We present the Copernicus Atmosphere Monitoring Service
TEMPOral profiles (CAMS-TEMPO), a dataset of global and European emission
temporal profiles that provides gridded monthly, daily, weekly and hourly
weight factors for atmospheric chemistry modelling. CAMS-TEMPO includes
temporal profiles for the priority air pollutants (NO
Spatially and temporally resolved atmospheric emission inventories are key
to investigate and predict the transport and chemical transformation of
pollutants, as well as to develop effective mitigation strategies (e.g.
Pouliot et al., 2015; Galmarini et al., 2017). During the last decade,
global and regional inventories have substantially increased spatial
resolution from
Using global and regional emission inventories in atmospheric chemistry
models requires the original aggregated annual emissions to be broken down
into fine temporal resolutions (ideally hourly) using emission temporal
profiles (e.g. Borge et al., 2008; Bieser et al., 2011; Mues et al., 2014).
In practice, temporal profiles are normalized weight factors for each hour
of the day, day of the week and month of the year. At the global scale, the
most commonly used emission temporal profiles are the monthly factors
provided by the air pollutant and greenhouse gas Emission Database for
Global Atmospheric Research inventory (EDGARv4.3.2; Janssens-Maenhout et
al., 2019) and the Evaluating the Climate and Air Quality Impacts of
Short-Lived Pollutants inventory (ECLIPSEv5.a; Klimont et al., 2017). Also
at the global level, the Temporal Improvements for Modeling Emissions by
Scaling (TIMES) dataset was produced to represent the weekly and hourly
variability for global CO
At the European level, the temporal factors provided by the University of Stuttgart (Institute of Energy Economics and Rational Energy Use, IER) as part of the Generation of European Emission Data for Episodes (GENEMIS) project are still considered as the main reference (Ebel et al., 1997; Friedrich and Reis, 2004). The original GENEMIS profiles were later used as a basis to derive two independent datasets: (i) the EMEP temporal profiles, which provide monthly, weekly and hourly weight factors that vary per emission sector, country and pollutant (Simpson et al., 2012), and (ii) the Netherlands Organisation for Applied Scientific Research (TNO) temporal profiles, which provide monthly, weekly and hourly weight factors that vary per emission sector (Denier van der Gon et al., 2011). These two sets of profiles have become over time the reference datasets under the framework of several European air quality modelling activities, including the earlier Monitoring Atmospheric Composition and Climate (MACC) project and the current Copernicus Atmosphere Monitoring Service (CAMS), among others. Other widely used regional temporal profile datasets include the North American profiles provided by the Environmental Protection Agency (EPA) Clearinghouse for Inventories and Emissions Factors (CHIEF) (US EPA, 2019a) and the monthly profiles provided by the Multi-resolution Emission Inventory for China (MEIC; Li et al., 2017).
Our goal is to provide a new set of global and European temporal profiles. Current datasets typically use the same temporal profiles for certain sectors and/or regions. For example, ECLIPSE and EMEP share the same monthly profiles for the energy sector in Europe and Russia. Similarly, TNO and EDGAR share the same monthly profiles for residential combustion and road transport (Friedrich and Reis, 2004), as well as for the energy industry (Veldt, 1992) and agriculture (Asman, 1992). In these two datasets, temporal profiles are mostly assumed to be both country- and meteorology-independent. The only exceptions are, in the case of EDGAR, for the residential and agricultural sectors, which are approximated as a function of the geographical zone: the seasonality assumed in the Northern Hemisphere is shifted by 6 months in the Southern Hemisphere, and a flat profile is assumed along the Equator. In the case of EMEP, the reported monthly and weekly profiles do consider differences across countries but are primarily based on old sources of information from the 1990s and beginning of the 2000s and subsequently neglect behavioural changes that may have happened over the last years. Similarly, road transport weekly and hourly factors reported by TNO are based on long time series of Dutch data registering the traffic intensity between 1985 and 1998. Moreover, variable climate conditions and changes in meteorology that may cause differences in the temporal weight factors within a country are not accounted for. In order to overcome this limitation, the ECLIPSE monthly profiles for the residential combustion sector were computed using global gridded temperature data and provided as monthly shares for each grid cell (Klimont et al., 2017).
This work presents the Copernicus Atmosphere Monitoring Service TEMPOral
profiles (CAMS-TEMPO), a new dataset of global and European emission
temporal profiles for atmospheric chemistry modelling. The development of
CAMPS-TEMPO comes from the need to overcome the aforementioned limitations
of current profiles (i.e. use of an outdated source of information and
neglection of the temporal variation of emissions across species and
countries or regions) and to improve the representation of the emission temporal
variations, which was defined as a priority task within the Copernicus
global and regional emissions service (CAMS_81) directly
supporting the CAMS production chains (
Main characteristics of the temporal profiles developed in this work compared to those reported in other datasets including TNO (Denier van det Gon et al., 2011), EMEP (Simpson et al., 2012), EDGARv4.3.2 (Janssens-Maenhout et al., 2019) and EDGARv5 (Crippa et al., 2020) regarding spatial coverage, temporal and spatial resolution, and year or meteorology dependency.
The CAMS-TEMPO profiles were created following the domain descriptions (resolution and geographical area covered) and emission sector classification system defined in the CAMS global anthropogenic inventory (CAMS-GLOB-ANT) and CAMS regional inventory for air pollutants and greenhouse gases (CAMS-REG_AP/GHG), also developed under CAMS_81 (Granier et al., 2019).
The CAMS-GLOB-ANT dataset (Elguindi et al., 2020a) is a
global emission inventory developed for the years 2000–2020 at a spatial
resolution of 0.1
The CAMS regional emissions are being prepared for air pollutants and
greenhouse gases (CAMS-REG_AP/GHG), in support of the CAMS
regional production systems and policy tools. The inventory is built up
largely using the official reported emission data from individual countries
in Europe for each source category, which has the main advantage that it
takes into account country-specific information on technologies, practices
and associated emissions. Where these data were either unavailable or not
fit for purpose, these were replaced with other estimates. Then, a
consistent spatial distribution is applied across Europe at a resolution of
0.1
The paper is organized as follows. Section 2 describes, for each sector, the approaches and sources of information used to develop the CAMS-TEMPO profiles. Section 3 discusses the obtained temporal profiles and compares them to currently available datasets. Section 4 provides a description of the data availability, and finally Sect. 5 presents the main conclusions of this work.
