2.a), non-ferrous metals (NFR 1.A.2.b), chemicals
and plastics (1.A.2.c), paper and paper pulp (NFR 1.A.2.d), food pro-
cessing, beverages, and tobacco (NFR 1.A.2.e), non-metallic minerals (1.
A.2.f) and other industries (1.A.2.g) (Fig. 1 – grey circles). The atmo-
spheric emissions covered by the other industries are those resulting
from combustion activities in metal products and machinery, motor
vehicles and other transport equipment, wood and wood products,
textiles, rubber products, mining, and mobile machinery in industry
construction (APA, 2022). The atmospheric emissions from manufac-
turers and construction activities were quantied using equations (5)
and (6).
Manufacuresu,f,t,pg
h=EFf,pg
Gj×Au,f,t(GJ) × SPMC ( − ) × TP ( − ) 5
Constructionu,f,t,pg
h=EFf,pg
Gj×Au,f,t(GJ) × SPMC ( − ) × TP ( − ) 6
where, Manufactures
u,f,t,p
and Construction
u,f,t,p
are the emitted
pollutant p (e.g., PM
10
) by the fuel f at unit u at time t. EF is the pollutant
emitted p by the fuel f and A
f,t
is the consumption of the fuel f at time t.
SP
MC
is the used spatial proxy (area of the facility, share capital of the
company, annual business volume, and residential area), and TP (in-
dustrial production index, TNO_MACC-III dataset, holidays, and noise
law) is the considered temporal proxy.
The annual energy consumption was obtained from E-PRTR datasets
(EEA, 2022) and Portuguese energy balance, which includes the fuel
consumption (>15 different fuels) for the following industries: i)
metallurgical; ii) steel; iii) metal-electro-mechanical; iv) chemical and
plastic; v) paper; vi) food, drinks, and tobacco; vii) ceramics; viii) glass;
ix) cement and lime; x) extraction; xi) textiles; xii) wood; xiii) rubber;
xiv) construction; and xv) other manufacturing industries (DGEG,
2020). The annual values were spatially distributed over the Portuguese
land area according to the industrial area of the facility (i.e., m
2
) (DGT,
2018; OpenStreetMap contributors, 2017), share capital of the company
(i.e.,
€
) (Racius, 2022) and annual business volume (i.e.,
€
) for each
economic activity (i.e., extractive industries, manufacturing industries,
and construction) by municipality (PORDATA, 2020).
The geographical locations of the industrial facilities were obtained
from E-PRTR datasets (EEA, 2022), the Portuguese environmental
licenses database (APA, 2021b), OpenStreetMap database (Open-
StreetMap contributors, 2017) while for the construction activities, the
residential areas from the mainland Portuguese land use data (DGT,
2018), as well as the Urban Atlas (Copernicus, 2021) and Global
(Buchhorn et al., 2020) land cover of the Copernicus program, were
considered.
The hourly energy consumption for manufacturing was estimated
using a similar approach as the reneries (Section 3.1.2). The weekly
and hourly were also obtained from the TNO_MACC-III dataset, while
the monthly proles were based on the monthly Portuguese industrial
production index by activity. This dataset includes the monthly indus-
trial index for twenty-ve types of activities, which comprise, for
example, extractive, food processing, beverages, textile, paper, chemi-
cal, plastic, and metallurgic industries. Regarding the construction ac-
tivities, following the Portuguese noise law, it was assumed that there
was a constant hourly energy consumption during weekdays (except on
national and municipality holidays) between 8 a.m. and 8 p.m.
3.2. Statistical analysis
To assess the signicance of the difference between the obtained
emission proles and the available datasets (Section 4.2), some statis-
tical tests were performed using IBM SPSS statistics software (version
28.0.1.0). The emission datasets were tested for normality and homo-
geneity of variance with, respectively, Shapiro–Wilk and Levene’s tests.
When these criteria were fullled (i.e., normality, and homogeneity),
D. Lopes et al.
Atmospheric Environment 312 (2023) 120026
6
comparisons between emission data were performed by one-way
ANOVA followed by Dunnett’s post hoc test. When the assumptions
for normality and homogeneity were not met, the logarithmic trans-
formation of the variable was performed to proceed with the same
parametric analysis of variance. In this work, a p-value <0.05 was
considered statistically signicant. Statistical relationships were sought
using the parametric Pearson correlation coefcient (including its sta-
tistical signicance).
4. Results and discussion
The results and discussion section are based on the estimated annual
values, INE, 2020, for PM
10
, PM
2.5
, NOx, CO, SOx, NH
3
, NMVOC, and
CH
4
by activity sector (Section 4.1), temporal variation of emissions
(monthly, weekly, and hourly proles) (Section 4.2) and total spatial
distribution by atmospheric pollutant (Sections 4.3 and 4.4).
4.1. Annual emissions
Fig. 3 shows a comparative analysis of the annual atmospheric
emissions (the year 2020) by activity (public power, reneries, manu-
facturers, and construction) for PM
10
, PM
2.5
, NOx, CO, SOx, NH
3
,
NMVOC, and CH
4
estimated by the BigAir approach and reported by the
Portuguese Environment Agency (APA) (APA, 2022) in Portugal (i.e.,
Portuguese mainland, Madeira, and Azores). The comparison was per-
formed with APA results since this is the only dataset that provides the
total emissions for each activity sector under analysis.
The atmospheric emissions values estimated by the BigAir approach
and the Portuguese Environment Agency (APA) were higher for
manufacturing and construction activities. The emissions from these
sectors were on average 120 ±286 (BigAir) and 26 ±53 (APA) times
higher than public power, while compared with the reneries, the re-
sults were higher on average 171 ±148 (BigAir) and 87 ±69 (APA)
times. Lower values were recorded for the NH
3
due to the reduced
number of emission factors available in the EMEP dataset, and since the
NH
3
emissions are not caused by a combustion process; the emissions
result from the incomplete reaction of the NH
3
additive in the
denitrication process (EEA, 2019). For public power, the APA values
were outside the 95% condence interval of the BigAir emissions for
PM
10
(−0.42 kton), PM
2.5
(−0.31 kton), CO (+3.93 kton), NH
3
(+0.02
kton), and NMVOC (+1.62 kton). This means that values outside the
95% condence interval are unlikely to be true values. According to
Fig. 4, biomass fuel was the main source of PM
10
(83%), PM
2.5
(88%),
NMVOC (49%), and CH
4
(85%), while hard coal was responsible for
62% of total NOx and 89% total of SOx from public power. In Portugal,
seven out of thirty public power plants are using biomass as fuel for
energy production (power =133 MW), while hard coal is being used by
two units (power =785 MW).
