Carbon monoxide plume tracking from industrial fires and accidents
Satellite retrievals of tropospheric carbon monoxide can trace pollution plumes from industrial fires and accidents across national borders within 24–48 hours. This page explains which sensors do it, what they miss, and what an analyst can honestly deliver.
Sensors
- Sentinel-5P TROPOMI: Retrieves tropospheric CO total columns in the near-infrared (2.3 µm SWIR band) at 7 × 5 km nadir resolution (reprocessed to 5.5 × 3.5 km from August 2019 onward), daily global coverage, with operational Level-2 products typically available within 3 hours of overpass via Copernicus Data Space.
- MOPITT (Terra): Multispectral thermal infrared and near-infrared CO retrievals at 22 km nadir resolution, operating since 2000. Provides vertical profile information (surface to upper troposphere) via simultaneous TIR/NIR channels, though sensitivity to the boundary layer is limited under stable atmospheric conditions. Revisit roughly every 3 days at mid-latitudes.
- IASI (MetOp-A/B/C): Thermal infrared sounder retrieving CO profiles at approximately 12 km nadir resolution, twice-daily global coverage per satellite (morning and evening equatorial crossings). Most sensitive to the mid-troposphere (around 3–6 km altitude); near-surface CO in stable nocturnal boundary layers is routinely underestimated by 20–40% relative to in-situ measurements in published validation studies.
- Aura MLS: Microwave Limb Sounder measuring CO in the upper troposphere and stratosphere (roughly 215 hPa and above). Not useful for boundary-layer or lower-tropospheric plumes from surface industrial events, but valuable for tracking long-range transport of elevated plumes that have been lofted by pyroconvection above 8–10 km.
What a CO column actually measures, and what it does not
Carbon monoxide absorbs solar backscatter in the shortwave infrared near 2.3 µm and emits thermally in the mid-infrared near 4.7 µm. TROPOMI exploits the SWIR path, which gives it sensitivity down to roughly 1–2 km altitude in clear sky. IASI and MOPITT TIR channels are most sensitive to the 3–6 km layer, where the thermal contrast between the gas and the surface is greatest. That distinction matters enormously after an industrial accident.
When a refinery fire or chemical plant explosion injects CO into a well-mixed or convective atmosphere, TROPOMI's column retrieval captures a large fraction of the enhancement. When the same release happens on a calm, cold night into a shallow stable boundary layer, the CO stays below 500 m, and IASI may underestimate the surface concentration by a factor of two or more. No satellite sensor currently provides reliable near-surface CO concentrations in stable nocturnal conditions without independent profile constraints from radiosondes or chemical transport models. That is an honest limit, not a caveat to be buried.
TROPOMI as the operational workhorse for accident response
TROPOMI's daily global swath of approximately 2,600 km means that any point on Earth is observed at least once per day under cloud-free conditions. After the 2019 Philadelphia Energy Solutions refinery fire and the 2020 Beirut port explosion, researchers demonstrated that TROPOMI CO columns showed statistically significant enhancements within one overpass of the event, with plumes traceable downwind for 48–72 hours across multiple overpasses. The 5.5 × 3.5 km pixel footprint (post-August 2019 reprocessing) is fine enough to separate adjacent industrial point sources in most petrochemical clusters.
Cloud cover is the principal operational constraint. TROPOMI's CO retrieval requires cloud radiance fraction below roughly 0.5 for usable retrievals; in tropical or frontal weather systems, consecutive days of cloud can create gaps of 48–96 hours. Analysts should always flag cloud-contaminated pixels explicitly rather than interpolating across them. Combining TROPOMI with IASI (which has some cloud-penetrating capability in TIR) partially mitigates this, though IASI's coarser 12 km footprint blurs the plume boundary.
Reading the vertical structure: when profiles matter
A column enhancement tells you that CO is present somewhere between the surface and the tropopause. It does not tell you at what altitude the mass is concentrated. For emergency response, altitude matters: CO lofted above the planetary boundary layer by pyroconvection is transported faster, further, and at different wind directions than surface-hugging plumes. MOPITT's simultaneous TIR and NIR retrieval produces a partial vertical profile with roughly 2–3 degrees of freedom for signal in the troposphere, enough to distinguish whether a plume is predominantly in the lower or middle troposphere.
Aura MLS becomes relevant only when a fire is intense enough to drive pyroconvective injection into the upper troposphere or lower stratosphere, which occurs with large boreal or tropical megafires rather than typical refinery incidents. For most industrial accident scenarios, MOPITT profiles combined with TROPOMI columns and a chemical transport model (CAMS, GEOS-Chem) provide the most complete picture of plume altitude and trajectory.
