2.1 Bottom-up emission calculation

Annual country-specific emissions are calculated using international activity data and emission factors, updated according to the latest scientific knowledge and following IPCC (2006a) methods. Emissions (EMs) from a given sector i in a country C accumulated during a year t for a chemical compound x are calculated with the country-specific activity data (AD), quantifying the activity for sector i, with the mix of j technologies (TECH) and with the mix of k (end-of-pipe) abatement measures (EOP) installed with share k for each technology j, the emission rate with an uncontrolled emission factor (EF) for each sector i and technology j and relative reduction (RED) by abatement measure k, as summarised in the following formula:

The activity data are very sector dependent and vary from fuel consumption in energy units (TJ) of a particular fuel type, to the amount (ton) of products manufactured, and to the number of animals or the area (hectares) and yield (ton) of cultivated crops. The technology mixes, (uncontrolled⁵) emission factors and end-of-pipe measures are determined at different levels: country-specific, regional, country group (e.g. Annex I/non-Annex I), or global. Technology-specific emission factors are used to allow a tier-2 approach, taking into account the different management/technology processes or infrastructures (e.g. different distribution networks) under specific “technologies”, and modelling explicitly abatements/reductions, e.g. the CH₄ recovery from coal mine gas at country level under the “end-of-pipe measures”. Just as with national inventories, EDGAR v4.3.2 starts from accounting over a period of time, 1 calendar year, and over a well-defined region, the country in which the emissions took place.

The sector-specific total emissions of substance x for country C in year t are then distributed in time and space using sector- and even technology-specific monthly shares m and spatial proxy datasets f. The proxy datasets are expressed as a function of coordinates (longitude, latitude) weighted at country level and with the Heaviside function equalling 1 when the grid cell belongs to the country area according to the following formula:

While the monthly shares are more specified in a generic way (only varying with the latitudinal band and with the sectors), the spatial proxy datasets take into account point-source information at sub-sector level (facilities) that can change from year to year.

2.2 Sector definition and data sources

Table 1a provides a structured overview of all the emission sources included in the EDGA Rv4 database. The energy-related sector (with the largest share in total GHG emissions) requires less detailed information on “technologies” than the agriculture- and waste-related sectors require on the “practices” applied⁶. This imbalance of the requirements for a higher level of detail for less important sources in terms of contribution to the national total is against the normal expectation (and time efficiency) of expending more efforts on those sources with the largest impact on the national totals (Pulles, 2018).

Table 1(a) Main category with all source/sink categories conforming to the IPCC (1996a, 2006b) Guidelines. Note that neither large-scale biomass burning nor land-use, land-use change and forestry emissions are included, although we do include biofuel combustion and agricultural activities (such as livestock and milk production, crop and rice production, agricultural waste burning, field burning, histosols and liming). (b) Data sources for activity statistics and emission factors for the main categories of emission sources defined in (a) (cf. references in the paper or the Supplement).

^a We note that (1) hard coal and brown coal data for 1970–1978 were split using the 1979 shares of the fuel types. (2) For the countries of the former Soviet Union and former Yugoslavia, the pre-1990 data was allocated to the countries using the same sector-specific country shares of the new countries from 1990. (3) We used “Serbia-Montenegro” in the dataset, which includes Kosovo and Montenegro. (4) For the lumped sum IEA regions “Other America”, “Other Africa” and “Other Asia”, the sector- and fuel-specific activity data have been disaggregated following the IEA definition of these regions and using the total production and consumption figures per country of coal, gas and oil from energy statistics reported by the US Energy Information Administration (EIA, 2014).
^b We note that (1) for Iceland, Israel and Mexico this is supplemented with the biofuel consumption reported by EIA (2014). (2) For Japan, Argentina, Brazil, China, India, Indonesia, Malaysia, Peru, Philippines and Thailand the biofuel data are supplemented with the data from US DA (2014).
^c We note that (1) abandoned and closed mines are taken up with very different shares to the total CH₄ emissions in 2012 from coal mining in UK (with 24 %), Romania, China, USA, Czech Republic, Germany and Ukraine (0.3 %). (2) CH₄ recovery from coal mining was estimated following IPCC (2006a) for the 11 countries with largest coal mining in the past. These are in decreasing order of the share of the total CH₄ emission from this sector (with absolute CH₄ recovery in 2012): Czech Republic (60 % with 41.1 kt CH₄ yr⁻¹), Spain (36 % with 5.0 kt yr⁻¹), Poland (33 % with 157.0 kt yr⁻¹), USA (29 % with 739.9 kt yr⁻¹), UK (25 % with 19.1 kt yr⁻¹), Germany (24 % with 45.0 kt yr⁻¹), Ukraine (16 % with 97.7 kt yr⁻¹), Australia (15 % with 186.0 kt yr⁻¹), China (9 % with 1974.1 kt yr⁻¹), Russia (3 % with 75.5 kton yr⁻¹), Kazakhstan (2 % with 10.0 kt yr⁻¹).
^d According to Peng et al. (2016) and Liu et al. (2015), Chinese underground coal mines are characterised by low quality coal and, as such, low EF, corresponding to the lower end of the range of EFs recommended by EMEP/EEA for coal mines in Europe.
^e The Common Agriculture Policy Regional Impact model (https://www.capri-model.org/dokuwiki/doku.php?id=start) provided tier-3 input to derive “technological parameters” and implied emission factors, representing country-specific practices.

Download XLSX

All the sources of Table 1a defined under the sectors and codes used in the IPCC (1996a, 2006b) guidelines, chap. 1 of vol. 1 Reporting Instructions and converted into the new IPCC (2006b) guidelines, chap. 8 of vol. 1 Guidance and Reporting, are considered, except the Land-Use, Land-Use Change and Forestry (LULUCF) sector.⁷ In contrast to the other sectors, LULUCF is not covered by annual, statistical assessments of the goods (“trees”), but needs geostatistical and/or remote sensing information as AD. For the emission sources and sinks related to carbon stock changes in the subcategory “Forest-land-remaining-forest-land”, we refer the reader to Petrescu et al. (2012), and for the large-scale biomass burning (including forest fires, savannah burning, grassland and woodland fires), we refer the reader to GFED (van der Werf et al., 2010), GFAS (Kaiser et al., 2012) or FINN (Wiedinmyer et al., 2011).

