Tropospheric ozone column mapping for air-quality assessment
Satellites measure total ozone columns easily; isolating the tropospheric fraction that governs ground-level air quality is the hard part. This page explains how TROPOMI, OMI and companion sensors do it, and where the method breaks down.
Sensors
- TROPOMI / Sentinel-5P: UV-Vis-NIR-SWIR push-broom spectrometer. Daily global coverage. Nadir pixel of 3.5 × 5.5 km (improved to 3.5 × 5.5 km after August 2019 processor update, from an original 3.5 × 7 km). Retrieves total and tropospheric ozone columns via direct DOAS fitting in the Huggins bands (310–340 nm) and a cloud-slicing approach.
- OMI / Aura: Hyperspectral UV-Vis nadir sounder. Operational since 2004; pixel size 13 × 24 km at nadir, degraded by row-anomaly masking since 2009. Tropospheric ozone retrieved via the residual method (total column minus stratospheric climatology) and via the DOAS cloud-slicing technique. Long archive underpins trend studies.
- GOME-2 / MetOp series: UV-Vis-NIR grating spectrometer on MetOp-A, -B and -C. Pixel size 80 × 40 km in standard mode, 40 × 40 km in narrow swath. Global coverage in 1.5 days. Contributes tropospheric ozone via the residual method and is cross-calibrated with TROPOMI to extend the long-term record back to GOME on ERS-2 (1995).
- IASI / MetOp series: Thermal infrared Fourier-transform sounder. Circular footprint of roughly 12 km diameter at nadir. Sensitive to ozone in the mid-troposphere (around 3–8 km altitude) via the 9.6 µm absorption band, complementing UV-Vis sensors that are blind under thick cloud. Twice-daily global coverage per satellite.
Why tropospheric ozone is hard to see from space
Total ozone columns observed from space are dominated by the stratospheric ozone layer, which holds roughly 90 percent of all atmospheric ozone. The tropospheric fraction, typically 20 to 50 Dobson Units against a total column of 250 to 350 DU, sits on top of a large, variable stratospheric background. Extracting a signal that is 10 to 20 percent of the total, with meaningful spatial and temporal resolution, requires either a precise independent estimate of the stratospheric contribution or a retrieval geometry that is selectively sensitive to the lower atmosphere.
Two main approaches are in operational use. The residual method subtracts a stratospheric ozone column estimate, derived from assimilation models or from overpass pairs at the same latitude, from the total column. Its accuracy is limited by how well the stratospheric field is known, and errors of a few DU in the stratospheric estimate propagate directly into the tropospheric product. The direct UV-Vis DOAS approach exploits the fact that photons scattered back from cloud tops at different altitudes sample different ozone path lengths. By comparing retrievals above high and low clouds within the same scene, it is possible to infer the ozone between those cloud levels. This cloud-slicing technique is elegant but requires suitable cloud fields and fails in cloud-free scenes or under optically thin cloud.
What TROPOMI actually measures, and what it cannot
TROPOMI's tropospheric ozone product uses a combination of the Optimal Estimation method applied to the Huggins absorption bands and cloud-top pressure information derived from the O2 A-band at 760 nm. The 3.5 × 5.5 km pixel is fine enough to resolve city-scale pollution plumes and to distinguish industrial point sources from regional background. Daily global coverage means a continuous time series is available for any location, though the useful daily fraction is reduced by cloud: pixels with cloud radiance fraction above roughly 0.5 are typically flagged or excluded.
The vertical sensitivity of a UV-Vis nadir measurement is poor near the surface. The averaging kernel for TROPOMI's tropospheric ozone peaks in the middle and upper troposphere; ozone in the boundary layer, where human exposure occurs, contributes weakly to the column signal. This is not a flaw in TROPOMI's design but a consequence of Rayleigh scattering physics: UV photons rarely penetrate to the surface before scattering back. IASI partially compensates by being sensitive at 3 to 8 km altitude via the thermal infrared, but it too has limited skill below 2 km. Neither sensor replaces a surface monitoring network for exceedance assessment.
Cloud-top pressure ambiguity and the ghost signal problem
Cloud-slicing retrieves ozone between two cloud layers by treating the upper cloud as a mirror that reflects UV light and the lower cloud as a floor. The method assumes the cloud-top pressure is known accurately. In practice, cloud-top pressure from the O2 A-band has an uncertainty of roughly 20 to 50 hPa under broken or semi-transparent cloud, which translates to an altitude ambiguity of 200 to 500 m. Over a troposphere that is 10 to 12 km deep, that ambiguity is manageable for column totals but significant for layer attribution.
A subtler problem is the ghost column: ozone below the cloud top is not sampled by the UV measurement, so the retrieval must fill this gap with a climatological profile. If the actual sub-cloud ozone departs from the climatology, for instance during a pollution episode close to the surface, the retrieval will be biased. Published validation studies comparing TROPOMI tropospheric ozone columns against ozonesonde profiles show biases of roughly minus 3 to plus 5 DU depending on region and season. Users building emission inventories or health-exposure estimates should treat these products as indicative of the free-tropospheric burden rather than as a surface concentration proxy.
