Black carbon and soot plume mapping from industrial combustion sources
Black carbon from coking plants, steel mills and open industrial fires carries a distinct absorbing-aerosol signature that satellite retrievals can separate from scattering dust, turning a physics difference into an attribution tool for regulators.
Sensors
- Sentinel-5P TROPOMI: Provides the Absorbing Aerosol Index (AAI) at 3.5 × 5.5 km nadir resolution (upgraded from 7 × 3.5 km in August 2019) with daily global coverage. AAI distinguishes UV-absorbing aerosols (black carbon, mineral dust, smoke) from non-absorbing species using the 340/380 nm band pair. Cannot independently separate BC from dust without ancillary data.
- POLDER/PARASOL (CNES): Multi-angle polarimetric imager that flew on PARASOL (2004–2013). Acquired radiance and polarisation at up to 16 viewing angles across nine spectral bands (443–1020 nm) at roughly 6 km resolution. Polarimetric retrievals constrain particle size, shape and refractive index, enabling separation of absorbing carbonaceous aerosols from non-spherical mineral dust. Archive only; no live feed.
- MISR (Terra): Nine fixed cameras spanning ±70.5° from nadir, 275 m to 1.1 km resolution depending on band and mode, roughly 9-day repeat at mid-latitudes. Multi-angle stereo gives plume injection height independently of radiative assumptions, which is critical for dispersion modelling and source attribution.
- MODIS (Terra and Aqua): Aerosol Optical Depth products (MOD04/MYD04) at 3 km and 10 km resolution, twice-daily combined. Deep Blue algorithm extends retrievals over bright surfaces including industrial areas. Useful for contextual AOD and fire radiative power co-location, but spatial resolution limits single-facility attribution.
What absorbing aerosol index actually measures
The Absorbing Aerosol Index is a ratio of observed to modelled UV reflectance at two wavelengths where a pure Rayleigh atmosphere would produce a predictable signal. When UV-absorbing particles are present above the boundary layer, measured reflectance departs from the Rayleigh baseline in a characteristic direction. A positive AAI indicates absorption; a negative value indicates scattering aerosols such as sea salt or sulphate. Black carbon from combustion produces strongly positive AAI values, typically above 1.0 for dense industrial plumes, though the index is dimensionless and non-linear.
The complication is that mineral dust also absorbs in the UV, producing positive AAI that overlaps with the black carbon signal. Iron oxides in dust are the main culprit. Over the Sahel, Arabian Peninsula or central Asian industrial zones, a coking plant plume and a regional dust event can produce indistinguishable AAI values on a single overpass. Resolving this ambiguity requires either polarimetric data (which constrains particle shape, dust being non-spherical) or temporal analysis that separates episodic industrial emission from background dust climatology.
Multi-angle polarimetry: the shape test for soot
POLDER on PARASOL was the only operational satellite instrument designed specifically to exploit polarisation across multiple viewing angles simultaneously. Soot aggregates and carbonaceous particles are broadly spherical at the sizes relevant to combustion plumes (sub-micron accumulation mode), whereas mineral dust particles are angular and non-spherical. Non-spherical particles depolarise light differently at oblique angles. POLDER's polarised reflectance at 865 nm across its angular range encodes this shape information, and inversion algorithms such as GRASP (Generalised Retrieval of Aerosol and Surface Properties) can retrieve the fraction of absorbing spherical particles separately from non-spherical dust.
The practical limit is sensitivity. POLDER retrievals work best when aerosol optical depth exceeds roughly 0.2 at 550 nm. Below that threshold, the polarimetric signal is too weak relative to surface contributions to constrain the absorption reliably. Coking plant and steel mill plumes near the source typically exceed this threshold; diffuse regional haze from dispersed industrial activity often does not. PARASOL's archive ends in 2013, so current operational work relies on combining TROPOMI AAI with MISR multi-angle geometry and MODIS AOD rather than on live polarimetric retrieval.
Attributing a plume to a specific facility
Source attribution requires more than detecting an absorbing aerosol column. The analyst needs to trace the plume back to a point of origin against a wind field. The standard approach combines TROPOMI AAI or MODIS AOD anomalies with ERA5 reanalysis wind vectors at the relevant pressure levels, then applies back-trajectory modelling (HYSPLIT is the publicly documented standard) to identify candidate sources within the plume's upwind footprint. MISR stereo height retrievals are particularly valuable here because they fix the plume altitude, which determines which wind layer governs transport.
For coking plants and steel mills, co-emitted species provide corroborating evidence. These facilities emit NO2 and CO alongside black carbon. Cross-referencing a positive AAI anomaly with a co-located TROPOMI NO2 column enhancement narrows attribution considerably, since the NO2 and BC should share a common source geometry. This multi-species approach does not eliminate ambiguity entirely but it raises the evidentiary threshold enough to support regulatory referral. Open industrial fires, by contrast, emit high CO with relatively lower NO2, a ratio that differs from coke oven emissions and can be used to distinguish fire events from stack emissions in the same pixel.