In 2020, the atmospheric emissions from the Portuguese reneries
were mainly related (>50%) to the use of gas oil (other fuels in Fig. 4).
The exceptions are CO and NMVOC, where natural gas was the main
source of these pollutants (>60%). The emissions reported by APA were
outside the 95% condence interval of the values estimated in this study
for NOx (+0.76 kton), SOx (−5.73 kton), and NMVOC (+0.01 kton).
Since the annual consumption considered in the BigAir and APA ap-
proaches was obtained from the same data source (and thus is the same),
the obtained differences between the public power and renery
computed values by BigAir and APA were related to the application of
emission factors from different databases. In this work, the datasets from
the most recent European air pollution emission inventory guidebook
(EEA, 2019) and the IPCC guidelines for national gas inventories (IPCC
et al., 2006) were considered, while according to the APA submission
report, emission factors from the IPCC (IPCC et al., 2006; 2000, 1997),
EEA (EEA, 2016, 2009, 2002) and the United States Environmental
Protection Agency (USEPA) (USEPA, 1998a, 1998b, 1998c, 1996a,
1996b) were used.
For manufacturers and construction, the APA values were outside the
95% condence interval of the BigAir emissions for SOx (−10.1 kton)
and NH
3
(2.22 kton). The emissions from sulte liqueurs (used by the
Portuguese paper and paper pulp industry) represented more than 20%
of the total emissions from manufacturing and construction activities
(other fuels in Fig. 4). Table 2 presents the annual atmospheric emissions
Fig. 3. Annual atmospheric emissions by activity (public power, reneries,
manufacturers, and construction) estimated by the BigAir approach and APA
for the year 2020.
Fig. 4. The BigAir approach’s emissions share (in percentage) of atmospheric
pollutants (PM
10
, PM
2.5
, NOx, CO, SOx, NH
3
, NMVOC, and CH
4
) for public
power, reneries, manufacturing, and construction, by fuel: biomass, diesel oil,
hard coal, natural gas, and others.
D. Lopes et al.
Atmospheric Environment 312 (2023) 120026
7
(the year 2020) for the manufacturing and construction activities by
NFR (Nomenclature for Reporting) for PM
10
, PM
2.5
, NOx, CO, SOx, NH
3
,
NMVOC, and CH
4
computed under the BigAir approach (including the
95% condence interval) and reported by APA (APA, 2022). APA did not
report atmospheric emissions for the non-ferrous metals industries
(NFR1.A.2. b; IE – included elsewhere). According to the BigAir data-
base (Fig. 1), there are about 20 manufacturing units for non-ferrous
metals.
The highest atmospheric emissions for both methodologies (BigAir
and APA) were reported in the paper and paper pulp industries (NFR 1.
A.2. d). The emission from this NFR was on average 1551 ±6546
(BigAir) and 297 ±1283 (APA) times higher than the remaining NFRs.
The paper and paper pulp industry is tremendously important to the
Portuguese economy (representing around 1.2% of the gross domestic
product - GDP). INE, 2020, Portugal ranked as the third and second
largest European producer, respectively, of chemical pulp (9.3% of the
total European production) and uncoated paper and board (18.1% of the
total European production) (CELPA, 2021).
Analyzing the manufacturing and construction sectors by NFR, it is
possible to verify that the NFR 1.A.2. g (other industries) emissions re-
ported by APA values were for all the pollutants (the grey shadow rows
in Table 2) inside the 95% condence interval of the atmospheric
emissions quantied by the BigAir approach. For the remaining NFRs,
on average, 55% ±34% of the APA values were outside the 95% con-
dence interval of the BigAir emissions (white rows in Table 2). In
addition to the emission factors used by BigAir and APA, these differ-
ences could be explained by the consideration of different assumptions
during the calculation approach. For example, the emissions from NFR
1.A.2. b were considered in an unknown NFR by APA, while BigAir
provided the impact of this NFR (non-ferrous metal industries) on total
emissions. Another example is the NFR 1.A.2. b (non-ferrous metals) or
NFR 1.A.2. g (metal products) activities, where the Portuguese energy
balance provides the annual fuel consumption for the metallurgic in-
dustries. In the BigAir approach, these values were spatially distributed
by metallurgic facilities according to their industrial area, the share
capital of the company, and annual business volume by the municipal-
ity, and then, depending on metal type, linked to the NFR 1.A.2. b (non-
ferrous metals) or NFR 1.A.2. g (metal products). The assumption made
by the APA approach is unknown.
In summary, although the annual emissions estimated by the BigAir
and APA approaches were similar when compared with the total values
for the manufacturing and construction activities (Fig. 3) ue to the un-
availability of data, the required assumptions during the computing
process increase the uncertainty of the result, the more detailed the
analysis of the activity sector (i.e., NFR sectors).
4.2. Emission proles
The BigAir approach provides hourly atmospheric emissions for
public power, reneries, manufacturing, and construction activities. To
compare the obtained results, Fig. 5 shows the monthly, weekly,and
hourly (Fig. 5—c) emission proles computed by BigAir, CAMS-REG,
CAMS-COVID, EDGAR, and TNO_MACC-III. At the European level,
only the CAMS-REG provides monthly (from January to December),
Table 2
Annual atmospheric emissions by NFR (1.A.2.a, 1.A.2.b, 1.A.2.c, 1.A.2.d, 1.A.2.e, 1.A.2.f, and 1.A.2.g)
estimated by the BigAir approach and APA for the year 2020. Grey cells represent the APA values inside the
95% condence interval of the atmospheric emissions quantied in this study.
D. Lopes et al.
Atmospheric Environment 312 (2023) 120026
8
weekly (from Monday to Sunday), and hourly (from 1 a.m. to 24 a.m.)
emission proles for the year 2020 by pollutants with a horizontal
spatial resolution of 0.1◦×0.1◦(Guevara et al., 2020). It should be
noted that for the BigAir the monthly industries prole is the monthly
average of the Portuguese industrial production, while for the
CAMS-REG, the results shown in Fig. 4 resulted in an average pollutant
prole.
