Policy Research Working Paper 11237 Satellite-Based Measures for Tracking Atmospheric CO₂ and CH₄ at National, Subnational, and Urban Scales Brian Blankespoor Susmita Dasgupta David Wheeler Development Economics A verified reproducibility package for this paper is Development Data & Research Groups available at http://reproducibility.worldbank.org, October 2025 click here for direct access. Policy Research Working Paper 11237 Abstract A scalable method for estimating changes in local greenhouse gas production zones. Regional illustrations were provided gas emissions from satellite-based atmospheric composi- for 11 Southeast Asian countries, alongside a global over- tion measures is developed and applied in this paper. The view organized by World Bank regions and income groups. analysis employs large panels of spatially-referenced, time- Findings indicated that long-term carbon dioxide decreases stamped atmospheric carbon dioxide observations from the outnumbered increases, but recent changes (2024–25) National Aeronautics and Space Administration’s Orbiting revealed a reversal. By contrast, methane displayed large Carbon Observatory-2 and methane observations from the net decreases in both long- and short-term measures. The European Space Agency’s Sentinel-5P. The analysis com- results highlighted substantial variation across regions and putes monthly mean concentration anomalies, defined as income groups. Low-income countries showed the stron- deviations from global trends. Long- and short-term trend gest movement toward reductions, yet their contributions regressions were estimated for cells of high-resolution global remain overshadowed by high-income economies, where grids, and cell-specific results meeting the classical signifi- performance has been mixed. It is hoped that this meth- cance test (p ≤ 0.05) were identified as positive or negative odology will inform global policy dialogue by enabling trends. These high-resolution findings were aggregated to transparent and comparable emissions assessments. The generate performance scores for geographic areas of arbitrary approach also provides a practical tool for identifying emis- scale. The global scalability of the approach was demon- sions hotspots, supporting policy makers at the national strated with performance assessments for 242 countries and subnational levels in developing targeted mitigation and disputed areas, 3,242 provinces, 36,563 sub-provinces, strategies aligned with global climate objectives. 6,672 Functional Urban Areas, and 670 offshore oil and This paper is a product of the Development Data Group and 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 bblankespoor@worldbank.org. A verified reproducibility package for this paper is available at http://reproducibility.worldbank.org, click here for direct access. RESEA CY LI R CH PO TRANSPARENT ANALYSIS S W R R E O KI P NG PA 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 Satellite-Based Measures for Tracking Atmospheric CO₂ and CH₄ at National, Subnational, and Urban Scales Brian Blankespoor Susmita Dasgupta David Wheeler Authorized for distribution by Florence Kondylis, Research Manager, Development Research Group, World Bank Group Keywords: Emissions trends, Satellite-based monitoring, emission hotspots JEL Codes: Q53, Q54, Q58, C55, O13 Acknowledgements: We gratefully acknowledge funding from the Global Data Facility. We also thank the participants of the National Aeronautics and Space Administration (NASA) Orbiting Carbon Observatory-2/3 Science Team Meeting at the Cooperative Institute for Research in the Atmosphere (Fort Collins, USA – September 2025). 1. Introduction The median projection from widely-used climate models estimates that the current greenhouse gas (GHG) emissions trend will increase global temperature around 4o C by 2100 (NSF UCAR, 2024). This would far overshoot the 2-degree limit pledged by the 2015 Paris climate accord (COP 21). Several industrial nations reacted to the emerging danger by pledging very steep emissions reductions at the Leaders’ Summit on Climate in April 2021. Unfortunately, these pledges have confronted a striking information shortfall: the near-total absence of directly-measured local and regional GHG data for problem diagnosis, program design and performance assessment. In recent years, the advent of satellite-based GHG measurement has greatly expanded the potential for empirical assessment. High-resolution observations of atmospheric GHG concentrations are now available from NASA’s OCO-2 and OCO-3 instruments; the European Space Agency’s METOP-A and TROPOMI (Sentinel-5P) platforms; China’s TANSAT and DQ-1; and the Japan Space Exploration Agency’s GOSAT and GOSAT-2. Detailed technical