XCO2 anomaly detection above fossil fuel facilities
Detecting CO2 plumes from power stations and industrial sites via satellite is possible but genuinely difficult: a 1,000 MW coal plant lifts the local column by perhaps 1–3 ppm against a 420 ppm background, demanding sub-ppm retrieval precision and careful error budgeting.
Sensors
- OCO-2 (NASA): 1.29 km × 2.25 km ground footprint per sounding, eight cross-track footprints giving a ~10 km swath. Near-polar sun-synchronous orbit with a 16-day exact repeat. Retrieves XCO2 in the 1.61 µm and 2.06 µm CO2 bands plus the 0.76 µm O2 A-band for surface pressure correction. Single-sounding precision roughly 1 ppm after quality filtering; useful for detecting city- and facility-scale plumes when the orbit track crosses the source.
- OCO-3 (ISS-mounted): Same spectrometer heritage as OCO-2 but mounted on the International Space Station at 51.6° inclination, enabling coverage between roughly 52°S and 52°N at variable local solar times. The Snapshot Area Map (SAM) mode re-points the instrument to map a ~80 km × 80 km area in about two minutes, producing spatially contiguous XCO2 fields over a target facility. Revisit is irregular, typically days to weeks depending on ISS orbital phasing.
- GOSAT-2 (JAXA): Thermal And Near-infrared Sensor for carbon Observation (TANSO-FTS-2) with a ~9.7 km diameter circular footprint and 3-day revisit. Retrieves XCO2 and XCH4 simultaneously. Coarser spatial sampling than OCO-2 limits its ability to resolve individual facility plumes, but its higher revisit and sensitivity improvements over GOSAT-1 make it useful for regional background characterisation and trend validation.
- Copernicus CO2M (forthcoming, ESA/EU): A constellation of two (potentially three) satellites designed specifically for anthropogenic CO2 monitoring. Target XCO2 precision of 0.7 ppm per sounding at ~2 km × 2 km resolution, with a swath wide enough for daily coverage of Europe and regular global coverage. Co-registered NO2 and cloud imagers are included to aid source attribution and cloud screening. Launch targeted for the late 2020s; no operational data yet available.
Why CO2 is harder to see from orbit than methane
Methane attribution from space benefits from a high signal-to-background ratio: a large point source can produce a local enhancement of hundreds of parts per billion against a background near 1,900 ppb, a relative anomaly of several per cent. CO2 offers no such luxury. A 1,000 MW coal plant burning at full load emits roughly 6–9 million tonnes of CO2 per year, yet the resulting column enhancement above the facility is only 1–3 ppm on top of a background that now sits around 420 ppm globally. That is a relative signal below 0.7 per cent, smaller than many retrieval error sources.
The dominant error sources are aerosol scattering, which biases the apparent photon path length, and surface pressure uncertainty, which changes the dry-air column denominator. A 1 hPa surface pressure error translates to roughly 0.25 ppm XCO2 bias. Aerosol errors of comparable magnitude are common in industrial areas where the very facilities being monitored also emit particulates. This creates a frustrating circularity: the dirtier the source, the harder it is to retrieve a clean CO2 signal above it.
What OCO-3 snapshot maps actually show
OCO-3's Snapshot Area Map mode was designed partly to address the orbital coincidence problem that limits OCO-2. Rather than relying on a narrow sun-synchronous track happening to cross a facility, SAM mode aggressively re-points the instrument to build a spatially contiguous XCO2 map over a pre-selected target. Published results from OCO-3 SAM campaigns over cities including Los Angeles, Mumbai and Riyadh show detectable XCO2 enhancements of 1–4 ppm downwind of urban-industrial cores, consistent with bottom-up emission inventories within reported uncertainties.
Individual power station attribution is harder. A single large plant's plume may span only a few OCO-3 footprints before diluting into the broader urban signal. Wind speed and direction at the time of overpass are critical: a 5 m/s wind disperses the plume rapidly, while calm conditions concentrate it. Published studies using OCO-2 glint-mode data over isolated industrial sites in regions with low background variability have demonstrated facility-level detection, but these are favourable cases. Cloudy skies, which block the solar backscatter signal entirely, are the most common reason for data loss.
The retrieval chain: from radiance to XCO2
OCO-2 and OCO-3 measure reflected sunlight in three spectral bands: the O2 A-band near 0.76 µm, a weak CO2 band near 1.61 µm, and a strong CO2 band near 2.06 µm. The O2 A-band constrains surface pressure and photon path length. The two CO2 bands together allow retrieval of the column-averaged dry-air CO2 mole fraction via full-physics radiative transfer inversion, fitting thousands of spectral channels simultaneously. NASA's operational product (B11 and later versions) reports per-sounding XCO2 with an estimated single-sounding precision of approximately 0.5–1 ppm after bias correction and quality filtering.
That quality filtering is aggressive. Roughly 20–30 per cent of soundings over land pass all quality flags in clear-sky conditions; over industrial areas with elevated aerosol optical depth, the fraction is lower. Analysts working with OCO data for facility attribution typically apply additional screening for aerosol optical depth from co-located MODIS or VIIRS retrievals, and use meteorological reanalysis fields (ERA5 is standard) to model expected plume geometry before interpreting residual XCO2 anomalies.
