Policy Research Working Paper 10391 Identifying and Monitoring Priority Areas for Methane Emissions Reduction Susmita Dasgupta Somik Lall David Wheeler Development Economics Development Research Group April 2023 Policy Research Working Paper 10391 Abstract This paper identifies high-priority areas for methane are outnumbered four to one by cases with increasing emissions reduction and estimates recent emissions changes emissions. The paper also analyzes trends in high-priority in those areas using atmospheric concentration data from areas for seven major methane source sectors (agricultural the European Space Agency’s Sentinel-5P satellite platform. soils, livestock, gas, oil, coal, landfills, and wastewater) The modeling approach is illustrated with three case studies: and finds only two where emissions decreases outnumber landfills in Spain (Madrid), irrigated rice production in increases (gas and oil). Among World Bank income groups, India (Karnal district, Haryana state), and oil production decreases outnumber increases in high-income economies in Iraq (Al Amarah district, Maysan governorate). For each but increases are hugely dominant in the other three groups. case, the paper estimates two change models by fixed effects: The paper concludes with a presentation of summary the monthly trend in methane concentration from January emissions trend reports for all 775 high-priority areas, with 2019 to November 2022, and the difference between mean accompanying maps and an Excel file. As satellite-based concentration in 2022 and the previous three years. The monitoring becomes more widely employed, such reports paper estimates the change models for 775 high-priority will provide a useful template for judging further progress areas and finds that cases with decreasing methane emissions toward fulfillment of the Global Methane Pledge. This paper is a product of the Development Research Group, Development Economics. It is part of a larger effort by the World Bank to provide open access to its research and make a contribution to development policy discussions around the world. Policy Research Working Papers are also posted on the Web at http://www.worldbank.org/prwp. The authors may be contacted at sdasgupta@worldbank.org. The Policy Research Working Paper Series disseminates the findings of work in progress to encourage the exchange of ideas about development issues. An objective of the series is to get the findings out quickly, even if the presentations are less than fully polished. The papers carry the names of the authors and should be cited accordingly. The findings, interpretations, and conclusions expressed in this paper are entirely those of the authors. They do not necessarily represent the views of the International Bank for Reconstruction and Development/World Bank and its affiliated organizations, or those of the Executive Directors of the World Bank or the governments they represent. Produced by the Research Support Team Identifying and Monitoring Priority Areas for Methane Emissions Reduction Susmita Dasgupta* Somik Lall David Wheeler World Bank The research is funded by the Knowledge for Change Trust Fund, administered by the World Bank. The findings, interpretations, and conclusions expressed in this paper are entirely those of the authors. They do not necessarily represent the views of the International Bank for Reconstruction and Development/World Bank and its affiliated organizations, or those of the Executive Directors of the World Bank or the governments they represent. * Authors’ names in alphabetical order. 1. Introduction The World Meteorological Organization forecasts that the current greenhouse gas (GHG) emissions trend will increase global temperature 3-5 degrees C by 2100 (Reuters 2018). This would far overshoot the 2-degree limit pledged by the 2015 Paris Climate Accord (COP 21) and might have a catastrophic impact (Steffen et al. 2018; World Bank 2012). In response, several industrial nations pledged very steep emissions reductions at the Leaders’ Summit on Climate in April 2021. The response deepened at COP 26 in Glasgow with the introduction of the Global Methane Pledge, whose 122 current participating nations have joined a collective effort to reduce global methane (CH4) emissions at least 30 percent from 2020 levels by 2030. The Pledge highlights the particular importance of methane reduction: “Methane is a powerful but short- lived climate pollutant that accounts for about half of the net rise in global average temperature since the pre-industrial era. Rapidly reducing methane emissions from energy, agriculture and waste … is regarded as the single most effective strategy to keep the goal of limiting warming to 1.5˚C within reach while yielding co-benefits including improving public health and agricultural productivity.”1 The Global Methane Pledge is a laudable commitment by the international community, and broad-based participation will build political support for emissions reduction. At the same time, a rapid transition may be assisted by an evidence-based global effort that identifies the highest- priority CH4 emissions sites and mobilizes the latest technology to monitor their progress. This paper attempts to contribute with a multi-level approach. First, we perform a spatial analysis of global CH4 emissions to assess potential gains from priority site identification. Then we use the World Bank’s gridded database of satellite-measured methane concentrations at 5-km resolution to establish a template for tracking methane emissions reduction at priority sites. We demonstrate its potential effectiveness with case studies for large sources of methane emissions from energy, agriculture and waste. Then we extend the template to create and analyze emissions trend reports for 775 high-priority global sites. The remainder of the paper is organized as follows. Section 2 analyzes the global distribution of methane emissions sources using “bottom-up” estimates from the Emissions Database for Global Atmospheric Research (EDGAR).2,3 In Section 3, we use the results to identify high-priority sites where rapid reduction of methane emissions would accelerate the global transition. Section 4 develops a strategy for sampling satellite-measured CH4 concentrations in sites that account for 50% of global methane emissions. Section 5 introduces our monitoring approach with illustrative case studies for landfills (Spain), irrigated rice production (India) and oil production (Iraq). In Section 6, we extend our approach to 775 high-priority areas and use the results to analyze the distribution of emissions trends globally, by source sector, and by World Bank income group. Section 7 presents the priority-area reports in detailed tables, maps and an accompanying Excel file, while Section 8 summarizes and concludes the paper. 