assessments of measures from several of these platforms have verified that they provide useful and comprehensive information for global carbon emissions analysis (Han et al. 2024; Zhang et al. 2024; Weir et al. 2021; Nassar et al. 2021; Pan et al. 2021; Wu et al. 2020; Hakkarainen et al. 2019; Labzovskii et al. 2019). Recent research at the World Bank has highlighted the potential contributions of satellite-based GHG measures to the World Bank’s climate change research, policy analysis and program design. This paper contributes by developing and implementing a standardized processing system that translates the latest satellite-based GHG measures into open, easily-accessed formats for emissions tracking by national, regional and urban policy analysts and GHG performance monitoring by World Bank staff and other global stakeholders. Our emissions trend estimation employs time series data for CO₂ from NASA’s Orbiting Carbon Observatory (OCO-2) and CH₄ from the European Space Agency’s TROPOMI (Sentinel-5P) platform. We also provide emissions-weighted trend estimates by integrating the OCO-2 and Sentinel-5P results with spatial information from EDGAR (Emissions Database for Global Atmospheric Research). Our approach employs a methodology for “trend scoring” based on earlier work by Dasgupta, Lall and Wheeler (2023). After adjusting the data for global trends, this approach estimates grid-level trends in local atmospheric concentration anomalies. For the current exercise, we employ a global grid with 5-km resolution and estimate CO₂ and CH₄ trends for terrestrial grid cells and marine cells that include offshore fossil fuel emissions sites. We overlay the grid cells with global maps of country administrative regions and urban areas. For each geographic area, we compute its “trend score” as the difference in counts for cell-level negative and positive significant trends, normalized to the range 0-100. We also compute a weighted trend score that weights each cell trend by the EDGAR estimate of emissions within the cell. The remainder of the paper is organized as follows. Section 2 describes the database that we use for trend estimation, while Section 3 introduces our trend models and trend scoring methodology. Section 4 focuses on 11 countries in Southeast Asia to provide a detailed illustration of our approach. In Section 5, we present a more general global assessment. Section 6 summarizes and concludes the paper. 2 2. Trend Estimation Data Our satellite platform for CO₂ trend estimation is NASA’s OCO (Orbiting Carbon Observatory)-2, which offers open access; a long panel of consistently measured, daily observations (beginning on September 6, 2014); and a spatial resolution of 1.29 × 2.25 km. OCO-2 follows a sun- synchronous near-polar orbit, crossing the equator in ascending mode around 1330 hours local time. OCO-2 has an observation repeat time of 16 days. We have downloaded georeferenced measures of XCO₂ (the column-averaged dry air mole fraction of CO₂) from JPL/NASA (2025). For CH₄ trend estimation we use data from ESA’s Sentinel-5P (S5P), which operates in a sun- synchronous near-polar orbit with an equatorial crossing at 1330 local solar time. It completes 14 orbits of the Earth per day, with a site revisit time of one day and a 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 methane), corrected for bias associated with XCH₄ dependence on surface albedo. For both CO₂ and CH₄, we filter the data for local concentration anomalies, or differences between observed and background concentrations at each point. We calculate the background concentration using the methodology of Hakkarainen et al. (2019), which incorporates both temporal and geographic elements. As Hakkarainen notes, the available data are insufficient for estimating daily medians at resolutions higher than 10 degrees of latitude. We compute the daily median XCO₂ for each 10-degree latitude band and linearly interpolate the result to each OCO-2 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 the 5-km grid cells in our database. 