Co-emitted methane and CO as combustion health-exposure proxies
Where ground-level PM monitors are absent, co-emitted CH4 and CO columns from TROPOMI and MOPITT reveal incomplete combustion intensity and let analysts map downwind health-exposure corridors, with honest limits on vertical resolution.
Sensors
- Sentinel-5P TROPOMI: Measures tropospheric CO columns at 5.5 × 7 km pixel resolution (upgraded from the original 7 × 7 km after 2019 reprocessing) with daily global coverage. Also retrieves CH4 total columns at the same resolution. Near-real-time products are available within 3 hours of overpass; offline products within 5 days. Minimum detectable CO column enhancement is roughly 5–10 ppb above background in cloud-free conditions.
- MOPITT (Terra): Measures tropospheric CO in thermal-infrared and near-infrared channels at 22 × 22 km resolution with 3-day global revisit. Uniquely provides vertical CO profiles at up to 7 pressure levels, allowing partial separation of boundary-layer from free-tropospheric contributions, though vertical resolution remains coarse (roughly 2–3 km per layer).
- VIIRS (Suomi-NPP / NOAA-20): Fire Radiative Power (FRP) retrievals at 375 m resolution, with twice-daily polar overpass. FRP is a proxy for combustion rate and, by extension, emission flux. Used to anchor the emission-ratio calculation spatially and temporally. VIIRS Nightfire also detects sub-pixel flares in shortwave infrared bands.
- MODIS (Terra / Aqua): Fire counts and FRP at 1 km resolution, twice-daily. The long archive (2000 to present) supports multi-year combustion climatologies that contextualise acute episodes detected by TROPOMI.
What incomplete combustion tells you before the PM monitor does
Open burning, whether agricultural residue, informal waste incineration or industrial flaring, rarely achieves complete oxidation. The products are a mixture: CO2 and water from complete combustion, and CO, CH4, volatile organic compounds and particulate matter from the incomplete fraction. The ratio of CO to CO2 in a fire plume is called the modified combustion efficiency (MCE); lower MCE means more smouldering, more CO, and critically more fine particulate matter. This relationship is documented across fuel types in published emission-factor compilations, including the GFED4 dataset and field studies from the SAFARI and ARCTAS campaigns.
The practical consequence is that a satellite measuring CO column enhancement over a burning region is not just counting molecules. It is reading a signal that correlates, imperfectly but usefully, with the particulate burden in the same plume. In regions where no PM2.5 ground station exists, which covers most of sub-Saharan Africa, large parts of South and Southeast Asia and almost all small island states, this proxy is often the only available evidence of an acute combustion event.
The emission-ratio method: how the conversion works
The core technique compares column enhancements of CO and CH4 against a background reference, typically a clean-air pixel upwind of the source, and computes the ratio delta-CO / delta-CH4 or delta-CO / delta-CO2 (the latter requiring a CO2 column product such as OCO-2 or TROPOMI's experimental retrieval). Published emission ratios for specific fuel types, flaring versus smouldering biomass versus coal combustion, allow the analyst to attribute the source class and estimate total carbon flux. Andreae and Merlet (2001) and subsequent updates remain the standard reference for fuel-specific emission factors.
Once combustion intensity is estimated, a forward trajectory model such as HYSPLIT, driven by ERA5 or GFS reanalysis winds, propagates the plume downwind. Population grids (GPWv4 or WorldPop at 100 m) intersect the trajectory corridor to produce an exposed-population count. The result is a ranked list of settlement clusters by estimated inhalation burden, not a precise dose, but a defensible triage tool for public health prioritisation.
VIIRS FRP anchors the spatial and temporal location of the source. TROPOMI provides the column chemistry. The meteorological model provides transport. Each layer carries its own uncertainty, and the combined error in estimated PM exposure is substantial, typically a factor of two to three relative to co-located ground measurements where comparisons have been published. That is an honest number, and it matters.
What the column cannot tell you: the vertical ambiguity problem
TROPOMI CO is a total tropospheric column measurement. It integrates everything from the surface to the tropopause. A strong CO enhancement over a city could reflect surface-level combustion affecting the people below, or it could reflect a residual layer from a distant fire that has been lofted to 3 km altitude by convection and poses little immediate inhalation risk. MOPITT's multi-level retrieval partially resolves this, but its 22 km footprint smears point sources considerably.
Separating surface-relevant from aloft-irrelevant CO requires either a boundary-layer height product (available from ERA5 or MERRA-2 reanalysis, with their own errors) or coincident lidar data from CALIPSO or the CATS instrument, neither of which provides daily global coverage. The practical rule: treat column-derived exposure estimates as upper bounds on surface-layer burden, and flag cases where the planetary boundary layer height is shallow (below roughly 500 m) as higher-confidence surface-exposure events. Nighttime retrievals and winter inversions in mid-latitude cities are the cases where the column-to-surface translation is most credible.
Flaring versus open burning: the same chemistry, different regulatory contexts
Industrial gas flaring and agricultural open burning both produce CO and CH4 column enhancements, but they sit in entirely different regulatory and health frameworks. Flaring is a point source, typically at a known facility, and VIIRS Nightfire can detect individual flares down to roughly 1 MW radiant heat output. The World Bank's Global Gas Flaring Reduction Partnership publishes annual flaring volumes by country derived from VIIRS, providing a public benchmark against which facility-level TROPOMI anomalies can be compared.