Most AD for EDGAR v4.3.2 are taken from international statistics and screened for completeness and consistency by EDGAR routines, removing outliers (clerical errors, wrong units) and gaps in time (missing single year) with a linear interpolation of the previous and following years. Preference is given to international statistics such as those of IEA (2014) and FAOSTAT (2014) over regional offices, such as EuroStat or national statistical bureaux, in order to profit from international definitions (e.g. for fuel types by IEA), inter-comparability amongst countries and the data quality and control by IEA or FAO. For China and the USA, national data from the Chinese Bureau for statistics and the US Energy Industry Administration respectively are consulted to assess and fill possible gaps in AD with consumption of fuels (fossil and bio) and of products (mainly metals and non-metallic minerals such as cement, chemicals, or solvents). For EU28, the biofuel statistics of EuroStat are used as they are updated more quickly than the IEA fuel statistics.

Where possible, GHG emission factors are selected from the IPCC (2006c) to ensure consistent and complete time series which are comparable across countries. The representativeness of default emission factors and the effectiveness of implemented control measures for the different regions are assessed based on expert judgement and by consulting annual Inventory Reports of Annex I countries to the UNFCCC (2014, 2016) or National Communications and Update Reports from some of the most important non-Annex I countries to UNFCCC (2014, 2012, 2017). Clean Development Mechanism projects (UNEP DTU, 2011) are taken into account in non-Annex I countries to account for abatement measures of CH₄ and N₂O emissions via CH₄ recovery from coal mining and landfills and N₂O reduction in nitric and adipic acid production.

Industrial process emissions have been calculated with the mineral production statistics of the US Geological Survey (USGS, 2014). For the agricultural activities we consulted the EU's Common Agricultural Policy Regionalised Impact (CAPRI) model and derived implied (weighted) emission factors which represent country-specific technologies and practices. For the waste sector we applied the IPCC First Order Decay (FOD) model of the IPCC (2006d) that is driven by the annual per capita generated municipal solid waste, the fraction deposited in landfills, and the fraction degradable organic carbon for the solid waste disposal emissions, whereas the chemical and biochemical oxygen demands are used to calculate the wastewater emissions.

Table 1b details the applied sector-specific and, where needed, region-specific data sources (activity statistics and emission factor with model parameters) on fuel balances, traded industrial products, crops, livestock and waste. For the agriculture and waste, a more detailed description with the model parameters is given under the “emission factor” heading. EDGAR v4.3.2 aims to collect all underlying human activity statistics and not to model the emissions directly as a function of the income, population or other proxy data. Table 1b uses the same main categories of Table 1a: energy, fugitive, industrial processes, solvents and products use, agriculture, waste and other (indirect N₂O emissions and fossil fuel fires). All emissions data can be downloaded also at subcategory level and are unambiguously identified with the IPCC (1996a) code.

2.3 Temporal profiles for the monthly distribution of the annual emissions

The legal reporting obligations under UNFCCC require time series of annual inventories, in line with the output of most national statistics infrastructures with accurate, annual accounting. For the atmospheric models, a higher temporal resolution is essential. Temporal profiles in EDGAR v4 were developed in 2010 under the European Commission's 6th Framework Programme Research projects CIRCE (Climate change and Impact Research: the Mediterranean Environment)⁸, because the global air quality models needed monthly disaggregated air pollutant emission grid maps as input. The temporal profiles are a bottom-up estimate of the monthly variations for major sectors, based on the insights of regional air quality models. Recently, the temporal profiles have been revised and extended, as documented by Crippa et al. (2019).

Table S4a summarises the sector-specific monthly profiles applied to the aggregated sectors for each GHG in the Northern Hemisphere. The largest variation is found in the temporal profiles for the agricultural sector (see Fig. S2a in the Supplement), then in the emissions from residential heating, and the smallest variation is present for the road transport and power generation sector. Covering regions from all over the world, a reverse profile is applied to the Southern Hemisphere, reflecting the opposite seasonality. No seasonal pattern is used for the equatorial region, defined within the range of [30^∘ S, 30^∘ N] latitude. For more refined time profiles (hourly) and in-depth analysis of the temporal distribution, we refer the reader to Crippa et al. (2019). Comparison of the EDGAR v4.3.2 monthly profiles and those used for other global emission products (Andres et al., 2011; Hoesly et al., 2018; Janssens-Maenhout et al., 2015) is given in Fig. S2b.

2.4 Proxy data for the spatial distribution of the country total emissions

For visualisation and as an input to atmospheric chemistry transport and climate models, the EDGAR v4.3.2 database distributes anthropogenic pollutant emissions over a uniform, global $\mathrm{0.1}{}^{\circ }\times \mathrm{0.1}{}^{\circ }$ grid defined with lower left coordinates. In emission inventories the emissions can be emitted either from a single point source or distributed over a line source (e.g. roads) or over an area source (e.g. agricultural fields), depending on the source sector or subsector. The line and area sources are distributed over the grid cells with the proxy data covering the globe entirely or partially, whereas the point sources are allocated to individual grid cells and reported as the area average of the sum of the point sources for that grid cell.

The proxy datasets that are used to grid different sector-specific sources are given in Table S4b. A detailed description is available in the EDGAR gridding manual (Janssens-Maenhout et al., 2013).

The spatial grid maps are graphical representations of the country totals, making use of spatial proxy data. EDGAR tries to allocate as much as possible the human activity to the places where it is likely located: to the place of the industrial facilities (several point source databases), or using the road network or the housing. Alternatives can be night light satellite data, as used by Oda et al. (2018) for those emission sources that were not yet covered with point sources (such as power plants) or the population data as proposed by Andres et al. (2016). We feel our proxy data to be more in line with our BU approach of allocating the sectoral emissions to the place of the emission source. We do not recommend an uncertainty analysis of the proxy data themselves, but a sensitivity assessment of the representativeness of the selected proxy data using atmospheric transport modelling. EDGAR v4.3.2 grid-map uncertainties are currently the subject of scrutiny and are being further investigated under European (Horizon 2020) research projects CO₂ Human Emissions (CHE, https://www.che-project.eu/) and the Observation-based System for monitoring and verification of greenhouse gases (VERIFY, http://verify.lsce.ipsl.fr/).

2.5 Uncertainty assessment of the greenhouse gas emissions

Uncertainties associated with emission of greenhouse gases stem from several sources, broadly described in vol. 1, chapt. 3, Sect. 3.1.5 of the IPCC (2006a) Guidelines. The uncertainties in this section are those caused by “statistical random sampling error”, which can primarily be thought of having an epistemic nature (lack of knowledge, thus reducible by gathering more data) but also including an aleatory component (uncertainty due to intrinsic randomness, and therefore uncompressible) (see e.g. Beven, 2016). As already pointed out by e.g. Gütschow et al. (2016), the heterogeneity of reporting, lack of documentation, differences, ranges of uncertainties, and sector aggregation all factor to make it difficult to compare, compile, and combine the multiple sources of information and to convey to a robust, coherent, estimate of uncertainty.