CO2M and the step change it is designed to deliver
The Copernicus CO2 Monitoring mission addresses the two structural weaknesses of current instruments: swath and precision. OCO-2's ~10 km swath means most facilities are never observed on any given day. CO2M's design specification targets a swath sufficient for daily European coverage, with per-sounding XCO2 precision near 0.7 ppm at roughly 2 km × 2 km spatial resolution. At that precision and density, plume inversion methods that are currently applied to city-scale signals could, in principle, be applied to large individual facilities.
The co-registered NO2 imager is a significant design choice. NO2 is co-emitted with CO2 during fossil fuel combustion, and its column can be measured at much higher signal-to-noise from space (as TROPOMI already demonstrates at 3.5 km × 5.5 km resolution). Using NO2 as a combustion tracer to constrain the spatial structure of the CO2 plume, then solving for the CO2/NO2 emission ratio, is a published method that CO2M's instrument complement is explicitly designed to support. The approach does not eliminate the aerosol problem, but it reduces dependence on absolute XCO2 precision alone.
Honest limits for buyers and regulators
No current operational satellite system can reliably attribute XCO2 anomalies to a single facility smaller than a large power station (roughly 500 MW or above) with the frequency needed for routine compliance monitoring. OCO-2 and OCO-3 are research instruments; their data are invaluable for validating emission inventories at regional scale and for demonstrating the physics of the retrieval, but their revisit and swath are not suited to operational enforcement cadences. GOSAT-2 has better revisit but coarser spatial resolution.
Cloud cover is not a minor inconvenience. In tropical and mid-latitude industrial regions, cloud-free observation rates can fall below 30 per cent of days at any given location. Combining multiple sensors and using partial-cloud retrievals where validated can improve temporal sampling, but gaps remain. Buyers should treat current XCO2 satellite products as inventory-validation tools rather than real-time enforcement instruments. CO2M, if it meets its design specifications, changes that calculus for facilities in Europe and other well-observed regions, but it is not yet operational.
Satellize monitors pre-operational CO2M data releases and integrates OCO-2, OCO-3 and GOSAT-2 products into its emissions analytics stack alongside co-located NO2 from TROPOMI. Clients interested in benchmarking current retrieval capability against a specific facility portfolio can request a feasibility assessment through the Overhead analysis service.
Typical figures
| Spatial resolution (OCO-2) | 1.29 km × 2.25 km per sounding; ~10 km swath, 8 footprints cross-track |
| Spatial resolution (OCO-3 SAM mode) | ~2.2 km × 1.6 km per sounding; ~80 km × 80 km area map per acquisition |
| Spatial resolution (CO2M, design target) | ~2 km × 2 km per sounding; swath designed for daily European coverage |
| XCO2 single-sounding precision (OCO-2, post-QC) | ~0.5–1.0 ppm (bias-corrected, cloud-free, low-aerosol conditions) |
| XCO2 precision target (CO2M) | 0.7 ppm per sounding (design specification; not yet validated in orbit) |
| Spectral bands used for XCO2 retrieval | O2 A-band (~0.76 µm), weak CO2 (~1.61 µm), strong CO2 (~2.06 µm) |
| Revisit (OCO-2) | 16-day exact repeat; sub-km track offset each day gives denser sampling over time |
| Revisit (OCO-3) | Irregular; ISS orbital phasing gives days to weeks between SAM acquisitions at any site |
| Minimum detectable facility signal (current instruments) | Approximately 1–3 ppm enhancement; reliable detection typically requires sources ≥500 MW under favourable meteorology and low aerosol loading |
| OCO-2 archive depth | September 2014 to present; freely available via NASA GES DISC |
Analytics Satellize can run
| Facility-level XCO2 anomaly time series | Quality-filtered OCO-2/OCO-3 soundings within a defined radius of facility centroid, background subtracted using upwind reference soundings; method follows published cross-track bias correction approaches | Quarterly CSV report of overpass-by-overpass XCO2 anomaly values with confidence flags, cloud-cover fraction and wind metadata per acquisition |
| Plume inversion emission estimate | Gaussian plume or cross-sectional flux inversion using ERA5 wind fields co-located with OCO-3 SAM or OCO-2 glint-mode soundings; published method class used by Nassar et al. and similar OCO-2 facility studies | Per-acquisition CO2 emission rate estimate (tonnes/hour) with uncertainty range, delivered as a structured data record alongside the source sounding subset |
| NO2-constrained CO2 attribution layer | TROPOMI NO2 column co-located with XCO2 soundings; emission ratio approach to improve spatial attribution of CO2 source; method consistent with CO2M instrument design rationale published by ESA | GIS polygon layer showing probable CO2 source footprint with NO2-derived combustion flag per facility, updated on each valid overpass |
| Inventory consistency assessment | Comparison of satellite-derived XCO2 anomalies against reported facility emissions from national registries (e.g. EU ETS, UNFCCC national inventories); statistical consistency test across all valid overpasses in a calendar year | Annual PDF report scoring consistency between observed satellite signal and reported emissions, with flagged discrepancy cases for regulatory review |
| Cloud-gap analysis and observability forecast | ERA5 cloud fraction climatology and MODIS cloud occurrence statistics over target facility locations; Monte Carlo sampling of OCO-2 track geometry to estimate annual clear-sky overpass frequency | Site-specific observability report stating expected number of usable OCO-2 and OCO-3 acquisitions per year, informing feasibility of satellite-based monitoring for a given facility |
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.