The following subsections describe the input data and methodologies used to compute the CAMS-TEMPO emission temporal profiles for each targeted sector: (i) energy industry (Sect. 2.2), (ii) residential and commercial combustion (Sect. 2.3), (iii) manufacturing industry (Sect. 2.4), (iv) road transport (Sect. 2.5), (v) aviation (Sect. 2.6), and (vi) agriculture (Sect. 2.7).
The CAMS-TEMPO dataset consists of a collection of global and regional
temporal factors that follow the domain description and sector
classification reported by the CAMS-GLOB-ANT and
CAMS-REG_AP/GHG emission inventories. In order to better
distinguish between the two sets of profiles, we refer to them as
CAMS-GLOB-TEMPO (
Tables 2 and 3 summarize the characteristics of each temporal profile included in the CAMS-GLOB-TEMPO and CAMS-REG-TEMPO datasets, respectively. The sector classification for each case corresponds to those used in CAMS-GLOB-ANT and CAMS-REG_AP/GHG. The specificity of the computed profiles depends upon the degree of sectoral disaggregation used in the original CAMS inventories. For example, the CAMS-GLOB-ANT dataset reports emissions from power and heat plants and refineries under the same sector (“ene”, see Table 2), and therefore a common set of temporal profiles had to be assumed for the two types of facilities. In contrast, the CAMS-REG_AP/GHG inventory reports power and heat plants under the GNFR_A (Gridding Nomenclature for Reporting) category (public power) and refineries under sector GNFR_B (industry), together with all manufacturing industries (Table 3). All the assumptions made regarding this topic are clearly stated in each subsection.
Main characteristics of the CAMS global temporal profiles
(CAMS-GLOB-TEMPO) reported by sector and temporal resolution (monthly,
daily, weekly and hourly). The text between brackets gives information on the
spatial resolution and pollutant and year dependency of each profile.
Gridded indicates that the profile varies per grid cell within a country;
per country indicates that the profile varies only per country; fixed
indicates that the profile is spatially invariant; year-independent
indicates that the profiles does not vary per year; year-dependent
indicates that the profile varies per year; pollutant-independent indicates
that the same profile is proposed for all pollutants (NO
Main characteristics of the CAMS regional temporal profiles
(CAMS-REG-TEMPO) reported by sector and temporal resolution (monthly, daily,
weekly and hourly). The text between brackets gives information on the spatial
resolution and pollutant and year dependency of each profile. Gridded
indicates that the profile varies per grid cell within a country; per
country indicates that the profile varies only per country; fixed
indicates that the profile is spatially invariant; year-independent
indicates that the profiles does not vary per year; year-dependent
indicates that the profile varies per year; pollutant-independent indicates
that the same profile is proposed for all pollutants (NO
For both CAMS-GLOB-TEMPO and CAMS-REG_AP/GHG, the sum of all temporal weight factors is equal to 12 for monthly profiles, 7 for weekly profiles, 365 or 366 (in the case of a leap year) for daily profiles, and 24 for hourly profiles. Note that the hourly temporal profiles in CAMS-TEMPO are provided in local standard time (LST). The conversion from LST to coordinated universal time (UTC) as a function of time zones is a process that needs to be performed by the final user. Time zone adjustments is a process typically performed by the emission processing systems or tools designed to adapt emission data to the air quality modelling requirements (e.g. Guevara et al., 2019).
The temporal profiles computed for the energy industry are reported under the ene sector in CAMS-GLOB-TEMPO and the GNFR_A category in CAMS-REG-TEMPO. The temporal variability of emissions from this sector was estimated from electricity production statistics under the assumption that it largely depends upon the combustion of fossil fuels in power and heat plants. This approximation is consistent with the definition of the GNFR_A sector in the CAMS-REG_AP/GHG dataset. The representativeness of the computed profiles is likely lower in CAMS-GLOB_TEMPO because the ene sector also includes other facilities such as refineries.
As shown in Tables 2 and 3, the profiles reported for this sector
include pollutant- and country-dependent monthly, weekly and hourly factors.
The electricity production dataset compiled to derive profiles for this
sector were as follows:
Figure 1a illustrates the spatial coverage of the compiled dataset by source of information (i.e. ENTSO-E, US EPA, IEA and MBS). Overall, main emission producers (e.g. China, India, Europe and North America) are covered, while most of the countries with no information available are located in South America and Africa. For those countries with no data, the TNO profiles reported under the energy sector (Denier van der Gon et al., 2011) are used.
Representation of the spatial coverage of the datasets used to
derive temporal profiles for the energy industry
The compiled data were first analysed to assess whether interannual variability is important for this sector. Seasonal cycles were computed for different years (2010–2017) and countries using the IEA statistics. In the majority of the countries analysed, the monthly profiles were found to be consistent through the different years and to present small interannual variations (Fig. S1 in the Supplement). Although some studies have pointed out a temperature dependence of the monthly electricity generation in power plants (Thiruchittampalam, 2014), we neglected it at present. Consequently, we assume the monthly temporal profiles for this sector to be the average over all the available years of data.
For European countries, monthly profiles were derived using the ENTSO-E
dataset. The analysis of the data showed that the seasonality of electricity
production varies significantly by fuel type (Fig. S1). The different use of
energy sources (i.e. lignite, hard coal, natural gas, biomass and oil)
implies that temporal patterns will also vary from one pollutant to another. For each month, country and pollutant, profiles were calculated
following Eq. (1):
Emission factors [kg TJ
For other countries, monthly factors by pollutant could not be developed, as
both the IEA and the MBS datasets do not report electricity production split
by fuel type. Hence, monthly factors were derived by averaging the available
production data per month and relating them to the total production in the
year. For the US, NO
Weekly profiles were developed for Europe using the hourly electricity production data reported by ENTSO-E. As in the case of the monthly profiles, weekly scale factors were found to significantly vary according to the type of fuel (Fig. S1). These results are in line with the conclusions of Adolph (1997), which identified three generic weekly profiles – base, medium and peak load – as a function of the type of power plant. Pollutant-related weekly profiles were developed following the same methodology applied for obtaining the monthly weight factors (Eq. 1).