This section presents an analysis of the relative uncertainty per country grouping and gas, calculated using Eq. (1) and the parameters reported in Table 2, which also identifies a few countries as examples of GHG emissions reporting. (expressed in CO_2eq)⁹:

In accordance with the IPCC tiered approach to infer uncertainties to emission factors as well as to activity data, the analysis here assumes that countries belonging to the 24 member countries of the OECD in 1990 (24OECD90)¹⁰ were economically stable and that they would already have, or be able to build, a good statistical infrastructure and have the lowest uncertainties in their inventories. On the same line, the 16 countries with Economies in Transition of 1990 (16EIT90)¹¹ have experienced greater economic instability, and their inventories are more uncertain than those of the 24OECD90 but less uncertain than those from the other remaining non-Annex I countries. Exceptions to the country grouping are made for the following new or historic trading nations, China, Russia and India, because of global proliferation of emission-regulated goods, as Crippa et al. (2016b) analysed for air pollution.

Table 2Relative uncertainty of the GHG inventory for countries/country types (a) with the uncertainties per gas (b).

Download Print Version

All uncertainties are reported in Tables 3, 4 and 5 within twice the standard deviation (±2σ) of the mean value, corresponding to a 95 % confidence interval of the sample. This is a larger uncertainty range than the ±1σ selected by the Global Carbon Budget 2017 (Le Quéré et al., 2018), but is in line with IPCC recommendations. For comparative shares and trends in biofuel or non-CO₂ GHG emissions, data on gases and sources are much more uncertain than for fossil fuel CO₂. While Denier van der Gon et al. (2015) indicate that the biofuel combustion activity (and corresponding short-cycle carbon CO₂) is difficult to estimate for the different countries in Europe, Tian (2015) estimate the large uncertainties in CH₄ and N₂O budgets. The uncertainties in these emissions are caused by the scarcity and limited accuracy of the corresponding international activity statistics combined with the use of less representative country-wide emission factors (Olivier, 2002; Olivier et al., 2010). Using Eq. (1), the uncertainty estimate in the global total anthropogenic CO₂ emissions is ±9.0 %, that is, slightly higher than the estimate of 8.4 % by Andres et al. (2014), most probably since EDGAR v4.3.2 also includes the highly uncertain waste incineration, urea and liming activities (IPCC, 2006b, reports an uncertainty associated with the default emission factors for CO₂ of 40 %, for waste incineration), which are not part of the analysis by Andres et al. (2014).

Table 3Intercomparison of eight global CO₂ datasets (GCP, Le Quéré et al., 2016; PKU-FUEL, Wang et al., 2013; ODIAC2016, Oda et al., 2018; CDIAC, Andres et al., 2014; EIA, 2014; IEA, 2014; BP, 2017) with regard to their spatial and temporal coverage and their estimate of the global total Pg CO₂ per source for 2010 (and 2007 for PKU-FUEL).

^a GCP, ODIAC and BP have used more recent energy statistics than EDGAR and IEA (2014), which explains the major difference in global CO₂ emissions between them. ^b The difference in the calculation for IEA and EDGAR are mainly the different carbon factors used: IPCC (1996a) for IEA and IPCC (2006a) for EDGAR. In addition, EDGAR v4 supplements the charcoal production activity with fuelwood data of FAO, the venting and flaring activity with satellite data and the fossil fuel mine gas recovery with UNFCCC data, and EDGAR v4 calculates the transformation losses which IEA neglects. The main difference meanwhile disappeared as IEA updated the carbon factors with IPCC (2006a) values.

Download Print Version | Download XLSX

Table 4Intercomparison of the global total Pg CH₄ in 2010 by EDGAR v4.3.2 and by four other global emission inventories: USEPA (2012), GAINS-ECLIPSEv5 CH₄ of Höglund-Isaksson et al. (2015), Kirschke et al. (2013) and the global methane budget of Saunois et al. (2016). Note that the sector-specific global total is given in Tg CH₄ yr⁻¹ for 2010 and in brackets for 2000. The USEPA 2010 value is projected. For Kirschke et al. (2013), instead of 2010 (2000) we used the maximum (minimum) of the 2000–2009 range. For Saunois we used instead of 2010 (2000) the 2012 value (mean value of the 2000–2009 range). The 2010 values are bold, and the 2000 values are in italics.

Download Print Version | Download XLSX

Table 5Intercomparison of the global (EU) total Tg N₂O in 2005 by EDGAR v4.3.2 and by other European and global inventories: the European N Assessment of Leip et al. (2011) for EU27, GAINS Europe of Höglund-Isaksson et al. (2010) and GAINS global of Winiwarter et al. (2018), global total of US EPA (2012). The global values are bold, and the European values are given in italics between brackets.

Download Print Version | Download XLSX

CO₂ uncertainty can vary significantly among countries (Marland et al., 1999; Olivier et al., 2014). Larger uncertainties of about ±15 % are obtained for non-Annex I countries, whereas uncertainties of less than ±5 % are obtained for the 24OECD90 countries for the time series from 1990 (Olivier et al., 2015) reported to UNFCCC. For emissions of CH₄ and N₂O, we estimate uncertainties of ±32 % and ±42 % respectively for 24OECD90 countries and ±57 % and ±93 % for the other countries. These are based on the default uncertainty estimates of IPCC (2006a) and are in line with Bun et al. (2010). These are higher than the estimates of ±25 % and ±30 % by UNEP (2012) but justified by the large uncertainties reported by Tubiello et al. (2015) for the FAO activity statistics of ±30 % and ±50 % for crop and livestock.

