The information contained in MethaneSAT L2 and L4-plume files is illustrated in Fig. . A more in-depth analysis of this acquisition is provided here. Figure a shows a XCH₄ map derived from MethaneSAT over an area of about 300×300 km² at the central part of the Appalacchian basin (USA). The equivalent map derived from a TROPOMI acquisition over the same area about 70 min earlier than the MethaneSAT overpass is shown in Fig. b. The visual comparison of the two XCH₄ maps confirms the consistency between the two datasets, and the higher level of detail enabled by MethaneSAT's higher spatial resolution (100×400 m² for MethaneSAT, and 7×5.5 km² for TROPOMI). In addition, a number of methane plumes and methane concentration enhancements can be seen across the scene. The plumes that were detected and quantified from this MethaneSAT dataset (i.e. the L4-plume product) are indicated in Fig. c on top of a map of near-infrared suface albedo extracted from the same dataset. It comprises a total of 16 plumes, with emission rates ranging between 800 and 7000 kg h⁻¹ (see Fig. ). These emissions from the Appalachian region are due to a combination of O&G, coal mining and waste facilities across the area, with the last two types of sources including a number of highly persistent point sources . A massive emission of more than 100 metric tonnes per hour (t h⁻¹) was also detected in this dataset (marked by a red cross in Fig. c), but we assume it was short-lived because we were not able to detect it in a follow-up study using geostationary observations by the GOES-ABI mission, with a continuous coverage over North America every 5–10 min . We ruled out the possibility of this plume being a retrieval artifact by comparing the XCH₄ map with the corresponding near-infrared radiance map, where no similar pattern is found.

The lower end of the Appalachian point source emissions (about 700 kg h⁻¹, see Fig. ) displayed in Fig.  represents a best-case detection limit for MethaneSAT, which is facilitated by the low winds across that scene. We have further explored the drivers of plume detection limits for MethaneSAT using the Qmin formalism described in Sect. . Figure  shows rasters (in sensor coordinates) of XCH₄, near-infrared at-sensor radiance (proxied by surface albedo at 1606 nm multiplied by the cosine of the sun zenith angle), the standard deviation of XCH₄ within a 3 km-side box (σ(XCH4)), and the resulting Qmin from four MethaneSAT acquisitions: the 5 September 2024 Appalachian scene presented in Fig. , and three other scenes over sites in the Delaware sub-basin of the Permian Basin, Algeria and West Siberia. These four acquisitions have been selected in order to sample a wide range of acquisition conditions, in particular of at-sensor radiance (from very low in the Appalachian and West Siberia scenes to very bright in the Permian and Algeria acquisitions) and of wind intensity (from very low in the Appalachian scene to very strong in the Permian scene). At-sensor radiance and wind speed are expected to be the main drivers of MethaneSAT's plume detection ability: at-sensor radiance drives the retrieval noise level, whereas wind speed determines the number of pixels showing methane concentration enhancements above that noise level.

As shown by the panels on the third column of Fig. , σ(XCH4) typically ranges between 20 and 40 ppb (parts-per-billion). This combines contributions from the per-pixel retrieval precision error, the effect of surface clutter on the retrieval, and the natural variability of XCH₄ across the scene. The smaller σ(XCH4) is found for bright and spatially homogeneous surfaces, which enable the maximum measurement signal-to-noise ratio and the minimum surface clutter. This is the case of the Algeria site, where σ(XCH4) is around 22.6 ppb, and the Permian scence, with an average σ(XCH4) of 26.6 ppb. The at-sensor radiance of the Permian scene is actually a bit larger than the one in the Algeria scene because of the smaller zun zenith angle during the Permian acquisition, which took place in mid June. However, σ(XCH4) in the Permian acquisition is greater than that in the Algeria one because the surface is less homogeneous, and also because there is a higher horizontal variability in the XCH₄ field from the multiple sources present in the Permian. On the other hand, the σ(XCH4) is largest for the Appalachian and West Siberia sites (28.3 and 33.4 ppb on average, respectively), which are covered by vegetated surfaces with a low near-infrared albedo. On the other hand, we observe a low occurrence of false XCH₄ enhancements caused by surface structures, as opposed to the case of high spatial resolution missions with a coarser spectral resolution, for which it is often challenging to disentangle the methane signal from that of surface features .