For the US, the CEMS data were used to compute pollutant-dependent profiles following the same procedure as described in Sect. 2.2.1. Measurements from all individual plants were averaged per day of the week and then normalized to sum 7. For countries with no information on daily electricity production data, we used the weekly profile reported in the TNO dataset for the energy sector.
Hourly profiles were developed for Europe and the US using the hourly electricity production data reported by ENTSO-E and the measured emissions reported by CEMS, respectively. As previously seen, large differences are observed between fuels. Profiles related to the so-called base peak load power plants (i.e. annual useful life of more than 4000 h) present a rather flat distribution, whereas in other cases the change in energy production between day and night is relatively high (Fig. S1).
Pollutant-related hourly profiles were developed following the same procedure shown in Eq. (1). For countries with no information on hourly electricity production data, we assumed the hourly profile reported in the TNO dataset. Some studies have suggested that the hourly variation of power plant activities may vary according to the season of the year (Thiruchittampalam, 2014). This feature is not considered in the present version of the CAMS-TEMPO profiles and will be addressed in future releases.
The temporal profiles computed for the residential and commercial sector are reported under the res sector in CAMS-GLOB-TEMPO and the GNFR_C category in CAMS-REG-TEMPO. The temporal variability of emissions for this sector is assumed to be dominated by the stationary combustion of fossil fuels in households and commercial and public service buildings. These categories are also assumed to be the main contributors to the total emissions reported by CAMS-GLOB-ANT and CAMS-REG_AP/GHG. Other combustion installations activities included under this sector (i.e. plants in agriculture, forestry and aquaculture and other stationary facilities including military) are assumed to follow the same temporal profile.
The temporal weight factors developed for this sector include monthly, daily and hourly profiles. The monthly and daily profiles depend upon year and region and were derived using meteorological parametrizations (Sect. 2.3.1). The hourly profiles depend upon pollutant and region (Sect. 2.3.2).
Gridded daily temporal profiles were derived according to the heating-degree-day (HDD) concept, which is an indicator used as a proxy variable to reflect the daily energy demand for heating a building (Quayle and Diaz, 1980). This method has been proven to be successful in previous emission modelling work (e.g. Mues et al., 2014; Terrenoire et al., 2015).
The heating-degree-day factor (
A challenge when using this method is to set the threshold or comfort
temperature (
Gridded daily temporal profiles were developed for 8 years (2010 to
2017). A climatological daily profile based on the average of each day over
all the available years was also produced. Monthly gridded factors were
derived from the daily profiles for all the years available. We interpolated
the estimated gridded daily factors from the ERA5 working domain (approx.
The hourly distribution of residential and commercial combustion activities has
typically been described following the profile A presented in Fig. 2a, used
in both EMEP and TNO datasets. This hourly distribution presents one peak in
the morning and another one in the afternoon, when energy consumption is
supposedly higher due to increased space heating or cooking activities. We
evaluated this profile with real-world measurements of natural gas
consumption for residential houses in the UK (Retrofit for the Future
project,
We created a second hourly profile (Fig. 2a, profile B) linked to the combustion of residential wood for space-heating purposes using as a basis information derived from citizen interviews performed in Norway and Finland (Finstad et al., 2004 and Gröndahl et al., 2010) as well as from long-term measurements of the wood-burning fraction of black carbon in Athens (Athanasopoulou et al., 2017). As shown in Fig. 2a, the resulting profile B presents an intense peak during the evening hours, but not during the morning, in contrast to profile A. It is actually a common practice in developed countries to use fireplaces and other types of wood-burning appliances mainly in the evening.
As reported by the World Health Organization (WHO), most developing
countries use wood not only for heating space purposes but also for cooking
activities (Bonjour et al., 2013). We created a third profile that
represents these activities (Fig. 2a, profile C) based on information
derived from continuous indoor PM
The results summarized in Fig. 2a indicate that the hourly behaviour of
residential combustion emissions varies according not only to the fuel type
but also to the type of end use (i.e. space heating or cooking). Both
CAMS-GLOB-ANT and CAMS-REG_AP/GHG report
total residential and commercial emission as a unique sector, without
discriminating by type of fuel or end use. Therefore, several decisions were
made in order to assign the three proposed profiles to different pollutants
and regions:
profile A: NO profile B: PM profile C: all pollutants in rural areas of developing countries.
The assumptions made behind this assignment are the following:
Natural gas and diesel heating combustion is the main contributor to total
NO Wood combustion is the main contributor to total PM In the urban and rural areas of developed countries wood is mainly used for
heating purposes. In the urban areas of developing countries wood is mainly used for heating
purposes. In the rural areas of developing countries all fuels are used both for
heating and/or cooking purposes (i.e. the two activities occur at the same
time).
The list of developing countries was obtained from the World Bank country
classifications (World Bank, 2014). The discrimination of human settlements
between urban and rural areas was derived from the Global Human Settlement
Layer (GHSL) project (Florczyk et al., 2019; Pesaresi et al., 2019). The
GHSL provides a global classification of human settlements on the basis of
the built-up settlement and population density at a resolution of 1 km
The temporal variability of industrial emissions is reported under the sectors indu (CAMS-GLOB-TEMPO) and GNFR_B (CAMS-REG-TEMPO). Both in the CAMS-GLOB-ANT and the CAMS-REG_AP/GHG inventories, all industrial manufacturing emissions are reported under these single categories. Hence, the same temporal pattern has to be assumed for all types of facilities (e.g. cement plants, iron and steel plants, and food and beverage). For this sector, only country-dependent monthly profiles were developed due to the lack of more detailed data.
Country-specific monthly profiles were estimated using the Industrial Production Index (IPI), which measures the monthly evolution of the productive activity of different industrial branches, including manufacturing activities. The IPI as a monthly surrogate for industrial emissions has been used in previous studies (e.g. Pham et al., 2008; Markakis et al., 2010).
The IPI data were obtained from the MBS database (MBS, 2018), which provides monthly information per country and general industrial branch (i.e. mining, manufacturing, electricity, gas and steam, and water supply) for the year 2015. The manufacturing branch, which includes several divisions such as iron and steel industries, chemical industries, and food and beverage products, was used to derive country-specific monthly profiles. Figure 1b shows the spatial coverage of the compiled dataset. As in the case of the energy industry sector (Fig. 1a), the lack of information mostly affects Africa and South America. For those countries without available information, the monthly profile reported in the TNO dataset under the industry sector was used (Denier van der Gon et al., 2011). In the case of China, the monthly profile reported in the MIX inventory under the industry sector was used (Li et al., 2017).
