Estimation of methane and carbon monoxide emissions in megacities by using simultaneous greenhouse gas and air pollutant observations by “IBUKI-2” (GOSAT-2)

1. Background and Objectives of the Study

The Greenhouse Gases Observing Satellite-2 "Ibuki-2" (GOSAT-2), a joint project by the Ministry of the Environment, Japan; the National Institute for Environmental Studies; and the Japan Aerospace Exploration Agency, was launched in October 2018 and has been continuously acquiring data to this day. By analyzing the shortwave infrared spectrum (Note 1) observed by the Thermal And Near infrared Sensor for carbon Observation-Fourier Transform Spectrometer-2 (TANSO-FTS-2) onboard GOSAT-2, it is possible to estimate the column-averaged concentrations (Note 2) of greenhouse gases such as carbon dioxide (CO₂) and methane (CH₄), just as was done by analysis of the data from its predecessor, GOSAT ("Ibuki," launched in January 2009). In addition to these gases, by using the same sensor, GOSAT-2 can simultaneously observe the column-averaged concentrations of carbon monoxide (CO), an air pollutant.

Since emissions from urban areas account for a significant proportion of global emissions of greenhouse gases and air pollutants, accurately quantifying urban emissions is crucial for implementing and evaluating emission reduction measures. Because emission databases are regularly reviewed, the emission estimates for the same year can change as information is updated. However, the updated values are not always more accurate than previous ones. Therefore, it is important to evaluate the accuracy of emission databases by using data from other independent sources.

When considering emissions from an entire urban area, analyzing the ratio of CO to CO₂ released into the atmosphere can yield insights into the area’s emission sources, because the ratio varies depending on the types of fuel used and the efficiency of combustion. In particular, in urban areas where the influence of photosynthesis is small, the column-averaged concentrations of CO₂, CH₄, and CO in the atmosphere increase, or are enhanced, relative to the background concentrations (Note 3) because of emissions from fossil fuel combustion, etc. If we denote these enhancements as ∆CO₂, ∆CH₄, and ∆CO, where each ∆ represents the difference between the urban and background concentrations in nearby rural areas, then the enhancement ratios (∆CO/∆CO₂, ∆CO/∆CH₄, and ∆CH₄/∆CO₂) are equal to the molar ratios of the corresponding emissions (Note 4). Hereinafter, "molar ratios of emissions" is simply referred to as "emission ratios." In this study, we attempted to evaluate the accuracy of emission databases by comparing the enhancement ratios calculated using GOSAT-2 data with the emission ratios calculated from emission databases for each city. Furthermore, we estimated the CH₄ and CO emissions for each city based on the calculated enhancement ratios.

2. Methods

This study used the GOSAT-2 TANSO-FTS-2 SWIR L2 Column-averaged Dry-air Mole Fraction Product version 02.00 for the period March 2019 to February 2023. The analysis targeted the 40 most populous cities in the world, and the spatial extent of each city was determined based on population grid data of approximately 1 km x 1 km resolution and the United Nations’ report on the total population of each city. The differences between the concentrations of CO₂, CH₄, and CO at the GOSAT-2 observation points within the city boundaries and their respective background concentrations in rural areas (Note 3) around each city were denoted as ∆CO₂, ∆CH₄, and ∆CO. The enhancement ratios were obtained as the slopes of linear regressions among ∆CO₂, ∆CH₄, and ∆CO (Note 5). Note that only data from periods with minimal photosynthetic influence were used in the analysis (Note 6). Next, CH₄ and CO emissions were estimated by multiplying the city-specific enhancement ratios (∆CH₄/∆CO₂ and ∆CO/∆CO₂), derived from GOSAT-2 data, by the CO₂ emissions in a reference emissions database (Note 7), either the Open-Data Inventory for Anthropogenic Carbon dioxide (ODIAC) or the Emissions Database for Global Atmospheric Research (EDGAR) (Note 8). CO₂ was used as a reference because it has the lowest uncertainty in the databases among CO₂, CH₄, and CO.

