Absorbing aerosol index and black carbon column retrieval
The absorbing aerosol index flags light-absorbing particles from space using UV backscatter, but cannot alone separate black carbon from mineral dust. Ancillary data and transport modelling are needed to make the signal actionable.
Sensors
- Sentinel-5P TROPOMI: UV spectrometer covering 270–500 nm; AAI product at 3.5 × 5.5 km nadir pixel (upgraded from original 7 × 3.5 km in August 2019); daily global coverage; near-real-time latency of roughly 3 hours after sensing.
- Aura OMI: Ozone Monitoring Instrument; AAI at 13 × 24 km nadir pixel; daily global coverage; archive from 2004 onward, providing the longest continuous UV aerosol record from a single modern sensor, though row-anomaly data gaps have affected coverage since 2009.
- NOAA-20 OMPS: Ozone Mapping and Profiler Suite nadir mapper; AAI at approximately 50 × 50 km; daily global coverage; part of the JPSS series intended to continue the long-term aerosol record beyond OMI.
- GOME-2 on MetOp-A/B/C: Global Ozone Monitoring Experiment-2; AAI at 40 × 80 km (MetOp-A/B) and 40 × 40 km (MetOp-C in tandem mode); three satellites provide sub-daily sampling at mid-to-high latitudes; archive from 2007.
What the UV backscatter signal actually measures
When sunlight enters an aerosol-laden atmosphere, absorbing particles reduce the amount of UV radiation scattered back to the sensor relative to what a purely scattering atmosphere would return. The absorbing aerosol index quantifies this deficit by comparing observed backscattered radiance at two UV wavelengths, typically around 340 nm and 380 nm for TROPOMI and OMI, against a modelled Rayleigh-scattering baseline. A positive AAI indicates absorption; the larger the value, the more absorbing the column.
The physics is straightforward but the interpretation is not. Both carbonaceous aerosols from combustion and mineral dust from arid regions absorb UV, and both produce positive AAI values. Sea salt and sulphate aerosols scatter rather than absorb, so they return near-zero or slightly negative AAI. The index is therefore a qualitative flag for absorbing aerosol presence, not a quantitative mass concentration. Treating a high AAI pixel as a direct black carbon measurement without additional constraints is a common and consequential error.
The dust-carbon ambiguity and how analysts resolve it
Mineral dust from the Sahara, Arabian Peninsula or Taklamakan Desert can produce AAI values exceeding 3 or 4, comparable to dense biomass-burning smoke. Separating the two requires ancillary data. The most common approach pairs AAI with aerosol optical depth from a multispectral imager such as MODIS or VIIRS: dust tends to produce spectrally flat or slightly increasing optical depth from blue to red, while carbonaceous smoke shows stronger blue absorption. Colour ratio products from MODIS have been used operationally for this purpose for over a decade.
A second disambiguation route uses transport model back-trajectories. If the air mass origin is a known fire region, the absorbing signal is more likely carbonaceous. If it originates over a desert source, dust is the default hypothesis. Neither approach is definitive over mixed-source regions such as West Africa during the boreal winter, where Saharan dust and savanna smoke co-exist in the same column. In those cases, honest analysts assign probability ranges rather than categorical labels.
Black carbon column amounts cannot be retrieved directly from AAI alone. Dedicated retrievals, such as those from the POLDER polarimeter aboard PARASOL, attempted to separate fine-mode absorbing aerosol, but PARASOL ceased operations in 2013. Current best practice combines AAI as a qualitative indicator with model assimilation schemes, such as the Copernicus Atmosphere Monitoring Service CAMS reanalysis, which ingests multiple satellite observations to produce gridded black carbon fields.
Bright surfaces: where retrievals become unreliable
The AAI algorithm assumes that the surface below the aerosol layer is dark in the UV. Desert sand, snow and ice violate this assumption badly. Over the Sahara or Greenland, the surface itself contributes substantially to the backscattered UV signal, and the modelled Rayleigh baseline becomes a poor reference. Retrievals over these surfaces carry uncertainties large enough to render individual pixel values misleading. ESA and EUMETSAT product documentation for both TROPOMI and GOME-2 explicitly flag retrievals over high-albedo surfaces as having reduced confidence.
Cloud contamination is a related problem. Optically thick cloud below an aerosol layer shields the surface signal but also blocks the aerosol from view. Thin cloud above an aerosol layer can enhance the apparent AAI by increasing the effective scattering path. TROPOMI's cloud fraction product is used to screen pixels with cloud radiance fractions above a threshold, typically 0.5, but this removes a significant fraction of tropical retrievals during convective seasons.
Transport modelling: what AAI contributes and what it cannot anchor
Long-range transport of black carbon from large fire events, such as the 2019–2020 Australian fires that injected smoke into the stratosphere, or Siberian fire seasons that deposit carbonaceous aerosol over the Arctic, is one of the highest-value applications of AAI data. The index provides a spatially continuous, daily constraint on plume position and approximate intensity that no surface network can match at continental scale.