The time profiles are based on IPI information from 2015 and are assumed to be representative for other years. Our assumption is supported by the low interannual variability observed in the IPI values collected from different national statistical offices including Italy (ISTAT, 2018), Norway (SSB, 2018), Spain (INE, 2018) and the UK (ONS, 2018) (Fig. S3). Another implicit assumption made is that the constructed monthly profiles can be equally applied to all the different industrial activities reported under the ind and GNFR_B sectors. The national IPI values collected for Italy and the US (Board, 2020) were used to compare the seasonality of individual industrial divisions to the general manufacturing IPI monthly profile. For both countries, as well as up to a certain extent, it was found that all the industrial divisions (except food and beverages and the petrochemical industry in the case of Italy) follow the seasonality of the general manufacturing profile (Fig. S4), which allows for concluding that the assumption made is reasonable. A similar result is reported for Thailand in Pham et al. (2008).
Share of final energy consumption in the residential sector by fuel and type of end use in Europe (Eurostat, 2018).
Due to the lack of country-specific data, the fixed weekly and hourly temporal profiles provided in the TNO dataset for industry sector are used. The weekly profile assumes a flat distribution during the working days and a slight decrease during weekends (Table A2). On the other hand, the hourly profile includes an increase of the activity during the central hours of the day (Tables A3 and A4).
Temporal profiles for road transport emissions are reported under the tro
sector in CAMS-GLOB-TEMPO and the GNFR_F1 (exhaust gasoline), GNFR_F2 (exhaust diesel),
GNFR_F3 (exhaust LPG gas) and GNFR_F4 (non-exhaust) categories in CAMS-REG-TEMPO. The fact
that CAMS-REG_AP/GHG traffic-related emissions are classified
into four different categories (discriminated by fuel and type of process)
allows for considering specific temporal features associated with each one of
them, including temperature dependence of CO and NMVOC gasoline exhaust
emissions (GNFR_F1), of NO
The temporal weight factors developed for this sector include monthly,
weekly and hourly profiles. As summarized in Table 2, the CAMS-GLOB-TEMPO
monthly and weekly profiles constructed for this sector are
region-dependent, whereas the hourly profiles vary per region and day of the
week (i.e. weekday, Saturday and Sunday). In the case of CAMS-REG-TEMPO, the
constructed profiles can vary by region, pollutant, day of the week and/or
year, as a function of the source sector and temporal resolution (Table 3).
Depending on the dataset and sector category, temporal emission variability
is assumed to be either exclusively driven by the traffic activity data
(e.g. CAMS-GLOB-TEMP, all cases) or by a combination of traffic activity
data and changes in ambient temperature (i.e. CAMS-REG-TEMP, CO and NMVOC
GNFR_F1, NO
List of traffic activity datasets and corresponding sources of information compiled.
Considering that for each traffic count dataset the reference years are different (Table 6), we analysed the data in view of differences in the resulting profiles as a function of the year. We took the Paris city traffic data as an example, since it covers a wide range of years (2013 to 2017). For each year, monthly, weekly and hourly (i.e. Wednesday and Saturday) profiles were constructed (Fig. S5). The results suggest that temporal patterns in vehicle activity do not change much over long timescales. Consequently and following the assumptions made for the energy and manufacturing industry sectors, we assumed that the interannual variability can be negligible. Hence, all profiles developed only as a function of traffic count data were constructed by averaging the values (per month, day of the week or hour of the day) over all the available years.
Some of the compiled datasets (e.g. Germany and California) report the traffic counts classified by vehicle type (i.e. light-duty vehicles, LDVs, and heavy-duty vehicles, HDVs). The monthly, weekly and hourly profiles as a function of the vehicle type showed significant differences, especially for weekly and hourly profiles (Fig. S6). HDV traffic presents a larger decrease on the weekend than LDV traffic. Moreover, the hourly LDV profile exhibits two distinct (morning and evening) commuter-related peaks, whereas HDV shows a single midday peak. These results highlight the importance of applying separate temporal profiles to characterize traffic and associated emissions for LDVs and HDVs. However, in the present work this disaggregation was not considered, since both the CAMS-GLOB-ANT and CAMS-REG_AP/GHG inventories report LDV- and HDV-related emissions under the same pollutant sector. When only vehicle-type temporal profiles were available (i.e. California), the information reported for LDVs is used, as this type of vehicle dominates the temporal distribution of total traffic flow.
A comparison between monthly variation in traffic patterns at urban and rural locations (i.e. urban streets and highways) was performed for selected countries or regions including California, Germany, Spain and the UK. For the UK and California, the original traffic statistics were already discriminated by type of location. For the German and Spanish datasets, each traffic station was classified as urban or rural considering its geographical location and the GHSL human settlement classification dataset (Sect. 2.3.2). As shown in Fig. 3, while there is little seasonal variation in German urban locations, rural areas tend to exhibit a stronger seasonality, with a peak occurring during summertime, presumably due to increased recreational and vacation-related driving. The results derived from California, Spain and the UK are consistent with these patterns (Fig. S7).
Map representing urban and rural human settlements in Germany as
reported by the Global Human Settlement Layer (GHSL; Pesaresi et al., 2019)
and the location of urban and rural German traffic count stations
The datasets collected from several cities (i.e. Athens, Barcelona, Berlin, Copenhagen, London, Madrid, Mexico City, Milan, Oslo and Paris) were also used to construct monthly profiles (Fig. 4a). For comparison purposes, the TNO road transport profile is also included (Denier van der Gon et al., 2011). The three southern European cities (i.e. Athens, Madrid and Milan) together with Paris present a similar pattern, with a significant decrease of the activity during the month of August due to the summer holidays. Similarly, northern European cities (i.e. Copenhagen and Oslo) also present a decrease in summer but of a lower intensity and during July. On the other hand, the seasonality observed in London, Berlin and Mexico City is rather flat and closer to the TNO profile.