3. Trend Scoring Methodology 3.1 Tracking Models We compute two tracking models to capture long and short trends in the data. Model (1) estimates the time trend for the mean concentration anomaly during the entire measurement period. Model (2) gauges recent performance by estimating the size and significance of changes in the most recent year. Technically, (2) replaces the trend term in (1) with a dummy variable (DF) for observations in the final year. (1) = 0 + 1 + (2) = 0 + 1 + where, for grid cell i in month t: Git = Monthly mean anomaly for CO₂ or CH₄ (after Hakkarainen pre-filtering) DF = Dummy variable for most recent 12 months t = Time from initial period in months εit = Random error term 3 3.2. Trend Scoring for Geographic Units Unweighted Scoring Our methodology uses three steps to produce summary CO₂ and CH₄ trend indicators for any area of interest. (1): For all grid cells with regression results, divide trend estimates into three categories: Emissions declining (D): negative and statistically significant (ρ ≤ .05); Emissions unchanged (U): statistically insignificant (ρ > .05); Emissions Increasing (I): positive and statistically significant (ρ ≤ .05). (2): Count the regression grid cells in each category and convert the counts to percents of the total. (3): Compute the trend score by subtracting percent in D from percent in I. Score values can vary from -100 (emissions significantly declining in all cells) to +100 (emissions significantly increasing in all cells). The indicators have three useful features. First, they provide comparable measures regardless of area size. Second, they indicate the degree to which emissions in an area are decreasing or increasing. Third, their absolute magnitude depends on the relative numbers of cells with significant changes. For example, in an area where 90% of the grid cells are in category U (no significant change), the indicator value is limited to the range -10 to +10. In contrast, the indicator value can vary from -90 to +90 in an area where only 10% of the cells are in category U. 1 Weighted Scoring The methodology can also assign additional weight to trends in cells that have greater emissions. Our database includes indicators weighted by 2023 emissions estimates from the Emissions Database for Global Atmospheric Research (EDGAR). 2 To calculate weighted indicator values, steps (1) and (3) remain the same. In step (2), category percents are calculated from total estimated emissions rather than grid cell counts. For CO₂, the weighted indicator incorporates the EDGAR/IEA estimates of fossil fuel CO₂ emissions. For CH₄, we use EDGAR CH₄ emissions estimates for all activities. 3 The databases published with this paper also include the information needed for identifying the direction and significance of emissions changes in individual grid cells of interest. All cells are assigned id numbers in a raster (TIFF) map file that covers the globe, making it easy to find the id for a grid cell of interest. This id identifies the cell’s change estimation results in the CO₂ and CH₄ data files, which also include the number of observations used for measurement and the radius of the bounding area for the observations. The id also identifies the cell’s estimated CO₂ and CH₄ emissions volumes in the data files for EDGAR emissions. 1 We should note that very small areas with one grid cell can only have one of three scores, -100, 0 or +100. 2 EDGAR data are available online at https://edgar.jrc.ec.europa.eu. 3 The EDGAR data are gridded at 10 km. To match with our CH4 data, we subdivide each EDGAR grid cell into 5 km cells and assign the same emissions to each cell. 4 4. An Illustration for Southeast Asia 4.1 Overall Results We illustrate our approach with an in-depth exploration of model estimates for Southeast Asia. 4 Table 1 summarizes weighted and unweighted long-run results for CO₂, trend model (1). In the case of Malaysia, for example, Table (1a) shows that our scoring results are derived from 10,351 trend regressions. Using the classical significance criterion (p<=.05), 8,481 regressions (81.9%) have no significant trend, 498 (4.8%) have a significant decreasing trend, and 1,372 (13.3%) have a significant positive trend. Our composite score for Indonesia is 8.4 ([Increasing %] – [Decreasing %]). Potential scores range from -100 (all trends negative and significant) to 100 (all trends positive and significant). In Malaysia’s case, increasing trends are nearly three times more numerous than decreasing trends, but the absolute score size is moderated by the fact that 81.9% of Malaysia’s regressions reveal no significant trends. Table (1b) weights a cell’s regression results by EDGAR’s estimate of its CO₂ emissions. In this accounting, the importance of a cell’s trend result for overall scoring is directly proportional to its importance as a source of CO₂ emissions. Total CO₂ emissions from the Malaysian regression cells are 697.4 million kg (mkg): 524.9 mkg (75.3%) from cells with no significant trend; 13.6 mkg (1.9%) from cells with significant decreasing trends; and 158.9 mkg (22.8%) from cells with significant increasing trends. In Malaysia, emissions from areas with significant increasing trends are 12 times the emissions from areas with significant decreasing trends and the resulting composite score (20.8) is 2.5 times higher than the unweighted score (8.4). Table 1 shows that country experiences