The limitation is vertical. AAI is a column-integrated quantity with no native vertical resolution. It cannot distinguish whether the absorbing aerosol is in the boundary layer, the free troposphere or, in extreme pyroconvection events, the lower stratosphere. Pairing AAI with spaceborne lidar, specifically CALIPSO or the CATS instrument on the ISS (now concluded), or with CALIOP extinction profiles, adds the vertical dimension needed to constrain deposition modelling. Without that vertical information, model assimilation of AAI alone can produce plume heights that are wrong by several kilometres.
For black carbon transport specifically, the CAMS global reanalysis assimilates MODIS aerosol optical depth and MOPITT carbon monoxide alongside AAI-derived constraints. The resulting black carbon fields have been used to assess Arctic deposition rates and radiative forcing at the surface, though the uncertainty in modelled black carbon mass remains large, often a factor of two or more, reflecting both retrieval uncertainty and poorly constrained emission inventories.
Operational products and their honest shelf life
TROPOMI AAI is delivered by ESA and processed by KNMI as part of the Sentinel-5P data stream, available through the Copernicus Data Space Ecosystem. Near-real-time products arrive within roughly three hours of sensing; offline reprocessed products, which apply updated calibration and improved cloud correction, follow with a delay of several weeks. Users relying on near-real-time data for rapid response should be aware that absolute values can shift slightly in the reprocessed record.
The OMI archive, extending from 2004 to the present despite its row anomaly, remains the most important long-term dataset for AAI trend analysis. Studies using OMI have documented multi-year trends in absorbing aerosol loading over South Asia and sub-Saharan Africa, though separating the dust and carbonaceous contributions in trend attribution requires careful auxiliary analysis. OMPS on JPSS satellites is designed to extend this record, with cross-calibration between sensors an active area of work at NOAA and NASA.
Satellize runs AAI time-series analysis from TROPOMI and OMI archives for clients tracking fire-season aerosol loading and its downstream effects on solar resource and air quality. The methodology is the same class of retrieval used in the CAMS operational service, applied to client-defined regions of interest with outputs formatted for integration into existing environmental monitoring workflows.
Typical figures
| Best nadir spatial resolution (TROPOMI) | 3.5 × 5.5 km (since August 2019 upgrade) |
| Spatial resolution (OMI) | 13 × 24 km nadir; degrades toward swath edge |
| Spatial resolution (GOME-2 MetOp-C tandem) | 40 × 40 km |
| Revisit (TROPOMI, OMI, OMPS) | Daily global coverage at nadir; TROPOMI swath 2600 km |
| Key spectral bands | UV: approximately 340 nm and 380 nm (pair used for AAI calculation) |
| Near-real-time latency (TROPOMI) | Approximately 3 hours after sensing |
| Archive depth | OMI from 2004; GOME-2 from 2007; TROPOMI from May 2018 |
| Reliable retrieval conditions | Dark UV surface, cloud radiance fraction below ~0.5; degraded over snow, ice, bright desert |
| AAI detection threshold (qualitative) | Values above ~1 indicate absorbing aerosol presence; values above ~3 indicate dense plumes |
| Delivery formats | NetCDF-4 (ESA/Copernicus standard); GeoTIFF and CSV on client request |
Analytics Satellize can run
| Daily AAI anomaly maps | TROPOMI Level-2 AAI product differenced against multi-year seasonal climatology from OMI archive | GeoTIFF layer and PDF summary for client-defined region, delivered within 6 hours of TROPOMI overpass |
| Fire-season absorbing aerosol loading index | Time-integrated AAI over defined fire seasons, with cloud-screened pixel weighting following KNMI product documentation | Annual report with inter-annual trend charts and comparison against CAMS reanalysis fields |
| Dust versus smoke probability flag | AAI combined with MODIS/VIIRS aerosol optical depth spectral ratio and HYSPLIT back-trajectory classification | Gridded probability layer (dust / carbonaceous / mixed) at TROPOMI resolution, delivered as NetCDF |
| Plume transport alert | Threshold exceedance on AAI combined with CAMS forecast trajectory to identify downwind receptor regions within 48–72 hours | Automated alert message with plume boundary polygon and estimated arrival window |
| Long-term AAI trend analysis | OMI and TROPOMI cross-calibrated time series with Mann-Kendall trend test, following published EUMETSAT multi-sensor harmonisation approach | Decade-scale trend report with confidence intervals and identified breakpoints |
| Solar resource impact assessment | AAI-derived absorbing aerosol optical depth proxy correlated with surface shortwave irradiance reduction, using published single-scattering albedo assumptions for regional aerosol types | Monthly irradiance reduction estimate for client solar assets, delivered as CSV with uncertainty bounds |
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.