Results in Figs. 3b and 4a showed that (i) monthly variations can significantly differ among countries and (ii) within a country, traffic regimes show differences according to the location (urban or rural). Considering all of the above, country- and region-specific (urban or rural) monthly profiles were constructed based on the traffic information compiled. For countries without any available temporal factors, assumptions were made considering geographical proximity. For instance, the urban profile for Scandinavian countries without data (i.e. Finland, Sweden and Iceland) was constructed by averaging the profiles of Oslo and Copenhagen. On the other hand, the rural profile constructed for Spain was assigned to other southern European countries (i.e. Italy, Greece, Malta, Croatia, Bosnia and Herzegovina, Montenegro, Albania, Slovenia, Cyprus, and Portugal). Similarly, the seasonality in Canada was assumed to be equal to the one observed in the US. For all the countries not listed in the table, the urban and rural profiles developed for Germany were assumed. This approach may be further improved as more traffic count data become available. In the case of China and India, profiles were derived from the MIX emission inventory (Li et al., 2017).
Two main assumptions underlie these profiles. First, the differences among cities within a country are assumed to be small, and therefore we use a unique urban profile therein. The second hypothesis was to assume that all the streets and highways located in urban and rural areas present the same seasonality. While this is a reasonable assumption overall, individual traffic count stations can show particular features. For instance, on certain highways near city entrances or crossing urban areas, traffic intensity shows a flat distribution without the typical summer anomalies. This level of detail, which would require a specific temporal profile per road segment, is out of the scope of this dataset but may be explored in future work.
Monthly
The seasonality of traffic emissions can be affected by temperature. As
shown by Zheng et al. (2014), during winter months vehicles in China produce
19 % more CO and 11 % more NMVOC than in the summer due to the higher
contribution of cold-start emissions. This study also showed that the
monthly pattern of emissions differs remarkably by latitude, which is
explained by the large contribution of cold-start emissions and the
relationship between latitude and temperature. More recently, Keller et al. (2017) and Grange et al. (2019) identified strong temperature dependence for
diesel vehicle NO
We used available parametrizations (Eqs. 6 to 8) to account for the
meteorological drivers of the seasonality of CO and NMVOC gasoline-related
emissions (GNFR_F1) (US EPA, 2015) and NO
The meteorology-dependent monthly profiles were then combined with the
meteorology-independent ones (Sect. 2.5.1) so that
the resulting seasonality accounts for both temperature influences and
traffic activity. For CO, we used a weight factor of 45 % for the
temperature-dependent profiles and of 55 % for the traffic activity ones,
following the UK National Atmospheric Emissions Inventory (NAEI), which
reports road transport annual emissions and distinguishes between cold starts
and hot exhaust. Due to the lack of information, we assumed the same share
for all countries. Likewise, for NMVOC profiles we assumed a 33 % weight
for the temperature-dependent temporal factors (i.e. emissions related to
cold starts) and 67 % for the ones derived from traffic counts (i.e.
emissions related to hot exhaust). Finally, for NO
In addition to the meteorological-dependent profiles described above, we
created a specific monthly profile for NMVOC evaporative emissions
(GNFR_F4) based on recent results obtained with the High-Elective Resolution
Modelling Emission System (HERMESv3) (Guevara et al., 2020c). The HERMESv3
model computes hourly gasoline evaporative emissions from standing cars
(diurnal losses) using the “Tier 2” approach reported in the EMEP/EEA emission
inventory guidelines 2016. Summer and winter temperature-dependent emission
factors are defined for each type of vehicle as a function of the 2 m
outdoor temperature obtained from the ERA5 reanalysis. The HERMESv3 model
was run over Spain for the year 2016 at a spatial resolution of
Country- and region-specific (urban or rural) weekly profiles were constructed based on the traffic information summarized in Table 6. In contrast to urban areas, rural traffic activity is lower during weekdays and decrease relatively less during the weekend, especially on Sundays (Figs. 3c and S7). Figure 4b shows how, depending on the location, the intensity of the weekend decrease is relatively higher (e.g. Madrid) or lower (e.g. Mexico City), which is likely due to different sociodemographic patterns. We used the weekly profile provided in the TNO dataset as the urban profile for the countries where local information is not available. Similarly, we used an average profile including data from Germany, Spain and the UK as the rural profile for countries without data. The resulting profiles were assigned equally to all the traffic-related categories of both CAMS-GLOB-TEMPO and CAMS-REG-TEMPO, with the exception of NMVOC gasoline evaporation (GNFR_F4), for which a flat profile is proposed.
The analysis of hourly profiles constructed per day of the week for six cities (i.e. Berlin, Madrid, Milan, Oslo, Paris and Utrecht) clearly highlights the need to create hourly profiles per day type (Fig. S9). Weekdays (i.e. Monday to Friday) tend to exhibit strong similarities and reflect commuting patterns that are typically bimodal with morning and afternoon volume peaks. Saturday and Sunday generally show the traffic activity to plateau between late morning and early evening, typically due to a decrease in commuting activity. Some studies have developed distinct hourly traffic profiles for Monday to Thursday, Friday, Saturday and Sunday (e.g. McDonald et al., 2014), and others have discriminated between weekdays and weekends (e.g. Zheng et al., 2009). CAMS-TEMPO includes hourly profiles that vary among weekdays, Saturdays and Sundays, following other studies such as Menut et al. (2012).
Hourly variations during weekdays at urban and rural locations were compared for selected countries and regions (Figs. 3d and S8). Morning traffic peaks associated with commuting are found in and near cities but not in rural areas (see California and Guangzhou in Fig. S8). Lunchtime peaks tend to be higher in rural areas, mainly due to the activity of HDVs (Spain and Germany). In contrast, hourly variations on Saturday and Sunday were found to be very similar in urban and rural areas for all the available datasets (not shown). Consequently, only for weekdays we differentiated the hourly profiles of urban and rural areas. Also, in this case, the GHSL dataset was used to assign the respective profiles to either urban or rural grid cells.
Weekday hourly profiles constructed for different cities are shown in Fig. 4c. Two groups of profiles (showed in red and blue) with similar behaviours
were identified. For the first group (in red), the rush hours in the morning
and in the evening can be clearly identified. The occurrence of the peaks
varies from one city to another due to different sociodemographic patterns.
In the second group (in blue), a maximum level of activity is reached in the
morning (between 07:00 and 08:00) that largely remains for the rest of the
daytime period (i.e. 07:00 to 19:00) and through part of the nighttime
(i.e. 19:00 to 21:00). The hour when traffic activity reaches the maximum
level also varies from one city to another. Besides the potential effect
of different social habits, the difference between the two groups of
profiles could be also associated with differences in the vehicle densities.