3. Results and Discussion

Figure 1 shows two-dimensional histograms (Note 9) and enhancement ratios derived from regression analysis for each gas component in Tokyo and Beijing. For example, in the case of ∆CO/∆CO₂, the enhancement ratio for Beijing is approximately 4.3 times that of Tokyo (=20.21/4.75). The implication is that when the same amount of CO₂ is emitted in Tokyo and Beijing, the amount of CO emitted together with CO₂ in Beijing is approximately 4.3 times that emitted in Tokyo and hence likely to exacerbate air pollution. The high ∆CO/∆CO₂ ratio in Beijing is likely due to the greater reliance on coal as a fossil fuel and the use of less efficient combustion equipment in China. Next, prior to comparing the enhancement ratios from GOSAT-2 with the emission ratios from emission databases, we examined the variability across different versions of the emission databases. We investigated the differences of reported emissions between EDGAR v5.0 and the latest available versions at the time of analysis (v7.0 for CO₂ and CH₄, and v6.1 for CO). The average differences for CO₂ and CH₄ for different versions of EDGAR were within 5%, while the CO emissions differed by about 32%. The indication was that CO emissions in EDGAR varied more significantly between versions than did emissions of CO₂ and CH₄. When comparing the ∆CO/∆CO₂ from GOSAT-2 with the emission ratios from EDGAR for cities where the difference of CO emissions between versions was less than 30% (mainly cities in developed countries, where the database accuracy was presumed to be high), good agreement was found between the two (average difference and standard deviation were −2.4% and 42.3%, respectively). In contrast, in cities with more than a 30% difference of CO emissions between versions (mainly cities in developing countries), the discrepancy between the ∆CO/∆CO₂ from GOSAT-2 and the emission ratios from EDGAR were larger (average difference and standard deviation were 33.5% and 142.1%, respectively). These results suggested that the accuracy of CO emissions in EDGAR was relatively low in developing countries. In addition, while the update to EDGAR version 6.1 improved the consistency between its CO emissions and our results in many cities, some cities—such as Buenos Aires, Mumbai, and Tehran—showed excessive declines in their CO emissions. The implication is that there is room for further improvement in EDGAR.

Next, we compared the city-wise ratios of CH₄ and CO emissions from GOSAT-2 data to those reported in EDGAR (Figure 2). The GOSAT-2 estimates were derived using the ∆CH₄/∆CO₂ and ∆CO/∆CO₂ ratios, along with CO₂ emissions data from the ODIAC v2022 dataset. The CH₄ emissions for 2019 in EDGAR V7.0 were smaller than the GOSAT-2 results for 74% of the analyzed cities (cities with values greater than 1 on the vertical axis in the upper panel of Figure 2). Similarly, the CO emissions for 2018 in EDGAR V6.1 were found to be underestimated compared to the GOSAT-2 results for 76% of the cities (cities with values greater than 1 on the vertical axis in the lower panel of Figure 2). Recent studies (Note 10) have suggested that the underestimation of CH₄ emissions in EDGAR when the database was compiled was likely due to insufficient consideration of leaks from natural gas supply networks and emissions during end use in households. In the case of CO, the large variability between versions of databases, as discussed above, suggests a high degree of uncertainty in the data. Although the uncertainties in GOSAT-2 observations and the varying number of usable data points across the cities may be influencing the differences from EDGAR emissions, in cities where emissions have been estimated using in-situ data (directly observed data), the tendencies of GOSAT-2-based CO estimates to overestimate or underestimate relative to EDGAR are generally consistent with those reported in prior studies.

In this way, by using the enhancement ratios of concentrations of atmospheric components that are co-emitted and transported from urban areas, it becomes possible to estimate emissions with relatively simple calculations, without the need for complex and time-consuming calculations such as atmospheric transport simulations; thus, emission estimates based on GOSAT-2 enhancement ratios can be used to evaluate emission databases.