For the public power activity, no signicant differences (p-value
>0.05) were observed between the monthly, weekly, and hourly proles
developed by BigAir and the compared datasets (i.e., CAMS-REG, CAMS-
COVID, EDGAR, and TNO_MACC-III) (Fig. 5). The BigAir quantied that
the highest values were recorded in September (Public Power =1.50)
while for CAMS-REG (Public Power =1.16), CAMS-COVID (Public
Power =1.18), and EDGAR (Public Power =1.21), the maximum energy
production was estimated in July. On the other hand, with the exception
of the TNO_MACC-III prole, the analyzed approaches estimated that
the lowest emissions from the energy sector were registered in April
(prole between 0.60 and 0.84) when the Portuguese state of emergency
caused by the COVID pandemic situation was in effect (between March
19 and May 3, 2020) (Decree-Law n
o
10-A/2020, 2020; Decree-Law n
o
20/2020, 2020). During this period, the Portuguese population stayed at
home, and only basic services (e.g., hospitals, health care centers,
pharmacies, and supermarkets) were allowed to remain open. The
proles estimated that public power emissions were higher on weekdays
(BigAir =1.05 ±0.04; CAMS-REG =1.04 ±0.01; CAMS-COVID =1.04
±0.01; EDGAR =1.05 ±0.01; TNO_MACC-III =1.06 ±0.0) than on
weekends (BigAir =0.88 ±0.05; CAMS-REG =0.89 ±0.06; CAMS--
COVID =0.89 ±0.06; EDGAR =0.88 ±0.05; TNO_MACC-III =0.85 ±
0.0). The hourly variation estimated in the BigAir approach was similar
than CAMS-REG (r =0.937; p-value <0.001), EDGAR (r =0.875;
p-value <0.001), and TNO_MACC-III (r =0.875; p-value <0.001) pro-
les. The atmospheric emissions rose in the early morning (between 4 a.
m. and 5 a.m.) in all analyzed proles, with their maximum value
reaching around 8 pm in the BigAir and CAMS-REG and at 10 a.m. in the
EDGAR and TNO_MACC-III proles. The statistical analysis revealed no
signicant differences (p-value ≥0.453) between the manufacturers’
monthly proles estimated by the BigAir approach and compared ap-
proaches (i.e., CAMS-REG, CAMS-COVID, and TNO_MACC-III). The
BigAir (Manufactures =0.71) and CAMS-COVID (Manufactures =0.77)
approaches quantied the lowest manufacturer emissions in April dur-
ing the Portuguese State of Emergency, while for CAMS-REG (Manu-
factures =0.87) and TNO_MACC-III (Manufactures =0.90), the
minimum value was estimated, respectively, in August (when several
facilities are closed due to the European holiday period) and June. It
should be noted that the BigAir approach also recorded an emission
reduction during the European holiday period (Manufactures =0.87).
For the weekly prole, the EDGAR and TNO_MACC-III approaches
assumed a at distribution during the working days (Manufactures =
Fig. 5. Monthly, weekly,and hourly emission proles computed by BigAir, CAMS-REG, CAMS-COVID, EDGAR, and TNO_MACC-III for the public power, reneries,
manufacturers, and construction activities.
D. Lopes et al.
Atmospheric Environment 312 (2023) 120026
9
1.08) and a slight decrease during weekends (Manufactures =0.8) while
the hourly prole included an increase in the emissions during the
central hours of the day (working hours period).
For the reneries, the monthly prole estimated by the BigAir
approach (Fig. 5 – Reneries) is equal to the one presented for the
manufacturers (Fig. 5 – Manufactures) since for this activity the monthly
average of the Portuguese industrial production index was used. The
EDGAR considered a weekly and hourly at distribution, while the
TNO_MACC-III assumed weekly and hourly proles equal to the
manufacturers.
Regarding the construction, its proles were related to the Portu-
guese holiday period and noise law regulations. The lowest emissions
were recorded in June (the month with the largest popular festivals in
the country), and no emissions were estimated during the weekend and
resting hours (between 8 p.m. and 8 a.m.).
4.3. Spatial distribution of emissions
To compare the obtained horizontal spatial distribution with data
from other approaches, Fig. 6 shows the total annual average normal-
ized differences (see Textbox S2 of the supplemental material) between
the atmospheric emissions estimated in BigAir and CAMS-REG (public
power, reneries, and manufacturing emissions) (Kuenen et al., 2021).
To facilitate the comparison between the obtained results using the
different approaches, Fig. 6 presents the differences using the horizontal
spatial resolution provided by CAMS-REG (0.05◦×0.1◦). The normal-
ized differences by pollutant and computed standard deviation (STD)
are also shown in the supporting material section (Figs. S4–S12). It
should be noted that for the manufacturing activities, the BigAir
approach only quantied the combustion emissions (i.e., NFR 1.A.2)
while the CAMS-REG provides emissions for the combustion (i.e., NFR 1.
A.2) and the process itself (i.e., NFR 2 – industrial processes).
The BigAir emissions are higher than CAMS-REG values in 30% of
the cells, meaning that the CAMS-REG inventory distributed the public
power, reneries, and manufacturing emissions over a larger area due to
the considered proxies (Fig. 6). According to Kuenen et al. (2022), the
E-PRTR datasets, a database developed by the CAMS-REG research
team, and outdated CORINE land cover (for the year 2012), were used to
spatially distribute the annual public power and industrial emissions
with a horizontal spatial distribution of 0.05◦by 0.1◦.
For the Azores archipelago, the CAMS-REG European inventory only
provides atmospheric emissions from the public power unit located at
S˜
ao Miguel Island (slightly red color in Fig. 6). According to the BigAir
database, there are nine power units (one per island) (EDA, 2020) and
several other facilities that include chemical, plastic, food, drinks, to-
bacco, ceramics, cement, extraction, and other manufacturing industries
(DGEG, 2020). In addition, it should be noted that the domain of the
CAMS-REG European inventory does not include values for the two
Azores islands located further west (i.e., Flores and Corvo islands).
Regarding the Madeira archipelago, the CAMS-REG inventory considers
both public power and manufacturing emissions; however, it comprises
emissions for the same cells for both activities. For example, the BigAir
approach includes public power emissions in 3 cells (there are 3 public
power units in Madeira), while in CAMS-REG, there are values in 9 cells
(as in manufacturing emissions).
The standard deviation of the normalized differences between the
BigAir and CAMS-REG approaches ranged from 0 to 0.39, with the
highest value recorded in the southwest of mainland Portugal (Fig. S12
of the supplement material), where the difference varied between −0.17
(NH
3
, Fig. S9 of the supplement material) and +1 (PM10, PM2.5, and
CO; Figs. S4, S5, and S7 of the supplement material). The maximum
average difference was also veried at this location (0.61, Fig. 6), where
there are key industrial hubs with two public powers, three industrial
plants for the production of paper and pulp, as well as ve chemical
facilities.