As for the uncertainty of the emission grid maps, Fig. 9 of Andres et al. (2016)¹² reports a population map's uncertainty in excess of 150 % for Europe, the western USA, China, etc. Such uncertainty, when propagated into the emission calculations, will likely outweigh the combined uncertainty of activity data and emission factors, especially for CO₂. Also according to Hogue et al. (2016) is the uncertainty on CO₂ mapping that “with $\mathrm{1}{}^{\circ }\times \mathrm{1}{}^{\circ }$ grid cells for the United States is typically on the order of ±150 %”. In light of the high impact of spatial proxies on the overall uncertainty, the authors wanted to focus on a complete uncertainty assessment of the emission grid maps in collaboration with the atmospheric modelling community, evaluating carefully a useful covariance matrix, and refer to the ongoing sensitivity assessments in the CHE project mentioned in Sect. 2.4. Observation-based verification of European CH₄ and N₂O emissions using inverse modelling (e.g. Bergamaschi et al., 2015, 2018) indicates that the relatively low uncertainty estimates for some countries are not consistent with the relatively large uncertainty estimates of others, and for CH₄ a common uncertainty band in the upper range is considered more appropriate.

3.1 The global CO₂ budget

Table 3 summarises the main features of eight global CO₂ atlases and/or inventories, EDGAR v4.3.2, GCB (Le Quéré et al., 2016, 2018), the PKU-FUEL (Wang et al., 2013) and the ODIAC2016 (Oda et al., 2018; Oda and Maksyutov, 2011), CDIAC (Andres et al., 2014), EIA (2014), IEA (2014) and BP (2017) in temporal and spatial characteristics, sector break-down, methodology and CO₂ totals for major source categories in 2010 (which, for PKU-FUEL, was approximated by the latest available year, 2007).

Despite the substantially different levels of detail for the fuel use calculations, the global totals are relatively similar. At global level the differences in CO₂ emissions between IEA (2014) and EDGAR v4.3.2 are around 4 %, which can be explained largely by the difference in overall emission factors used (differences due to different default values for the carbon content and oxidation factors in IPCC, 2006a, and IPCC, 1996a). This yields 2 %, 1 % and 0.5 % higher CO₂ emissions from coal, oil and gas combustion respectively and increases overall fossil fuel emissions by about 1.3 %. In addition, the latest IEA statistics for recent years show more updated values for fuel consumption than for years further in the past. Marland et al. (1999) compared for the first time the EDGAR and CDIAC datasets. Andres et al. (2012) followed this further with a more detailed analysis of the differences between the global CO₂ datasets available in 2012, including the 2012 version of CDIAC, IEA, EIA and EDGAR v4.2 (EC-JRC/PBL, 2011). One of the remaining differences is that the flaring in EDGAR v4.3.2 is twice as high as in CDIAC and EIA, which is explained by the different estimation method for the activity data (reported energy statistics in CDIAC and EIA versus satellite night lights of flaring from NOAA-NGDC (2015) and Elvidge et al. (2016) in EDGAR). Although the different EDGAR datasets deviate by less than 0.5 % for Annex I countries, this deviation becomes 3.4 % for non-Annex I countries (see Fig. S3).

Larger differences are seen for the non-combustion CO₂ emissions. Figure 6a examines the most important ones comparing process emissions of the non-metallic sector (cement, lime, dolomite limestone, ceramics and glass production) of EDGAR v4.3.2, Le Quéré et al. (2016) and Xi et al. (2016). CO₂ from cement production in EDGAR v4.3.2 is 13 % (19 %) lower than in Xi et al. (2016) (based on CDIAC) because of the correction for the fraction of clinker in the cement produced. The EDGAR v4.3.2 data provide cement production emission estimates very close to the estimates of Andrew (2018) as reported in Figs. 3 and 4. A further large difference is found for developing countries, especially those with emerging economies. Figure 6b zooms in with the total CO₂ emissions regionally on China and compares EDGAR v4.3.2 estimates per sector with those of Guan et al. (2012) and Liu et al. (2015), who brought a large underestimation or overestimation in the Chinese CO₂ inventory to the broad attention of scientists and the media. Guan et al. (2012) indicated the 1.4 Gt CO₂ gap in the national total compared to the sum of the provincial statistics and proposed 9.1 Gt CO₂ in 2010. The EDGAR v.4.3.2 estimate of 8.8 t CO₂ for 2010 differs only by −3 %, which is composed of a difference of −19 % for the fossil fuel combustion emissions and of +27 % for the process emissions. In 2015 China revised its coal statistics with lower coal carbon content and the energy consumption was considerably decreased (for coal power plants with −12 %). Liu et al. (2015) published coal carbon content for 4200 Chinese mines and analysed the impact on the total CO₂ from combustion in China. EDGAR v4.3.2 revised the activity data for 1990–2012 and obtained for 2010 an emission reduction for power generation of −8 % and for the CO₂ total of −2 % only. Although Liu et al. (2015) reported 14 % lower emissions compared to EDGAR, this is effectively only 6 % (below the uncertainty range for China's CO₂ emissions) when correcting for the flaring, coke production, chemical production and limestone which were not accounted for in their study. This illustrates the importance of clearly documented datasets for data comparisons and further understanding the sources of discrepancies. The higher estimate of Liu et al. (2015) can be understood by his 3 % lower average net calorific value¹³ than the default of IPCC (2006a) used by EDGAR.

Figure 3Time series 1970–2012 of fossil fuel CO₂, CH₄ and N₂O global emissions from human activities excluding the LULUCF sector with global total (a) and European total (b). The stacked bars use AR4 GWP-100 values, whereas the dashed line and full line indicate the total CO_2eq of the three gases in the case where the SAR and AR5 GWP-100 values are respectively used.

Download

Figure 4Annual greenhouse gas time series 1970–2012 of EDGAR v4.3.2 with periodic error bar indication for the different types of countries with top emitters: (i) non-Annex I countries with China, India, Brazil and the rest of the non-Annex I countries, (ii) 24OECD90 countries with USA, EU15 and the remaining eight OECD countries of 1990, and (iii) 16EIT90 countries with Russia, EU13 and the remaining two newly independent Eurasian states. For the figures per gas we refer the reader to Fig. S4a–c.

Download

3.2 The global CH₄ budget

Table 4 compares the EDGAR v4.3.2 global CH₄ estimates of 0.34(±0.16) Pg CH₄ yr⁻¹ with four other global datasets (the bottom-up inventories of US EPA, 2012, and GAINS¹⁴ Eclipse v5 of Höglund-Isaksson, 2012); and the global budgets of Kirschke et al. (2013) and Saunois et al. (2016). Even though the global total CH₄ emissions for the bottom-up inventories vary by less than 4 %, global annual emissions from the agricultural and fossil fuel production sectors vary with ±22 % and ±17 % respectively. The top-down inventory estimates are 16 %–29 % larger than the bottom-up ones.

