Satellite aerosol optical depth retrieval and particulate mapping
Satellite sensors retrieve aerosol optical depth by separating aerosol signal from surface reflectance across multiple wavelengths and angles. The method is powerful and globally consistent, but column measurements alone cannot tell you where in the atmosphere the aerosol sits.
Sensors
- MODIS (Terra and Aqua): 36-band imager at 250 m to 1 km resolution depending on band; daily global coverage from the two-satellite pair. The Dark Target and Deep Blue algorithms produce the Collection 6.1 AOD product at 3 km and 10 km grid cells, with reported uncertainty of ±0.05 ± 0.15τ over land under favourable surface conditions.
- VIIRS (NOAA-20 and Suomi-NPP): Successor to MODIS heritage; 375 m and 750 m bands. The NOAA Enterprise AOD product applies Dark Target logic adapted to VIIRS spectral response. Provides continuity of the long-term aerosol climate data record into the 2020s and beyond.
- MISR (Terra): Nine cameras at fixed angles from 70.5° forward to 70.5° aft, four spectral bands. The multi-angle geometry allows surface reflectance and aerosol contribution to be separated with fewer surface assumptions than single-view sensors. Spatial resolution 275 m to 1.1 km per camera; global coverage every nine days.
- Sentinel-3 OLCI and SLSTR: OLCI provides 21 bands at 300 m; SLSTR adds thermal and shortwave-infrared channels. Synergy products combining both instruments improve aerosol characterisation over complex surfaces. Revisit is roughly one day at mid-latitudes with the two-satellite Sentinel-3A/3B tandem.
What aerosol optical depth actually measures
AOD is a dimensionless number: the fraction of incoming solar radiation extinguished by aerosol particles along a vertical column through the entire atmosphere. An AOD of 0.1 is a fairly clean atmosphere; values above 1.0 indicate dense smoke, dust or pollution. The number integrates everything from the surface to the top of the atmosphere, which is both its strength and its central limitation.
Satellites retrieve AOD by measuring top-of-atmosphere reflectance and then subtracting the estimated contribution of the surface and the molecular (Rayleigh) atmosphere. What remains is attributed to aerosol. The accuracy of that subtraction depends almost entirely on how well the algorithm characterises the surface beneath. Over dark, vegetated land and open ocean, the problem is tractable. Over bright desert, snow, or dense urban surfaces, it is genuinely difficult, and retrieval uncertainty roughly doubles.
Dark Target, Deep Blue, and why two algorithms exist
The Dark Target algorithm, developed for MODIS Collection 5 and refined through Collection 6.1, exploits the empirical relationship between surface reflectance in the mid-infrared and the visible bands over vegetated and dark-soiled land. The surface is estimated from the mid-infrared, and the aerosol signal is isolated in the blue and red. It works well over dense vegetation and dark urban surfaces but fails over bright soils where that empirical relationship breaks down.
Deep Blue was designed specifically for bright desert and arid surfaces. It uses the 0.412 µm channel, where the surface is relatively dark even over sand, allowing aerosol signal to dominate. Combining both algorithms into a merged product extends global land coverage substantially, though the two retrievals carry different uncertainty profiles and should not be treated as interchangeable over transition zones between surface types.
MISR sidesteps some of this by using angular diversity rather than spectral assumptions. Because surface reflectance varies with angle differently from aerosol scattering, the nine-camera geometry can partially decouple the two signals without relying on a prescribed surface model. The trade-off is a nine-day revisit, which limits its use for tracking fast-moving pollution events.
The vertical distribution problem
A passive column measurement cannot tell you whether an AOD of 0.4 comes from a thin layer of smoke at 8 km or a thick boundary-layer haze at 200 m. That distinction matters enormously for surface air quality estimation, aviation hazard assessment, and climate forcing calculations. Resolving it requires either active lidar (CALIPSO, ICESat-2 in limited modes) or ground-based sun-photometer networks such as AERONET, which Satellize does not cover on this page.
The practical consequence is that translating satellite AOD into surface PM2.5 concentrations requires an auxiliary model for the aerosol vertical profile, typically drawn from chemical transport models such as GEOS-Chem or MERRA-2 reanalysis. Published studies report correlations between MODIS AOD and ground PM2.5 in the range of r² 0.5 to 0.8 depending on region and season, meaning the satellite column explains roughly half to two-thirds of the variance in surface concentrations. That is useful for spatial pattern analysis and long-term trend monitoring; it is not a substitute for surface monitoring networks in regulatory contexts.