in the region have differed greatly during the past decade. Indonesia, the Philippines and Timor-Leste have negative scores for both weighted and unweighted regression results. The converse is true for Lao PDR, Malaysia, Singapore, Thailand and Viet Nam, which have uniformly positive scores. Despite this sign uniformity, however, weighting by emissions can significantly alter the magnitudes of scores. Indonesia, for example, has a change from -3.3 to -19.3 when results are emissions-weighted. For Viet Nam, emissions weighting changes the score from 7.7 to 36.2. Emissions weighting has sufficient impact to change the direction of results in some cases. Without emissions weighting, Cambodia’s score suggests an overall emissions trend that is negative. Once emissions are weighted, however, the score turns positive and jumps in magnitude. The converse is true for Papua New Guinea, which has a positive unweighted score that reverses to a negative score when emissions are taken into account. We believe that emissions-weighted scores are the appropriate measure for this exercise because they are more useful for policy analysis and priority-setting. Logically, attention should focus on areas with high emissions that are increasing over time. In the same vein, areas with high but decreasing emissions may provide useful lessons that can be applied elsewhere. 4 Our results are global, so any region could be used for this exposition. 5 Table 1: Southeast Asia - weighted and unweighted regression scoring for CO₂, trend model 1 (1a) Unweighted Regression Results Grid Cell Count No Significant Significant Significant Decreasing Increasing No Decreasing Increasing Composite Country Total Trend Trend Trend Trend % % % Score Brunei Darussalam 186 186 0 0 100.0 0.0 0.0 0.0 Cambodia 5,996 4,061 1,074 861 67.7 17.9 14.4 -3.6 Indonesia 58,544 45,414 7,527 5,603 77.6 12.9 9.6 -3.3 Lao PDR 7,858 5,685 727 1,446 72.3 9.3 18.4 9.1 Malaysia 10,351 8,481 498 1,372 81.9 4.8 13.3 8.4 Papua New Guinea 14,586 12,187 872 1,527 83.6 6.0 10.5 4.5 Philippines 9,073 7,303 1,194 576 80.5 13.2 6.3 -6.8 Singapore 15 4 4 7 26.7 26.7 46.7 20.0 Thailand 17,136 12,914 1,976 2,246 75.4 11.5 13.1 1.6 Timor-Leste 466 354 102 10 76.0 21.9 2.1 -19.7 Viet Nam 10,920 8,293 894 1,733 75.9 8.2 15.9 7.7 (1b) Emissions-Weighted Regression Results CO2 Emissions (million kg) No Significant Significant Significant Decreasing Increasing No Decreasing Increasing Composite Country Total Trend Trend Trend Trend % % % Score Brunei Darussalam 21.8 21.8 0.0 0.0 100.0 0.0 0.0 0.0 Cambodia 58.2 33.0 8.4 16.9 56.6 14.4 29.0 14.6 Indonesia 2,084.9 1,526.3 480.8 77.8 73.2 23.1 3.7 -19.3 Lao PDR 85.3 75.2 3.3 6.8 88.2 3.8 8.0 4.2 Malaysia 697.4 524.9 13.6 158.9 75.3 1.9 22.8 20.8 Papua New Guinea 14.3 11.8 1.8 0.7 82.5 12.8 4.6 -8.2 Philippines 375.7 326.8 26.6 22.3 87.0 7.1 5.9 -1.2 Singapore 104.9 43.7 19.0 42.1 41.7 18.2 40.2 22.0 Thailand 1,023.0 718.9 87.2 216.8 70.3 8.5 21.2 12.7 Timor-Leste 1.5 1.3 0.2 6 0.0 86.0 11.3 2.7 -8.6 Viet Nam 984.9 572.5 27.8 384.7 58.1 2.8 39.1 36.2 Table 2: Southeast Asia: countries by change category, CO₂ trend model 1 Score Change Category Country Unweighted Weighted Indonesia -3.3 -19.3 Negative / Negative Philippines -6.8 -1.2 Timor-Leste -19.7 -8.6 Lao PDR 9.1 4.2 Malaysia 8.4 20.8 Positive / Positive Singapore 20.0 22.0 Thailand 1.6 12.7 Viet Nam 7.7 36.2 Negative / Positive Cambodia -3.6 14.6 Positive / Negative Papua New Guinea 4.5 -8.2 No Change Brunei Darussalam 0.0 0.0 4.2 Country-Level Changes To illustrate the implications of our weighting methodology, this section narrows the focus progressively from countries to provinces, sub-provinces and individual urban areas. Figures 1- 5 display CO₂ and CH₄ results for model (1) (long-run trends) and model (2) (recent changes). Figure 6 extends the presentation to long- and short-term changes in CH₄ emissions from offshore oil and gas production facilities. Figure 1 provides a geographic display of the information presented in the weighted scores column of Table 2. Scores on the maps are consistently color-coded: dark blue [-100, -10], blue [-9.9, -5.0], green [-4.9, 0], yellow [0.1, 5.0], orange [5.1, 10], red [10.1, 100]. For CO₂, weighted long-term emissions trends (panel (a)) are increasing in continental states (Viet Nam, Cambodia, Lao PDR, Thailand, Malaysia, Singapore) but decreasing in large island states (Indonesia, the Philippines, Timor-Leste, Papua New Guinea). Brunei Darussalam exhibits no long-term trend. However, recent changes in CO₂ emissions (panel (b)) have been uniformly increasing across the region. The positive scores for large island states are generally lower than those for the continental states, but they indicate increases nonetheless. For CH₄ emissions, two aspects of the long-term results (panel (c)) are particularly noteworthy. First, the regional split between continental and large island nations remains: Emissions are increasing in the majority of continental states, but decreasing in all the large island states. Second, however, the within-region difference is less extreme than for CO₂. For recent changes in CH₄ emissions (panel (d)), the trends are quite different. While all countries have recent 7 increases for CO₂, the majority (both continental and large-island) have recent decreases for CH₄. Where increases remain, they are in the lowest (yellow) category. 