Honest limits of the method
TROPOMI's 3.5 × 5.5 km pixel is large relative to individual stacks. A single coking battery might occupy less than one pixel, meaning the retrieved AAI reflects a mixture of the plume, background aerosol, and surface reflectance. In urban-industrial zones where multiple sources cluster, pixel-level attribution to a specific facility is often not possible without independent dispersion modelling or higher-resolution ancillary data.
Cloud cover is a hard block. TROPOMI requires cloud radiance fraction below roughly 0.5 for valid aerosol retrievals, and industrial regions in northern China, central Europe and the Great Lakes corridor frequently experience cloud cover exceeding that threshold for days at a time. Monthly compositing recovers spatial patterns but loses the event-level temporal resolution needed for enforcement. Finally, the AAI is sensitive to aerosol layer height: the same mass of absorbing aerosol produces a larger AAI signal when elevated than when confined to the surface layer, introducing uncertainty in mass loading estimates that cannot be resolved from AAI alone.
What a monitoring programme looks like in practice
A credible black carbon surveillance programme for an industrial zone combines three data streams. Daily TROPOMI AAI provides the detection trigger. MISR overpasses, roughly every nine days at mid-latitudes, supply plume height and multi-angle texture that help separate dust from combustion aerosol. MODIS fire radiative power flags open burning events that might otherwise be misclassified as stack emissions. The three streams are reconciled against a facility inventory, and anomalies that persist across multiple TROPOMI overpasses and align with known facility locations are escalated for regulatory review.
Satellize runs this kind of multi-source aerosol attribution on open constellations for government clients. The workflow is similar in structure to the crop-estimation analytics we operate for the Kingdom of Tonga, adapted for atmospheric rather than surface signals. Outputs are delivered as georeferenced GIS layers with per-event confidence scores, not raw imagery, so that an environmental enforcement team without remote sensing expertise can act on them directly.
The dust co-occurrence problem in enforcement contexts
Regulators need to understand that a positive AAI finding is not, by itself, a violation notice. In any region with episodic dust transport, the false-positive rate from dust contamination is non-trivial. The Saharan dust plumes that reach southern Europe in spring, or the Gobi dust events that cross the Korean Peninsula in March and April, produce AAI values that can exceed those from nearby industrial sources. An enforcement programme that ignores this will generate contested findings.
The defensible approach is to build a local dust climatology from years of TROPOMI and MODIS data, then flag only those AAI exceedances that occur outside the expected dust-season envelope or that are spatially inconsistent with the prevailing dust transport direction. Polarimetric data from future missions, including the Multi-Viewing Multi-Channel Multi-Polarisation Imager (3MI) on Metop-SG, which is designed as a POLDER successor, will eventually restore the shape-discrimination capability that was lost when PARASOL ended. Until those data are operational and validated, dust disambiguation remains the most labour-intensive part of any black carbon attribution workflow.
Typical figures
| TROPOMI AAI spatial resolution | 3.5 × 5.5 km (nadir, post-August 2019 upgrade) |
| TROPOMI revisit | Daily global coverage (approximately 14 orbits per day) |
| TROPOMI AAI spectral basis | 340 nm and 380 nm UV band pair |
| MISR spatial resolution | 275 m (red band, nadir camera) to 1.1 km (multi-angle modes) |
| MISR revisit at mid-latitudes | Approximately 9 days |
| POLDER/PARASOL resolution | Approximately 6 km; archive 2004–2013 only |
| MODIS AOD products | 3 km and 10 km; twice-daily combined Terra/Aqua |
| Minimum detectable AAI anomaly | AAI > ~0.5 considered significant; dense industrial plumes typically > 1.0 |
| Cloud cover limit for AAI retrieval | Cloud radiance fraction must be below approximately 0.5 |
| TROPOMI data archive depth | From May 2018 to present; freely available via Copernicus Data Space |
Analytics Satellize can run
| Daily AAI anomaly map per industrial zone | TROPOMI AAI time-series anomaly detection against rolling 30-day baseline | GeoTIFF layer with flagged pixels exceeding threshold, delivered within 24 hours of overpass |
| Plume back-trajectory and source attribution report | HYSPLIT back-trajectory modelling driven by ERA5 wind fields, anchored to MISR plume height where available | Per-event PDF report with candidate facility list and confidence score |
| Dust contamination filter | Comparison of AAI anomaly against multi-year TROPOMI dust climatology and MODIS dust optical depth; spatial consistency check against transport direction | Boolean dust-flag field appended to anomaly GIS layer |
| Multi-species corroboration layer | Spatial and temporal co-location of TROPOMI AAI with TROPOMI NO2 and CO column anomalies to distinguish stack emissions from open fire | Classified plume-type GIS layer (stack combustion / open fire / dust / ambiguous) |
| Monthly facility-level exceedance summary | Aggregation of daily anomaly flags against a geo-coded industrial facility inventory; persistence filtering to require at least two independent overpasses | Tabular report per facility with exceedance count, mean AAI, and dust-cleared event count |
| Regulatory-grade event dossier | Compilation of TROPOMI, MISR and MODIS evidence for a single high-confidence event, with meteorological context and uncertainty characterisation | Structured PDF with imagery, trajectory maps, data provenance and caveats, formatted for submission to an environmental enforcement body |
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.