Carbon monoxide plume tracking from industrial fires and accidents
Tropospheric CO columns retrieved from TROPOMI, IASI and MOPITT let analysts trace pollution plumes from industrial fires and chemical accidents across hundreds of kilometres within 24 to 48 hours. This page explains the physics, the sensors, and the honest limits of what orbit can and cannot see.
Sensors
- Sentinel-5P TROPOMI: CO total column at 7 × 5 km nadir pixel (improved from the original 7 × 3.5 km after 2019 reprocessing), daily global coverage, near-real-time product available within 3 hours of overpass. The primary workhorse for event-driven plume tracking.
- MOPITT (Terra): Thermal infrared and near-infrared CO profiles at 22 km spatial resolution, operating since 2000. Provides the longest continuous CO record from orbit, useful for pre-incident baseline and multi-day plume evolution, though coarse resolution limits point-source attribution.
- IASI (MetOp-A/B/C): Thermal infrared sounder retrieving CO profiles at approximately 12 km footprint, twice-daily global coverage per satellite. Sensitive to the mid-troposphere (roughly 3 to 8 km altitude); three MetOp satellites in orbit give four daily overpasses combined, improving temporal sampling during fast-moving events.
- Aura MLS: Microwave limb sounder measuring CO in the upper troposphere and stratosphere, not the boundary layer. Relevant only when pyroconvective events from very large fires inject CO above 8 km; not useful for typical ground-level industrial incidents.
What the sensors actually measure, and why that matters for enforcement
Carbon monoxide absorbs solar radiation near 2.3 micrometres and emits thermally near 4.7 micrometres. TROPOMI exploits the near-infrared window; IASI and MOPITT use both channels to retrieve vertical profiles. What you get from any of these instruments is a column-averaged mixing ratio, not a ground-level concentration. The distinction is consequential. A column measurement integrates CO through the full atmospheric depth, weighting each layer by how much sunlight or thermal emission passes through it.
For thermal infrared retrievals, sensitivity peaks in the mid-troposphere, typically between 3 and 8 km altitude. In a stable nocturnal boundary layer, where cold dense air traps pollutants near the surface, the satellite may see a modest column enhancement while a worker at fence-line level breathes air well above safe thresholds. TROPOMI's near-infrared channel has somewhat better surface sensitivity than thermal IR in cloud-free conditions, but it too loses skill when a dense smoke plume or low cloud deck sits between the sensor and the ground. This is not a flaw to be engineered away; it is a physical constraint that any honest analysis must carry forward into risk assessments.
How fast a plume travels, and what 24 to 48 hours means operationally
Mid-latitude westerly winds routinely move air masses at 20 to 60 km per hour at the altitudes where CO accumulates after a large fire. A refinery incident in one country can deposit a detectable column enhancement across a national border within a day. Published studies using TROPOMI data have traced CO plumes from large industrial fires across distances exceeding 1,000 km while maintaining column enhancements distinguishable above background noise, which TROPOMI's product documentation places at roughly 10 parts per billion by volume for individual pixels under favourable retrieval conditions.
The operational implication is that a single TROPOMI overpass, arriving in near-real-time within three hours of acquisition, can establish plume geometry, transport direction and approximate emission magnitude before a regulatory authority has finished drafting its first press release. Successive daily overpasses allow trajectory reconstruction. When cloud cover interrupts TROPOMI, IASI's twice-daily cadence per satellite, combined across three MetOp platforms, can partially fill the gap, though at coarser resolution.
Turning a column map into an emission estimate
The standard published approach is a mass-balance or cross-sectional flux method: multiply the measured column enhancement (in molecules per square centimetre) across a transect perpendicular to the plume by the wind speed at plume height, integrate across the transect width, and you have an approximate emission flux in grams per second or tonnes per hour. Wind data come from meteorological reanalysis products such as ERA5 or operational NWP output. The method is sensitive to wind speed uncertainty, which typically contributes 20 to 40 percent error in published estimates, and to the assumed plume height, which affects both the wind speed chosen and the column-to-surface-concentration conversion.
A complementary approach is atmospheric inversion, fitting a chemical transport model to observed column fields to back-calculate surface fluxes. This is more computationally demanding and introduces model uncertainty, but it can separate overlapping sources and extend attribution over multi-day events. Neither method replaces ground-based monitoring; both methods provide independent, spatially continuous evidence that ground networks cannot.
Cloud cover, retrieval flags, and the honest detection floor
TROPOMI's CO product is delivered with a quality assurance flag that rejects pixels where cloud radiance fraction exceeds roughly 0.5. In practice, this means that in persistently overcast regions, or during the first days of a fire when dense smoke itself degrades retrieval quality, data gaps are common. IASI retrievals degrade similarly under thick cloud. There is no satellite-based workaround that restores CO columns through opaque cloud; the physics simply do not permit it.