For instance, Oslo, Utrecht and Berlin (first group) have vehicle densities
of 600, 1100 and 1300 vehicles km
Figure 4d shows the constructed Saturday hourly profiles for selected cities. As before, two groups of profiles showing similar patterns are highlighted. The profiles related to the first group (in red) tend to present larger activity levels during daytime (between 09:00 and 18:00), whereas in the second group (in blue) weight factors are higher during nighttime (between 21:00 and 03:00). A similar pattern is observed with Sunday hourly profiles (not shown).
The resulting profiles were assigned equally to all the traffic-related categories with the exception of NMVOC evaporative emissions (GNFR_F4), in which we use a specific hourly profile based on HERMESv3 (Sect. 2.5.2). The resulting profile is shown in Fig. 4c and d (yellow lines) and Tables A3 and A4.
For European and North American countries without any available temporal factors, assumptions were made considering geographical proximity as described in Sect. 2.5.1. For China, the profiles were computed as an average of the weight factors reported for Beijing and Guangzhou. For all the rest of the cases (mainly Africa, Latin America and Asia), the urban and rural profiles developed for Germany were assumed to be the default, as they were based on the largest number of traffic count stations (more than 1500). This approach may, of course, be improved but was constrained in this study by the traffic count data availability.
The temporal profiles developed for air traffic emissions during landing and take-off (LTO) cycles in airports are reported under the GNFR_H category in the CAMS-REG-TEMPO dataset. Country-dependent monthly temporal profiles were constructed using airport traffic data, as described below. Due to the lack of country-specific data, a fixed hourly temporal profile is proposed. We could not consider this sector for CAMS-GLOB-TEMPO, since it is excluded in the CAMS-GLOB-ANT inventory. Aviation emissions are reported in a separate inventory called CAMS-GLOB-AIR, in which emissions from LTO cycles are reported together with climbing, descent and cruise aircraft operations.
We collected monthly airport traffic data by reporting airport for the years 2011 to 2017 from the Eurostat statistics (Eurostat, 2019). The year 2010 was excluded from the data gathering process due to the air travel disruption in northern and central Europe caused by the Eyjafjallajökull eruption. An analysis of the seasonality observed in several airports for each individual year allowed for confirming a low interannual variability (Fig. S10). Consequently, the constructed temporal profiles were based on the average data of all the available years. Country-dependent monthly profiles were derived by aggregating the respective national airports available in the Eurostat dataset.
We assumed flat weekly profiles for this sector, as no clear patterns could be found in the available datasets. We use a fixed hourly profile based on airport traffic from the Madrid–Barajas and Barcelona–El Prat airports (Aena, Aeropuertos Españoles y Navegación Aérea, personal communication, 2018). The computed fixed profile (Tables A3 and A4) was found to be broadly consistent with the hourly variations reported by other studies (e.g. Unal et al., 2005; Zhou et al., 2019).
Global and regional temporal profiles for the agricultural emissions are reported in two separate sectors: mma and agr in CAMS-GLOB-TEMPO and GNFR_K and GNFR_L in CAMS-REG-TEMPO. In both cases, the former category only includes emissions from livestock (enteric fermentation, manure management), whereas the latter reports emissions from several activities but mainly fertilizer applications and agricultural-waste burning. For both sectors, monthly and daily region-dependent profiles were constructed considering specific meteorological parametrizations (Sect. 2.7.1 to 2.7.3). For the hourly profiles, only fixed weight factors are proposed due to a data limitation issue (Sect. 2.7.4).
For the livestock sector (mma and GNFR_K), both in CAMS-GLOB-TEMPO and
CAMS-REG-TEMPO, we assumed NH
For the livestock sector (mma and GNFR_K), the temporal variation of NH
Since both the CAMS-GLOB-ANT and CAMS-REG_AP/GHG inventories report livestock emissions under a unique category, with
no discrimination by manure management practice, the computed
The resulting gridded daily temporal profiles were developed for 8 years (2010 to 2017). Using these time series as a basis and following the procedure described in Sect. 2.3.1, a climatological daily profile and monthly gridded factors were also produced.
The seasonality of NH
With the objective of computing daily variations, the gridded monthly
profiles obtained using the aforementioned mosaic approach were combined
with the daily meteorological parametrizations reported by Gyldenkærne
et al. (2005) and Skjøth et al. (2011) (Eq. 11):
For all the other criteria pollutants (i.e. NO
The hourly distribution of agricultural emissions has typically relied on
the profile reported by both the TNO and EMEP datasets (Fig. 2b). The profile
is constructed on the idea that NH
Regarding the other criteria pollutants (i.e. NO
In this section, we discuss the obtained temporal profiles for CAMS-GLOB-TEMPO and CAMS-REG-TEMPO. In Sect. 3.1 the profiles are compared to independent observational datasets and in Sect. 3.2 to other existing sets of temporal profiles currently used under the framework of CAMS.
Figure 5 shows the 0.1
CAMS-GLOB-TEMPO (0.1
CAMS-GLOB-TEMPO (0.1
Figure 6 shows two examples of CAMS-GLOB-TEMPO gridded daily profiles for the residential and commercial sector along with the times series at four geographically or climatically different locations (i.e. Athens, Barcelona, Buenos Aires and Oslo) for the years 2010 and 2017. As expected, the largest factors occur in winter, and the lowest ones occur in summer at all four locations. According to the results, emissions in Athens, Barcelona or Buenos Aires can be 3 to 5 times higher during the cold periods (i.e. January in Barcelona and Athens and June in Argentina) than during warm periods (August in Barcelona and Athens and January in Argentina). In Oslo both the seasonal cycle and daily variability are less pronounced than in the other locations because the differences between daily and annual mean temperatures are generally lower. There is a large interannual variability in the four locations. In the winter of 2010, Barcelona experienced three cold outbreaks of similar intensity (in January, February and March), whereas in 2017 only one significant episode can be observed (mid-January). Similarly, in 2010 three major peaks are observed in Athens in mid-January, the beginning of February and mid-December, whereas in 2017 only one episode stands out above the rest. Results clearly highlight that extreme weather events can strongly affect the temporal profiles and thereby the resulting emissions.
