Carbon monoxide column tracking for biomass-burning plume transport
Carbon monoxide is the most useful long-range tracer of incomplete combustion, surviving long enough to cross ocean basins. Thermal-infrared sounders and near-infrared spectrometers each see different altitude slices, and combining them gives the most complete picture of plume transport.
Sensors
- MOPITT / Terra: Multispectral thermal-infrared and near-infrared CO sounder. Nadir pixel footprint 22 km. Retrieves CO vertical profiles in roughly 3-km-thick layers from the surface to ~25 km, though sensitivity peaks in the mid-troposphere (500–300 hPa). Near-daily global coverage. Archive from March 2000, making it the longest continuous satellite CO record.
- IASI / MetOp-A, -B, -C: Thermal-infrared Fourier-transform sounder. Nadir footprint 12 km. Retrieves total CO column and partial columns; sensitivity profile similar to MOPITT, peaking in the mid-troposphere. Three satellites in the series provide twice-daily global coverage per satellite. Spectral range 645–2760 cm⁻¹ at 0.5 cm⁻¹ apodised resolution.
- TROPOMI / Sentinel-5P: Near-infrared shortwave CO band centred near 2.3 µm. Ground pixel 5.5 × 7 km (upgraded from original 7 × 7 km in 2019). Near-daily global coverage. Provides better sensitivity to the lower troposphere and boundary layer than thermal-infrared sounders under clear-sky conditions, because the retrieval uses reflected solar radiation rather than emitted thermal radiance. Minimum detectable column enhancement roughly 5–10 ppbv in favourable conditions.
- CrIS / Suomi NPP and NOAA-20: Cross-track Infrared Sounder. Nadir footprint 14 km. Retrieves CO profiles in the thermal infrared using similar physics to IASI. Suomi NPP and NOAA-20 together provide twice-daily global coverage. Useful for cross-validation with IASI and for extending the record into the JPSS era.
Why CO and not the fire itself
Fire radiative power products from geostationary imagers capture the active combustion event with sub-hourly temporal resolution. That is the right tool for the fire. Carbon monoxide is the right tool for everything that happens afterwards. Its atmospheric lifetime is roughly two to three weeks in the troposphere, long enough for a plume originating in the African savannah to reach South America, or for Siberian fire smoke to cross the Pacific and appear in North American surface-air-quality records.
CO is produced by incomplete combustion of biomass, meaning smouldering fires and agricultural burns generate it in large quantities even when visible smoke has dispersed. That makes it a tracer of emission history rather than just of active burning. The ratio of CO to other species, particularly aerosols and nitrogen oxides, also carries information about combustion efficiency and fire type, which matters when trying to attribute air-quality episodes to specific source regions.
What thermal-infrared sounders actually see, and what they miss
MOPITT and IASI both retrieve CO by measuring the thermal emission from the atmosphere in spectral bands where CO absorbs, primarily around 4.7 µm. The retrieval is an inverse problem: given the measured radiance spectrum and a prior estimate of the CO profile, find the profile that best fits the observations. The sensitivity of this approach is not uniform with altitude. It peaks in the mid-troposphere, roughly 500 to 300 hPa, and falls off sharply towards the surface. A dense low-level plume sitting below 850 hPa may be substantially underestimated.
Thermal contrast is the physical reason. The retrieval depends on a temperature difference between the CO layer and the surface below it. At night, or over surfaces with low thermal emission, that contrast collapses and the surface sensitivity degrades further. Over cold, high-latitude surfaces in winter, MOPITT and IASI are effectively blind to the lowest kilometre of the atmosphere. This is not a sensor deficiency; it is a consequence of the underlying physics, and any honest transport analysis must account for it.
Where TROPOMI fills the gap, and where it does not
TROPOMI's CO retrieval uses reflected sunlight in the 2.3 µm shortwave infrared band. Because the signal originates at the surface and passes through the full atmospheric column, the retrieval is sensitive to CO throughout the troposphere, including the boundary layer. This makes TROPOMI substantially better than thermal-infrared sounders at detecting fresh, low-altitude plumes close to the source region, and at quantifying surface-layer enhancements that matter most for air-quality applications.
The trade-off is cloud cover. TROPOMI requires clear-sky or thin-cloud conditions for a valid retrieval; cloud fractions above roughly 0.3 typically cause the pixel to be flagged. Over the tropics, where biomass burning is most intense during dry season, convective cloud can mask large fractions of the domain on any given day. Combining TROPOMI with IASI or MOPITT, which have some sensitivity through thin cloud in the thermal infrared, partially compensates, but there is no complete solution. Persistent cloud cover over a burning region will degrade any satellite CO estimate.