4. Future Prospects

Although we calculated the average enhancement ratios over the entire four-year analysis period in this study, we can also estimate annual enhancement ratios for cities where GOSAT-2 obtains at least 10 data points per year. In the future, as more observation data accumulate, satellite-based monitoring of the relative temporal variation in emissions from these cities is expected to become possible. In the first half of FY2025, a new satellite, GOSAT-GW, is scheduled for launch. Compared to GOSAT-2, GOSAT-GW will provide spatially denser observations of CO₂ and CH₄. Although GOSAT-GW will not measure CO, it will add nitrogen dioxide (NO₂) observations. NO₂ is produced by the oxidation of nitric oxide (NO), which is co-emitted with CO and CO₂ during fossil fuel combustion. Since NO₂ has a shorter atmospheric lifetime than CO or CO₂, it can more clearly capture the enhanced concentrations originating from large-scale emission sources such as thermal power plants and megacities. Similar to the CO/CO₂ ratio, the NO₂/CO₂ ratio is useful for examining the type of fuel used and combustion efficiency of emission sources. The GOSAT-GW data will also be utilized for evaluating NO₂-emission databases and quantifying NO₂ emissions.

5. Notes

Note 1: TANSO-FTS-2 observes wavelength bands around 0.76, 1.6, and 2.0 µm, referred to as Band 1, Band 2, and Band 3, respectively. These wavelength bands are collectively referred to as the shortwave infrared (SWIR) region. The distribution of light intensity across wavelengths is called a spectrum.

Note 2: In satellite observations of greenhouse gases using the shortwave infrared region, the column-averaged concentration is estimated as the ratio of the number of molecules of the target substance to the total number of molecules (excluding water vapor) in the atmospheric column through which sunlight passes and is reflected back by Earth’s surface before being observed by the satellite.

Note 3: Background concentrations refer to atmospheric concentrations representative of the surrounding region that are not affected directly by urban emissions. In this study, background values were defined as the monthly medians of data for the area within 10° latitude × 20° longitude centered around each city and classified as non-urban areas based on a land cover map derived from data from a U.S. satellite.

Note 4: As described in Note 2, the column-averaged concentration is expressed as a mole fraction, that is, the number of molecules of the target substance divided by the number of dry air molecules. To enable a meaningful comparison between the enhancement ratios (mole fraction ratios) with the mass ratios of emissions, the emissions of the target substances were first converted from mass units to mole units by dividing by their respective molecular weights. We then calculated the ratios between these mole-based emissions, which can be directly compared to the enhancement ratios.

Note 5: An analytical method to find the best-fit line representing the relationship between two datasets (x, y). In this study, we used a method called York fit, which calculates the optimal straight line considering the uncertainties of each data point when both variables (x and y) have uncertainties (σ_x, σ_y).

Note 6: We used only data from periods with minimal photosynthetic influence for the analysis by examining the seasonal correlation between ∆CO₂ and ∆CO and the seasonal average value of ∆CO₂.

Note 7: Using the CO₂ emission E_CO2 as a reference, the CH₄ emission E_CH4 and CO emission E_CO for each city were calculated from the enhancement ratios ∆CH₄/∆CO₂ and ∆CO/∆CO₂ calculated from GOSAT-2 data using the following equations:

E_CH4 = (∆CH₄/∆CO₂)•E_CO2•(m_CH4/m_CO2)

E_CO = (∆CO/∆CO₂)•E_CO2•(m_CO/m_CO2)

Here, m_CO2, m_CH4, and m_CO represent the molecular weight of each gas component.

Note 8: An emissions database of greenhouse gases and air pollutants created by the European Commission's Joint Research Centre. In this study, we used data provided at a global grid resolution of 0.1º latitude x 0.1º longitude. The latest versions at the time of analysis were v7.0 for CO₂ and CH₄, and v6.1 for CO, with v5.0 being the common older version for all components.

Note 9: A type of graph that represents the frequency distributions of two variables using color gradients.

Note 10: For example,

Plant, G., Kort, E. A., Floerchinger, C., Gvakharia, A., Vimont, I., and Sweeney, C.: Large fugitive methane emissions from urban centers along the U.S. East Coast, Geophysical Research Letters, 46, https://doi.org/10.1029/2019GL082635, 2019.

Ueyama, M., Umezawa, T., Terao, Y., Lunt, M., and France, J. L.: Evaluating urban methane emissions and their attributes in a megacity, Osaka, Japan, via mobile and eddy covariance measurements, EGUsphere [preprint], https://doi.org/10.5194/egusphere-2024-3926, 2025.