For context, high resolution imaging spectrometers being used for plume detection, such as EnMAP or PRISMA, can have similar retrieval precision errors over ideal observation conditions (namely, homogeneous surfaces and high solar irradiance) , but the precision errors and the occurrence of false positives depend strongly on the heterogeneity of the surface due to the coarse spectral sampling of those instruments .

On the other hand, we find a substantial impact of the wind speed on our ability to detect methane plumes with MethaneSAT. For stronger winds, plumes and local methane enhancements are more effectively dispersed by the atmosphere, and the number of pixels with XCH₄ enhancements above the noise level becomes smaller. Thus, the Qmin calculated with Eq. () is smallest for the Appalachian acquisition (wind speed is 1.0±0.4 m s⁻¹ across the scene), and greatest for the Permian Basin one (6.5±1.0 m s⁻¹), despite the higher retrieval precision found for the Permian Basin scene. The Algeria and West Siberia cases show intermediate values of Qmin, with smaller Qmin, wind speeds and σ(XCH4) in the Algeria case. The extreme values of Qmin shown in Fig.  for the Appalachian and Permian scenes in Fig.  (very low and very high Qmin, respectively) actually match what we find when analyzing these two datasets for plume detection: we find a total of 16 plumes in the Appalachian (see Sect. ), whereas only two plumes are detected in the Permian scene, which can be attributed to the strong winds on that day.

This is further illustrated in Fig. . It displays the same Permian Delaware scene included in Fig. , and a second one acquired over the same area about 3-weeks earlier. Large concentrations of high-emitting point sources are usually found in this region . The visual comparison of the two maps reveals the impact of wind on the spatial distribution of methane enhancements: a number of methane plumes and local enhancements are found on 22 May 2024, where low-to-medium wind speeds were present (wind speed of 3.0±1.5 m s⁻¹), whereas a much more uniform methane concentration field with lower XCH₄ values is found in the 14 July 2024 acquisition, with much stronger winds (6.9±0.6 m s⁻¹). This led to a total of 10 plumes being detected across the area on 22 May, with emission rates between 0.7 and 3.5 t h⁻¹, whereas only two large plumes of 4 and 24 t h⁻¹ were found on 14 July, as shown in Fig. c. These results are consistent with the analysis in , which found hundreds of plumes over this area from two flight campaigns with MethaneAIR (MethaneSAT's airborne version), but only one source was estimated to be above 2 t h⁻¹.

This dependece of detection limits on wind speed actually happens to all instruments capable of plume detection (e.g. GHGSat or EnMAP) , but it may be stronger for MethaneSAT because of its coarser spatial resolution, which reduces the number of plume pixels with XCH₄ values above the noise level. The fact that MethaneSAT's plume detection ability is greater for a vegetated site (Appalachian) than for a semi-arid surface (Permian) in this comparison is mediated by the lower wind speed in the Appalachian acquisition. This role of wind speed as a limiting factor for plume detection is a characteristic of MethaneSAT observations. In contrast, the surface type is usually the main driver of plume detection for high-resolution instruments.

The point at which a distribution of emission rates rolls off toward smaller values can be taken as an indicator of the system detection limits. We have used the full list of plume detections from MethaneSAT to estimate an average plume detection limit representing a range of sites and acquisitions sites. This plume lists contains 165 quality-controlled plumes, extracted from 40 MethaneSAT acquisitions over 15 different sites. The histogram and lognormal fit for this plume distribution is shown in Fig. . A peak value of 1300 kg h⁻¹ is found, which can be considered a conservative estimate of plume detection for most of the acquisitions included in this dataset. For example, assumed this histogram peak value to represent a 90 % POD. We note that the histogram deviates substantially from a lognormal distribution due to the reduced number of detections in the 10 000 kg h⁻¹ range, which may be due to a statistical artifact caused by the relatively small plume sample. We assume that this has not a strong impact on the 1300 kg h⁻¹ estimate.