1 https://www.globalmethanepledge.org/#about 2 https://edgar.jrc.ec.europa.eu/ 3 EDGAR’s “bottom-up” approach estimates emissions by combining sectoral activity data with broadly-calibrated emissions factors (Crippa et al. 2020; Solazzo et al. 2021). 3 2. The Global Distribution of Methane Emissions The global standard for “bottom-up” methane emissions reporting is the Emissions Database for Global Atmospheric Research (EDGAR (2022)), which combines detailed sectoral activity data with emissions factors that are adjusted for broad regional differences (Crippa et al. 2020; Solazzo et al. 2021). EDGAR reports estimated emissions4 by sector in raster files with a resolution of 0.10 degrees (approximately 10 km), yielding information for 6,480,000 grid cells. Table 1 reports global totals from these files by sector in 2018, the latest year for which complete estimates are available. Table entries are displayed in descending order of emissions, which total 375.2 million tons.5 For expositional clarity and simplicity, we confine our analysis to the seven sectors that account for 90% of global methane emissions. Two are in agriculture (Enteric Fermentation, Agricultural Soils), three in energy (Gas, Coal, Oil Exploitation) and two in the waste sector (Waste Water Handling, Solid Waste Landfills). Priority site selection is most effective when the spatial distribution of emissions is dominated by relatively few large sources. We assess the potential effectiveness of priority-setting in this case by calculating total methane emissions for countries’ Level 1 administrative units (GADM1s). We expand the dataset to include emissions from countries’ marine exclusive economic zones (EEZs), because many oil and gas operations are offshore. We tabulate estimated emissions in descending order, calculate the cumulative percentage of total emissions, and divide the results into deciles. For example, the first decile comprises the largest-emitting areas whose total emissions account for 10% of the global total. Figure 1 provides a summary table and graph for emissions from 3,812 areas that comprise Level 1 ADM1 and EEZ units. The tabIe reveals a highly-skewed spatial distribution; in the top decile, 7 areas account for 10% of global methane emissions. In the next four deciles, 10% of emissions are accounted for by 14,18, 28 and 40 areas, respectively. In summary, only 2.8% (107) of the 3,812 areas account for 50% of global methane emissions. Emissions deciles for all areas are mapped in Figure 2, which shows that the 39 areas in the top 3 deciles (accounting for 30% of total methane emissions) are concentrated in relatively few countries. Table 3 displays these areas in descending order of emissions, along with sectoral percent shares. Among the 11 countries represented, China accounts for 17 areas, followed by India and Brazil (6 each), Pakistan and the US (2), and Qatar, the Russian Federation, Saudi Arabia, the Islamic Republic of Iran, Nigeria and Indonesia (1). 4 For brevity, we refer to anthropogenic emissions as emissions in this paper. 5 As a greenhouse gas, methane is far more potent than CO2. Applying the standard conversion factor (25), total methane emissions are the equivalent of 9.4 Gt CO2. 4 Table 3 summarizes Table 2 in a cross-tabulation of dominant aggregate sectors by country. Of the 39 areas in the top 3 deciles, methane emissions are dominated by agriculture in 23, fuel in 15 and waste in 1. Agricultural emissions have important roles in 4 countries (China, Brazil, India and Pakistan), while fuel-based emissions dominate in the remaining 7 (United States, Indonesia, Islamic Republic of Iran, Nigeria, Qatar, Russia, Saudi Arabia). Table 1: Global methane emissions, 20186 Emissions Cum. EDGAR Sector (mt) 2018 % Enteric Fermentation 108.4 28.9 Waste Water Handling 44.4 40.7 Fuel Exploitation - Gas 43.6 52.3 Agricultural Soils 37.9 62.4 Fuel Exploitation - Coal 36.7 72.2 Solid Waste Landfills 34.9 81.5 Fuel Exploitation - Oil 32.2 90.1 Manure Management 12.3 93.4 Energy For Buildings 12.3 96.7 Oil Refining and Transformation Industry 6.3 98.3 Agricultural Waste Burning 2.0 98.9 Solid Waste Incineration 1.5 99.3 Road Transportation No Resuspension 1.0 99.5 Combustion For Manufacturing 0.7 99.7 Power Industry 0.5 99.8 Chemical Processes 0.3 99.9 Iron And Steel Production 0.2 100.0 Fossil Fuel Fires 0.2 100.0 Railways Pipelines Off-Road Transport 0.0 100.0 Total 375.2 CO2 Conversion Factor 25.0 C02 Equivalent 9,381.2 Source: EDGAR (2022) 6 For detailed sector definitions and descriptions, see: EDGAR Sector-Specific Gridmaps, 1970 – 2021 https://edgar.jrc.ec.europa.eu/dataset_ghg70#sources IPCC 2006 Reporting Guidance and Tables https://www.ipcc-nggip.iges.or.jp/public/2006gl/pdf/1_Volume1/V1_8_Ch8_Reporting_Guidance.pdf 5 Figure 1: Global CH4 emissions from level-1 administrative units and EEZs, by decile CH4 250 Areas in Decile (GADM1 + EEZ) Cum. (mt) Decile Areas Areas 2018 200 1 7 7 31.8 2 14 21 34.2 150 3 18 39 32.2 4 28 67 34.6 100 5 40 107 33.1 6 69 176 33.2 50 7 124 300 33.3 8 232 532 33.2 0 9 492 1024 33.3 10 2788 3812 33.2 10 20 30 40 50 60 70 80 Total Methane Emissions Decile 6 Figure 2: Global methane emissions by decile: Level-1 administrative units and exclusive economic zones 7 Table 2: Top 3 deciles, global CH4 emissions – Level-1 administrative units and EEZs Sector Shares (%) of CH4 Emissions Agriculture Fuel Waste CH4 Solid Waste (mt) Enteric Ag. Waste Water Decile State/Province Country 2018 Top Sectors Ferment. Soils Gas Oil Coal Landfills Handling 1 Shanxi China 8.93 Fuel 1.2 0.1 0.6 0.2 93.5 1.6 2.9 1 Texas United States 4.27 Fuel 21.0 0.5 59.6 10.2 0.8 5.8 2.1 1 Punjab Pakistan 4.05 Agriculture 66.8 13.6 1.5 0.1 0.0 5.8 12.3 1 Henan China 3.86 Fuel 13.3 6.2 0.6 0.0 54.6 9.9 15.4 1 Shandong China 3.74 Fuel 11.1 2.3 1.9 0.6 58.3 10.0 15.8 1 Uttar Pradesh India 3.62 Agriculture 56.5 13.9 0.9 0.3 0.8 2.6 24.9 1 Kalimantan Timur Indonesia 3.39 Fuel 0.2 1.5 4.3 2.4 90.5 0.2 0.9 2 Hunan China 3.16 Agriculture 8.6 58.0 0.2 0.0 8.2 8.2 16.9 2 Anhui China 3.11 Agriculture 7.4 35.0 0.5 0.0 36.8 7.3 12.9 2 Khuzestan Iran, Islamic Rep. 2.67 Fuel 1.5 0.4 42.7 53.8 0.0 0.3 1.3 2 Guangdong China 2.59 Agriculture 9.9 42.1 0.9 0.0 3.2 17.2 26.6 2 Sichuan China 2.58 Agriculture 17.2 37.6 1.0 0.0 8.6 12.4 23.1 2 Ash Sharqiyah Saudi Arabia 2.41 Fuel 1.2 0.0 36.9 56.2 0.0 2.6 3.1 2 Jiangsu China 2.39 Agriculture 3.8 47.6 1.1 0.1 13.6 13.6 20.3 2 Jiangxi China 2.28 Agriculture 7.8 59.9 0.6 0.0 8.7 7.7 15.3 2 São Paulo Brazil 2.24 Waste 36.2 0.1 1.7 1.4 0.1 21.7 38.8 2 Minas Gerais Brazil 2.19 Agriculture 73.8 0.3 0.9 0.2 0.0 9.6 15.4 2 Guangxi China 2.18 Agriculture 24.3 47.3 0.8 0.0 3.1 7.7 16.9 2 North Dakota United States 2.16 Fuel 6.8 0.0 77.1 15.1 0.9 0.0 0.1 2 Madhya Pradesh India 2.13 Agriculture 64.7 7.5 0.3 0.0 7.8 1.6 18.0 2 Kemerovo Russian Federation 2.09 