Atmospheric CO₂ and CH₄ column measurement for climate-treaty datasets
OCO-2, OCO-3, GOSAT-2 and TROPOMI retrieve column-averaged CO₂ and CH₄ from shortwave infrared spectra, giving treaty bodies an independent check on national greenhouse-gas inventories. The technique is powerful but cloud-biased and cannot directly invert surface fluxes without modelling.
Sensors
- OCO-2 (NASA, launched 2014): Three-band grating spectrometer covering O₂ A-band (0.765 µm), weak CO₂ (1.61 µm) and strong CO₂ (2.06 µm). Nadir and glint modes; 1.29 × 2.25 km footprints in a 10.6 km swath. Repeat cycle 16 days. Single-sounding XCO₂ precision approximately 1 ppm; regional bias target < 0.5 ppm. Cloud and aerosol screening rejects roughly 75–80 % of soundings globally.
- OCO-3 (NASA, ISS, launched 2019): Same optical design as OCO-2 but mounted on the International Space Station, giving non-sun-synchronous coverage between 52° N and 52° S with variable overpass times. Adds a Snapshot Area Map mode that tiles a ~80 × 80 km region in minutes, useful for city-scale flux studies. Revisit at any given point is irregular, roughly every few days in the tropics.
- GOSAT-2 (JAXA, launched 2018): Fourier-transform spectrometer (TANSO-FTS-2) measuring CO₂, CH₄ and CO in the shortwave infrared plus a thermal infrared channel. Circular footprint of approximately 9.7 km diameter; 3-day global repeat. Lower spatial density than OCO-2 but broader spectral coverage aids aerosol correction and adds CH₄ column sensitivity. Cloud imaging camera (CAI-2) screens cloudy scenes.
- TROPOMI CH₄ (ESA/Copernicus Sentinel-5P, launched 2017): Pushbroom imaging spectrometer; SWIR channel centred near 2.3 µm retrieves XCH₄. Pixel size 5.5 × 7 km (improved from original 7 × 7 km after 2019 processor update). Daily global coverage. Single-pixel XCH₄ precision approximately 0.6–1 %; particularly effective for detecting large point-source methane plumes from oil and gas infrastructure when scenes are cloud-free.
- GOSAT (JAXA, launched 2009): The original TANSO-FTS instrument; still operational and extending the XCO₂/XCH₄ record back to 2009, making it the longest continuous satellite GHG column dataset. Footprint and revisit similar to GOSAT-2. Valuable as an archive anchor for trend analysis over more than 15 years.
What the spectrum actually measures, and what it does not
All four instruments exploit the same physical principle. Sunlight reflected from the Earth's surface passes twice through the atmospheric column. Molecules of CO₂ or CH₄ absorb at characteristic near-infrared wavelengths, leaving depth-of-absorption fingerprints that the spectrometer resolves. The ratio of measured to expected absorption gives the column-averaged dry-air mole fraction: XCO₂ in parts per million, XCH₄ in parts per billion.
The critical word is 'column-averaged'. The retrieval integrates over the entire atmospheric depth, weighting most heavily toward the lower troposphere but not exclusively sensing the surface layer where emissions originate. Converting a measured XCO₂ anomaly into a surface flux requires an atmospheric transport model, prior assumptions about meteorology, and an inversion framework. The satellite does not see a smokestack; it sees the cumulative chemical signature of everything below it. That distinction matters enormously when a government asks whether a satellite measurement confirms or contradicts its reported inventory.
The aerosol problem: scattering distorts the optical path
Aerosol particles scatter photons sideways, effectively shortening the optical path that the retrieval algorithm assumes is a clean vertical column. If uncorrected, aerosol loading causes the algorithm to underestimate the true gas column. OCO-2 and OCO-3 address this partly through the O₂ A-band, which probes the same atmospheric path without a GHG signal; discrepancies between observed and modelled O₂ absorption flag scenes with heavy aerosol. GOSAT-2's thermal infrared channel provides additional aerosol constraint.