4.4. BigAir approach potentiality
The inventory developed in this work provides hourly atmospheric
emissions for the selected horizontal spatial resolutions, being able to
reduce the uncertainty of the air quality modelling applications. Fig. 7
shows two examples of high atmospheric emission resolutions (0.005◦
and 0.01◦) that the BigAir approach could provide for the annual PM
10
emissions in the Porto region (the remaining atmospheric pollutants are
presented in Fig. S13 of the supplement material). The Porto region was
selected as a case study since this area comprises emissions from all the
analyzed activities (public power, reneries, and manufacturers).
Fig. 7 shows the capability of the developed inventory to provide a
high spatial resolution inventory in the selected region, unlike CAMS-
REG, where the annual total emissions were distributed in areas
where there are no public power, reneries, or manufacturing activities.
In the BigAir approach, the highest emissions (except for NOx, and NH
3
)
were recorded in the south of the domain, while in CAMS-REG, the
location of the maximum values varied with the pollutant type (Fig. 7
and Fig. S13 of the supplement material). According to Table 3 (BigAir
approach), the paper and paper pulp industry (1.A.2. d) was the main
Fig. 6. Total annual average normalized differences (public power, reneries, and manufacturing emissions) between BigAir and CAMS-REG approaches in Madeira,
Azores, and mainland Portugal.
D. Lopes et al.
Atmospheric Environment 312 (2023) 120026
10
source of emissions in the Porto region (on average, 49% ±27% of the
total emissions). The second most relevant activity sector was other
industries (which include wood and textile products), followed by non-
metallic minerals (on average, 8.7% ±3.2% of total emissions). For the
remaining activities, the emission contribution to the total annual
emissions in the Porto region ranged from <0.01% (NH
3
emitted in the
reneries, iron, and steel industry) to 18.1% (NH3 emissions from food
and beverages).
5. How does BigAir approach help to improve emission
inventories?
The developed BigAir method and dataset can improve the accuracy
of the atmospheric emission inventories for public power, reneries,
manufacturing, and construction activities. These improvements could
be carried out in the global, European, and country-level inventories.
The analyses performed in this work identied that the up-to-date
European emission inventory did not include emission values for the
two European Island located further west (i.e., Flores and Corvo islands),
and inaccurate spatial and temporal distributions of the emissions were
veried that could be suppressed using the BigAir method and dataset.
For example, the CAMS-REG inventory provided atmospheric emissions
from one out of nine public power units located in the Azores archi-
pelago, and its inventory did not register the lowest industrial emissions
during the Portuguese State of Emergency as expected according to the
Portuguese production index dataset.
The replicability nature of the BigAir approach/guidelines could
help worldwide countries improve their atmospheric inventories as well
as support the Portuguese Environment Agency in improving the annual
reporting of atmospheric emissions under the National Emission
Reduction Commitments (NEC) Directive 2016/2284 and the UNECE
(United Nations Economic Commission for Europe) convention on long-
range transboundary air pollution. For example, the use of outdated
emission factors and assumptions (due to incomplete input datasets)
during the computing process increases the uncertainty of the reported
emissions.
The scalability of the BigAir approach could assist countries world-
wide (with a lack of data) to develop high-resolution gridded emissions.
The BigAir approach intends to use open-source datasets that could be
easily obtained. For example, in the industrial sector, the location of the
facilities was obtained using the E-PRTR datasets (EEA, 2022), the
Portuguese environmental licenses database (APA, 2021b), and the
OpenStreetMap database (OpenStreetMap contributors, 2017); while for
the construction activities, the residential areas with high spatial reso-
lution from the mainland Portuguese land use data (DGT, 2018), Urban
Atlas (Copernicus, 2021) and Global (Buchhorn et al., 2020) land cover
of the Copernicus program were considered. Regarding temporal pro-
les, the construction sector is a clear example of how its temporal
Fig. 7. Annual spatial distribution of PM
10
emissions (kton) in the Porto region with a horizontal spatial resolution of 0.005◦×0.005◦(left panel, BigAir approach),
0.01◦×0.01◦(middle panel, BigAir approach), and 0.1◦×0.05◦(right panel, CAMS-REG approach).
Table 3
Annual atmospheric emissions (kton and %) in the Porto region for the public power, reneries, and manufacturing activities obtained from the BigAir approach.
NFR PM
10
Pollutants (kton)
PM
2.5
NOx CO SOx NH
3
NMVOC CH
4
Public power
1.2.1.a 0.0054 (0.43%) 0.0045 (0.39%) 0.2475 (4.52%) 0.0812 (0.87%) 0.0698 (1.00%) 0.0003 (0.39%) 0.0060 (0.31%) 0.0027 (0.94%)
Reneries
1.2.1.b 0.0180 (1.46%) 0.0108 (0.93%) 0.7871 (14.4%) 0.4184 (4.47%) 0.1401 (2.00%) 0 (0%) 0.0262 (1.36%) 0.0124 (4.25%)
Manufacturers
1.A.2.a 0.0061 (0.50%) 0.0057 (0.50%) 0.0647 (1.18%) 0.0610 (0.65%) 0.0416 (0.59%) 0 (0%) 0.0188 (0.98%) 0.0007 (0.23%)
1.A.2.b 0.0004 (0.03%) 0.0004 (0.04%) 0.0255 (0.47%) 0.0087 (0.09%) 0.0006 (0.01%) <0.0001 (<0.01%) 0.0066 (0.35%) 0.0003 (0.11%)
1.A.2.c 0.0332 (2.69%) 0.0324 (2.80%) 0.6316 (11.5%) 0.2549 (2.73%) 0.0947 (1.35%) 0.0025 (3.30%) 0.1315 (6.83%) 0.0018 (0.61%)
1.A.2.d 0.7458 (62.3%) 0.6886 (59.6%) 1.3048 (23.9%) 5.9974 (64.1%) 5.7218 (81.8%) 0 (0%) 0.6282 (32.6%) 0.1945 (66.6%)
1.A.2.e 0.0623 (5.04%) 0.0612 (5.29%) 0.4381 (8.01%) 0.3264 (3.49%) 0.0319 (0.46%) 0.0138 (18.1%) 0.1932 (10.0%) 0.0153 (5.26%)
1.A.2.f 0.1249 (10.1%) 0.1170 (10.1%) 0.5703 (10.4%) 0.9548 (10.2%) 0.7765 (11.1%) 0.0044 (5.81%) 0.1859 (9.65%) 0.0057 (1.94%)
1.A.2.g 0.2403 (19.4%) 0.2357 (20.4%) 1.4013 (25.6%) 1.2494 (13.3%) 0.1212 (1.73%) 0.0550 (72.4%) 0.7294 (37.9%) 0.0585 (20.0%)
1.2.1.a =public power.