Fuel 0.6 0.0 1.5 0.1 94.4 2.4 1.0 3 Sind Pakistan 2.07 Agriculture 54.7 10.1 11.0 8.7 1.7 4.7 9.1 3 Maharashtra India 2.05 Agriculture 50.7 6.4 2.3 1.0 5.5 2.4 31.7 3 Hebei China 2.04 Fuel 16.0 4.0 1.9 0.2 40.2 14.2 23.4 3 Hubei China 2.01 Agriculture 9.9 56.3 1.6 0.1 1.5 10.6 20.1 3 Mato Grosso Brazil 1.98 Agriculture 94.9 0.9 0.2 0.0 0.0 1.7 2.3 3 Heilongjiang China 1.96 Fuel 10.9 22.0 2.1 0.8 41.5 7.0 15.8 3 Rajasthan India 1.95 Agriculture 73.6 0.7 2.3 0.8 0.3 1.6 20.6 3 West Bengal India 1.92 Agriculture 44.0 28.2 0.3 0.0 6.5 2.2 18.8 3 Qatar_EEZ Qatar 1.77 Fuel 0.0 0.0 80.3 19.7 0.0 0.0 0.0 3 Mato Grosso do Sul Brazil 1.70 Agriculture 93.9 0.2 0.3 0.0 0.0 1.6 3.9 3 Odisha India 1.68 Agriculture 36.1 25.0 0.6 0.1 18.7 1.1 18.4 3 Goiás Brazil 1.64 Agriculture 88.3 0.4 0.2 0.1 0.0 4.3 6.7 3 Liaoning China 1.64 Fuel 7.5 14.4 2.8 0.8 46.7 9.6 18.3 3 Yunnan China 1.57 Agriculture 33.6 25.9 1.3 0.0 3.8 11.2 24.3 3 Zhejiang China 1.56 Agriculture 2.5 58.7 1.0 0.1 0.9 14.8 22.0 3 Nigeria_EEZ Nigeria 1.56 Fuel 0.0 0.0 15.3 84.7 0.0 0.0 0.0 3 China_EEZ China 1.54 Fuel 0.0 0.0 69.4 30.6 0.0 0.0 0.0 3 Rio Grande do Sul Brazil 1.53 8 Agriculture 73.5 4.0 1.8 1.4 2.1 6.8 10.5 Table 3: Top 3 deciles – dominant sectors by country Country Agriculture Fuel Waste Total China 10 7 0 17 Brazil 5 0 1 6 India 6 0 0 6 Pakistan 2 0 0 2 United States 0 2 0 2 Indonesia 0 1 0 1 Iran, Islamic Rep. 0 1 0 1 Nigeria 0 1 0 1 Qatar 0 1 0 1 Russian Federation 0 1 0 1 Saudi Arabia 0 1 0 1 Total 23 15 1 39 Figure 3 displays emissions shares by decile for countries identified by World Bank income group. It reveals roughly equal roles for lower- and upper-middle-income countries after the 3rd decile. High- income countries rise to parity with the middle-income countries by the 5th decile, while low-income countries trend modestly upward from the 3rd decile to an emissions share of 13% in the 10th. Figure 3: CH4 Emissions shares of World Bank income groups by decile 9 3. The Political Geography of Methane Monitoring 3.1 The Spatial Distribution of Methane Emissions Methane emissions come primarily from fossil fuel extraction and processing, livestock grazing, irrigated rice production and waste disposal. Table 4 shows that the spatial distribution of methane emissions sites is hugely skewed. Of 6.5 million cells in the EDGAR 10 km global raster, only 142 (0.002% of total cells) account for 10% of global methane emissions and 21,057 (0.325%) account for 50%.7 Table 4: Spatial clustering of global methane emissions EDGAR Cum. Cum. Decile Cells Cells % 1 142 142 0.002 2 578 720 0.011 3 1,775 2,495 0.039 4 5,817 8,312 0.128 5 12,745 21,057 0.325 6 22,364 43,421 0.670 7 37,919 81,340 1.255 8 64,334 145,674 2.248 9 124,529 270,203 4.170 10 6,209,797 6,480,000 100.000 Figure 4 displays the global distribution of the 21,057 cells that account for 50% of methane emissions. The map reveals widely-scattered cell clusters in all global regions except South and East Asia, which also have very large clusters in some subregions. Figure 4 motivates a cluster-concentrated strategy, because the cost-effectiveness of emissions-reducing interventions rises with spatial clustering.8 Table 4 poses a corollary challenge for the global community: The Global Methane Pledge is only likely to succeed if significant reductions can be achieved in the spatial clusters that account for 50% of global methane emissions. Beyond those clusters, methane emissions sites become progressively more scattered and interventions more costly. 3.2 Local Accountability This paper highlights potential benefits from combining “top down” satellite monitoring of emissions changes with “bottom up” estimates of emissions magnitudes. However, these benefits can only be realized by effectively-administered interventions. Effective administration will in turn require the informed participation of governing institutions at the appropriate geographic scale. In this case the best match is provided by the second-level administrative units whose territories include important methane emissions sites. Accordingly, this paper develops and illustrates a strategy for satellite 7 We include the full global grid in these calculations, because many large methane sources are offshore oil and gas extraction sites. 8 Monitoring, regulatory enforcement and abatement investment all tend to be lower-cost for activities that are concentrated in relatively small areas. 10 Figure 4: Areas that account for 50% of global methane emissions 11 monitoring of methane emissions from the high-priority second-level administrative units that account for 50% of total emissions. Figure 5 provides a detailed illustration of our approach for an important methane emissions cluster in Iraq. Figure 5(a) overlays a map of Iraq governates with 10 km EDGAR cells whose methane emissions place them among 21,057 cells (Table 4) in the top five deciles (accounting for 50% of global emissions). The cells are color-coded by decile. For this exercise we highlight the emissions cluster in the east-central part of Maysan governate. Enlarging the provincial map (Figure 5(b)) reveals that the principal cluster is in one second-level administrative unit, Al Amarah district. Within this district the emissions cells are roughly clustered along a north-south central axis. The key to effective monitoring in this context is mobilization of high-resolution Sentinel-5P observations within a methane emissions cluster area. The measurement area should be large enough to ensure a robust sample of high-quality satellite measures,9 but small enough to ensure that area-average measures reflect cluster activities. For this exercise, we adopt the approach illustrated in Figure 5(c). Using the R programming language, we compute the convex hull that includes all EDGAR 5-decile sites in Al Amarah district. We construct a 20-km buffer around the hull to enlarge the sample of proximate Sentinel-5P observations. Then we intersect the buffered polygon with the boundaries of Al Amarah to define the sampling area (with blue boundary) for Sentinel-5P observations. 3.3 Global Monitoring Areas We extend the previous illustration to global coverage in a multi-step exercise. For terrestrial emissions, we adopt GADM level 2 as the default standard. For countries without GADM 2 units, we adopt the highest GADM level available (0 or 1). For emissions from offshore gas and oil sites, we create geographic units of comparable scale by intersecting countries’ exclusive economic zones (EEZs) with a global grid whose cells have sides of 250 km. In all, this combined terrestrial/offshore exercise yields 51,976 global geographic units (GGUs). We compute total EDGAR emissions in the latest available year for each GGU, order by total emissions, compute cumulative emissions, and divide units by decile. Table 5 displays the results. Table 5: Global geographic units by EDGAR emissions decile Cum. Decile GGUs GGUs 1 39 39 2 93 132 3 174 306 4 303 609 5 522 1,131 6 888 2,019 7 1,674 3,693 8 3,411 7,104 9 7,164 14,268 10 37,708 51,976 9 Image quality can be degraded by extensive cloud cover or random technical problems. 