Figure 7a illustrates the origin of the large variations in the estimated fugitive emissions of oil and gas production (including extraction, transmission and distribution). Large uncertainties in CH₄ from venting and flaring at oil and gas extraction facilities have been reported by e.g. Lyon et al. (2015) or Peischl et al. (2015). The CH₄ venting of oil and gas extraction facilities is, in particular during the times of the Soviet Union, now believed to be larger than previously thought (e.g. in EDGAR v4.2 or US EPA), after Höglund-Isaksson (2017) used ethane–methane ratios as indicators. Additionally, gas distribution is a relatively large source of uncertainty, in particular in countries with old gas distribution city networks using steel pipes now distributing dry rather than wet gas, with potentially more leakages. Based on IPCC (2006a), EMEP/EEA (2009, 2013) and Marcogaz (2013), the emission factors for steel and grey cast iron pipelines vary in the range of 0.1–7 t km⁻¹ yr⁻¹, whereas this is a factor of 2 lower for PVC and polyethylene pipelines. The difference in composition of the gas distribution networks is taken into account in EDGAR v4.3.2 with country-specific variations in emission factors. The high CH₄ emissions during the natural gas transmission in the Russian reporting to UNFCCC (2016) might also account for all or part of accidental CH₄ releases, which are not negligible according to Höglund-Isaksson (2017). These are not included in the EDGAR datasets.

China is currently the top emitter of CH₄ because it has become the largest coal producer and it is a major rice cultivator. While the fugitive CH₄ emissions from coal production in China are increasing, emissions from rice cultivation are decreasing, as shown in Fig. 7b. The emission factor CH₄ ha⁻¹ yr⁻¹ for irrigated rice fields has been reduced from 1970 to 2000 by $\sim \mathrm{1}/\mathrm{3}$ by changing farming practices, as reported by Li et al. (2002), resulting in 0.47 kg CH₄ ha⁻¹ yr⁻¹ for the last decade. A comparison with Peng et al. (2016) illustrates the large range of emission factors used: the emission factor in EDGAR v4.3.2 for rice cultivation is twice as high as in Peng et al. (2016). Also for the coal mining the CH₄ emission factor for China in EDGAR v4.3.2 is 9 % higher than in Peng et al. (2016). EDGAR v4.3.2 revised emission factors for coal mining with local data from Peng et al. (2016), weighted by coal mine activity per province. These emission factors are at the lower end of IPCC (2006c) recommendations and yield EDGAR v4.3.2 estimates of 17.2 Tg in 2008 and 21.2 Tg in 2012, which are comparable to estimates of Peng et al. (2016) within ±2 Tg.

Total CH₄ emissions in EDGAR v4.3.2 in 2005 are 2 % (3 %) lower than in the v4.2 (4.1) version, which has been used in global inverse modelling studies of Monteil et al. (2011), Bergamaschi et al. (2013, 2015, 2018), Ganesan et al. (2015), Kort et al. (2008), and Miller et al. (2013). Except for the Chinese coal mining, no other major shortcomings to v4.2 were indicated in these global studies. More regional inverse modelling studies are nowadays able to “verify”¹⁵ the CH₄ emissions better (such as Henne et al., 2016, for Switzerland), and first atmospheric model runs with EDGAR v4.3.2 CH₄ emissions started recently. Total emissions have not changed significantly for either EU28 or the USA, but there are changes in the patterns of emissions: the −2.5 % (−0.2 %) change in the EU28 estimates of v4.3.2 compared to those of v4.2 (v4.1) is still within the range of the inverse model simulations of Bergamaschi et al. (2018), while the −4.7 % (−3.4 %) change in the USA in EDGAR v4.3.2 compared to v4.2(v4.1) is not in line with the suggested +50 %–70 % higher anthropogenic emissions based on the inverse modelling study of Miller et al. (2013). The latter might be explained on the emissions side by delayed reporting of statistics on fracking for shale gas and oil and the not well-characterised and highly uncertain emission factors as indicated by the US EPA (2015) and on the modelling side by large uncertainties in inverse models and the potential contribution of natural sources. For China the EDGAR v4.3.2 estimate for fugitive emissions from coal mining yields a 38 % lower CH₄ emissions total in 2008, which is in line with Saunois et al. (2016), Brandt et al. (2014) and Kirschke et al. (2013), suggesting lower CH₄ emissions in particular in northern China where coal mining takes place.

3.3 The global N₂O budget

An overview of the global N₂O budget is not yet available as for CO₂ and CH₄. Recent efforts from the modelling community to provide input for the global N₂O budget by Tian et al. (2018) report anthropogenic emission estimates for 2006 of 10.8 Tg N₂O yr⁻¹, confirming the 2005 global total by US EPA (2012) of 10.9, but a full overview of the global nitrous oxide budget is still forthcoming. The bottom-up estimate of EDGAR v4.3.2 of 7.2(±3.7) Tg N₂O yr⁻¹ for 2005 differs from this with 34 %, which is still within the uncertainty range. The bottom-up estimate of GAINS by Winiwarter et al. (2018) differs in a similar way by 29 %. It is noted that the differences within each source category remain very large (see Table 5). A comparison at European level between EDGAR v4.3.2 and the N-budget of Leip et al. (2011) shows relative moderate discrepancies also at sector-specific level for the total and agricultural sectors, 26 % or 37 % smaller estimates by EDGAR compared to Leip et al. (2011), but for the non-agricultural sectors, 19 % larger estimates. Höglund-Isaksson et al. (2010) provided GAINS estimates for EU27 that are respectively 28 %, 42 % and 20 % larger than the total, agriculture and non-agricultural sector estimates of Leip et al. (2011).

Although in EDGAR v4.3.2 the agricultural sector contributes most to the anthropogenic direct and indirect N₂O emissions, the production of chemicals, such as nitric acid, glyoxal, caprolactam and adipic acid, and its use as anaesthesia or for aerosol spray cans also play an important role. In 1970 the chemicals sector contributed 20 % to the total, but this has been significantly reduced to less than 8 % because of technological developments. Figure 8 shows the impact of technological developments from old plants to higher-pressure plants or plants with non-selective catalytic reduction, reducing the N₂O emissions by factors of 2 and 10 respectively. The N₂O emissions of nitric and adipic acid plant facilities which EDGAR v4.3.2 estimated are in line with the estimates of US EPA (2012) and by GAINS for the year 2005. However, a discrepancy evolves when looking at the 2010 values, because of the relatively large reduction between 2007 and 2010 in EDGAR and the relative constant trend in GAINS. While EDGAR assumes abatement technologies for nitric and adipic acid plants in China following the reporting under the Clean Development Mechanism, Schneider et al. (2010) assume that abatement was not used at least for the new adipic acid plants. The latter assumption was followed by Winiwarter et al. (2018) and explains the differences in the global nitric and adipic acid N₂O emission estimates between GAINS and EDGAR.