The minimum detectable column enhancement for TROPOMI in a single pixel is approximately 10 ppbv against a background of roughly 80 to 100 ppbv over continental regions. Large industrial fires and refinery incidents routinely produce enhancements of 50 to several hundred ppbv at the source, making them detectable even at 7 × 5 km pixel size. Smaller accidental releases, or those that disperse rapidly in high-wind conditions, may fall below the single-pixel detection threshold but can sometimes be recovered by spatial averaging at the cost of spatial precision.
Building the evidentiary record for cross-border incidents
Regulatory bodies and legal teams handling transboundary pollution cases need more than a single overpass screenshot. A credible evidentiary package includes: the time series of column maps from first detection through plume dissipation; a documented provenance chain showing which version of the TROPOMI L2 product was used and what quality flags were applied; wind-trajectory analysis linking the observed plume to the source facility; and an explicit uncertainty statement covering retrieval error, wind uncertainty, and the boundary-layer sensitivity caveat described above.
Satellize can structure this kind of package for government clients. The methodology rests entirely on open, peer-reviewed retrieval algorithms and publicly archived satellite data, which means the analysis is reproducible by any competent third party. That reproducibility is what makes it defensible in an administrative or legal proceeding. For context on the kind of analytics Satellize can run on open constellations, the Kingdom of Tonga crop-estimation programme illustrates how the same underlying approach, applying published retrieval methods to open satellite data and delivering structured outputs to a government client, transfers across very different use cases.
Archive depth and what it enables retrospectively
TROPOMI has been operational since October 2017, with a reprocessed Level-2 CO dataset available from the Copernicus Data Space. MOPITT extends the record to March 2000. IASI data from MetOp-A begin in 2007. This archive depth matters for two reasons. First, it allows pre-incident baseline characterisation: what was the typical CO column over this facility and its downwind region before the accident? Second, it supports retrospective investigation of incidents that were not monitored in real time, which is common when a release occurs in a jurisdiction with limited ground infrastructure.
Latency for operational use is short. TROPOMI's near-real-time product is typically available within three hours of overpass; the offline product, with improved calibration, follows within a few days. IASI operational products are available within hours through EUMETSAT's data services. For emergency response, the near-real-time products are almost always the right choice.
Typical figures
| Spatial resolution (TROPOMI CO) | 7 × 5 km per pixel (nadir) |
| Spatial resolution (IASI CO) | ~12 km footprint diameter |
| Spatial resolution (MOPITT CO) | 22 km footprint |
| Revisit (TROPOMI) | Daily global coverage; near-real-time product within ~3 hours of overpass |
| Revisit (IASI, three MetOp satellites combined) | Up to 4 overpasses per day globally |
| Spectral channels used | Near-infrared (~2.3 µm) and thermal infrared (~4.7 µm) depending on instrument |
| Minimum detectable enhancement (TROPOMI, single pixel) | ~10 ppbv above local background (~80–100 ppbv continental) |
| Vertical sensitivity | Mid-troposphere peak (3–8 km); reduced near-surface sensitivity in stable boundary layers |
| Archive depth | TROPOMI from Oct 2017; MOPITT from Mar 2000; IASI from 2007 |
| Primary delivery formats | NetCDF L2/L3, GeoTIFF column maps, GIS-ready shapefiles, structured JSON event reports |
Analytics Satellize can run
| Plume detection alert | Threshold exceedance on TROPOMI L2 CO quality-flagged column fields, cross-checked against facility location registry | Automated alert with overpass timestamp, peak column value, and plume centroid coordinates; delivered within 4 hours of TROPOMI near-real-time product availability |
| Plume trajectory and extent map | Spatial clustering of enhanced pixels combined with ERA5 or operational NWP wind-field overlay to reconstruct transport path | GeoTIFF and GIS polygon layer showing plume footprint per overpass, with wind vectors and estimated downwind reach |
| Emission flux estimate | Cross-sectional mass-balance method applied to column transects perpendicular to plume axis, with explicit wind-speed uncertainty propagation | Tabular report giving flux in tonnes CO per hour, with stated uncertainty range (typically ±20–40%) and sensitivity assumptions |
| Multi-day plume time series | Sequential L2 product ingestion across TROPOMI and IASI overpasses, gap-flagged where cloud fraction exceeds retrieval threshold | Animated column-map series and summary CSV of peak enhancement, plume area and centroid per overpass for the incident period |
| Pre-incident baseline characterisation | Statistical summary of archived TROPOMI CO columns over facility and downwind region for the 12 months preceding the event | PDF report with percentile envelopes and anomaly magnitude relative to historical distribution, suitable for regulatory submission |
| Transboundary transport evidence package | Combined plume mapping, HYSPLIT or equivalent back-trajectory analysis, and documented retrieval provenance chain | Structured evidentiary dossier including data provenance, uncertainty statement, and reproducibility notes for administrative or legal use |
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.