12 Figure 5: Sentinel-5P sampling area for Al Amarah District, Iraq 13 Table 5 shows that of 51,976 GGUs, 1,131 units in the top 5 deciles account for 50% of global methane emissions. For these high-priority GGUs, we simplify the monitoring problem by confining our exercise to emissions-intensive (5-decile) EDGAR cells in the set of 21,057 cells that account for 50% of global methane emissions (Table 4). Within each GGU, we compute the convex hull that includes all EDGAR 5-decile cells in the unit, construct a 20-km buffer around the hull, and intersect the buffered polygon with the GGU boundaries to define the sampling area for Sentinel-5P observations. This generates sampling areas for 1,001 high-priority GGUs.10 Figure 5 maps these units, revealing significant representation for on- and offshore GGUs in all world regions. 4. Sample Data We have constructed our atmospheric CH4 concentration database from observations by the ESA’s Sentinel-5P platform that typically number over 200,000 per day. We locate each observation in a 5-km global grid comprising 25,920,000 cells. The dominant component in each observation is the global stock of atmospheric CH4 molecules that have accumulated during the past decade. The second component is seasonal, determined by natural processes over the annual cycle. The third component is geographic: Emitted CH4 molecules tend to remain within their latitude bands of origin for significant periods before full global mixing. The fourth observation component is determined by local anthropogenic CH4 emissions, which remain near their point of origin long enough to be distinguished from the three “background” components. We term this the “concentration anomaly” since it measures the local deviation from the global background CH4 concentration. As previously noted, its primary sources are emissions from agriculture, fossil fuel production and waste disposal. Sentinel-5P (S5P) operates in a Sun-synchronous near-polar orbit with an equatorial crossing at 13:30 local solar time. It completes 14 orbits of the Earth per day, with a site revisit time of one day and a current spatial resolution of 5.5 x 3.5 km. We use the S5P L2 Offline georeferenced measures of XCH4 (the column average dry air mixing ratio of methane11), corrected for bias associated with XCH4 dependence on surface albedo.12 Recent global research by Sha et al. (2021) has shown that S5P CH4 measures correspond very closely to independent measures taken at ground stations maintained by TCCON (Total Carbon Column Observing Network).13 Sentinel-5P’s bias-corrected XCH4 data exhibit a difference of 10 The remaining 5-decile GGUs do not contain any 5-decile EDGAR cells. 11 The total atmospheric column between the surface and the top of the atmosphere normalized to the corresponding dry air column. 12 For more technical details, see the documentation at https://sentinel.esa.int/documents/247904/3541451/Sentinel-5P-Methane-Product-Readme-File 13 Current TCCON sites are located in the United States, China, Canada, Germany, Poland, France, Japan, Australia, New Zealand, the Republic of Korea, Réunion, and Ascension Island. 14 Figure 6: Global geographic units, EDGAR top 5 deciles EDGAR Decile 1 2 3 4 5 15 −0.26±0.56 % when compared with the corresponding TCCON data. Per the ESA’s recommendation, we use only pixels with quality values greater than 0.5. The data have been downloaded from the ESA’s AWS repository managed by MEEO (Meteorological Environmental Earth Observation).14,15 We filter the bias-corrected XCH4 data for local concentration anomalies, or differences between observed and background CH4 at each point. 16 We calculate background CH4 using the methodology of Hakkarainen et al. (2019), which isolates local anomalies from the three background components described above. We compute the daily median XCH4 for each 10- degree latitude band and linearly interpolate the result to each S5P observation with 1-degree resolution. Following Hakkarainen, we use the median as the representative value because it is not skewed by extreme observations. We subtract this background value to compute the local anomaly for each observation. Then we compute monthly mean values of concentration anomalies for each 5-km grid cell in our database. For this exercise, we overlay the 5-km grid with the sampling polygons for our high-priority GGUs. Numerous technical and atmospheric factors produce highly-varied sample counts. To ensure robust estimates, we confine our exercise to GGUs with at least 20 observations in each year. Table 6 provides summary statistics on sample counts for the 775 GGUs that meet this condition. The median GGU has 2,033 S5P observations for the period 2019 – 2022, while the least- represented GGU has 92 observations and the most-represented has 77,011. Table 6: Statistics for GGUs with minimum yearly S5P sample of 20 GGU S5P Observations, 2019 - 2022 GGUs Min P10 P25 P50 P75 P90 Max 775 92 464 980 2,033 3,861 8,123 77,011 5. Priority Area Monitoring – Illustrations This exercise aims to use methane data from the Sentinel-5P platform to track changes in emissions-intensive areas identified by EDGAR. In the previous section, we developed a methodology for sampling S5P data for those areas. We illustrate our approach for three areas dominated by important CH4 sources: Madrid Province, Spain [two large landfills]; Karnal District, Haryana State, India [irrigated rice production]; and Al Amarah District, Maysan Governorate, 14 https://www.meeo.it/ 15 The data are available in NetCDF format at the ESA’s AWS repository, https://meeo- s5p.s3.amazonaws.com/OFFL/L2__CH4, with filenames catalogued at https://scihub.copernicus.eu/catalogueview/S5P/ 16 For further discussion, see Dasgupta, Lall and Wheeler (2022, 2023). Iraq [oil production]. We previously used Al Amarah to illustrate our methodology for identifying sampling areas. 