4.3 Changes within Countries Figures 2-5 provide more disaggregated regional information for long- and short-term changes in CO₂ and CH₄. In each figure, the first panel reproduces the corresponding country result in Figure 1. The second panel displays results for level-1 administrative units (termed Provinces) in the World Bank’s most recent update. The third panel disaggregates to level-2 administrative units (Sub-Provinces), while the fourth panel provides color-coded results for Functional Urban Areas with populations greater than 250,000 in 2015. Figure 2 reveals a striking pattern in the transition to more disaggregated results. The Countries panel suggests a sharp disparity between continental and large-island states, but this distinction diminishes in the Provinces panel. Here, divergent patterns emerge in each state, with positive scores (increasing trend) in some provinces and negative scores (decreasing trend) in others. The pattern recurs in the transition from the Provinces panel to the Sub-Provinces panel, which reveals the same pattern of divergence within provinces. From a policy perspective, we believe that comparative scores provide useful information at each level of aggregation. For international resource allocation, the first panel provides a view of overall country performance. Within countries, the second panel provides useful information for inter- provincial resource allocation decisions, while the third and fourth panels can inform local targeting of emissions reduction efforts. Results for Functional Urban Areas (FUAs) in the fourth panel complement the sub-provincial information in the third panel. In some cases FUAs are in individual sub-provinces, while in others they are large metro areas that may overlap several sub- provinces. In summary, a broad spectrum of emissions performance emerges for each country as the view moves across panels. In major part, the broadened score spectrum reflects the spatially-skewed distribution of major CO₂ emissions sources. The third and fourth panels reveal the diverse performance of major emissions sources in local areas. Increases and decreases are both common, and aggregative scoring simply reveals the balance when the relative scale of emissions is taken into account. The first panel of Figure 3 displays our country results for recent changes in CO₂ emissions. The contrast with the first panel of Figure 2 is striking: At the country level, recent changes in CO₂ emissions yield positive (increasing) scores for all countries. As before, however, this general pattern fragments into broader score spectra in the transition through the Provinces panel to the Sub-Provinces and FUA panels. Comparisons with the corresponding panels in Figure 2 show that while red, orange and yellow sub-provinces are more prevalent in Figure 3, many provinces have significant green and blue areas in both continental and large island states. Figures 4 and 5 for CH₄ changes provide contrasting cases in the aggregate because their country-level patterns (previously-displayed in Figure 1) exhibit more decreases than their long- term counterparts for CO₂. However, as before, progressive fragmentation into broader score spectra accompanies the transition through provinces to sub-provinces and FUAs. Many 8 provinces in both figures exhibit diverse performance across local areas, with plentiful red-coded sub-provinces and FUAs as well as blue- and green-coded areas. As before, aggregative scoring at the provincial and national levels reflects the relative prevalence of increases and decreases across major emissions sources. Figure 6 extends our assessment for CH₄ emissions to oil and gas production areas in the exclusive economic zones (EEZs) of the Southeast Asian countries. The first panel of Figure 6 displays diverse long-term trends in each country’s EEZ, with an apparent prevalence of increasing trends for Indonesia and decreasing trends for the Philippines. However, the overall pattern in the second panel reveals a general shift toward decreased emissions in the most recent period. 