Sentinel-3 synergy and the push toward finer resolution
MODIS and VIIRS AOD products at 3 to 10 km grid cells are well suited to regional and global analysis but miss intra-urban gradients and localised industrial plumes. Sentinel-3 OLCI at 300 m native resolution, combined with SLSTR's thermal and shortwave-infrared channels, offers a path toward finer-scale retrievals over heterogeneous surfaces. ESA's Synergy Level-2 aerosol product applies a coupled surface-atmosphere inversion across both instruments.
The improvement is real but conditional. At 300 m, surface heterogeneity within a pixel is lower, which reduces the smearing of bright and dark surfaces that inflates uncertainty at coarser scales. However, the fundamental spectral constraints remain: Sentinel-3 is not a multi-angle instrument, so it relies on spectral rather than angular separation. Over bright urban rooftops and desert margins, retrieval confidence is still lower than over dark surfaces, regardless of spatial resolution.
Where the numbers hold and where they do not
Cloud contamination is the most pervasive practical limit. Aerosol retrievals are masked wherever cloud cover exceeds a threshold, which in persistently cloudy tropical regions can eliminate valid retrievals for weeks at a time. Thin cirrus, which is spectrally similar to some aerosol types, is a particular hazard. Collection 6.1 MODIS includes improved cirrus screening, but residual contamination remains a known source of high-AOD artefacts.
Retrieval latency is a separate consideration. Near-real-time MODIS and VIIRS AOD products from NASA LANCE are typically available within three hours of observation, which is fast enough for air quality nowcasting and wildfire smoke tracking. Sentinel-3 near-real-time products are available through Copernicus within roughly three hours as well. Reprocessed, quality-controlled products take days to weeks and carry lower uncertainty, so the choice between them depends on whether the application needs speed or accuracy.
Satellize runs AOD time-series analysis and anomaly detection on MODIS, VIIRS, and Sentinel-3 open data for clients monitoring industrial emissions, cross-boundary dust transport, and ambient air quality trends. The approach is the same one underpinning the Tonga crop-estimation programme: open constellation data processed against validated algorithms, with explicit uncertainty bounds in every deliverable.
Typical figures
| Spatial resolution (MODIS AOD product) | 3 km and 10 km grid cells (Collection 6.1 Dark Target / Deep Blue) |
| Spatial resolution (Sentinel-3 OLCI native) | 300 m (synergy aerosol product at 1 km) |
| Revisit (MODIS Terra + Aqua combined) | 1–2 observations per day globally |
| Revisit (MISR) | 9 days global; narrower swath than MODIS |
| Revisit (Sentinel-3A/3B tandem) | Approximately 1 day at mid-latitudes |
| Near-real-time latency (MODIS/VIIRS LANCE) | Typically within 3 hours of observation |
| Reported AOD uncertainty (Dark Target, land) | ±0.05 ± 0.15τ at 550 nm over dark surfaces; higher over bright surfaces |
| Spectral bands used in retrieval | Primarily 0.412, 0.47, 0.55, 0.66, 2.1 µm (varies by algorithm) |
| Archive depth (MODIS Terra) | March 2000 to present |
| Primary delivery formats | HDF4/HDF5, NetCDF, GeoTIFF (derived); accessible via NASA Earthdata and Copernicus Dataspace |
Analytics Satellize can run
| Daily AOD anomaly map | MODIS Collection 6.1 Dark Target / Deep Blue merged product, z-score anomaly against rolling 10-year climatology | GeoTIFF layer with anomaly classification, updated daily |
| Industrial plume fingerprinting | Time-series AOD elevation correlated with wind-field back-trajectories (HYSPLIT or equivalent) to attribute elevated AOD to candidate sources | PDF report with source attribution probability and confidence bounds |
| Cross-boundary dust transport alert | VIIRS Deep Blue AOD threshold exceedance combined with aerosol type index (absorbing vs. scattering) from OMPS or TROPOMI UV aerosol index | Email or API alert with affected corridor polygon and estimated duration |
| Long-term AOD trend analysis | Mann-Kendall trend test on annual mean AOD from MODIS/VIIRS merged record (2000 to present) at user-defined administrative or watershed boundaries | Time-series chart and tabular trend statistics, delivered as report or data feed |
| AOD-to-PM2.5 surface concentration estimate | Geographically weighted regression of MODIS AOD against MERRA-2 vertical correction factor; uncertainty explicitly propagated and stated | Gridded PM2.5 estimate layer with per-pixel uncertainty band; clearly labelled as model-assisted, not regulatory-grade |
| Wildfire smoke extent and persistence tracking | MODIS and VIIRS AOD combined with active fire radiative power (FIRMS) to map downwind smoke plume area and estimate daily exposure duration | Daily GIS layer and weekly summary report |
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.