To complement this estimate of the average plume detection limit, a preliminary assessment of MethaneSAT's minimum detection limit has also been carried out. For this, WRF-LES plumes have been embedded into real XCH₄ maps. Figure  shows the result of adding simulated plumes into a subset XCH₄ map from a MethaneSAT acquisition over Hassi Messaoud (Algeria) on 17 November 2024. The subset covers an area of about 50 km × 50 km, with wind speed of 1.0±0.2 m s⁻¹ and σ(XCH4) of 25 ppb (estimated as the local standard deviation from the XCH₄ map). This subset included 3 real plumes according to the corresponding MethaneSAT L4-plume product, with emission rates between 1.1 and 2.7 t h⁻¹ (Fig. a), whereas one likely plume may have been missed around pixel (450, 90). The simulated plumes are marked by a red cross in Fig. b–f. The simulations were performed for different wind speeds (either 1.6 or 3.3 m s⁻¹) and flux rates (500, 1500 or 2000 kg h⁻¹). From visual inspection of the simulated plumes (Fig. b–f), it is evident that a plume with flux rate around 500 kg h⁻¹ would likely be detected for this particular scene if the wind speed remained below 1.6 m s⁻¹ (Fig. b), but this may not be the case for the same flux rate range if the wind speed were greater than 3 m s⁻¹ (Fig. d–f), as less plume pixels show XCH₄ values above the noise level in this case.

We would like to note that these results only provide a rough estimate of the minimum detection limit of MethaneSAT, applies to very bright and homogeneous areas such as the one in Algeria used for this test, and to low winds as described in Sect. . A more rigorous analysis of minimum and average detection limits would require a probability-of-detection framework and a greater number and range of simulated source rates, wind speeds, plume morphologies and site characteristics.

MethaneSAT has a 100 m × 400 m spatial sampling at nadir. This spatial sampling is substantially coarser than that of the high-spatial resolution missions typically used for point source detection, which have squared pixels with a spatial sampling between 20 and 60 m. This may challenge MethaneSAT's ability to pinpoint single sources with respect to those higher resolution missions.

We have explored this issue using again WRF-LES plumes. A simulated plume from a 2000 kg h⁻¹ source and a 30 m spatial sampling has been convolved with MethaneSAT's 100 m × 400 m sampling assuming a boxcar point spread function. This spatial convolution was made so that the plume was aligned with either the across-track or the along-track direction of the observation. The results are represented in Fig. . Pixels with ΔXCH₄ values below 25 ppb have been filtered out. The simulated plume and a zoom on the source location, both at a 30 m sampling, are shown in Fig. a and b, and the spatially-resampled plume is displayed in Fig. c and d for the plume aligned with the across- and along-track direction, respectively. From these simulations, it can be observed that the direction of the plume with respect to the observation is important for the potential detection and attribution of the plume, because of MethaneSAT's rectangular pixels: if the plume is aligned with the along-track direction (Fig. c), the pixels contain a smaller fraction of the methane enhancement, which decreases the chances of those pixels to stand out above the noise level to enable detection. In the case of plumes aligned with the across-track direction (Fig. d), peak XCH₄ values are greater than in the previous case because a larger fraction of the enhancement is sampled by each pixel, which increases the probability of plume detection with respect to the along-track alignment case.

Regarding the attribution of the plume to a source, the limitations imposed by MethaneSAT's relatively coarse spatial resolution are also illustrated by the simulations in Fig. . Whereas the attribution to the source is straightforward for the 30-m sampling case, this becomes challenging at MethaneSAT's spatial sampling, especially in the case of the along-track alignment, in which the first upwind pixel above the noise level is shifted with respect to the source location. This effect is smaller in the case of across-track plume alignment, but even in this case the pixel with the largest concentration enhancement is not the one containing the source, which could lead to ambiguities in the plume attribution process. Again, this example is provided here only to illustrate the problem. More simulations and a robust statistical analysis would be needed for a proper assessment of the limitations of MethaneSAT observations for plume attribution.