Open burning is diffuse, seasonal and often legally ambiguous. Crop-residue burning in the Indo-Gangetic Plain, for example, drives acute PM2.5 episodes in Delhi that exceed WHO guidelines by an order of magnitude, and the TROPOMI CO signal during October and November is visually striking even at 5.5 km resolution. The emission-ratio approach is less precise for diffuse sources because the background reference is harder to define cleanly, but the population-exposure corridor calculation remains valid as a relative risk ranking across districts or provinces.
Building an operational alert: what a working system looks like
A practical combustion-health monitoring system has four components: daily TROPOMI CO and CH4 ingest, VIIRS FRP ingest for source attribution, a 48-hour forward trajectory ensemble, and a population-weighted exposure index aggregated to administrative units. Thresholds for alerting are necessarily empirical. A reasonable starting point is a CO column anomaly exceeding two standard deviations above the 5-year seasonal baseline for that grid cell, coincident with a VIIRS fire cluster, triggering a trajectory run and an exposure estimate.
Satellize runs this class of analysis on open constellation data and can configure alert thresholds to match a client government's existing public health warning tiers. The Tonga crop-estimation programme demonstrated that operational analytics pipelines built on open satellite data can be maintained affordably by a small sovereign team once the methodology is established. The same principle applies here: the satellite data are free, the value is in the calibrated interpretation and the connection to decision-ready outputs.
Latency is manageable. TROPOMI offline CO products are available within 5 days; the near-real-time product within 3 hours, though with slightly higher retrieval uncertainty. For acute episode response, the near-real-time product is adequate for trajectory modelling and population alerting.
Cloud cover and the gaps you must plan for
TROPOMI CO retrievals use thermal-infrared and near-infrared channels. Cloud cover degrades the near-infrared retrieval significantly; the thermal-infrared channel retains partial sensitivity above cloud but loses surface-layer information entirely. In persistently cloudy regions, such as the Congo Basin during the wet season or maritime Southeast Asia during monsoon, data gaps of 5 to 10 consecutive days are common. A monitoring system that depends on TROPOMI alone will miss events during precisely the periods when convective burning and post-storm agricultural burning often peak.
The mitigation is ensemble thinking: MODIS and VIIRS fire counts are less affected by thin cloud than passive spectrometers, and a fire detection without a corresponding CO column retrieval is still actionable as a combustion alert. Combining FRP-based source detection with climatological emission factors provides a degraded but usable exposure estimate when column chemistry is unavailable. Honest system design documents the gap rate explicitly and sets user expectations accordingly.
Typical figures
| TROPOMI CO spatial resolution | 5.5 × 7 km (post-2019 reprocessing); original 7 × 7 km for pre-2019 archive |
| TROPOMI CH4 spatial resolution | 5.5 × 7 km; retrieval requires cloud fraction below ~0.5 |
| TROPOMI global revisit | Daily (single overpass, ~13:30 local solar time) |
| TROPOMI CO product latency | Near-real-time: ~3 hours; offline (higher quality): ~5 days |
| MOPITT CO resolution and revisit | 22 × 22 km; ~3-day global revisit; vertical profiles at up to 7 pressure levels |
| VIIRS FRP resolution | 375 m active fire product; minimum detectable fire ~1 MW radiant power under clear sky |
| Minimum detectable CO column enhancement (TROPOMI) | Approximately 5–10 ppb above background in cloud-free conditions |
| Archive depth | TROPOMI: May 2018 to present; MOPITT: March 2000 to present; VIIRS: 2012 to present; MODIS fire: 2000 to present |
| Trajectory model inputs | ERA5 reanalysis (0.25° hourly) or GFS forecast (0.25°, 6-hourly); HYSPLIT or equivalent forward-trajectory framework |
| Population grid resolution | GPWv4 at ~1 km; WorldPop at 100 m for supported countries |
Analytics Satellize can run
| Daily CO column anomaly map | Z-score against 5-year seasonal baseline per TROPOMI grid cell; source attribution via coincident VIIRS FRP clusters | GeoTIFF and GeoJSON layer, daily, delivered to client GIS or API endpoint |
| Combustion-intensity estimate by source | Emission-ratio approach using delta-CO / delta-CH4 column enhancements against upwind background; fuel-type classification from VIIRS spectral fire temperature and land-cover context | Tabular report per identified source cluster, with confidence interval on estimated carbon flux |
| 48-hour downwind population-exposure corridor | HYSPLIT forward trajectory ensemble (multiple meteorological ensemble members) intersected with WorldPop or GPWv4 population grid; exposure index weighted by column anomaly magnitude | Ranked list of administrative units by estimated exposed population, with trajectory uncertainty envelope shown as polygon layer |
| Acute combustion episode alert | Threshold trigger: CO anomaly exceeding 2 standard deviations above seasonal baseline, coincident VIIRS fire cluster, boundary-layer height below 800 m from ERA5 | Push alert (email or webhook) with episode summary, source location, estimated exposed population within 200 km downwind |
| Multi-year combustion climatology | Annual and seasonal aggregation of TROPOMI CO columns and MODIS/VIIRS FRP over user-defined region; trend analysis against population growth and land-use change | Annual PDF report with time-series charts and anomaly maps; underlying data in NetCDF |
| Flare-specific methane and CO emission estimate | VIIRS Nightfire point-source detection combined with TROPOMI CH4 and CO column excess; comparison against World Bank GGFR published flaring volumes for validation | Facility-level emission table with monthly time series; flagging of facilities showing anomalous CH4/CO ratios relative to expected flaring efficiency |
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.