4.1 Global greenhouse gases 1970–2012

A country-based statistical analysis including 4 decades of GHG emissions (EDGAR v4.2) and GDP (Purchasing Power Parity data of the Penn World Tables 7 of Feenstra et al., 2013) was carried out to investigate the possible causality between emissions and income. The results, summarised in Paruolo et al. (2015), showed that no presence of causality could be statistically proven. This reflects a complex link between the very heterogeneous economic activities (ranging from manufacturing to services) and emissions, and justifies the meticulous bottom-up inventory compilation using statistics instead of modelling.

Figure 3 shows the global trend of GHG emissions in CO₂ equivalent (100-year time horizon), using the GWP-100 values of AR4 (IPCC, 2007)¹⁶. The GHG total is composed of all sources (excluding LULUCF) of CH₄ and N₂O but only CO₂ from long-cycle C fossil sources and excluding the short-cycle¹⁷ C for the CO₂ accounting, conforming to IPCC (2006a). The estimated global total GHG in 2010 of 44.7 Pg CO_2eq was shown to be 0.7 % lower than the estimates for the 2010 global total (without LULUCF) in the UNEP (2012, 2015) Emission Gap reports. The share of each gas to the total GHG is relatively stable and yields for CO₂ 76.8 % (+2.1 pp, −1.2 pp), for CH₄ 18.1 % (−2.5 pp, +1.9 pp) and for N₂O 5.1 % (+0.4 pp, −0.6 pp), where in between brackets the percent point impact of the evolution of the GWP-100 value from the SAR (IPCC, 1996b)¹⁸ to the AR5 (IPCC, 2014)¹⁹ is given.

In the global GHG emissions time series, the trend was shown to be dominated by CO₂ as it has the largest share and the largest increase. In the 1970s N₂O increased at the same rate as CO₂ (2.6 % yr⁻¹), while CH₄ was half as fast. In the 1980s and 1990s, N₂O and CH₄ increases were very small, while CO₂ continued albeit at a slower rate (1.6 % yr⁻¹). In the last decade, 2002–2012, CO₂ and CH₄ growth rates increased with respectively 3.2 % yr⁻¹ and 2.0 % yr⁻¹. While over the 4 decades (1970–2012) the global total GHG increased in line with global population (91 % versus 88 %), the inter-annual and regional emission variations do not always reflect the rates in population increase, but are instead better explained by the global fuel markets and economy, the 1973 and 1979 oil crises, the dissolution of the Soviet Union (1989–1991), the growth of the Chinese economy, after they joined the World Trade Organisation in 2002, and the 2008 global financial crisis.

4.2 Greenhouse gas trend analysis for regions and top emitting countries

Figure 4 shows the GHG trends for the major regions: 24OECD90 (split into USA, EU15 and the rest), 16EIT90 (with Russia and EU13 and the rest) and non-Annex I (for which China, India and Brazil are shown separately). The gas-specific GHG trend is also available per country in Janssens-Maenhout et al. (2017) and is downloadable from https://edgar.jrc.ec.europa.eu/overview.php?v=CO2andGHG1970-2016. To understand the trends of the total GHG (in CO_2eq), the decomposition with the trends of CO₂, CH₄ and N₂O for the same regions is given in Fig. S4a, b and c respectively and with a discussion per country group in the Supplement.

Focussing on the top four emitting countries and regions, Fig. 5 compares the reported UNFCCC (2004, 2012, 2014, 2016, 2017) emissions of China, USA, EU28, and Russia and the emission estimates of EDGAR v4.3.2. There is a very good agreement between the UNFCCC-reported values and the EDGAR v4.3.2 estimates for the EU28, whereas for USA and Russia the EDGAR v4.3.2 estimates are lower than those reported by UNFCCC (2016). For the USA this is explained by lower N₂O emissions in EDGAR v4.3.2, although N₂O emissions reported by USA to UNFCCC (2014, 2016) are within the large uncertainty range for the EDGAR v4.3.2 estimates. For Russia CH₄ emissions reported to UNFCCC (2016) are 37 % higher than those estimated by EDGAR v4.3.2, but this is also within the uncertainty range. The largest difference is found in the estimation of gas pipeline transmission emissions, which are 4 times higher in the UNFCCC inventory of Russia than in EDGAR v4.3.2. The relatively low emission factor for gas pipelines, used by EDGAR, is in line with the recommendations of Lelieveld et al. (2005). For China, a very good agreement between the EDGAR v4.3.2 estimate and the UNFCCC (2004, 2012, 2017) reported values is obtained, taking into account the importance of the coal statistics revision. In order to evaluate the latter effect, two time series of emission are calculated by EDGAR, with and without coal statistics revision. The revision includes a decrease in the 2010–2012 values and yields an increase for the 1990–2009 values of about +3 % for 2005 and 1994. It is evident that the previous estimates of the UNFCCC inventory in 2005 and 1994 would need to be revised in order to evaluate the emissions change from 2005 to 2012. Even if relative uncertainty in EDGAR estimates for China could be reduced, it is evident that the size of the Chinese inventory has a large impact on the global absolute uncertainty.

Figure 5GHG emissions of the largest emitting countries and regions (USA, EU28, Russia, China) of EDGAR v4.3.2 (solid line) with their uncertainty band compared to the reported UNFCCC time series of 2016 (dotted line). For China, two inventories were reported by national communications (1994, 2005), and a biennial update in 2017 added a new inventory value for 2012. The dashed yellow line gives the EDGAR v4.3.1 estimate of the Chinese GHG emissions using the energy statistics before the Coal Statistics Abstract (CSA) revision of October 2015.

Download

Figure 6Intercomparison of CO₂ emissions trends estimated by EDGAR and by others with (a) details for cement process emissions globally with data of Le Quéré et al. (2016) and Xi et al. (2016), and (b) details for China's sector-specific emissions with data of Guan et al. (2012) and Liu et al. (2015). The total is for all datasets subdivided into fossil fuel combustion and industrial process emissions (i.e. non-combustion industrial emissions, including cement).