Even with these corrections, scenes over dust-prone or smoke-affected regions carry higher retrieval uncertainty. Published validation studies comparing OCO-2 against TCCON (Total Carbon Column Observing Network) ground stations report single-sounding XCO₂ biases that can reach 1–2 ppm in aerosol-heavy conditions, against a background atmospheric gradient of roughly 400–425 ppm. That is a small fractional error but not negligible when the policy question is whether a country's emissions are 5 % higher or lower than declared.
Cloud bias: the regions that matter most are often hidden
Cloud screening is not optional. A cloudy scene scrambles the optical path and makes the retrieval meaningless, so algorithms discard it. The consequence is systematic: tropical and high-latitude regions, which include major emitters such as the Congo Basin, Southeast Asian peatlands, and Arctic methane sources, are disproportionately cloud-covered. OCO-2 retains roughly 20–25 % of soundings after quality filtering globally, but that fraction is considerably lower over persistently cloudy regions.
This is not a solvable instrument problem; it is a physical constraint of passive shortwave sensing. The practical implication for treaty verification is that satellite column data provide dense, reliable coverage over semi-arid and temperate regions and sparse, seasonally biased coverage elsewhere. Any national inventory comparison must account for where the satellite actually sampled, not just where it flew. Aggregating sparse soundings over months reduces random noise but does not eliminate the geographic sampling bias.
From column anomalies to inventory verification: the inversion step
The standard workflow for treaty-relevant analysis is an atmospheric inversion. A transport model (GEOS-Chem, TM5, or similar) simulates how surface fluxes propagate into column concentrations. The inversion adjusts the prior flux map until the modelled columns match the satellite observations. The residual between the inversion-derived flux and a country's self-reported inventory is the verification signal.
Several research groups, including those running the CarbonTracker system at NOAA and the Copernicus Atmosphere Monitoring Service (CAMS), publish regular global inversion products using OCO-2 and GOSAT data. These are not yet formally adopted as binding treaty instruments under the Paris Agreement's Enhanced Transparency Framework, but they are increasingly cited in scientific assessments and bilateral diplomatic discussions. The gap between 'scientifically credible' and 'politically accepted' remains wide, partly because inversion uncertainty is still large enough that a country can plausibly dispute a discrepancy within the error bars.
TROPOMI and the point-source opportunity
For methane specifically, TROPOMI's daily global coverage at 5.5 × 7 km resolution has opened a different application: direct detection of large point-source emitters. Coal mines, landfills, and oil and gas facilities with sufficiently high emission rates (published detection thresholds are roughly 1,000–5,000 kg CH₄ per hour for a single overpass, depending on wind and atmospheric conditions) produce plumes detectable as XCH₄ enhancements above background.
This is more operationally direct than the inversion approach. A detected plume above a specific facility is harder to dismiss than a regional inversion anomaly. Groups including the International Methane Emissions Observatory (IMEO) and academic teams have catalogued thousands of such events from TROPOMI data since 2018. The limitation is the detection floor: smaller facilities, distributed agricultural sources, and diffuse wetland emissions remain below TROPOMI's single-pass sensitivity. Combining TROPOMI with higher-resolution commercial sensors is an active research direction, though that falls outside the scope of what open operational instruments alone can provide.
Satellize runs TROPOMI-based methane anomaly screening as part of its analytics stack, applying the same published plume-detection methods to support clients building independent emissions-monitoring capacity. The Kingdom of Tonga crop-estimation programme is a separate engagement, but the underlying pipeline for ingesting and processing Sentinel-family data is shared infrastructure.
What a government buyer should realistically expect
Satellite GHG column data are a credible, independent cross-check, not a definitive audit. A country building climate-treaty compliance capacity should expect the following: regional XCO₂ anomalies detectable at roughly 0.5–1 ppm above background after multi-year aggregation; XCH₄ point-source detection for large industrial emitters with cloud-free overpass; and inversion-based flux estimates with uncertainties that are typically 20–40 % at the national scale for smaller emitters, narrowing as the archive lengthens.