1.2.1.b =reneries.
1.A.2.a =iron and steel.
1.A.2.b =non-ferrous metals.
1.A.2.c =chemical and plastic.
1.A.2.d =paper and paper pulp.
1.A.2.e =food processing, beverages, and tobacco.
1.A.2.f =non-metallic minerals.
1.A.2.g =other industries.
D. Lopes et al.
Atmospheric Environment 312 (2023) 120026
11
emissions variability could be improved using international laws (i.e.,
noise law). In addition, for the European countries without datasets to
improve their temporal proles, the BigAir database could be used as a
proxy since the temporal patterns of the European countries’ production
are similar due to their internal market.
6. Limitations and further improvements
Despite the potential of the obtained atmospheric emission inventory
using the BigAir approach, several limitations (mainly related to a lack
of data) and further improvements were identied.
The lack of information is mainly related to the unavailability of
hourly activity data for industrial facilities. For public power, it is only
available for the total hourly energy production (by fuel) for mainland
Portugal and Madeira (the exception is the data provided by the ENTSO-
E transparency platform, where hourly information of electricity gen-
eration data for two facilities in mainland Portugal was considered). To
obtain the hourly energy production by units, the installed power by fuel
in each facility was considered. However, for example, this proxy means
that the unit with the highest installed power using natural gas will al-
ways record the maximum atmospheric emissions using this fuel. Since
the energy production by a unit depends not only on the installed power
but also on the hourly energy needs in each of the regions (data not
available), this methodology could provide inaccuracy hourly values for
public power units.
Regarding the reneries and manufacturers, the monthly Portuguese
industrial production index was used, and due to the lack of country-
specic data, the weekly and hourly temporal proles provided in the
TNO_MACC-III dataset for industries were considered. In the future,
hourly consumption for different industrial activities (e.g., reneries)
should be requested to improve the accuracy of the BigAir approach.
In Portugal, there is no data available regarding the locations and
periods (beginning and ending date) of the construction activities. For
the manufacturing industries, the annual values were spatially distrib-
uted over the Portuguese land area according to the industrial area, the
share capital of the company, and the annual business volume by the
municipality. Since there are annual business volumes for a scarce type
of activity (i.e., manufacturing industries, construction, and extractive
activities), this means that a highly industrialized municipality will have
larger atmospheric emissions than a municipality with a key industry for
a specic type of production. For example, municipality 1 includes
several small wood industries, while municipality 2 comprises one key
wood industrial unit. However, since municipality 1 is a highly indus-
trialized area when compared with municipality 2, the atmospheric
emissions from wood industries will be higher in municipality 1 than in
municipality 2. In this sense, new win-win collaborations (e.g., with the
Portuguese Environment Agency) should be created for the continuous
development of the dataset.
For manufacturing activities, the emission factors could be another
source of uncertainty since the most recent database from the European
Environment Agency air pollution emission inventory guidebook (EEA,
2019) provides globally equal emission factors for several types of in-
dustrial activities. For example, the PM
10
(117 g/GJ), PM
2.5
(108 g/GJ),
NOx (173 g/GJ), CO (931 g/GJ), SOx (900 g/GJ), and NMVOC (88.8
g/GJ) emission factors for solid fuels are the same, although particular
production processes by industrial activity are applied (e.g., paper and
textile production) and, therefore, the emitted air pollutants should be
different.
7. Conclusion
The main goal of this work was to provide a methodological
approach and dataset to estimate anthropogenic emissions suitable to
different spatial scales), while, at the same time, minimizing the un-
certainties of emission inventories. Focused on combustion point sour-
ces (i.e., public power, reneries, manufacturers, and construction
activities), and using Portugal as a case study, this work corresponds to
the rst phase of the BigAir project.
The main outcomes of the current research are summarized as
follows.
•The highest Portuguese anthropogenic emissions were estimated for
manufacturing and construction activities. The obtained results were
similar to the annual values reported by APA. However, due to the
unavailability of data, the required assumptions during the
computing process increase the uncertainty of the result, the more
detailed the analysis of the activity sector;
•For public power, biomass fuel was the main source of atmospheric
emissions. The values from reneries were mainly related to the use
of gas oil, while for manufacturing and construction activities, the
sulte liqueurs used by the paper and pulp industry were the main
sources.
•No signicant differences were observed between the temporal
proles developed by BigAir and CAMS-REG. However, the country-
specic proxies from the BigAir database allowed us to better
represent the temporal variation of the Portuguese atmospheric
emissions.
•The BigAir emissions are higher than CAMS-REG values in 30% of
the cells, meaning that the CAMS-REG inventory distributed the
public power, reneries, and manufacturing emissions over a larger
area due to the coarse and outdated proxies. The developed multi-
scale inventory demonstrated a high capability to provide high
spatial resolutions in urban areas.
The combination of the BigAir database with a comprehensive and
standardized approach to emission estimation could help policymakers
dene mitigation and/or plan measures to reduce emissions from point
sources, support countries worldwide (with a lack of data) to develop
high-resolution emission inventories, and improve the current global
and European inventories. Further steps include the development of
methodologies to estimate the atmospheric emissions from other key
activities (e.g., land transport and residential sectors), as well as the
development of PM and VOC speciation proles to assess the air quality
modelling performance (i.e., Eulerian, Gaussian, and Lagrangian ap-
proaches) from regional to local scales using the developed BigAir
datasets.
CRediT authorship contribution statement
D. Lopes: Funding acquisition, Methodology, Data curation, Visu-
alization, Writing – original draft. D. Graça: Methodology, Data cura-
tion. S. Rafael: Formal analysis, Writing – original draft. M. Rosa:
Writing – review & editing. H. Relvas: Resources, Writing – review &
editing. J. Ferreira: Supervision, Writing – review & editing. J. Reis:
Resources. M. Lopes: Funding acquisition, Supervision.
Declaration of competing interest
The authors declare that they have no known competing nancial
interests or personal relationships that could have appeared to inuence
the work reported in this paper.
Data availability
The BigAir database is available on zenodo platform (https://doi.
org/10.5281/zenodo.7653237).