The long record and what it tells us about trends
The value of tropospheric ozone mapping from space is cumulative. OMI's archive stretches back to 2004, GOME-2 to 2006, and the combined GOME/SCIAMACHY/GOME-2 record to 1995. Cross-calibration between these instruments, carried out through international programmes including the Copernicus Climate Change Service, allows trend detection at the regional scale. Published analyses show statistically significant increases in tropospheric ozone over parts of South and East Asia over the past two decades, consistent with rising NOx and VOC emissions, while North American and European columns have shown more modest changes or slight declines in some seasons.
Trend detection requires careful handling of the row-anomaly in OMI (which reduces usable pixels), differences in pixel size between instruments, and changes in retrieval algorithms across the record. A trend of 1 DU per decade is near the detection limit of the residual method given these uncertainties. TROPOMI's finer pixel and improved signal-to-noise ratio should sharpen trend detection from the mid-2010s onward, though its record is still short for climatological analysis.
Practical use in air-quality assessment and what satellite data cannot replace
Governments and environmental agencies use tropospheric ozone columns primarily for three purposes: tracking regional-scale pollution transport, evaluating chemistry-transport model performance, and identifying areas where surface monitoring is absent. The satellite data is well suited to all three. A city with no surface ozone monitor can still be placed in regional context using TROPOMI daily composites. A model that systematically underestimates ozone over the North Atlantic in summer can be diagnosed against the satellite column.
What the satellite cannot provide is a reliable daily exceedance alert for a specific city block. The column measurement integrates over the full tropospheric depth and, as noted, is insensitive near the surface. For regulatory compliance and public health alerts, satellite data works best as a constraint on model assimilation systems, such as the Copernicus Atmosphere Monitoring Service (CAMS), which ingests TROPOMI ozone and produces gridded surface ozone analyses at roughly 10 km resolution. Satellize can operationalise TROPOMI and IASI data streams and combine them with CAMS outputs to produce regional air-quality assessments for clients who lack in-country monitoring infrastructure. The same analytical pipeline that supports our Tonga crop-estimation programme, which fuses multiple open constellations into a single decision product, applies directly to atmospheric column work.
The honest summary: tropospheric ozone from space is a powerful regional diagnostic. It is not a substitute for surface sensors where exposure thresholds matter.
Typical figures
| Best available spatial resolution | 3.5 × 5.5 km (TROPOMI, post-August 2019) |
| Revisit frequency | Daily global coverage (TROPOMI); 1.5 days (GOME-2 per satellite); 16 days (OMI, accounting for row anomaly) |
| Latency (near-real-time product) | TROPOMI NRT product available within 3 hours of sensing |
| Key spectral bands | UV Huggins bands 310–340 nm (DOAS); O2 A-band 760 nm (cloud pressure); thermal IR 9.6 µm (IASI ozone) |
| Tropospheric column precision (single pixel) | Approximately 6–8 DU for TROPOMI; 5–10 DU for OMI depending on scene |
| Vertical sensitivity | Middle to upper troposphere (3–10 km); boundary layer contribution weak below ~2 km for UV sensors |
| Cloud screening threshold | Pixels with cloud radiance fraction >0.5 typically excluded; cloud-slicing requires cloud fraction 0.2–0.8 |
| Archive depth | TROPOMI from October 2017; OMI from October 2004; GOME-2 from 2006; GOME from 1995 |
| Standard data formats | NetCDF-4 (Sentinel-5P); HDF-EOS5 (OMI); BUFR and NetCDF (IASI, GOME-2) |
| Typical bias vs. ozonesonde | Minus 3 to plus 5 DU depending on region, season and retrieval version |
Analytics Satellize can run
| Daily tropospheric ozone column map | TROPOMI L2 tropospheric ozone product ingestion, cloud-flag filtering, gridding to client AOI | GeoTIFF or NetCDF layer at 0.05° grid, updated daily |
| Monthly and seasonal anomaly maps | Baseline climatology constructed from OMI/GOME-2 archive; TROPOMI anomaly computed relative to baseline | PDF report with maps and time-series charts per region of interest |
| Long-term trend analysis | Merged GOME/SCIAMACHY/GOME-2/TROPOMI record; Mann-Kendall trend test on deseasonalised monthly means | Trend magnitude (DU per decade) with confidence intervals, delivered as a written assessment and GIS layer |
| Pollution episode detection and transport tracking | Day-to-day column difference maps combined with HYSPLIT or FLEXPART back-trajectory analysis on publicly available meteorological fields | Event bulletin with source attribution confidence rating and plume extent polygon |
| Model validation dataset | Co-location of TROPOMI and IASI retrievals with chemistry-transport model output at matching grid cells and times | Bias and RMSE statistics table, scatter plots, ready for inclusion in regulatory model evaluation reports |
| CAMS-fused surface ozone indicator | TROPOMI column assimilated into CAMS reanalysis; surface ozone field extracted and downscaled to national grid | Gridded surface ozone indicator layer (not a regulatory measurement) at 0.1° resolution, monthly |
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.