In order to further investigate the limitations interposed by MethaneSAT's relatively coarse spatial sampling, we have compared a MethaneSAT plume map with the equivalent data as provided by high-resolution imagers. We have chosen a MethaneSAT acquisition over Turkmenistan's South Caspian basin on 23 September 2024, which was also sampled by the Landsat-8 mission (30 m spatial sampling) on the same day. MethaneSAT overpass was at 10:42 UTC and Landsat's at 07:01 UTC, so there was a time difference of about 3.6 h between the two acquisitions. Wind speed for this subset region during the MethaneSAT overpass was 1.3±0.7 m s⁻¹. The coordinates and flux rate of the Landsat plumes were taken from the IMEO plume dataset . The results from this comparison are shown in Figs.  and . We have chosen this Landsat acquisition for comparison with MethaneSAT because it spans a number of plumes within a relatively broad range of emission rates. No match-ups with imaging spectroscopy missions, such as EnMAP or EMIT (more sensitive to methane than Landsat, but with a sparse spatio-temporal sampling), have been found for this comparison with MethaneSAT.

Figure  shows a spatial subset of a MethaneSAT XCH₄ map over the South Caspian basin, and the locations of the plumes obtained from both MethaneSAT's L4-plume product and the IMEO plume dataset. We find an overall good agreement in the source locations by the two datasets, although we also observe that one of the plumes detected by MethaneSAT corresponds to groups of several plumes in the finer-resolution IMEO dataset (i.e. the cluster of Landsat plume detections around 38.5° N, 54.2° E, see Fig. ). This indicates that the MethaneSAT L4-plume product is merging several sources into one single plume in this dataset because of the relatively coarse spatial sampling. The low wind conditions in this dataset may exacerbate this issue, as they cause the emitted gas to accumulate and linger near the source rather than dispersing downwind, complicating the separation of single sources.

A comparison between the emission rates provided by each dataset is shown in Fig. . The point with a yellow filling depicts the MethaneSAT plume for which several sources in the IMEO dataset have been merged into one, as discussed before. We find a relatively high linear correlation between the two datasets despite the difference in the observation time. The point with the largest deviation between the two datasets is the one from the merged Landsat plumes, for which we can expect the largest uncertainty. The consistency of emission rate estimates from Landsat was assessed by using controlled releases. Considering this, the good correspondence between the MethaneSAT and Landsat plume datasets shown in Fig.  represents a first proof of the reliability of MethaneSAT's L4 plume quantifications. However, it must be stated that this comparison does not represent a validation of the MethaneSAT emission rates: first, the Landsat emission rates may have substantial errors ; second, the emission rates from these super-emissions can vary strongly within the 3.6 h time difference between the MethaneSAT and Landsat acquisition.

Comparisons between the matched-filter and the CO₂-proxy retrievals are shown in Fig.  for two selected cases: a desert area in Hassi Messaoud (Algeria) with a number of relatively weak sources (see also Fig. ), and a Permian Basin acquisition affected by scattered clouds and strong winds (see also Fig. ). The XCH₄ raster maps from the CO₂-proxy retrieval are presented in raw and quality-masked formats (left and center column, respectively). In the second case, the quality mask included in MethaneSAT's L2 products has been applied to screen out low-quality pixels (including cloud-contaminated pixels) from the raw XCH₄ maps. Relative ΔXCH₄ maps from the raw matched-filter output (no quality mask applied) are shown on the right column for the two areas.

The visual comparison of the results from the two retrievals confirms that the matched-filter retrieval does offer an added-value for the detection of methane plumes, because of its enhanced signal-to-noise ratio and its lower sensitivity to clouds and artifacts. The lower noise in the ΔXCH₄ maps from the matched-filter is visible in the Hassi Messaoud maps. The Hassi Messaoud case also shows that the raw XCH₄ map from the CO₂-proxy retrieval presents some artifacts resembling methane plumes (Fig. a) that are removed by the quality mask (Fig. b), whereas these do not appear in the raw ΔXCH₄ map from the matched-filter (Fig. c). With respect to clouds and cloud shadows, we observe that the matched-filter retrieval is more robust against cloud contamination, leading to the detection of two plumes in the bottom-right quarter of panel Fig. f which would not be detectable in the L2 product, with or without the quality mask (Fig. d and e).

