Aerosol optical depth mapping from multispectral imagers
Aerosol optical depth (AOD) quantifies how much sunlight a column of atmosphere scatters or absorbs. Satellites retrieve it at multiple wavelengths, separating dust, smoke, and sulphate pollution by spectral fingerprint and Ångström exponent, with known limits over bright surfaces.
Sensors
- MODIS (Terra / Aqua): 36 spectral bands, 250 m to 1 km depending on band. AOD products (MOD04/MYD04) delivered at 10 km resolution using Dark-Target and Deep Blue algorithms. Terra overpass ~10:30 local, Aqua ~13:30, giving two daily snapshots. Archive runs from 2000 (Terra) and 2002 (Aqua).
- VIIRS (NOAA-20 / Suomi-NPP): 22 spectral bands from 412 nm to 12 µm. VIIRS Deep Blue AOD delivered at 6 km resolution, improving on MODIS spatial detail. NOAA-20 overpass ~13:30 local. Useful for continuity as MODIS instruments age.
- TROPOMI (Sentinel-5P): UV-VIS-NIR-SWIR push-broom spectrometer. AOD retrievals at 5.5 × 3.5 km pixel (post-2019 update). Daily global coverage. Particularly strong for absorbing aerosols and for co-locating aerosol with trace-gas columns (NO₂, SO₂) in a single overpass.
- MISR (Terra): Nine cameras at fixed angles from 70.5° forward to 70.5° aft, four spectral bands (446–866 nm). Multi-angle geometry allows aerosol type discrimination and height estimation without assuming surface reflectance. 275 m per-camera resolution, 9-day global repeat.
What aerosol optical depth actually measures
AOD is a dimensionless number: the integrated extinction of a vertical atmospheric column at a given wavelength. An AOD of 0.1 at 550 nm is a clean maritime atmosphere. Values above 1.0 indicate dense smoke or dust; readings above 3.0 occur in severe wildfire plumes and near major dust source regions during storm events. The number itself says nothing about altitude, particle size, or composition. Those inferences come from combining AOD at multiple wavelengths.
The Ångström exponent, derived from the slope of AOD across wavelengths, is the first discriminator. Coarse particles such as mineral dust produce exponents near zero or negative, because they scatter long and short wavelengths almost equally. Fine-mode particles from combustion or industrial sulphate produce exponents between 1.5 and 2.5. A single MODIS overpass carrying AOD at 470, 550, and 660 nm is enough to compute a rough exponent and flag the dominant aerosol type, though the separation is probabilistic, not definitive.
Dark-Target and Deep Blue: two algorithms, one awkward truce
The Dark-Target algorithm, developed by Kaufman and Tanré and applied operationally to MODIS since the early 2000s, exploits the empirical relationship between surface reflectance in the mid-infrared and in the visible. Over dense dark vegetation and dark ocean surfaces, the surface contribution to top-of-atmosphere reflectance is small and predictable, so the aerosol signal can be isolated. Over bright urban or arid surfaces, the assumption collapses. Dark-Target masks pixels where surface reflectance at 2.1 µm exceeds roughly 0.25, which excludes large fractions of the Sahara, Arabian Peninsula, and Central Asian deserts.
Deep Blue, developed specifically for bright surfaces by Hsu and colleagues at NASA GSFC, uses short blue wavelengths (412 nm) where mineral surfaces are relatively dark even when they appear bright in the visible. It recovers AOD over desert regions where Dark-Target is silent, but carries higher uncertainty, typically ±0.03 + 15% of AOD over land versus ±0.05 + 15% for Dark-Target. Neither algorithm works well over snow and ice, where surface reflectance is high across all visible wavelengths. The operational MODIS product merges both retrievals, flagging which algorithm was used per pixel.
TROPOMI's advantage: chemistry alongside opacity
TROPOMI on Sentinel-5P retrieves AOD as part of a broader atmospheric composition product, and its real value is co-location. A single overpass delivers AOD, NO₂ column, SO₂ column, and formaldehyde in the same pixel at the same moment. That combination lets an analyst distinguish industrial sulphate aerosol (high SO₂, high fine-mode AOD) from secondary organic aerosol over a forest fire (high formaldehyde, moderate AOD, low SO₂). MODIS cannot do this alone.
TROPOMI's 5.5 × 3.5 km pixel is coarser than some applications would prefer, but the daily global repeat and the spectral richness compensate. The instrument's UV channels also allow retrieval of the UV Aerosol Index, a qualitative flag for absorbing aerosols that remains valid even where formal AOD retrieval fails, including over bright surfaces. This is not a quantitative AOD product, but it preserves spatial coverage where the algorithms above go blind.
Where retrievals fail, and what to do about it
Cloud contamination is the dominant source of missing data. MODIS Dark-Target requires a minimum of 12 cloud-free pixels in a 20 × 20 km box to report a retrieval; in persistently cloudy regions such as the Congo Basin or the maritime tropics, daily coverage can be below 20%. Temporal compositing over 8-day or monthly periods recovers spatial completeness but loses the ability to track individual pollution events.