Constructing a transport analysis from multiple sensors
A plume transport analysis typically begins with fire emission inventories, such as the Global Fire Emissions Database or FINN, which use fire radiative power observations to estimate CO emission rates. Those emissions are then injected into an atmospheric transport model, most commonly HYSPLIT, FLEXPART, or a global chemistry-transport model such as GEOS-Chem, at an assumed injection height. The simulated CO field is compared against satellite retrievals to assess whether the model captures the observed plume location, shape, and column amount.
The comparison is not straightforward. Satellite retrievals report column amounts convolved with the instrument's averaging kernel, a vertical weighting function that encodes the altitude-dependent sensitivity described above. A fair model-to-satellite comparison requires applying the averaging kernel to the model profile before comparing to the retrieved column. Skipping this step can produce apparent discrepancies of 20 to 40 percent even when the model is performing well. Published validation studies for MOPITT Version 9 and TROPOMI CO report retrieval uncertainties of roughly 10 to 15 percent in the free troposphere under favourable conditions, rising substantially near the surface.
Assimilation of satellite CO retrievals into transport models, as done operationally by the Copernicus Atmosphere Monitoring Service, produces reanalysis fields that are more accurate than either the model or the satellite alone. These fields are the appropriate starting point for attribution of a specific air-quality episode to a distant fire source.
Practical limits a buyer should know before commissioning an analysis
Spatial resolution is a persistent constraint. MOPITT's 22 km footprint and IASI's 12 km footprint are adequate for basin-scale transport but cannot resolve individual fire clusters or urban corridors. TROPOMI's 5.5 km pixel is better, but still coarser than the scales at which surface receptors experience the highest concentrations. For a government seeking to attribute a specific pollution episode to a specific country's agricultural burning, the satellite column data provides strong evidence of transport pathway but cannot alone establish ground-level concentration at a receptor point without auxiliary dispersion modelling.
Revisit is once per day per sensor in the best case, which means a fast-moving plume may be sampled at very different stages by different sensors. Combining IASI's morning and afternoon overpasses from three MetOp satellites with TROPOMI and CrIS improves temporal sampling but does not eliminate the gap. Geostationary CO sounders do not yet exist operationally, though the planned MTG-IRS instrument is expected to change this later in the decade.
Satellize runs CO transport analyses on open TROPOMI and IASI data streams, combining retrieval products with trajectory modelling to produce attribution reports for government clients. The methodology is the same class used in the peer-reviewed literature, applied operationally.
Typical figures
| Spatial resolution (MOPITT) | 22 km nadir footprint |
| Spatial resolution (IASI) | 12 km nadir footprint |
| Spatial resolution (TROPOMI CO) | 5.5 × 7 km (post-August 2019) |
| Revisit (combined fleet) | 1–2 times per day per sensor; 3–4 overpasses per day possible with IASI triplet plus TROPOMI |
| Spectral bands | Thermal IR ~4.7 µm (MOPITT, IASI, CrIS); near-IR ~2.3 µm (MOPITT, TROPOMI) |
| Altitude sensitivity peak (TIR) | Mid-troposphere, approximately 500–300 hPa; low surface sensitivity |
| Minimum detectable column enhancement (TROPOMI) | Approximately 5–10 ppbv under clear-sky, low-aerosol conditions |
| Retrieval uncertainty (free troposphere) | Approximately 10–15 percent under favourable conditions; higher near surface |
| Archive depth | MOPITT from March 2000; IASI from 2007; TROPOMI from May 2018 |
| Cloud sensitivity (TROPOMI NIR) | Pixels flagged at cloud fraction above approximately 0.3 |
Analytics Satellize can run
| Plume origin attribution report | Backward trajectory analysis (HYSPLIT or FLEXPART) initialised at receptor location, compared against TROPOMI and IASI column maps | PDF report with trajectory ensemble maps, confidence bounds, and contributing fire-region breakdown |
| Daily CO column anomaly map | TROPOMI Level-2 CO product bias-corrected against MOPITT climatology; anomaly computed relative to 5-year baseline | GeoTIFF layer at 5.5 km resolution, updated daily, delivered via SFTP or cloud bucket |
| Plume transport forecast overlay | CAMS reanalysis CO field combined with 72-hour forward trajectory ensemble | Animated GIF and GeoJSON polygon set showing forecast plume corridor with uncertainty envelope |
| Emission flux estimate for a burning episode | Top-down inversion using TROPOMI CO columns and a prior from GFED fire emission database | Tabular emission estimate (Tg CO) with uncertainty range, disaggregated by fire cluster and day |
| Multi-year fire-season CO trend summary | MOPITT Version 9 Level-3 monthly means processed over user-defined region; Mann-Kendall trend test applied | Time-series chart and CSV with trend magnitude, p-value, and seasonal decomposition |
| Cross-sensor consistency check | Collocated TROPOMI and IASI retrievals compared using averaging-kernel correction; bias and standard deviation computed per season | Validation table and scatter plots, flagging periods or regions where sensor disagreement exceeds 20 percent |
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.