5.1 Landfills in Madrid Province, Spain Recent technical research has used satellite data from Sentinel-5P and other sources to detect methane flares from large emissions sources (Maasakkers et al. 2022; Ouerghi et al. 2021). Particular attention has focused on emissions flares from two large landfills – Valdemingómez and Pinto -- southeast of Madrid, Spain (Tu et al. 2022; ESA 2021). Figure 7 provides summary information about these two sites. Figure 7(a) displays the GGU (Madrid Province) identified for this exercise, along with the sampling area for Sentinel-5P data (bordered in blue).17 Both landfill sites are identified in the figure. Figure 7 (b) displays satellite images of the two sites, along with the results of methane plume detection on August 20, 2021 (ESA 2021). Methane concentrations in the plume are colored from red (highest) to blue (lowest). Figure 7 (c) draws on our 5-km global database to display mean methane concentration anomalies for the GGU sampling area from January 2019 to November 2022. Our data corroborate the recent studies, identifying the landfill cells and their immediate neighbors as the highest-intensity areas in the Madrid GGU. The data also indicate that prevailing winds typically disperse emissions from the two sites toward the southwest. Our 5-km database contains 3,300 observations for the GGU sampling area. Figure 8 plots the observations against time, with mean monthly observations superposed. The plot reveals significant annual variation in 2019 – 2021, with peaks in the first quarter of each year. We see no sign of a “covid effect” in these data, since the most notable variation is in 2019 and the annual series for 2020 looks very similar to the series in 2021. Although some monthly fluctuation remains in 2022, the previous first-semester “peaking” is not present. This exercise aims to mobilize global GGU-level information like the Madrid data for monitoring methane emissions trends in high-priority areas. To date, satellite-based estimation of physical emissions volumes from methane plumes has relied on relatively sophisticated models that utilize local atmospheric data for specific periods (Maasakkers et al. 2022; Varon et al. 2019; Cusworth et al. 2021; Zhang et al. 2020.). Such tailored exercises are undoubtedly critical for monitoring the performance of individual methane emitters that commit to specific reductions in emissions. However, their data- and model-intensities have precluded global scale-up to date. 17 Per Section 3.3, this is the 20-km-buffered convex hull surrounding all GGU EDGAR points in the top 5 deciles for global methane emissions. 17 Figure 7: Methane emissions from Madrid landfills Satellite Imagery Source: ESA (2021) 18 Figure 8: Madrid GGU, monthly CH4 anomalies Our approach is more modest technically, but it permits immediate scale-up to the global arena for monitoring progress at many sites under the Global Emissions Pledge. Using the 5-km global database, we develop indicators of the direction, size, significance and sectoral locus of changes in atmospheric methane concentration anomalies for high-priority GGUs that satisfy the sampling conditions for robust estimation. We use the Madrid data to introduce two econometric trend models: (1) 4 = 0 + 1 + (2) 4 = 0 + 1 + where, for GGU i in month t: CH4ijt = Mean CH4 anomaly (ppb) in 5-km sampling cell j t = Time from initial period in months DF = Dummy variable for the final observation year (2022) εijt = Random error term In model (1), overall change during the period January 2019 – November 2022 can be inferred from the sign, size and statistical significance of the econometric estimate for the trend 19 parameter (γ1). Model (2) provides more immediate feedback: The difference between the average CH4 anomaly in the most recent year and the average for all preceding years can be inferred from the sign, size and significance of parameter (δ1).18 Although the Sentinel-5P platform provides the best available CH4 data, its coverage for our 5- km grid cells is not necessarily complete for all sample months. From an econometric perspective, this implies an unbalanced panel for estimation. Within the Madrid GGU, typical CH4 concentration anomalies can differ substantially across cells (see Figure 7c). To cite one possible consequence, a naive trend analysis could generate spuriously-positive results in cases where early-period observations are more numerous in lower-anomaly cells and later observations are more concentrated in higher-anomaly cells. We control for this possibility with a panel estimator that incorporates cell-specific fixed effects. Table 7 reports estimates for the Madrid GGU from OLS, Random Effects and Fixed Effects models. In this case the two trend models align with very high significance in all cases. The monthly trend (γ1) is negative and highly significant, yielding an estimated decline of about 0.19 ppb/month (or about 9 ppb over four years). The final-year deviation is also negative and highly significant, yielding an estimated difference of about 3.5 parts per billion between 2022 and the previous three years. Table 7: Trend model results, Madrid GGU Dependent variable: Monthly Mean CH4 Anomaly (1) (2) (3) (4) (5) (6) Panel Panel Panel Panel OLS RE FE OLS RE FE Month -0.186*** -0.191*** -0.195*** (-9.88) (-10.32) (-10.46) [Year = 2022] -3.343*** -3.503*** -3.635*** (-7.07) (-7.56) (-7.81) Const. 5.929*** 6.101*** 6.194*** 1.641*** 1.710*** 1.743*** (9.96) (9.53) (10.53) (5.89) (4.64) (6.39) N 3,300 3,300 3,300 3,300 3,300 3,300 t statistics in parentheses ="* p<0.05 ** p<0.01 *** p<0.001" 18 Results for models (1) and (2) can diverge in cases where a large discontinuity emerges in the most recent year. For example, a modest uptrend from 2019 to 2021 followed by a large decrease in 2022 can produce a small, positive, statistically-insignificant estimate for γ1 and a negative, statistically-significant estimate for δ1. 20 In summary, our results for the Madrid GGU indicate significant reduction in methane emissions during the past four years, coupled with continuing progress during the most recent year. While the Madrid landfills undoubtedly remain high methane emitters among Western European sites, the evidence from our database suggests that their emissions are at least headed in the right direction. 5.2 Irrigated Rice Production in Karnal, Haryana, India Our second case focuses on irrigated production of rice, the staple crop for the majority of the world’s population (Adhya et al. 2014). Table 1 shows that Agricultural Soils account for about 10% of global methane emissions, and irrigated rice production accounts for most of those (Smith, Reay and Smith 2021). In 2013, rice was harvested on 165 million hectares of land in 100 countries, with 90 percent of global production in Asia. Irrigated fields occupy about 80 million hectares and produce 75 percent of the global crop (FAO 2014; Fischer et al. 2014). Rice production in flooded fields produces methane because oxygen does not penetrate the soil when it is blocked by water. This promotes the growth of methane-producing bacteria. We identify rice paddy flooding months using RiceAtlas, a spatial database of global rice calendars developed by Laborte et al. (2017). RiceAtlas distinguishes up to three rice cultivation seasons and identifies the first flooding and harvest days for each season by day-of-year. We have chosen Karnal district in India for this case because it is identified by RiceAtlas as an area with