On the other hand, we have found the matched-filter retrieval to have strong variations in the absolute ΔXCH₄ values with respect to the proxy retrieval, ranging from a 15 % underestimation with respect to the CO₂-proxy retrieval for some pristine acquisitions (like the Hassi Messaoud dataset used in this test) up to a 2× difference in others with less favourable acquisition conditions (e.g. regions with vegetated surfaces and high cloud contamination). For this reason, we opt for using arbitrary units with the matched-filter ΔXCH₄ maps in Fig. . The matched-filter ΔXCH₄ maps are in general very useful for plume detection, but we do not use them to produce quantitative estimates of emission rates. Further work is required to understand this low-bias behavior of the matched-filter with MethaneSAT, which was not observed for MethaneAIR as reported by . The relatively low number of along-track samples available in MethaneSAT (about 500 pixels) could partly explain this low bias . A similar ΔXCH₄ underestimation of MethaneSAT matched-filter retrievals and potential approaches to solve it are discussed in .

Turkmenistan's South Caspian basin is considered the largest hotspot of O&G emissions in the world . MethaneSAT's XCH₄ maps provide a particular view of methane super-emissions in this basin thanks to MethaneSAT's wide coverage and relatively fine spatial resolution.

Figure  shows a time series of XCH₄ maps over the region between September 2024 and May 2025. Red crosses mark the sources that were detected in the 23 September 2024 acquisition, but are plotted in all 6 panels to illustrate the persistency of the sources in the region within this time period, for which the 23 September acquisition is taken as a reference. A greater number of sources are found in the southern part of the South Caspian basin, although these sources are in general weaker (typically ranging between 5 and 25 t h⁻¹) than the two detected in the north, which are estimated to be between 25 and 30 t h⁻¹ on 23 September, and between 35 and 40 t h⁻¹ on 18 November (see Fig. ). In general, we observe a large degree of consistency in the sources that are active between September and November 2024 despite the variability introduced by changes in wind speed and direction. This confirms that the emissions in the South Caspian basin are not only very strong, but in general also very stable over long periods of time. However, we do observe some changes in the location of the sources located in the northernmost side of the region in the May 2025 acquisition, with the stronger sources being shifted westwards with respect to the September acquisition, which would indicate an evolution in the O&G extraction activity in the region between November 2024 and May 2025.

We acknowledge that some of these plumes may combine emissions from several sources, as also discussed for the same 23 September acquisition in Sect. , so it is not clear whether it is always the exact same sources that are active, or neighboring sources within the same area. Adding plume detections from higher-resolution satellite instruments, such as those included in the IMEO plume database , would help to better understand these patterns.

The Permian Basin, located between New Mexico and Texas, in the USA, is another major O&G methane super-emission hotspot. Even though O&G operations in the Permian Basin are much more regulated than in Turkmenistan, methane super-emissions in the Permian Basin are also widespread because of the high production activity in the region.

The Permian Basin is comprised of two main sub-basins, namely the Delaware, in the western portion, and the Midland, in the eastern portion. High-resolution airborne surveys and inversion data from TROPOMI show spatial emission hotspots in both sub-basins, although the Delaware sub-basin is the one for which higher levels of methane emissions are expected according to previous reports based on airborne and satellite data . MethaneSAT XCH₄ maps for the Delaware sub-basin have been discussed earlier in this study, including a medium-wind acquisition over this sub-basin from which up to 10 plumes with emission rates between 700 and 3500 kg h⁻¹ where detected (see Figs. a and ).