Acknowledgements
This work was nancially supported by the project “BigAir - Big data
to improve atmospheric emission inventories”, PTDC/EAM-AMB/2606/
2020, funded by national funds through FCT - Foundation for Science
D. Lopes et al.
Atmospheric Environment 312 (2023) 120026
12
and Technology. We acknowledge nancial support to CESAM by FCT/
MCTES (UIDP/50017/2020+UIDB/50017/2020+LA/P/0094/2020),
through national funds. Thanks are due for the nancial support to the
PhD grant of Daniel Graça (2022.11105. BD). Thanks are due to FCT/
MCTES for the contracts granted to Joana Ferreira (2020.00622. CEE-
CIND), and H´
elder Relvas (2021.00185. CEECIND).
Appendix A. Supplementary data
Supplementary data to this article can be found online at https://doi.
org/10.1016/j.atmosenv.2023.120026.
References
Alves, C.A., Lopes, D.J., Calvo, A.I., Evtyugina, M., Rocha, S., Nunes, T., 2015. Emissions
from light-duty diesel and gasoline in-use vehicles measured on chassis
dynamometer test cycles. Aerosol Air Qual. Res. 15, 99–116. https://doi.org/
10.4209/aaqr.2014.01.0006.
APA, (Agˆ
encia Portuguesa do Ambiente), 2021a. Emiss˜
oes de poluentes atmosf´
ericos por
concelho - 2015, 2017 e 2019. Amadora.
APA, (Agˆ
encia Portuguesa do Ambiente), 2021b. Licenças ambientais emitidas [WWW
Document]. URL. https://ladigital.apambiente.pt/, 6, 6.
APA, (Agˆ
encia Portuguesa do Ambiente), 2022. National Informative Inventory Report
2022 Portugal (Amadora).
Baklanov, A., Zhang, Y., 2020. Advances in air quality modeling and forecasting. Glob.
Transitions 2, 261–270. https://doi.org/10.1016/j.glt.2020.11.001.
Borrego, C., Amorim, J.H., Tchepel, O., Dias, D., Rafael, S., S´
a, E., Pimentel, C.,
Fontes, T., Fernandes, P., Pereira, S.R., Bandeira, J.M., Coelho, M.C., 2016. Urban
scale air quality modelling using detailed trafc emissions estimates. Atmos.
Environ. 131, 341–351. https://doi.org/10.1016/j.atmosenv.2016.02.017.
Buchhorn, M., Smets, B., Bertels, L., Lesiv, M., Tsendbazar, N.-E., Masiliunas, D.,
Linlin, L., Herold, M., Fritz, S., 2020. Copernicus Global Land Service: Land Cover
100m: Collection 3: Epoch 2019. https://doi.org/10.5281/zenodo.3939050. Globe
(Version V3.0.1) [Data set]. Zenodo.
Celpa, 2021. Boletim Estatístico da CELPA 2020 (Lisboa, Portugal).
contributors, OpenStreetMap, 2017. Planet dump [Data le from $date of database dump
$]. Retrieved from. https://planet.openstreetmap.org ([WWW Document]).
Copernicus, 2021. Urban Atlas 2018 [WWW Document]. URL. https://land.copernicus.
eu/local/urban-atlas/urban-atlas-2018?tab=download, 4.21.21.
Crippa, M., Guizzardi, D., Muntean, M., Schaaf, E., Dentener, F., van Aardenne, J.A.,
Monni, S., Doering, U., Olivier, J.G.J., Pagliari, V., Janssens-Maenhout, G., 2018.
Gridded emissions of air pollutants for the period 1970–2012 within EDGAR v4.3.2.
Earth Syst. Sci. Data 10, 1987–2013. https://doi.org/10.5194/essd-10-1987-2018.
Crippa, M., Oreggioni, G., Guizzardi, D., Muntean, M., Schaaf, E., Lo Vullo, E.,
Solazzo, E., Monforti-Ferrario, F., Olivier, J.G., Vignati, E., 2019. Fossil CO2 and
GHG Emissions of All World Countries - 2019 Report Publications Ofce of the EU,
JRC Science for Policy Report. https://doi.org/10.2760/687800.
Crippa, M., Solazzo, E., Huang, G., Guizzardi, D., Kof, E., Muntean, M., Schieberle, C.,
Friedrich, R., Janssens-Maenhout, G., 2020. High resolution temporal proles in the
emissions database for global atmospheric research. Sci. Data 7, 1–17. https://doi.
org/10.1038/s41597-020-0462-2.
Decree Law n
o
10-A/2020, 2020. Decreto-Lei n.o 10-A/2020, de 13 de março [WWW
Document]. URL. https://dre.pt/dre/detalhe/decreto-lei/10-a-2020-130243053,
9.14.22.
Decree Law n
o
20/2020, 2020. Decreto-Lei n.o 20/2020, de 1 de maio [WWW
Document]. URL. https://dre.pt/dre/detalhe/decreto-lei/20-2020-132883356,
9.14.22.
Degraeuwe, B., Pisoni, E., Thunis, P., 2020. Prioritising the sources of pollution in
European cities: do air quality modelling applications provide consistent responses?
Geosci. Model Dev 13, 5725–5736. https://doi.org/10.5194/gmd-13-5725-2020.
Deng, F., Lv, Z., Qi, L., Wang, X., Shi, M., Liu, H., 2020. A big data approach to improving
the vehicle emission inventory in China. Nat. Commun. 11, 1–12. https://doi.org/
10.1038/s41467-020-16579-w.
Denier van der Gon, H., Hendriks, C., Kuenen, J., Segers, A., Visschedijk, A., 2011. TNO
Report Description of Current Temporal Emission Patterns and Sensitivity of
Predicted AQ for Temporal Emission Patterns (Utrecht).
DGEG, 2020. (Direcç˜
ao-Geral de Energia e Geologia). In: Balanços Energ´
eticos - 2020
[WWW Document]. URL. https://www.dgeg.gov.pt/pt/estatistica/energia/balancos
-energeticos/, 5.18.22.
DGT, (Direç˜
ao Geral do Territ´
orio), 2018. Cartograa de Uso e Ocupaç˜
ao do Solo (COS,
CLC e Copernicus) [WWW Document]. URL. https://snig.dgterritorio.gov.
pt/rndg/srv/por/catalog.search#/metadata/b498e89c-1093-4793-ad22-635
16062891b, 10.19.18.
Dias, D., Amorim, J.H., S´
a, E., Borrego, C., Fontes, T., Fernandes, P., Pereira, S.R.,
Bandeira, J., Coelho, M.C., Tchepel, O., 2018. Assessing the importance of
transportation activity data for urban emission inventories. Transport. Res.
Transport Environ. 62, 27–35. https://doi.org/10.1016/j.trd.2018.01.027.
ECCAD, 2022. Emissions of Atmospheric Compounds and Compilation of Ancillary Data
[WWW Document]. URL. https://eccad.aeris-data.fr/, 9.30.22.