only one of three potential rice cultivation seasons. This permits a clear illustration of the utility of the 5 km methane database for tracking methane emissions across seasons and years. In Karnal, fields are typically flooded from early summer through October. Figure 9 provides summary information about Karnal. Figure 9(a) displays the GGU (Karnal District, Haryana) along with the sampling area for Sentinel-5P data (bordered in blue). Figure 9(b) draws on our 5-km global database to display mean methane concentration anomalies for the GGU sampling area from January 2019 to November 2022. Our 5-km database contains 2,747 observations for the Karnal GGU sampling area. Figure 10 plots the observations against time, with mean monthly observations superposed. The plot reveals the close correspondence between rice paddy flooding months and annual peak periods for mean CH4 anomalies. Table 8 reports estimates for the Karnal GGU from OLS, Random Effects and Fixed Effects models. Results for the two models align in all cases. For Karnal the monthly trend (γ1) is positive and highly significant, yielding an estimated increase of about 0.40 ppb/month (or about 19.2 ppb over four years). The final-year deviation is also positive and highly significant, yielding an estimated difference of about 7.8 parts per billion between 2022 and the previous three years. 21 Figure 9: Methane emissions from Karnal District, Haryana State, North India 22 Figure 10: Karnal Haryana GGU, monthly CH4 anomalies Table 8: Trend model results, Karnal Haryana GGU Dependent variable: Monthly Mean CH4 Anomaly (1) (2) (3) (4) (5) (6) Panel Panel Panel Panel OLS RE FE OLS RE FE Month 0.397*** 0.397*** 0.400*** (15.86) (15.86) (15.80) [Year = 2022] 7.803*** 7.803*** 7.891*** (10.94) (10.94) (10.96) Conts. 13.66*** 13.66*** 13.56*** 21.73*** 21.73*** 21.71*** (19.18) (19.18) (18.82) (58.74) (58.74) (58.21) N 2,747 2,747 2,747 2,747 2,747 2,747 t statistics in parentheses ="* p<0.05 ** p<0.01 *** p<0.001" 23 5.3 Oil Production in Al Amarah District, Maysan Governorate, Iraq Section 3.2 and Figure 5 provide a detailed introduction to Al Amarah district that illustrates our methodology for identifying global geographic accountability units and appropriate areas within those units for sampling data from Sentinel-5P. Figure 11 displays mean CH4 anomalies from January 2019 to November 2022 for the Al Amarah GGU sampling area. In the figure, the areas of greatest CH4 intensity are in the southern and western parts of the sampling area. Figure 11: Mean CH4 anomalies, Al Amarah District, 2019 – 2022. Our 5-km database contains 3,775 observations for the Al Amarah sampling area. Figure 12 plots the observations against time, with mean monthly observations superposed. In contrast to the Madrid and Karnal GGUs, the Al Amarah series have no apparent trend or regular annual periodicity. These impressions are confirmed by Table 9, in which neither trend model has any significance for any estimator. 24 Figure 12: Al Amarah Maysan GGU, monthly CH4 anomalies Table 9: Trend model results, Al Amarah Maysan GGU Dependent variable: Monthly Mean CH4 Anomaly (1) (2) (3) (4) (5) (6) Panel Panel Panel Panel OLS RE FE OLS RE FE Month -0.0203 -0.00725 -0.00725 (-1.54) (-0.60) (-0.60) [Year = 2022] 0.112 0.385 0.404 (0.29) (1.08) (1.14) Conts. 7.475*** 7.086*** 7.147*** 6.933*** 6.782*** 6.853*** (19.88) (12.19) (20.80) (33.76) (13.60) (37.13) N 3,775 3,775 3,775 3,775 3,775 3,775 t statistics in parentheses ="* p<0.05 ** p<0.01 *** p<0.001" 25 6. Global Trends for High-Priority GGUs In this section, we extend our sampling and modeling approach to global scale. Among 51,976 Global Geographic Units, we have identified 1,131 GGUs that account for 50% of global methane emissions in the EDGAR database. After limiting our exercise to GGU sampling areas with at least 20 Sentinel-5P observations per year, we estimate trend models (1) and (2) by fixed effects for 775 high-priority GGUs. 6.1 Overall Results Table 10 provides a summary that aggregates the econometric results into four categories: No Trend (statistically-insignificant results for both models); Mixed (statistically-insignificant results for one model or oppositely-signed but significant results for two models); Decreasing (significant negative results for both models); and Increasing (significant positive results for both). We find No Trend for 56 GGUs because neither model attains classical significance (p=0.05). Results are Mixed for 150 GGUs, Decreasing for 106 and Increasing for 453. In summary, a substantial majority of 775 GGUs have experienced unambiguous increases in CH4 emissions during the past four years, whether judged by the overall trend or change in the most recent year. The same picture emerges if we weight the regression results by EDGAR emissions. Table 10: GGU estimation results (fixed effects) Emissions- Weighted Results Count % % No Trend 56 7.2 6.5 Mixed 160 20.6 23.3 Decreasing 106 13.7 13.1 Increasing 453 58.5 57.1 Total 775 100 100.0 6.2 Sectoral Results Table 11 explores the sectoral distribution of results weighted by EDGAR emissions. Soils and Livestock exhibit enormous disparities between Increasing and Decreasing cases (90.2 % vs. 0.6% for the former; 63.6% vs. 4.6% for the latter). The fuels sectors reveal striking differences, with a huge imbalance toward Increasing for Coal (66.7% vs. 5.1%) while Gas and Oil are both dominated by Decreasing (44.7% vs. 25.3% for the former; 23.8% vs. 10.6% for the latter). The two waste sectors have large imbalances toward Increasing (52.4% vs. 18.0% for Landfills; 61.0% vs. 18.3% for Wastewater). In summary, Oil and Gas exhibit more Decreasing results but the two sectors only account for 27.2% of EDGAR emissions. The remaining sectors all exhibit huge dominance for Increasing results. 