Figure 7 shows examples of the 0.1
CAMS-REG-TEMPO (0.1
CAMS-GLOB-TEMPO (0.1
Figure 8 shows examples of CAMS-GLOB-TEMPO profiles for the different
agricultural emission sources, including April and November (monthly)
weights for fertilizer NH
We compared the CAMS-TEMPO profiles to independent observational datasets. The comparison is mainly performed at the European level, as sufficient data could not be gathered to provide a robust comparison for other regions.
Figure 9 shows the CAMS-TEMPO monthly weight factors for SO
In Fig. 10 we compare our profiles for the residential sector with daily
factors derived from minute-resolved measurements of natural gas from a
residential house in Canada during 2013 (Makonin et al., 2016). We compared
our daily factors estimated using the HDD approach for the grid cell closer
to the house and considering two different values of
In Fig. 11 we compare the CAMS-REG-TEMPO road transport profiles to the
temporal variation of air pollutant concentrations measured in Madrid,
Milan, Barcelona and Berlin. CO and NO
Comparison between monthly variations of CAMS-REG-TEMPO profiles
for NO
The monthly variability of CO and NO
Similarly, the hourly profiles proposed for urban locations in Spain and
Germany during weekdays and Sundays were compared to the hourly variation of
NO
We compared the CAMS-TEMPO profiles to the profiles reported by other existing datasets. The comparison is focussed on the profiles that are currently being used for air quality modelling purposes under the framework of CAMS, namely the global EDGARv4.3.2 (Janssens-Maenhout et al., 2019) monthly profiles (used in the CAMS global production) and the European EMEP (Simpson et al., 2012) and TNO (Denier van der Gon et al., 2011) profiles (used in CAMS regional production).
The comparison of monthly, weekly and hourly profiles for the energy industry sector in selected countries is shown in Fig. 12. It is worth noting that both TNO and EDGAR report the same profile for all countries. At the monthly level, significant differences are observed between CAMS-TEMPO and EDGAR in China and the USA. In both cases, the summer peak observed in CAMS-TEMPO (presumably due to the intensive use of air-conditioning systems) does not show up in the EDGAR profile. In the case of Romania, all profiles show important decreases but at different times of the year: April in CAMS-TEMPO (presumably due to the low use of heating and cooling devices), July in TNO and EDGAR, and September in EMEP. In the UK, the patterns between the different datasets are more similar, with all of them reproducing a V shape, except for the relatively flatter NMVOC emission profiles in CAMS-TEMPO. Concerning the weekly and hourly profiles, important discrepancies are observed between CAMS-TEMPO and the factors proposed by TNO and EMEP for certain countries. In the CAMS-TEMPO weekly profiles for Austria the intensity of the weekend decrease is relatively higher, while the hourly profiles for Spain are flatter than the ones reported by TNO and EMEP.
Comparison of monthly
Comparison of monthly profiles for residential and commercial combustion emissions developed in the present work (CAMS-TEMPO) with profiles from EDGAR, EMEP and TNO for selected countries. The CAMS-TEMPO profiles represent the climatological weight factors (clim) based on the average of each month over all the available years (2010–2017).
The comparison of monthly weight factors for the residential and combustion sector indicates generally low discrepancies between the different datasets (Fig. 13). The largest difference is observed in Greece, where both CAMS-TEMPO and EMEP allocate most of the emissions during wintertime, while TNO and EDGAR propose a smoother transition between this season of the year and spring and autumn. Both the EDGAR and CAMS-TEMPO datasets propose an almost flat profile for India, where residential fuel is mainly used for cooking activities, an activity that can be considered constant throughout the year. For the daily temporal disaggregation of emissions, TNO and EMEP propose a fixed weekly profile (not shown) which disregards the daily dynamics inferred by the heating-degree-day approach considered in CAMS-TEMPO (Fig. 6).
For road transport, the differences between CAMS-TEMPO and other datasets
are quite significant. The monthly profiles reported by TNO, EMEP and EDGAR
are almost flat, while the pollutant- and meteorology-dependent profiles
developed in the present work suggest important decreases during summertime,
especially for the case of CO (Fig. 14a). At the weekly level, the TNO
profile is in line with most of the city-level constructed profiles,
although in some cases differences in the intensity of the weekend decrease
are observed (Fig. 4b). At the hourly level, the main discrepancies are
observed when comparing TNO with the Saturday and NMVOC evaporative profiles
of CAMS-TEMPO (Fig. 4d). It is worth noting that the hourly weekday profile
constructed for Utrecht is almost perfectly correlated with TNO (
Comparison of monthly profiles for road transport emissions
developed in the present work (CAMS-TEMPO) with profiles from EDGAR, EMEP
and TNO for selected countries and categories: CO gasoline exhaust
(GNFR_F1,
Comparison of monthly profiles for agricultural emissions
developed in the present work (CAMS-TEMPO) with profiles from EDGAR, EMEP
and TNO for selected countries and categories: NH
Figure 15 shows a comparison of the monthly profiles for the agricultural
sector. All the datasets selected for comparison, except for TNO, propose a
unique profile for all the different agricultural activities (i.e.
livestock, use of fertilizers and agricultural-waste burning), which is equally
applied to all emissions. The CAMS-TEMPO profiles constructed for NH
Gridded maps with all the temporal factors (monthly, weekly, daily and hourly)
per sector and year are available as NetCDF (Network Common Data Format) files for the global domain at a
resolution of 0.1
This paper presents the CAMS-TEMPO dataset, a collection of monthly, weekly,
daily and hourly emission temporal profiles for the priority air pollutants
(NO
The CAMS-TEMPO profiles were designed to be combined with the global and
regional anthropogenic emission inventories developed under the framework of
Copernicus (CAMS-GLOB-ANT and CAMS-REG_AP/GHG,
respectively) and to break down the original aggregated annual emissions to
finer temporal resolutions (up to hourly). In order to ensure this
combination, the developed temporal weight factors were constructed at a
global 0.1
There are several features that makes the CAMS-TEMPO profiles a major step
when compared to the datasets currently being used for air quality modelling
under the framework of CAMS:
Several CAMS-TEMPO profiles are analysed in this paper to illustrate their
main characteristics and potential. Moreover, an intercomparison exercise
of independent observational datasets (e.g. real-world measurements of
natural gas consumption, emissions and pollutant concentrations) and other
existing sets of temporal profiles was also performed. The comparison
between CAMS-TEMPO temporal weight factors and the measurement-based
profiles showed in general a high degree of correlation. Despite the
scarcity of independent measurements, our comparison suggests a high level
of representativeness of the developed profiles. On the other hand, the
comparison to other sets of temporal profiles showed important
discrepancies, especially for the traffic and agricultural sectors. This
comparison highlights some shortcomings of the global and regional profiles
currently used in the framework of CAMS, namely the omission of
meteorological influences and the neglection of the temporal variation of
emissions across sectors, species, and/or countries or regions.