However, our analysis of MethaneSAT data over both sub-basins shows that the Midland sub-basin presents an outstanding concentration of high-emitting point sources in September and October 2024. This is illustrated in Fig. , which presents XCH₄ maps over an area of about 100×100 km² sampling the Midland sub-basin on four different days (two pairs of consecutive days). A high number of strong point sources is found in three of the four acquisitions (see Fig. ). For example, 14 plumes with emission rates between 0.5 and 15 t h⁻¹ (5 of which have an emission rate greater than 5 t h⁻¹) can be found in the 28 September 2024 map (Fig. a). These numbers are substantially greater than what has been found for the Delaware sub-basin using MethaneSAT data, and they also exceed those in other studies of super-emitters in the Delaware sub-basin. For example, detected hundreds of plumes in the Delaware sub-basin with MethaneAIR in 2021 and 2023, but only one exceeded 2 t h⁻¹; used space-based imaging spectrometers detected 37 plumes across the Delaware sub-basin combining several dates in 2020, but only 3 of them had emission rates greater than 5 t h⁻¹. A very small fraction of plumes above 5 t h⁻¹ are also reported in the airborne survey by covering both the Delaware and Midland sub-basins.

Considering how clear those plumes appear in the MethaneSAT maps, we have confidence that the high rates that we have estimated are real. Also, zooming in over single plumes, we do not find evidence of the merging of several sources. The fact that many of the plumes are from the same sources on subsequent days with different patterns in the winds gives us some confidence that the reported plumes are from single sources are not coincidental aggregation of flux from multiple sources.

The dynamics of these super-emissions in the Midland sub-basin between September and October 2024 are also intriguing: whereas the sources appear to be very stable between the two consecutive acquisitions on 25 and 26 October, the opposite happens for those in September, when the 14 sources detected on 28 September are reduced to only four in the next day. This variability can also be seen in the total point source emissions reported in Fig. : the maximum total point source emission is found on 28 September (87.8 t h⁻¹), which is reduced by a factor ∼5x on the next day (16.3 t h⁻¹). Intermediate and more similar values are found on 25 and 25 October (50.4 and 57.6 t h⁻¹). The strong reduction in emissions from 28 to 29 September are also found in the corresponding total regional fluxes, which are included in the L4 products available in MethaneSAT's Data Portal : the total emissions are 180 t h⁻¹ for 28 September, and 140 t h⁻¹ for 29 September. It must be remarked that this total emission estimate combines the detected point source emissions with the total area emissions estimated through a full atmospheric model inversion. Interestingly, the total emission rate estimated from the 26 October scene, 192 t h⁻¹, is slightly higher than the one obtained for 28 September, unlike what happens when only point sources are computed. There is no total emission estimate from the 25 October scene, as the L4-total flux product from this date did not pass the quality control.

The strong variability found between the two September scenes can not be related to an increase in the plume detection limits associated to wind speed, as this is actually lower on 29 September than in the day before (see the mean Q values in Fig. ), so the abrupt change in emissions may be explained by the particular operations taking place in the area between those two days. Prior research has shown variation in Permian emissions connected with takeaway capacity and concurrent variation in oil prices . Before additional pipeline capacity came online in November 2024, west Texas natural gas spot prices were frequently negative throughout the summer and early fall of 2024, creating an environment where Permian operators lost money to transport natural gas . We note that 76 % (23/30) of the marked plumes (Fig. ) originated from midstream facilities, including compressor stations, processing plants, and gathering pipelines. The heighted rate of plumes from midstream facilities may be indicative of the beforementioned limitations in takeaway capacity for the broader Permian at that time of the year.

We combined the information on methane enhancements included in the IMEO plume database with our own analysis of the MethaneSAT data archive in order to select the O&G regions in the world with a persistent presence of super-emissions (Fig. ). Sample XCH₄ enhancement maps from the 6 regions where we found the strongest presence of high-emitting point sources in the world are shown in Fig. . The 9 panels show XCH₄ enhancements represented with the same colorbar and cover the same area (1.5°×1.5°), which enables the visual comparison of the different maps.