9 Figure 1: Southeast Asia country trend weighted scores 10 Figure 2: Southeast Asia CO₂ long trends 11 Figure 3: Southeast Asia CO₂ recent changes 12 Figure 4: Southeast Asia CH₄ long trends 13 Figure 5: Southeast Asia CH₄ recent changes 14 Figure 6: Southeast Asia offshore oil and gas production areas 15 5. Global Trends Our illustration for Southeast Asia has provided an in-depth example that could be replicated for any region because our scoring database is global. Satellite tracking of changes in CO₂ and CH₄ emissions permits consistent, objective measurement of trends for countries, provinces, sub- provinces, FUAs and EEZs anywhere in the world. 5.1 Mapping Global Changes To illustrate, Figures 7-10 replicate the Southeast Asia figures at global scale for emissions- weighted indicators of long- and short-term changes in CO₂ and CH₄ emissions. Figure 7(a) reveals diverse and roughly balanced results for long-trend CO₂ emissions change at the country level, while Figure 8(a) shows a higher incidence of increases in the most recent period. Figure 9(a) shows that decreases have dominated long-term trends in CH₄ emissions, while Figure 10(a) displays more balanced country representation in the most recent period. As in Southeast Asia, shifting to the sub-provincial level in Figures 7(b) – 10(b) reveals a fragmentation into broader score spectra in many countries, with ample representation for increases and decreases. Again, the primary driver of this fragmentation is undoubtedly the skewed spatial distribution of emissions sources. In each country, the result of aggregation from sub-provincial and FUA scores depends on the direction of change in large local emissions sources. 5.2 Global Changes by Region and Income Group Figures 11 and 12 display graphical summaries of the sub-provincial maps in Figures 7(b) -10(b) for World Bank regions and income groups. They provide global views of long-term and recent changes as the risks associated with CO₂ and CH₄ emissions have gained widespread attention. Each chart uses percents to display the relative numbers of sub-provinces with significant long- and short-term changes. 5 Both figures show the same broad global patterns, although overall counts are somewhat lower for the income group assessment because group assignments are missing for some countries. In Figure 11, the long-trend regional results for CO₂ (panel (a)) are somewhat encouraging, with 13,179 sub-provinces exhibiting significant decreases while 9,616 have significant increases. Unfortunately the overall pattern reverses in the most recent period 2024-2025 (panel (b)), with more increases than decreases. Figure 11 reveals the extent to which changes in different regions have reflected the overall trends. For long-trend CO₂ (panel(a)), the largest net decreases 5 For interpretive clarity, we exclude cases with no significant change. The charts display counts as percents of all cases, including cases with no significant change. 16 (increase % - decrease %) are in South Asia (SAS) and Sub-Saharan Africa (SSA), followed by net decreases in North America (NA) and the Middle East, North Africa, Afghanistan and Pakistan (MENA). Some net decreases occurred in Latin America & Caribbean (LAC) and Europe and Central Asia (ECA), while East Asia and Pacific (EAP) has actually had a small net increase. The chart for recent CO₂ changes in CO₂ (panel (b)) is much less encouraging, with net increases in EAP, LAC, MENA and SAS. Net decreases remain in ECA, NA and SSA, but they are very small. The picture is more encouraging overall for CH₄, which has exhibited large net decreases globally for both long-term changes (panel (c)) and recent changes (panel (d)). However, these large global differences mask great regional variation. The dominant net decreases are in MENA, which exhibits huge long- and short-term declines. Among long-term results (panel (c)), net decreases have also occurred in EAP, ECA, LAC and SSA, while increases have occurred in NA and SAS. Among recent changes (panel (d)), ECA is the only region with net increases. MENA, EAP, NA, SAS and SSA have all had large net decreases, while LAC has had a modest decrease. Figure 12 displays variations across World Bank income groups. For CO₂, the most striking result is the large net decreases for low-income countries in both the long term (panel (a)) and short term (panel (b)). The other three income groups all exhibit long-term net decreases (panel (a)), but recent changes (panel (b)) have been net increasing for lower- and upper-middle-income countries while high-income countries have had a slight net decrease. For CH₄, the four income groups exhibit uniformly net decreasing changes in both the long and short