Cross-border attribution and the 24–48 hour window
Mid-tropospheric winds routinely move air masses at 20–60 km per hour. A large industrial fire releasing tens of thousands of tonnes of CO can produce a detectable column enhancement 500–2,000 km downwind within two days. TROPOMI's near-real-time Level-2 product, available within 3 hours of overpass through Copernicus Data Space, enables analysts to begin downwind tracking on the same calendar day as the event.
Attribution, however, is harder than detection. A CO enhancement over a downwind country is consistent with the accident but not proof of it: background CO from other combustion sources, biomass burning, and traffic must be separated using trajectory models (HYSPLIT, FLEXPART) and isotopic or co-emitted tracer ratios where available. Regulators and legal teams should understand that satellite CO columns provide strong circumstantial evidence of transboundary transport, not a court-ready mass-balance attribution without additional modelling and ground-truth.
Practical limits an analyst must state upfront
Minimum detectable CO column enhancement for TROPOMI is approximately 10–15 ppbv in the tropospheric column under favourable conditions, corresponding to a total column enhancement of roughly 1–2 × 10¹⁷ molecules cm⁻². Smaller releases, or releases into deep stable boundary layers, may fall below this threshold. The sensor is not designed for point-source flux quantification at the precision that methane sensors like GHGSat achieve; it is a plume-mapping and transport-tracking tool.
Archive depth for TROPOMI extends to May 2018, which provides a meaningful baseline for comparing background CO over an industrial region against post-accident columns. MOPITT's archive runs from March 2000, making it the longest continuous satellite CO record available. Neither archive is a substitute for ground-level air-quality monitoring networks, which remain the authoritative source for human-health exposure assessment near the accident site.
Satellize runs TROPOMI and IASI CO products operationally for clients requiring near-real-time plume alerts, combining automated column-anomaly detection with HYSPLIT back-trajectory attribution. The Overhead column has covered several published industrial-plume case studies using this workflow.
Typical figures
| TROPOMI CO spatial resolution | 5.5 × 3.5 km (from August 2019 reprocessing); 7 × 5 km prior |
| TROPOMI revisit | Daily global coverage; swath ~2,600 km |
| TROPOMI Level-2 latency | ~3 hours after overpass (near-real-time product via Copernicus Data Space) |
| IASI CO resolution / revisit | ~12 km nadir footprint; twice-daily global coverage per MetOp satellite |
| MOPITT CO resolution / revisit | 22 km nadir; ~3-day repeat at mid-latitudes; archive from March 2000 |
| Spectral bands used | SWIR ~2.3 µm (TROPOMI NIR-CO); TIR ~4.7 µm (IASI, MOPITT); microwave limb (Aura MLS, upper troposphere only) |
| Minimum detectable column enhancement (TROPOMI) | ~10–15 ppbv tropospheric column under clear-sky conditions; ~1–2 × 10¹⁷ molecules cm⁻² |
| Vertical sensitivity | TROPOMI SWIR: surface to ~2 km in clear sky. IASI/MOPITT TIR: peak sensitivity 3–6 km; underestimates stable boundary-layer CO by 20–40% |
| Cloud constraint | TROPOMI retrieval requires cloud radiance fraction <0.5; heavy cloud can create 48–96 hour data gaps |
| Archive depth | TROPOMI: May 2018 to present. MOPITT: March 2000 to present. IASI: 2007 to present |
Analytics Satellize can run
| Plume onset alert | Automated column-anomaly detection against 30-day rolling background (TROPOMI Level-2 CO product) | Email or API alert within 6 hours of first detectable overpass, with affected pixel map as GeoTIFF |
| Downwind transport trajectory | HYSPLIT or FLEXPART forward-trajectory modelling driven by NOAA GFS or ECMWF ERA5 wind fields, initialised at source location and time | 48–72 hour ensemble trajectory GIS layer (GeoJSON or Shapefile) with uncertainty envelope |
| Multi-day plume animation | Time-series compositing of consecutive TROPOMI overpasses, gap-filled with IASI where cloud fraction exceeds threshold | MP4 animation and frame-by-frame PNG sequence for regulatory or public communications use |
| Column enhancement time series at receptor locations | Pixel extraction and area-averaging over user-defined receptor polygons (downwind cities, protected areas, national borders) | CSV time series with overpass timestamps, mean column enhancement, cloud-flag status, and detection-limit annotations |
| Vertical profile characterisation | MOPITT TIR/NIR joint retrieval analysis to estimate plume altitude distribution; comparison with CAMS reanalysis CO profiles | PDF technical report with altitude-resolved concentration estimates and stated uncertainty ranges |
| Transboundary exposure summary | Integration of column enhancement maps with population gridded data (GPWv4) to estimate exposed population counts by country | Tabular summary by administrative unit, suitable for regulatory notification or insurance loss assessment; stated caveats on surface-concentration inference |
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.