Sun-glint over ocean, high solar zenith angles at high latitudes in winter, and residual cloud contamination in broken cloud fields all introduce retrieval artefacts. MISR's multi-angle approach is more tolerant of some of these conditions, particularly thin cirrus, because the angular signature of ice cloud differs from aerosol. However, MISR's 9-day repeat makes it unsuitable for tracking fast-moving events. The practical answer for most operational programmes is to fuse MODIS, VIIRS, and TROPOMI retrievals, accepting that no single sensor covers every condition.
From AOD to air quality: the translation problem
Governments and health agencies want PM2.5 concentrations at ground level, not column-integrated optical depth. The conversion requires a model of aerosol vertical distribution, typically from reanalysis products such as MERRA-2 or CAMS, and an assumed mass extinction efficiency that varies by aerosol type and relative humidity. Published studies using MODIS AOD and ground-truth PM2.5 from surface networks report correlations of 0.7 to 0.9 over well-characterised regions, but the relationship degrades in complex terrain, near strong local sources, and in regions with limited ground-truth for calibration.
The column-to-surface translation is an area of active research rather than a solved problem. Satellize's analytics pipelines treat the AOD-to-PM2.5 step as a separately flagged uncertainty layer in any delivered product, rather than presenting a single PM2.5 estimate as though the physics were settled. That distinction matters when the output informs a regulatory or public health decision. For clients building long-term air quality baselines, MODIS's 20-plus-year archive, consistent reprocessing under Collection 6.1, and free availability through NASA Earthdata make it the starting point regardless of its resolution limits.
Putting it to work
Practical AOD programmes tend to fall into two categories: event monitoring and baseline climatology. Event monitoring, such as tracking a wildfire smoke plume or a Saharan dust outbreak crossing the Atlantic, needs daily or sub-daily data, rapid delivery, and aerosol-type flags. Baseline climatology needs a long, consistently processed archive and careful attention to algorithm version changes across the record.
Satellize runs AOD analytics on open constellations including MODIS, VIIRS, and TROPOMI, combining retrievals into fused daily mosaics with per-pixel quality flags and Ångström exponent layers. The Overhead column has covered AOD methodology in the context of seasonal dust and biomass-burning cycles. Clients wanting to assess a specific region, such as an industrial corridor or an agricultural burning season, can request a scoped feasibility review that maps data availability, cloud climatology, and retrieval uncertainty before committing to a full programme.
Typical figures
| Spatial resolution (MODIS AOD product) | 10 km (standard Dark-Target / Deep Blue grid) |
| Spatial resolution (VIIRS Deep Blue AOD) | 6 km |
| Spatial resolution (TROPOMI AOD) | 5.5 × 3.5 km per pixel |
| Revisit (MODIS Terra + Aqua combined) | Twice daily globally, ~10:30 and ~13:30 local overpass |
| Revisit (TROPOMI / Sentinel-5P) | Daily global coverage |
| Key spectral bands for AOD retrieval | 412, 470, 550, 660, 870 nm (MODIS); 412–2314 nm (TROPOMI); 446–866 nm (MISR) |
| Retrieval uncertainty (Dark-Target over land) | ±0.05 + 15% of AOD (one-sigma, Collection 6.1) |
| Archive depth (MODIS Terra) | February 2000 to present |
| Cloud-free coverage (daily, tropics) | Often below 30–50%; 8-day composites improve to 70–90% |
| Delivery formats (standard NASA products) | HDF4/HDF-EOS (MODIS), NetCDF-4 (TROPOMI), GeoTIFF via processing |
Analytics Satellize can run
| Daily fused AOD mosaic | Pixel-level merging of MODIS Dark-Target, Deep Blue, and VIIRS retrievals with quality-flag weighting | GeoTIFF layer with per-pixel AOD at 550 nm, quality flag, and algorithm-source band; delivered within 6 hours of last overpass |
| Aerosol type classification | Ångström exponent computation from multi-wavelength AOD; threshold-based separation of dust, smoke, and fine-mode industrial aerosol | Classified raster (dust / biomass-burning / anthropogenic / mixed) with confidence score, updated daily |
| Monthly AOD climatology baseline | MODIS Collection 6.1 archive aggregation with outlier screening; percentile distributions per grid cell | Gridded NetCDF climatology report covering user-defined region and period, with anomaly threshold maps |
| Pollution event alert | Threshold exceedance detection on daily AOD mosaic against rolling 30-day background; TROPOMI SO₂ cross-check for industrial attribution | Automated alert with event boundary polygon, peak AOD value, estimated aerosol type, and TROPOMI co-location summary |
| AOD-to-PM2.5 indicative estimate | MERRA-2 aerosol vertical profile combined with MODIS AOD; mass extinction efficiency lookup by aerosol type; uncertainty explicitly propagated | Gridded PM2.5 estimate layer with separate uncertainty band layer; flagged as indicative, not regulatory-grade |
| Ångström exponent time series | Multi-wavelength AOD retrieval from MODIS bands 1, 3, 4 and MISR where available; log-linear regression across wavelengths | Station or grid-cell time series CSV with exponent, AOD, and retrieval-source metadata; suitable for seasonal trend analysis |
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.