26 Table 11: GGU estimation results by sector (fixed effects) % of Emissions-Weighted Estimation Results % of Priority Sectora No Trend Mixed Decreasing Increasing Emissions Soils 0.0 9.2 0.6 90.2 20.4 Livestock 2.9 28.8 4.6 63.6 19.9 Coal 5.7 22.5 5.1 66.7 25.1 Gas 4.5 25.5 44.7 25.3 14.5 Oil 26.7 38.9 23.8 10.6 12.7 Landfills 13.0 16.5 18.0 52.4 3.1 Wastewater 1.0 19.7 18.3 61.0 4.2 a Sectoral abbreviations (from Table 1): Soils [Agricultural Soils]; Livestock [Enteric Fermentation]; Coal [Fuel Exploitation – Coal]; Gas [Fuel Exploitation – Gas]; Oil [Fuel Exploitation – Oil]; Landfills [Solid Waste Landfills]; Wastewater [Waste Water Handling] 6.3 Results by World Bank Income Group Our results by World Bank income group (Table 12) suggest that global accountability has begun to affect methane emissions intensity in high-income countries. They have an imbalance toward Decreasing (34.6% vs. 24.0%), but they account for only 14.5 % of EDGAR emissions for our priority GGUs. The other three income groups have very large imbalances toward Increasing. The greatest imbalance (65.2% vs. 9.0%) is in upper-middle-income countries, which also dominate methane emissions (56.5% of the total). Table 12: GGU estimation results by World Bank income group % of Emissions-Weighted Estimation Results % of Priority Income Group No Trend Mixed Decreasing Increasing Emissions Low 4.9 27.5 15.1 52.6 1.7 Lower Middle 8.0 23.8 10.4 57.8 27.3 Upper Middle 6.2 19.6 9.0 65.2 56.5 High 5.0 36.4 34.6 24.0 14.5 27 7. Reporting GGU Emissions Changes The models and results presented in this paper can offer potentially valuable information to participants in the Global Methane Pledge. Accordingly, we provide illustrative reports in tabular and graphical formats. Table 13 reports results for EDGAR Decile 1 GGUs, which are dominated by Increasing in Coal (10 districts in China and Russia) and Livestock (4 districts in Pakistan and China). There are 2 other GGUs with Increasing (Soils in Chongqing, China; Landfills in Durango, Mexico). Only 2 GGUs have Decreasing, both in Gas (Aksu District, China and Al Gharbia District, UAE). Among the remaining GGUs in Table 13, 7 have Mixed results and 1 has No Trend. Complete results for GGUs in all five EDGAR deciles are provided in the accompanying Excel file19. Figures 13 and 14 map our results for the top 4 EDGAR deciles.20 Map icons are centered over their GGUs, color-coded by emissions change categories and bordered by color codes for dominant GGU sectors. To cite two examples: All Algerian icons are Decreasing with Gas- dominated GGUs; all Sudanese icons are Increasing with Livestock-dominated GGUs.21 Overall, the results can be roughly summarized as follows: Europe: Almost all Decreasing [Landfills, Wastewater] Russia: Rough parity in Increasing, Mixed, Decreasing [Oil, Gas, Coal] North Africa: Almost all Decreasing [Oil, Gas] Sahelian Africa: Rough parity in Increasing, Mixed [Livestock] Southern Africa: Mostly Mixed [Coal] Gulf Region: Decreasing or Mixed [Oil, Gas] Central Asia: Majority Decreasing [Gas] Pakistan: Majority Mixed [Livestock] India: Almost all Increasing [Livestock, Soils, Coal] Other South and Southeast Asia: Almost all Increasing [Soils, Livestock] China: Dominated by Increasing [Soils, Livestock, Coal, Wastewater] North America: Majority Increasing, along with Mixed [Oil, Gas, Landfills] Mexico: Increasing [Landfills] South America: Decreasing [Gas, Wastewater]22 19 https://datacatalog.worldbank.org/int/search/dataset/0064309/methane_emissions_changes_by_edgar_decile 20 The maps are visually cluttered by spatial “overcrowding” when 5th-decile GGUs are also included. 21 To minimize clutter in the maps, we include only the names of the countries with EDGAR 1 st-decile results. 22 One important caveat: Our regional coverage is biased by the relative scarcity of clear imagery because of cloudiness in the West African coastal region, the Amazon and Central African rain forest regions, and large swathes of Southeast Asia. Inspection of Figure 1 shows that numerous GGUs would be included for these areas if images were more plentiful. 28 Table 13: CH4 emissions changes, EDGAR Decile 1 Country State/Province District Sector Model_1 Model_2 Overall Canada Saskatchewan Division No. 1 Oil Decreasing Increasing Mixed China Beijing Beijing Livestock Increasing Increasing Increasing China Chongqing Chongqing Soils Increasing Increasing Increasing China Henan Zhengzhou Coal Increasing Increasing Increasing China Shandong Jining Coal Increasing Increasing Increasing China Shandong Tai'an Coal Increasing Increasing Increasing China Shanxi Yangquan Coal Increasing No Trend Mixed China Shanxi Changzhi Coal Increasing Increasing Increasing China Shanxi Shuozhou Coal Increasing Increasing Increasing China Shanxi Linfen Coal Increasing Increasing Increasing China Shanxi Taiyuan Coal Increasing Increasing Increasing China Shanxi Jincheng Coal Increasing Increasing Increasing China Shanxi Luliang Coal Increasing Increasing Increasing China Xinjiang Uygur Aksu Gas Decreasing Decreasing Decreasing Iran, Islamic Rep. Ilam Dehloran Oil No Trend No Trend No Trend Iran, Islamic Rep. Khuzestan Ahvaz Oil No Trend Increasing Mixed Iraq Diyala Khanaqin Oil No Trend No Trend No Trend Mexico Durango Durango Landfills Increasing Increasing Increasing Pakistan Punjab Dera Ghazi Khan Livestock Increasing Increasing Increasing Pakistan Punjab Bahawalpur Livestock Increasing Increasing Increasing Pakistan Punjab Multan Livestock Increasing Increasing Increasing Pakistan Sind Hyderabad Livestock Increasing Decreasing Mixed Russian Federation Kemerovo Novokuznetskiy Coal Increasing Increasing Increasing South Africa Mpumalanga Nkangala Coal Decreasing No Trend Mixed United Arab Emirates Abu Dhabi Al Gharbia Gas Decreasing Decreasing Decreasing United States North Dakota McKenzie Gas Decreasing Increasing Mixed 29 Figure 13: CH4 Emissions Changes, Top 4 EDGAR Deciles – Europe, Africa, Asia Figure 14: CH4 Emissions Change, Top 4 EDGAR Deciles – North, Central, South America 8. Summary and Conclusions This paper has identified high-priority sites for methane emissions reduction and introduced a global monitoring approach that employs data from the ESA’s Sentinel-5P (S5P) satellite platform. Drawing on geolocated S5P observations that typically number over 200,000 per day, we compute monthly mean methane concentration anomalies for each cell of a 5-km global grid. Our data span the period from January 2019 to November 2022. To highlight local accountability, our monitoring approach estimates trends in S5P atmospheric concentration data for district- level (GADM2) administrative units and scale-equivalent areas within national exclusive economic zones (EEZs) where offshore oil and gas extraction occur. These global geographic units (GGUs) comprise 51,976 areas. The spatial distribution of methane emissions is so highly skewed that 2.2% of these areas account for 50% of anthropogenic methane emissions in “bottom -up” estimates provided the Emissions Database for Global Atmospheric Research (EDGAR). We focus on these 1,131 high-priority GGUs, drawing on EDGAR to identify their dominant sources of emissions from seven sectors (Agricultural Soils, Livestock, Gas, Oil, Coal, Landfills and Wastewater). We illustrate our monitoring approach with three GGU case studies: landfills in Spain (Madrid); irrigated rice production in India (Karnal district, Haryana state); and oil production in Iraq (Al Amarah district, Maysan governorate). Each case employs S5P data from the GGU subregion that includes all EDGAR emissions points in the point set that accounts for 50% of global emissions. These points are often spatially clustered, and our approach ensures that our trend models focus on areas where emissions are most highly concentrated. In each case, we estimate two trend models by fixed