terms. Among long trends (panel (c)), the net decrease is largest for upper-middle-income countries and smallest for high-income countries. For recent changes (panel (d)), the net decrease is largest for low- and lower-middle-income countries but still substantial for upper-middle- and high-income countries. In summary, our results for both CO₂ and CH₄ are somewhat encouraging globally. But they are also paradoxical because the greatest shift toward decreases has been in low-income countries, whose emissions are dwarfed by emissions from high-income countries. Additional insight is afforded by comparison of South Asia and Sub-Saharan Africa, the two regions with the largest poverty populations. Sub-Saharan Africa has net decreases for both pollutants and both time intervals. On the other hand, South Asia has net decreases for long-trend CO₂ and recent CH₄, but net increases for recent CO₂ and long-trend CH₄. At the other end of the income spectrum, North America has net decreases in long-trend CO₂ and recent CH₄, but almost no net change in recent CO₂ and a net increase in long-trend CH₄. 17 Figure 7: Global CO₂ Long Trends 18 Figure 8: Global CO₂ Recent Changes 19 Figure 9: Global CH₄ Long Trends 20 Figure 10: Global CH₄ Recent Changes 21 Figure 11: Global significant sub-provincial changes by World Bank region 22 Figure 12: Global significant sub-provincial changes by World Bank income group 6. Summary and Conclusions In this paper, we have developed and applied a scalable method for estimating local greenhouse gas emissions changes from satellite-based measures of atmospheric gas composition. Our approach is based on large panels of spatially-referenced, time-stamped measures of atmospheric CO₂ from NASA’s OCO-2 and CH₄ from the ESA’s Sentinel-5P. We use these measures to compute monthly mean concentration anomalies (deviations from global trends) and estimate long- and short-term trend regressions for the 23 cells of high-resolution global grids. From the cell-specific results, we identify positive and negative trend estimates that meet the classical significance test (p<=.05). These high-resolution results enable us to estimate greenhouse gas emissions performance scores for geographic areas of arbitrary scale. Within an area, we count results in unweighted and weighted modes. In the unweighted mode, we assign parity to all regression results, calculate increasing and decreasing cases as percents of all cases (including statistically-insignificant results), and compute a performance score as [Increasing % - Decreasing %]. We replicate this approach in the weighted mode, but use cell- specific CO₂ and CO4 emissions estimates from the EDGAR database as counting weights for regression results. Both scores also normalize by case totals to incorporate the relative importance of statistically- significant changes. The greater the proportion of statistically-significant cases the higher the score, other things equal. For any area, the potential score varies from -100 (all cases have statistically-significant trend decreases) to +100 (all significant increases). The paper demonstrates the global scalability of this approach by summarizing results for long-and short- term CO₂ and CH₄ performance scores for 242 countries and disputed areas; 3,242 provinces (our general term for level-1 administrative units; 36,563 sub-provinces (level-2 administrative units); 6,672 Functional Urban Areas; and 670 offshore oil and gas production areas that are within national EEZs. We provide an in-depth illustration for 11 countries in Southeast Asia in a set of figures that track the scoring transition from countries through provinces to sub-provinces. The Countries panels suggest a sharp disparity between continental and large-island states in the region, but this distinction diminishes in the Provinces panels. Here, divergent scoring patterns emerge in each country, with positive scores (increasing trend) in some provinces and negative scores in others. The pattern recurs in the transition from the Provinces panel to the Sub-Provinces and FUA panels, which reveal the same pattern of score divergence within provinces. From a policy perspective, we believe that emissions-weighted scores provide the most useful information at each level of aggregation. For international resource allocation, the Countries panels provide useful views of overall country performance. Within countries, the Provinces panels provide useful information for inter-provincial resource allocation decisions, while the Sub-Provinces panels can help with precise local targeting of emissions reduction programs. The Functional Urban Areas (FUAs) panels complement the Sub-Province panels. In some cases FUAs are in