EDA, (Electricidade dos Açores), 2020. Produç˜
ao e Consumo - 2020 [WWW Document].
URL. https://www.eda.pt/, 3.18.22.
EEA, (European Environmental Agency), 2002. EMEP/CORINAIR Emission Inventory
Guidebook, third ed. Environment European Agency. October 2002 UPDATE
Technical report No 30.
EEA, (European Environmental Agency), 2009. EMEP/EEA air pollutant emission
inventory guidebook 2009. In: Technical Guidance to Prepare National Emission
Inventories. https://doi.org/10.2800/23924.
EEA, (European Environmental Agency), 2016. EMEP/EEA air pollutant emission
inventory guidebook 2016. In: Technical Guidance to Prepare National Emission
Inventories. https://doi.org/10.2800/247535.
EEA, (European Environmental Agency), 2019. EMEP/EEA air pollutant emission
inventory guidebook 2019. In: Technical Guidance to Prepare National Emission
Inventories. https://doi.org/10.2800/293657.
EEA, (European Environmental Agency), 2022. Industrial Reporting under the Industrial
Emissions Directive 2010/75/EU and European Pollutant Release and Transfer
Register Regulation. EC) No 166/2006 [WWW Document]. URL. https://www.eea.
europa.eu/data-and-maps/data/industrial-reporting-under-the-industrial-6,
5.20.22.
EEM, (Empresa de Eletricidade da Madeira), 2020. Mix de produç˜
ao - 2020 [WWW
Document]. URL. https://www.eem.pt/pt/gracos/, 6, 2.
Emep, 2017. European Monitoring and Evaluation Programme). European Monitoring
and Evaluation Programme [WWW Document]. URL. http://www.emep.int/,
12.9.17.
EMEP, 2018. European monitoring and evaluation Programme). In: Transboundary
Particulate Matter, Photo-Oxidants, Acidifying and Eutrophying Components. EMEP
Status Report 1/2018.
Entso-E, 2023. ENTSO-E transparency platform [WWW Document]. URL. https://tran
sparency.entsoe.eu/, 5, 8.
Ferreira, J., Lopes, D., Rafael, S., Relvas, H., Almeida, S.M., Miranda, A.I., 2020.
Modelling air quality levels of regulated metals: limitations and challenges. Environ.
Sci. Pollut. Res. 27, 33916–33928. https://doi.org/10.1007/s11356-020-09645-9.
Guevara, M., Jorba, O., Tena, C., Gon, H.D. Van Der, Kuenen, J., Elguindi, N., Darras, S.,
Granier, C., Studies, A., 2020. CAMS-TEMPO: Global and European Emission
Temporal Prole Maps for Atmospheric Chemistry Modelling. https://doi.org/
10.5194/essd-2020-175.
Guevara, M., Petetin, H., Jorba, O., Denier Van Der Gon, H., Kuenen, J., Super, I.,
Jalkanen, J.P., Majamaki, E., Johansson, L., Peuch, V.H., Perez Garcia-Pando, C.,
2022. European primary emissions of criteria pollutants and greenhouse gases in
2020 modulated by the COVID-19 pandemic disruptions. Earth Syst. Sci. Data 14,
2521–2552. https://doi.org/10.5194/essd-14-2521-2022.
Holnicki, P., Nahorski, Z., 2015. Emission data uncertainty in urban air quality
modeling—case study. Environ. Model. Assess. 20, 583–597. https://doi.org/
10.1007/s10666-015-9445-7.
Huang, Z., Zhong, Z., Sha, Q., Xu, Y., Zhang, Z., Wu, L., Wang, Y., Zhang, L., Cui, X.,
Tang, M.S., Shi, B., Zheng, C., Li, Z., Hu, M., Bi, L., Zheng, J., Yan, M., 2021. An
updated model-ready emission inventory for Guangdong Province by incorporating
big data and mapping onto multiple chemical mechanisms. Sci. Total Environ. 769,
144535 https://doi.org/10.1016/j.scitotenv.2020.144535.
INE, (Instituto Nacional de Estatística), 2020. ´
Indices de Produç˜
ao Industrial [WWW
Document]. URL. https://www.ine.pt, 9.14.22.
IPCC, (Intergovernmental Panel on Climate Change), 1997. Revised 1996 IPCC
Guidelines for National Greenhouse Gas Inventories. In: Workbook. The
Intergovernmental Panel on Climate Change (IPCC), ume 2. the Organization for
Economic Co-operation and Development (OECD) and the International Energy
Agency (IEA).
IPCC, (Intergovernmental Panel on Climate Change), 2000. Good practice guidance and
uncertainty management in national greenhouse gas inventories. In: The
Intergovernmental Panel on Climate Change (IPCC).
IPCC, (Intergovernmental Panel on Climate Change), 2006. In: Eggleston, H.S.,
Buendia, L., Miwa, K., Ngara, T., Tanabe, K. (Eds.), 2006 IPCC Guidelines for
National Greenhouse Gas Inventories, Prepared by the National Greenhouse Gas
Inventories Programme. IGES, Published (Japan).
Janssens-Maenhout, G., Pagliari, V., Guizzardi, D., Muntean, M., 2013. Global Emission
Inventories in the Emission Database for Global Atmospheric Research (EDGAR)–
Manual (I), Gridding: EDGAR Emissions Distribution on Global Gridmaps. https://
doi.org/10.2788/81454.
Kuenen, J.J.P., Visschedijk, A.J.H., Jozwicka, M., Denier Van Der Gon, H.A.C., 2014.
TNO-MACC-II emission inventory; A multi-year (2003-2009) consistent high-
resolution European emission inventory for air quality modelling. Atmos. Chem.
Phys. 14, 10963–10976. https://doi.org/10.5194/acp-14-10963-2014.
Kuenen, J., Dellaert, S., Visschedijk, A., Jalkanen, J.-P., Super, I., Denier van der Gon, H.,
2021. Copernicus Atmosphere Monitoring Service Regional Emissions Version 4.2
(CAMS-REG-v4.2) Copernicus Atmosphere Monitoring Service. https://doi.org/
10.24380/0vzb-a387 [publisher] ECCAD [distributor].
Kuenen, J., Dellaert, S., Visschedijk, A., Jalkanen, J.P., Super, I., Denier Van Der Gon, H.,
2022. CAMS-REG-v4: a state-of-the-art high-resolution European emission inventory
for air quality modelling. Earth Syst. Sci. Data 14, 491–515. https://doi.org/
10.5194/essd-14-491-2022.