The technology is improving. The planned CO2M constellation (Copernicus Expansion Mission, two satellites expected in the late 2020s) is designed specifically for national inventory verification, with 2 × 2 km pixels and daily revisit over Europe. Until that generation arrives, OCO-2, GOSAT-2 and TROPOMI represent the operational ceiling. They are genuinely useful. They are not yet precise enough to resolve a 5 % discrepancy in a mid-sized country's annual inventory with statistical confidence. Knowing that limit is the beginning of using the data well.
Typical figures
| XCO₂ spatial resolution (OCO-2/OCO-3) | 1.29 × 2.25 km per sounding; 10.6 km swath |
| XCH₄ spatial resolution (TROPOMI) | 5.5 × 7 km per pixel (post-2019 processor) |
| GOSAT-2 footprint | ~9.7 km diameter circular; 3-day global repeat |
| OCO-2 repeat cycle | 16 days (sun-synchronous); ~1:30 pm local overpass |
| TROPOMI global revisit | Daily (one overpass per day at any given point) |
| Single-sounding XCO₂ precision (OCO-2) | ~1 ppm; regional bias target < 0.5 ppm vs TCCON |
| XCH₄ point-source detection floor (TROPOMI) | ~1,000–5,000 kg CH₄ hr⁻¹ (published range, wind-dependent) |
| Spectral channels | O₂ A-band (~0.765 µm), weak CO₂/CH₄ (~1.61 µm), strong CO₂/CH₄ (~2.06–2.3 µm) |
| Clear-sky data yield | ~20–25 % of soundings pass quality filter globally (OCO-2) |
| Archive depth | GOSAT from 2009; OCO-2 from 2014; TROPOMI from 2017; OCO-3 from 2019 |
Analytics Satellize can run
| National XCO₂ anomaly time series | Spatial aggregation of OCO-2 or GOSAT-2 quality-filtered soundings over defined administrative boundaries, bias-corrected against TCCON anchor stations | Annual and quarterly CSV/GeoJSON report with uncertainty ranges, suitable for inclusion in Enhanced Transparency Framework submissions |
| TROPOMI methane point-source screening | Published cross-sectional flux method applied to TROPOMI XCH₄ plumes; wind-field integration from ERA5 reanalysis to estimate emission rate | Facility-level alert feed (JSON) flagging statistically significant plume events above a client-defined emission threshold |
| Seasonal cloud-bias assessment | Sounding-density mapping per administrative region per month, quantifying where and when the satellite record is sparse relative to the emission calendar | Sampling-bias GIS layer (GeoTIFF + report) advising which months and regions require supplementary ground or aircraft data |
| Inversion-ready satellite data package | Pre-processed, quality-filtered and georeferenced OCO-2/GOSAT-2 soundings formatted to standard inversion model input specifications (e.g. CarbonTracker-compatible) | Structured NetCDF dataset with metadata, ready for client atmospheric modelling teams |
| Multi-year XCH₄ trend analysis | Linear and seasonal decomposition of TROPOMI column retrievals over defined regions, using published SRON/ESA processor output | Trend report (PDF + data tables) showing annual rate of change with confidence intervals, referenced to global background from NOAA surface network |
| Inventory cross-check summary | Comparison of satellite-derived regional flux estimates (from published CAMS or CarbonTracker inversion products) against client's self-reported national inventory figures | Structured briefing document identifying sectors and regions where satellite signal is consistent or inconsistent with reported totals, with explicit uncertainty caveats |
Who does the work
We can get this done for you. Satellize runs its own analyst desk and a strong science team. You do not buy a data feed and work out what it means; our people source the imagery, run the analysis described on this page, and hand you the answer with its confidence limits stated. Discuss this requirement.