It is important to highlight that the continuous growth of the open-data
movement has been a key element for the successful development of the
CAMS-TEMPO data. Services such as the European Open Data Portal (
The CAMS-TEMPO dataset provides an updated global and regional (European) picture of the temporal characterization of emissions by aggregated sectors. Despite all the efforts, there are, however, some limitations associated with the current version of the dataset that potential users of the CAMS-TEMPO should take into account.
First, the specificity of the computed profiles depends upon the degree of
sectoral disaggregation used to report the original CAMS inventories. For
instance, monthly profiles per industrial divisions could not be considered,
since all manufacturing emissions are reported under a unique sector in both
CAMS-GLOB-ANT and CAMS-REG_AP/GHG. Similarly,
fuel-weighted profiles that are spatially aggregated to the national level
were considered to be in the energy industry sector, as original
emissions are not split by fuel type. On the other hand, the split by fuel
type of traffic emissions in the CAMS-REG_AP/GHG inventory
allowed for considering meteorological influences associated with e.g. CO gasoline
and NO
A second limitation is related to the assumptions made when applying the
heating-degree-day approach for the residential and commercial combustion
sector. The computation of daily temporal factors was done considering a
threshold temperature and a fraction of non-space-heating activities
homogenous for all the world (15.5
Another important shortcoming of the current CAMS-TEMPO dataset is related to the scarcity of available information in developing regions (i.e. Africa, South America and Asia) to construct detailed profiles for the energy industry, manufacturing industry and road transport sectors. In the current version of the CAMS-TEMPO dataset, a simple gap-filling method has been implemented, which consists of mainly using the TNO European-based profiles when no national or local datasets are available. The rationale behind this choice is that the TNO profiles have been largely used and tested over the last decade in multiple international modelling exercises such as the Air Quality Model Evaluation International Initiative (AQMEII; Pouliot et al., 2015). However, TNO profiles are mostly based on western European data, and therefore the degree of representativeness for other world regions is a source of uncertainty. To address this constraint, the current gap-filling procedure will be reviewed in the future by constructing world region profiles for countries with geographical, climatological and/or behavioural similarities, following the approach presented by Crippa et al. (2020).
The CAMS-TEMPO dataset represents an effort to improve the temporal
characterization of emission data to be used for atmospheric chemistry
modelling. Future work will include the evaluation of these temporal
profiles when used for modelling activities. Through close cooperation with
air quality modellers, we expect to obtain feedback on the dataset as well as
suggestions for future improvements. Besides that, a number of future
updates have also been identified during the present work, including the following:
the extension of the year-dependent temporal profiles to be in line with the
current time span considered in the CAMS-GLOB-ANT (2000–2018)
and CAMS-REG_AP/GHG (2000–2017) inventories; the development of new temporal profiles for certain sectors and pollutants that
have not currently been considered, including CH the refinement of the heating-degree-day approach considering
region-dependent threshold temperature values.
Besides the identified updates, the investigation of monthly and seasonal
changes of emissions using satellite data will also be explored in future
work. Global and high-resolution observations of the atmospheric
composition provided by missions such as the Copernicus Sentinel-5 Precursor
(S5P) can be of great value to improve the description of the
spatiotemporal distribution of emissions (Lorente et al., 2019). Finally,
understanding and quantifying the impact of the COVID-19 lockdowns upon the
temporal distribution of emissions will also be an important topic to be
studied.
The CAMS-TEMPO profiles are distributed free of charge through the Emissions
of atmospheric Compounds and Compilation of Ancillary Data (ECCAD) system
(
Fixed monthly temporal profiles per domain, sector and pollutant. The definition of the sectors can be found in Tables 2 and 3.
Fixed weekly temporal profiles per domain, sector and pollutant. The definition of the sectors can be found in Tables 2 and 3.
Fixed hourly temporal profiles per domain, sector and pollutant (part 1: 00:00–01:00 to 11:00–12:00). The definition of the sectors can be found in Tables 2 and 3.
Fixed hourly temporal profiles per domain, sector and pollutant (part 2: 12:00–13:00 to 23:00–00:00). The definition of the sectors can be found in Tables 2 and 3.
The supplement related to this article is available online at:
MG conceived and coordinated the development of the CAMS-TEMPO profiles. CT helped construct the CAMS-TEMPO data files. NE and CG provided feedback on the construction of the global profiles and their combination with the CAMS-GLOB-ANT inventory. HDvdG and JK provided feedback on the construction of the European regional profiles and their combination with the CAMS-REG_AP/GHG inventory. SD processed and prepared the CAMS-TEMPO data files to make them available through the ECCAD system. OJ and CPGP helped conceive the CAMS-TEMPO dataset and supervised the work. MG prepared the paper with contributions from all co-authors.
The authors declare that they have no conflict of interest.
This article is part of the special issue “Surface emissions for atmospheric chemistry and air quality modelling”. It is not associated with a conference.
The present work was funded through the CAMS_81 (CAMS global
and regional emissions) contract, coordinated by the Centre National de la
Recherche Scientifique (CNRS, Claire Granier). The Copernicus Atmosphere
Monitoring Service (CAMS,
This research has been supported by the Copernicus Atmosphere Monitoring Service (CAMS), which is implemented by the European Centre for Medium-Range Weather Forecasts (ECMWF) on behalf of the European Commission (grant no. CAMS_81); the Ministerio de Ciencia, Innovación y Universidades (grant nos. RTI2018-099894-BI00, CGL2017-88911-R and RYC-2015- 18690); the Agencia Estatal de la Investigación (grant no. PID2019-108086-RA-I00 / AEI / 10.13039/501100011033); the AXA Research Fund; and the European Research Council (grant no. 773051, FRAGMENT).
This paper was edited by Mauricio Osses and reviewed by three anonymous referees.