individual sub-provinces, while in others they are large metro areas that may overlap several sub-provinces. We conclude the paper with a global overview of sub-provincial patterns for emissions-weighted scores. For CO₂ the long-trend results are somewhat encouraging, with significant decreases substantially more numerous than significant increases. Unfortunately, assessment of recent changes shows that the period 2024-2025 has witnessed a reversal, with more increases than decreases. The picture is more encouraging overall for CH₄, which has exhibited large net decreases globally for both long-term and recent changes. For both CO₂ and CH₄, broad global patterns mask substantial variation by region and income group. We find that the greatest shift toward decreases has been in low-income countries, whose emissions are dwarfed by emissions from high-income countries. Additional insight is afforded by comparison of South Asia and Sub-Saharan Africa, the two regions with the largest poverty populations. Sub-Saharan Africa has exhibited net decreases for both CO₂ and CH₄ in both the long and short terms. South Asia has had net decreases for long-trend CO₂ and recent CH₄, but net increases for recent CO₂ and long-trend CH₄. At the other end of the income spectrum, North America has had net decreases in long-trend CO₂ and recent CH₄, but almost no net change in recent CO₂ and a net increase in long-trend CH₄. 24 To conclude, this paper has developed and demonstrated a method for using satellite-based measures to develop consistently-derived performance metrics for changes in CO₂ and CH₄ emissions for arbitrarily- specified geographic areas. Our assessment of current global performance at the sub-provincial level has revealed a hopeful movement toward lower emissions in more areas. Unfortunately, however, this trend is most pronounced in the poorest regions whose emissions are dwarfed by the emissions of higher- income areas whose short- and long-term performance is mixed at best. Much better performance will be required to achieve reductions sufficient to avoid significant climate change. We hope that the methods developed in this paper will contribute by demonstrating that satellite-based, globally- comparable, comprehensive performance scoring can contribute to the global policy dialogue on these issues. In addition, the approach offers the potential to identify emissions hotspots more precisely, providing critical insights for policy making at the national and subnational administrative levels. This, in turn, can support the design of locally tailored mitigation strategies and strengthen the alignment of global targets with local action. 25 References Dasgupta, S., S. Lall and D. Wheeler. 2023. Scalable Tracking of CO₂ Emissions: A Global Analysis with Satellite Data. Policy Research Working Papers No. 10297. 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. Han, G., H. Zhang, Y. Huang et al. 2024. First Global XCO₂ Products from Spaceborne Lidar: Methodology, Results and Potentials. SSRN: https://ssrn.com/abstract=5067761 or http://dx.doi.org/10.2139/ssrn.5067761. JPL/NASA. 2025. OCO-2 Data Set. Jet Propulsion Laboratory, California Institute of Technology. https://ocov2.jpl.nasa.gov/ Labzovskii, L., S. Jeong and N. Parazoo. 2019. Working towards confident spaceborne monitoring of carbon emissions from cities using Orbiting Carbon Observatory-2. Remote Sensing of Environment. 233: 111359. Nassar, R., J. Mastrogiacomo, W. Bateman-Hemphill et al. 2021. Advances in quantifying power plant CO₂ emissions with OCO-2. Remote Sensing of Environment. 264: 112579. NSF UCAR Center for Science Education. 2024. Predictions of Future Global Climate. https://scied.ucar.edu/learning-zone/climate-change-impacts/predictions-future-global-climate Pan, G., X. Yuan and M. Jieqi. 2021. The potential of CO₂ satellite monitoring for climate governance: A review. Journal of Environmental Management, 277: 111423. Steffen, W., J. Rockström and K. Richardson. 2018. Trajectories of the Earth System in the Anthropocene. PNAS 115(33): 8252-8259. Weir, B., D. Crisp, C. O’Dell et al. 2021. Regional impacts of COVID-19 on carbon dioxide detected worldwide from space. Science Advances, 7(45). World Bank. 2012. Turn Down the Heat ~ Why a 4°C warmer world must be avoided. World Bank: Washington, D.C. Wu, D., J. Lin, T. Oda and E. Kort. 2020. Space-based quantification of per capita CO₂ emissions from cities. Environmental Research Letters, 15: 035004. Zhang, M., Y. Xie, S. Zhu, et al. 2024. "The status and development proposal of China's comprehensive environmental observation satellites system", Proc. SPIE 13189, Second Conference on Space, Atmosphere, Marine, and Environmental Optics (SAME 2024). 26