Download

5.1 CO₂ emissions and urban hot spots

The 2012 grid map of CO₂ emissions from both long-cycle and short-cycle carbon in Fig. 9 with the relative sectorial breakdown for selected world regions (Europe, North America, Latin America, Africa, Middle East, Oceania, Russia and China) clearly shows the fossil fuel combustion activities, representing 90.6 % of the total CO₂ emissions. In this section we include for completeness biofuel emissions, which were omitted from the comparisons with UNFCCC reporting, because UNFCCC assumes carbon neutrality for all agricultural and biofuel CO₂ emissions in a country for any individual year. In the 24OECD90 countries 75.2 % of CO₂ emissions are produced by the power, road transport and residential sectors, while these sectors represent only 60.9 % in non-Annex I countries. The share of the industrial combustion and production sectors (mining/manufacturing) of non-Annex I countries reaches 36.8 %. The CO₂ shares of the fuel combustion in the power generation, road transport, buildings and manufacturing sectors vary for the different regions from 16 % to 50 %, from 5 % to 27 %, from 6 % to 39 % and from 9 % to 22 % of total emissions (see Table 6a and b) respectively. Interestingly, agricultural waste burning²⁰ represents 10 % of CO₂ emissions in Latin America (mainly due to sugarcane crop residue burning), and 22 % of CO₂ emissions in Africa derives from the transformation industry (charcoal production using as input primary solid biomass). Industrial emissions are distributed at the point-source locations of the power/heat plants or industrial facilities (e.g. cement factories) using the capacity of the plants or facilities as a weighting factor.

Figure 7Intercomparison of CH₄ emissions trends estimated by different versions of EDGAR and by others with (a) details for the CH₄ venting for oil and gas extraction, transmission and distribution with data of Höglund-Isaksson (2017) and (b) details for China's sector-specific emissions with data of Peng et al. (2016).

Download

Figure 8Global N₂O emissions trends for chemical processes, which mainly originate from nitric and adipic acid production (apart from smaller contributions from glyoxal and caprolactam production). The coloured area illustrates the penetration of technology for nitric acid production (with high-pressure plants, medium-pressure plants, low-pressure plants, plants with non-selective catalytic reduction and old plants) to reduce the emissions.

Download

In the grid maps hot spots are particularly visible over cities, of which the top four emit 2.75 % of the global total²¹ and coincide with the cities of Shanghai, Huangshi, Shenyuang and Moscow. In fact, 5 % of the $\mathrm{0.1}{}^{\circ }\times \mathrm{0.1}{}^{\circ }$ grid cells emit more than 5 Mt $(\mathrm{0.5}{}^{\circ }\times \mathrm{0.5}{}^{\circ }{)}^{-\mathrm{1}}$ yr⁻¹ and account for 34.08 % of the global total. It is therefore interesting to look at the contribution of the various sectors in megacities, as shown in Fig. 10. Emissions from the road transport sector (Fig. 10a) for the 20 selected cities seem to be more important in suburban areas than in the centre of the megacity. For power plants more heterogeneity was found (Fig. 10b), with larger power plants typically located on the periphery of the city in the 24OECD90 countries, while for major cities of the 16EIT90 and non-Annex I countries, several larger power plants are located within the central city areas. The remaining share of CO₂ emissions was shown to be mainly from the building sectors and the industrial manufacturing emissions.

Figure 9CO₂ emission grid map and relative contribution of EDGAR sectors in world regions (pie charts) for 2012. The legend for the PIE charts relates to the EDGAR sectors defined in Table S3: AGS: agricultural soils, AWB: agricultural waste burning, MNM: manure management, ENF: enteric fermentation, ENE: power industry, PRO: fuel production, PRU: production and use of products, REF: oil refineries, TRF: transformation industry, RCO: residential, TRO: road transport, TNR: non-road transport, WWT: waste water, SWD: solid waste disposal, FFF: fossil fuel fires, IND: manufacturing industry, IRO: iron and steel, CHE: chemicals, NEU: non-energy use, NFE: non-ferrous metals, NMM: non-metallic minerals, SOL: solvents, IDE: indirect emissions. The represented CO₂ emissions also include those from short-cycle carbon (i.e. of e.g. biofuel combustion and agricultural waste burning).

Download

Figure 10(a) Zoom of CO₂ emission grid maps over cities, representing the share of the road transport within the cities. The represented CO₂ emissions also include those from short-cycle carbon. (b) Zoom of CO₂ emission grid maps over cities, representing the share of the power plants within the cities. The represented CO₂ emissions also include those from short-cycle carbon.

Download

Figure 11(a) Difference in CO₂ emissions from buildings between 2012 and 1970. The represented CO₂ emissions also include those from short-cycle carbon. The figures for the long-cycle and short-cycle carbon separately are taken up in Fig. S5. (b) Difference in CO₂ emissions from transport between 2012 and 1970. The represented CO₂ emissions also include those from short-cycle carbon. The figures for the long-cycle and short-cycle carbon separately are taken up in Fig. S5.

Download

Figure 12CH₄ emission grid map and relative contribution of EDGAR sectors in world regions (pie charts) for 2012. The legend for the PIE charts relates to the EDGAR sectors defined in Table S3: AGS: agricultural soils, AWB: agricultural waste burning, MNM: manure management, ENF: enteric fermentation, ENE: power industry, PRO: fuel production, PRU: production and use of products, REF: oil refineries, TRF: transformation industry, RCO: residential, TRO: road transport, TNR: non-road transport, WWT: waste water, SWD: solid waste disposal, FFF: fossil fuel fires, IND: manufacturing industry, IRO: iron and steel, CHE: chemicals, NEU: non-energy use, NFE: non-ferrous metals, NMM: non-metallic minerals, SOL: solvents, IDE: indirect emissions.

Download

The evolution over time from 1970 to 2012 shows a different pattern for the residential sector than for the road transport sector. Figure 11a shows that while the residential sector decreased over these 4 decades in America and Europe, it increased in Asia and Africa. The difference in CO₂ emissions from the road transport sector meanwhile presents in Fig. 11b a more homogeneous picture with increases from 1970 to 2012 in almost all regions. Please note that Fig. 11 includes both long-cycle and short-cycle carbon fuel use, but Fig. S5a–d presents these separately and shows e.g. the use of the vegetal waste and dung for residential heating in India and the biofuel use for car transport in Brazil.