Li, X., Lopes, D., Mok, K.M., Miranda, A.I., Yuen, K.V., 2019. Development of a road
trafc emission inventory with high spatial-temporal resolution in the world’s most
densely populated region — Macau. Environ. Monit. Assess. 191–239. https://doi.
org/10.1007/s10661-019-7364-9.
Lin, P., Gao, J., Xu, Y., Schauer, J.J., Wang, J., He, W., Nie, L., 2022. Enhanced
commercial cooking inventories from the city scale through normalized emission
factor dataset and big data. Environ. Pollut. 315, 120320 https://doi.org/10.1016/j.
envpol.2022.120320.
D. Lopes et al.
Atmospheric Environment 312 (2023) 120026
13
Lopes, D., Rafael, S., Ferreira, J., Relvas, H., Almeida, S.M., Faria, T., Martins, V.,
Diapouli, E., Manousakas, M., Vasilatou, V., Fetfatzis, P., Miranda, A.I., 2022.
Assessing the levels of regulated metals in an urban area: a modelling and
experimental approach. Atmos. Environ. 290, 119366 https://doi.org/10.1016/j.
atmosenv.2022.119366.
Ma, D., Wu, X., Sun, X., Zhang, S., Yin, H., Ding, Y., Wu, Y., 2022. The characteristics of
light-duty passenger vehicle mileage and impact analysis in China from a big data
perspective. Atmosphere 13, 1–11. https://doi.org/10.3390/atmos13121984.
Maxim, L., van der Sluijs, J.P., 2011. Quality in environmental science for policy:
assessing uncertainty as a component of policy analysis. Environ. Sci. Pol. 14,
482–492. https://doi.org/10.1016/j.envsci.2011.01.003.
Pordata, 2020. Volume de neg´
ocios das empresas n˜
ao nanceiras: total e por sector de
actividade econ´
omica - 2020 [WWW Document]. URL. https://www.pordata.
pt/Municipios/Volume+de+neg´
ocios+das+empresas+n˜
ao+nanceiras+total+e+
por+sector+de+actividade+econ´
omica-589, 6, 3.
Racius, 2022. Empresas em Portugal - Em atividade [WWW Document]. URL. htt
ps://www.racius.com, 3.31.22.
Relvas, H., Miranda, A.I., 2018. An urban air quality modeling system to support
decision-making: design and implementation. Air Qual. Atmos. Heal. 11, 815–824.
https://doi.org/10.1007/s11869-018-0587-z.
Relvas, H., Lopes, D., Ferreira, J., Silva, A., Rafael, S., Lopes, M., Almeida, S.M.,
Martins, V., Diapouli, E., Korhonen, A., H¨
anninen, O., Lazaridis, M., Miranda, A.I.,
2022. Scenario analysis of strategies to control air pollution. Urban Clim. 44, 101201
https://doi.org/10.1016/j.uclim.2022.101201.
REN, (Rede Energ´
etica Nacional), 2022. Electricidade - Balanço di´
ario [WWW
Document]. URL. https://datahub.ren.pt/pt/eletricidade/balanco-diario/, 3.18.22.
Russell, A., Dennis, R., 2000. NARSTO critical review of photochemical models and
modeling. Atmos. Environ. 34, 2283–2324. https://doi.org/10.1016/S1352-2310
(99)00468-9.
Russo, M.A., Gama, C., Monteiro, A., 2019. How does upgrading an emissions inventory
affect air quality simulations? Air Qual. Atmos. Heal. 12, 731–741. https://doi.org/
10.1007/s11869-019-00692-x.
Shami, A. Al, Aawar, E. Al, Baayoun, A., Saliba, N.A., Kushta, J., Christoudias, T.,
Lakkis, I., 2022. Updated national emission inventory and comparison with the
emissions database for global atmospheric research (EDGAR): case of Lebanon.
Environ. Sci. Pollut. Res. 29, 30193–30205. https://doi.org/10.1007/s11356-021-
17562-8.
Sun, X., Tian, Z., Malekian, R., Li, Z., 2018. Estimation of vessel emissions inventory in
Qingdao port based on big data analysis. Symmetry (Basel) 10, 1–11. https://doi.
org/10.3390/sym10100452.
Thunis, P., Clappier, A., Pisoni, E., Bessagnet, B., Kuenen, J., Lopez-aparicio, S., 2021.
A multi-pollutant and multi-sectorial approach to screen the consistency of emission
inventories. Geosci. Model Dev. (GMD) 1–18. https://doi.org/10.5194/gmd-2021-
390.
USEPA, (United States Environmental Protection Agency), 1996a. AP-42, Compilation of
Air Pollutant Emission Factors, Section 8.4. U.S. Environmental Protection Agency,
Emissions Inventory Branch, Ofce of Air Quality Planning and Standards, Research
Triangle Park, New York,USA. Final Section, “Ammonium Sulphate,”.
USEPA, (United States Environmental Protection Agency), 1996b. AP-42, Compilation of
Air Pollutant Emission Factors, Section 1.2, Final Section, “Anthracite Coal
Combustion,”. U.S. Environmental Protection Agency, Emissions Inventory Branch,
Ofce of Air Quality Planning and Standards, Research Triangle Park, New Yor.
USEPA, (United States Environmental Protection Agency), 1998a. AP-42, Compilation of
Air Pollutant Emission Factors, Section 1.3, Final Section, “Fuel Oil Combustion,”. U.
S. Environmental Protection Agency, Emissions Inventory Branch, Ofce of Air
Quality Planning and Standards, Research Triangle Park, New York, USA.
USEPA, (United States Environmental Protection Agency), 1998b. AP-42, Compilation of
Air Pollutant Emission Factors, Section 1.4, Final Section, “Natural Gas
Combustion,”. U.S. Environmental Protection Agency, Emissions Inventory Branch,
Ofce of Air Quality Planning and Standards, Research Triangle Park, New York, U.
USEPA, (United States Environmental Protection Agency), 1998c. AP-42, Compilation of
Air Pollutant Emission Factors, Section 1.1, Final Section, “Bituminous and
Subbituminous Coal Combustion,”. U.S. Environmental Protection Agency,
Emissions Inventory Branch, Ofce of Air Quality Planning and Standards, Research
Tria.
WHO, (World Health Organization), 2021. New WHO Global Air Quality Guidelines Aim
to Save Millions of Lives from Air Pollution [WWW Document]. URL. https://www.
who.int/news/item/22-09-2021-new-who-global-air-quality-guidelines-aim-to-save-
millions-of-lives-from-air-pollution, 2, 9.
D. Lopes et al.