effects. The first model estimates the monthly trend for the period January 2019 – November 2022, while the second focuses on recent history by estimating the difference between mean concentrations in 2022 and the previous three years. We extend this approach to 775 high-priority GGUs that have sufficient S5P observations for robust trend estimation. For each GGU, we provide a prototype monitoring report that incorporates the direction and significance of parameter estimates for the two models, along with the dominant source sector from EDGAR. We aggregate the econometric results into four categories: No Trend (statistically-insignificant results for both models); Mixed (statistically- insignificant results for one model, or significant but opposite-signed results for the two models); Decreasing (significant negative results for both models); and Increasing (significant positive results for both). Overall, we find that about 30% of the GGUs have No Trend or Mixed results, about 13% are Decreasing and 57% are Increasing. These results hold for both “raw” estimates and estimates weighted by EDGAR-estimated emissions. We find striking differences across sectors, with Increasing emissions dominant in five sectors and Decreasing emissions in two (Gas and Oil). For World Bank income groups, we find moderate dominance for Decreasing GGUs in high-income economies and heavy dominance for Increasing GGUs in the other three groups. In the final section of the paper, we summarize our results and provide an accompanying Excel file with monitoring reports for 775 GGUs, separated into five deciles by estimated EDGAR emissions. For each GGU, we report separate results for the two models, the composite result in the four previously-defined categories, and the dominant source sector. We also illustrate our results with global maps, with each GGU represented by color-coded icons that display the composite results and dominant source sectors. These maps provide graphic illustrations of the challenges that confront the Global Methane pledge. Although some regions have many Decreasing results (e.g., Western Europe, North Africa, the Gulf Region), most are dominated by Increasing results across several sectors. We close with two observations. First, the current state of the art enables us to estimate the significance and direction of trends for GGUs, but not the magnitudes of emissions reductions. In consequence, even in areas where Decreasing results dominate, the levels of emissions may remain high for extended periods. Second, our results provide a cautionary reminder of the challenges ahead. We will know that the Global Methane Pledge is beginning to succeed when future versions of Figures 13 and 14 are dominated by Decreasing (Blue) icons. At present, that is far from the case. 33 References Adhya, T. K. et al. 2014. “Wetting and Drying: Reducing Greenhouse Gas Emissions and Saving Water from Rice Production.” Working Paper, Installment 8 of Creating a Sustainable Food Future. Washington, DC: World Resources Institute. Crippa, M., E. Solazzo, G. Huang, D. Guizzardi, et al. 2020. High resolution temporal profiles in the Emissions Database for Global Atmospheric Research. Sci Data 7: 121. Cusworth, D. et al. 2021. Multisatellite imaging of a gas well blowout enables quantification of total methane emissions. Geophysical Research Letters, 48. Dasgupta, S., S. Lall and D. Wheeler. 2022. Subways and CO2 Emissions: A Global Analysis with Satellite Data (under review for publication). Dasgupta, S., S. Lall and D. Wheeler. 2023. Scalable Tracking of CO2 Emissions: A Global Analysis with Satellite Data. World Bank Policy Research Working Paper No. WPS10297. EDGAR (Emissions Database for Global Atmospheric Research). 2022. Global Greenhouse Gas Emissions - EDGAR v6.0. Sector-Specific Gridmaps: N2O - IPCC 4C+4D - Agricultural soils. https://edgar.jrc.ec.europa.eu/gallery?release=v60ghg&substance=N2O§or=AGS European Space Agency (ESA). 2021. Satellites detect large methane emissions from Madrid landfills. Phys.org, November 11. https://phys.org/news/2021-11-satellites-large-methane-emissions-madrid.html FAO (Food and Agriculture Organization of the United Nations). 2014. FAOSTAT. Rome: FAO. http://faostat.fao.org Fischer, T., D. Byerlee, and G. Edmeades. 2014. “Crop Yields and Global Food Security: Will Yield Increases Continue to Feed the World?” ACIAR Monograph No. 58. Canberra, Australia: Australian Center for International Agricultural Research. GADM. 2022. Maps for the administrative areas of all countries. https://gadm.org/data.html Hakkarainen, J., I. Ialongo, S. Maksyutov and D. Crisp. 2019. Analysis of Four Years of Global XCO2 Anomalies as Seen by Orbiting Carbon Observatory-2. Remote Sensing 11, no. 7: 850. Laborte, A. et al. 2017. RiceAtlas, a spatial database of global rice calendars and production. Nature Scientific Data 4:170074. Maasakkers, J. et al. 2022. Using satellites to uncover large methane emissions from landfills. Science Advances, 8(32). 34 Ouerghi, E. et al. 2021. Detection of Methane Plumes in Hyperspectral Images From Sentinel- 5P by Decoupling Anomaly Detection and Pattern Recognition. ISPRS Annals of the Photogrammetry, Remote Sensing and Spatial Information Sciences, Volume V-3-2021. Reuters. 2018. Global temperatures on track for 3-5 degree rise by 2100: U.N. https://www.reuters.com/article/us-climate-change-un/global-temperatures-on-track-for-3-5- degree-rise-by-2100-u-n-idUKKCN1NY186?edition-redirect=uk Sha, M., et al. 2021. Validation of methane and carbon monoxide from Sentinel-5 Precursor using TCCON and NDACC-IRWG stations, Atmospheric Measurement Techniques, 14: 6249– 6304. Smith, P., D. Reay and J. Smith. 2021. Agricultural methane emissions and the potential for mitigation. Philosophical Transactions of the Royal Society. 379(2210). Solazzo, E., M. Crippa, D. Guizzardi, M. Muntean, M. Choulga and G. Janssens-Maenhout. 2021. Uncertainties in the Emissions Database for Global Atmospheric Research (EDGAR) emission inventory of greenhouse gases. Atmospheric Chemistry and Physics, 21: 5655–5683. Steffen, W., J. Rockström and K. Richardson. 2018. Trajectories of the Earth System in the Anthropocene. PNAS 115(33): 8252-8259. Tu, Q. et al. 2022. Quantification of CH4 emissions from waste disposal sites near the city of Madrid using ground- and space-based observations of COCCON, TROPOMI and IASI. Atmospheric Chemistry and Physics 22: 295–317. Varon, D. 2019. Satellite discovery of anomalously large methane point sources from oil/gas production. Geophysical Research Letters, 46: 13507–13516. World Bank. 2012. Turn Down the Heat ~ Why a 4°C warmer world must be avoided. World Bank: Washington, D.C. Zhang, Y. et al. 2020. Quantifying methane emissions from the largest oil-producing basin in the United States from space. Science Advances, 6. 35