5.2 CH₄ emission maps

Because CH₄ is mainly released from fermentation processes (enteric, manure, landfills or rice) or diffusion processes (coal mine leakage or gas distribution losses), the 2012 CH₄ emission grid map with sector contributions for major world regions (Fig. 12) does not mirror the same human activities as the CO₂ map. The CH₄ shares for enteric fermentation, fossil fuel production and transmission, and solid and water waste treatment range from 9 % to 59 %, from 8 % to 68 % and from 11 % to 37 % of the global total respectively, depending on the region. For 24OECD90 countries enteric fermentation (with 31.1 % share), fossil fuel production (28.1 %) and landfills (21.4 %) are the three dominant sectors, whereas in the 16EIT90 countries, CH₄ emissions are dominated by fossil fuel production (49.4 % share). The non-Annex I countries show a similar high share of enteric fermentation and fossil fuel production to the 24OECD90 countries, but rice cultivation and domestic wastewater together give much higher emissions than solid waste disposal. Rice cultivation was shown to contribute significantly to the total CH₄ inventory of China (21.5 % or 14.2 Tg in 2012), which is almost 11 times the CH₄ emissions of rice cultivation in India (3.8 Tg), despite the larger area for rice fields in India than in China (425 000 compared to 303 000 km²). This is explained by the fact that India typically has one harvest per year from 1∕3 rain-fed fields and 2∕3 irrigated fields, whereas China has multiple harvests per year from irrigated rice fields. Rain-fed rice fields in India are modelled with a 5 times lower emission factor than the irrigated fields in China. Figure 13a and b show the opposing trends with mainly positive 2012–1970 increments in enteric fermentation (mainly cattle) (a) and mainly negative increments in CH₄ emissions from rice cultivation (b). The CH₄ trend from rice cultivation in Asia was shown to be remarkably stable, with the exception of Thailand, where increased activity is noticed. The remaining non-Annex I countries of Africa and Latin America show similar high contributions from enteric fermentation (25.8 Tg versus 20.9 Tg respectively in 2012). However, the total CH₄ emissions from the African continent are higher than those of Latin America because of the 3.5 times larger CH₄ emissions from fossil fuel production (gas and oil production). Interestingly, both continents show significant CH₄ emissions from charcoal production, which compares to 16 % (Africa) and 15 % (Latin America) of their gas and oil production emissions of CH₄.

Figure 13(a) Difference in CH₄ emissions from enteric fermentation between 2012 and 1970. (b) Difference in CH₄ emissions from rice cultivation between 2012 and 1970.

Download

Figure 14CH₄ emissions from fossil fuel production in 2012 with zoom on areas with intense coal mining (within green frame) and gas and oil production activities with venting (within blue circle). The shipping lines are representing the CH₄ leakage during transmission of oil tanker transport as fugitive emissions from the fuel and not as combustion emissions from the tanker.

Download

Hot spots of CH₄ are estimated for fossil fuel production, typically at gas and oil production facilities or at coal mines, as shown in Fig. 14. In North America a shift over the period 2005–2012 from coal mining in the north-east (−21 %) to gas and oil production in particular in North Dakota, Montana and Texas (+65 %) took place. The USA is nowadays the largest producer of both shale gas and tight oil, which make up almost half of total US gas and oil production (EIA, 2015). In Europe a much larger decrease of −87 % in coal production happened earlier, while gas production increased by 30 %. Consequently the EU28 needed to rely on oil and gas imports and expanded its transmission and gas distribution network, with a corresponding increase in CH₄ leakages. Apart from the USA, the Middle East was also shown to be a global world player on the oil and gas market, shifting from oil production (with a decrease of 71 % over the period 1976–1985) to gas production (with a 9.3-fold increase from 1985 to 2012), mainly driven by Iran, Saudi Arabia and Qatar. The African countries with the highest CH₄ emissions from fossil fuel production were in decreasing order of importance Algeria and Nigeria (for oil and gas production) and South Africa (for coal mining). Similarly to Nigeria, which showed an approximate doubling of CH₄ emissions from oil (and gas) production over the last 4 decades, Mexico and Venezuela also showed similar levels of CH₄ emissions from oil and gas production (increasing by a factor of 1.6 over the 4 decades). For gas production, Russia has shown the largest CH₄ venting and leakage, overtaking the USA in 1985.

Coal mining has become important for China, which since 1982 has been the largest bituminous coal producer in the world, overtaking the USA. Moreover, China was shown to also be the largest coal importer since 2011 (overtaking Japan), as domestic coal produced in mainly the western and northern inland provinces of China faced a bottleneck in transportation, lacking southbound rail lines (Tu, 2012) towards the southern coast that has the highest coal demand. Not only did EDGAR v4.3.2 revise the country-specific coal mining emission factors, but the spatial distribution was also considerably updated with hot spots at the location of the mining activity. For coal mine activities in China (split into brown and hard coal), the coal mine database of Liu et al. (2015) provided over 4200 coal mine locations, which is 10 times more than that available for EDGAR v4.2. For Europe, the closure of mines since the 1990s has been taken into account using the European Pollutant Release Transfer Register (EPRTR, 2012).

Figure 15N₂O emission grid map and relative contribution of EDGAR sectors in world regions (pie charts) for 2012. The legend for the PIE charts relates to the EDGAR sectors, defined in Table S3: AGS: agricultural soils, AWB: agricultural waste burning, MNM: manure management, ENF: enteric fermentation, ENE: power industry, PRO: fuel production, PRU: production and use of products, REF: oil refineries, TRF: transformation industry, RCO: residential, TRO: road transport, TNR: non-road transport, WWT: waste water, SWD: solid waste disposal, FFF: fossil fuel fires, IND: manufacturing industry, IRO: iron and steel, CHE: chemicals, NEU: non-energy use, NFE: non-ferrous metals, NMM: non-metallic minerals, SOL: solvents, IDE: indirect emissions.

Download

Figure 16Difference between 2012 and 1970 in N₂O emissions from fertiliser use on agricultural soils.

Download

5.3 N₂O emissions including indirect sources

Unlike the CO₂ and CH₄ grid maps, the gridded N₂O emissions for the year 2012 in Fig. 15 with the share of different sectors for world regions showed a quite uniform global coverage distribution, due to the predominance of soil emissions and indirect emissions (distributed with the N-deposition map of Dentener et al., 2006), also from the sea's surface. Over land, most N₂O is emitted from the agricultural soils (the use of animal manure as fertiliser, the application of N-containing fertilisers and cattle in pasture), representing 35 % to 86 % of total N₂O emissions depending on the region. Fertilising farmland with pasture or animal waste as fertiliser or crop residues has not increased so much as the use of nitrogen fertilisers. Figure 16 shows the increased use (by the difference 2012–1970) of nitrogen fertiliser, in particular in Asia.