The South Caspian and Permian Midland maps were already presented in Figs.  and , and are reproduced in Fig. a and b as a reference for comparison with the other basins. Maps over the Permian Delaware sub-basin showing a number of plumes are represented to cover the three areas in the world where the highest concentration of super-emissions can be expected (Fig. c and d). In addition, maps over three other regions are included in the center and bottom row of Fig. : the Maturin basin in Venezuela and the Zagros-Foldbelt and the Widyan basins in Iran. Although these three basins are also depicted by the IMEO database as regions with strong methane emission activity (Fig. ), the regional-level snapshot of XCH₄ enhancements provided by MethaneSAT suggests that the super-emissions from those regions are outstanding. For example, five sources with emission rates higher than 5 t h⁻¹ are found to be simultaneously active across the Maturin basin (see Fig.  and ), whereas three sources with emission rates greater than 20 t h⁻¹ are found in the Zagros-Foldbelt basin on 19 December 2024 (Fig. ). Several very strong point sources are also identified in the Widyan scenes. The constant presence of strong super-emissions in these six regions is confirmed by several other MethaneSAT acquisitions over each of the basins. The very large emissions from Maturin, Zagros-Foldbelt and Widyan have received limited attention in previous studies, but the MethaneSAT XCH₄ data indicate that these areas warrant further targeted investigation.

Finally, methane concentration maps derived from MethaneSAT observations over other sites for which interesting methane emission events have been found, are presented in Fig. .Figure a shows a number of relatively small plumes (range between 500 and 2500 kg h⁻¹, see Fig. ) across the Haynesville-Bossier basin. This range of emissions fits the emission rates typically detectable with public point-source imagers (e.g. EnMAP, EMIT, and Sentinel-2) over semi-arid areas, but they would have much higher detection limits over vegetated surfaces like the ones in this Haynesville-Bossier site. The low-wind conditions for this Haynesville-Bossier observation (0.5±0.2 m s⁻¹) are similar to the ones over the Appalachian region shown in Fig. , leading to low detection limits despite the fact that the high fraction of vegetated cover makes this region to be less optimal for plume detection than arid and desert sites. Regarding other acquisition conditions, Fig. b shows plumes from three active sources between 1300 and 4000 kg h⁻¹ (see Fig. ) within about 30 km in the Amu Darya basin (Eastern Turkmenistan), which was also identified as a global hotspot with TROPOMI (see Fig. ). Figure c shows two plumes separate sources, in this case with very high emission rates (greater than 100 t h⁻¹) found at high latitudes, in West Siberia. No plumes from these sources were found from an overpass of the VIIRS instrument earlier on that day using the methods by , which suggests that this could have been a short-lived event. We speculate that those two concurrent, large and possibly transient plumes may be caused by pipeline maintenance works in the area. Another very strong emission from an O&G facility is shown in Fig. d. It corresponds to an O&G extraction platform near the coast of Campeche, in the Mexican side of the Gulf of Mexico. Interestingly, this plume is not from the nearby Zaap-C platform, from which persistent and very large emissions (sometimes greater than 100 t h⁻¹) have been detected in the past . A different MethaneSAT acquisition does show a plume from the Zaap-C platform (data not shown). The last two panels in Fig.  show examples of emissions from the waste sector: a large plume from a landfill in Saudi Arabia, close to Bahrain, is represented in Fig. e. An emission rate of 15.2 t h⁻¹ was estimated for this source, which is consistent with emission estimates for this landfill from high-resolution missions. Lastly, Fig. f shows a massive methane enhancement around Dhaka city, in Bangladesh, that we attribute to a combination of several landfills and dump sites across the city. Similar very large methane enhancements have been found over this site in other MethaneSAT observations. These are probably among the largest methane enhancements found in the entire MethaneSAT archive, and the site was also listed as one of the largest and most persistent methane emitting regions in the world by .

The XCH₄ maps from MethaneSAT shown in the previous figures offer a unique combination of wide spatial coverage, medium resolution and low sensitivity to surface-related artifacts. This complements the data from available point source imagers, which generally have a finer spatial sampling at the expense of a narrower coverage and a higher occurrence of false positive caused by surface structures.