Sentinel-5P TROPOMI sulphur dioxide column mapping for volcanic and industrial emissions
TROPOMI maps sulphur dioxide columns globally every day at 3.5 km resolution, distinguishing stratospheric eruption plumes from the faint boundary-layer signal of industrial point sources, with a detection floor near 0.5 Dobson units.
Sensors
- Sentinel-5P TROPOMI: Primary sensor. UV-visible spectrometer covering 270–500 nm, 675–775 nm and 2305–2385 nm. SO2 retrieval from UV backscatter in the 312–326 nm window. Nadir pixel 3.5 × 5.5 km (improved from 7 km in 2019). Daily global coverage, offline L2 product typically available within 3 hours of overpass.
- Aura OMI: Heritage UV spectrometer (2004–present), 13 × 24 km nadir pixel, daily global coverage. Long archive enables trend analysis back to 2004; row-anomaly artefact limits some swath positions. Used for multi-year volcanic and industrial SO2 climatologies.
- GEMS (GEO-KOMPSAT-2B): Geostationary UV-visible spectrometer covering East Asia at roughly 3.5–8 km resolution, hourly revisit. Captures diurnal SO2 evolution from industrial corridors in China, Korea and Japan that polar orbiters sample only once per day.
- TROPOMI offline L2 SO2 products (ESA/Copernicus): Three operational column retrievals shipped per scene: lower troposphere assumption (1 km box profile), middle troposphere (7 km), and stratospheric (25 km). Choice of product determines which source altitude is correctly represented; using the wrong one biases flux estimates by factors of two to five.
What the UV backscatter actually measures
TROPOMI does not image SO2 directly. It records the spectrum of sunlight scattered back from the atmosphere and surface in the 312–326 nm window, then retrieves a sulphur dioxide vertical column density (VCD) by fitting the SO2 absorption cross-section against that spectrum using differential optical absorption spectroscopy (DOAS). The result is a total slant column, corrected to a vertical column using an air-mass factor that depends critically on where in the atmosphere the SO2 sits.
That altitude dependence is the central complication. SO2 near the surface is harder for the instrument to see because ozone and Rayleigh scattering attenuate the UV signal before it reaches the lower troposphere. A 1 Dobson unit (DU) column at the surface produces a weaker signal than the same column at 20 km altitude. ESA's operational processing therefore ships three separate L2 products per scene, each assuming a different box-profile altitude: 1 km (boundary-layer industrial sources), 7 km (moderate volcanic injection) and 25 km (stratospheric eruptions). Analysts must choose the product that matches their source, or run their own radiative-transfer correction.
The detection floor and what it means for passive degassing
The published detection limit for TROPOMI SO2 is approximately 0.5 DU under favourable conditions: low solar zenith angle, dark ocean surface, cloud-free sky. In practice, over bright land surfaces or at high latitudes in winter, the effective floor rises to 1–2 DU. That matters enormously for passive volcanic degassing. Many persistently active vents, including small fumarolic systems and low-level open-vent degassing at basaltic volcanoes, emit fluxes that produce columns well below 0.5 DU at the pixel scale. TROPOMI will not reliably detect them.
The instrument is well-suited to explosive eruptions injecting SO2 above 10 km, where stratospheric columns routinely reach tens or hundreds of DU. The 2022 Hunga Tonga–Hunga Ha'apai eruption produced SO2 columns exceeding 400 DU in TROPOMI data. For industrial point sources such as copper smelters or coal-fired power stations, the signal is weaker but often detectable: large smelters can produce boundary-layer columns of 1–5 DU downwind, which TROPOMI's 3.5 km pixel can resolve if wind conditions disperse the plume slowly. Smaller facilities sit below the noise floor.
Converting column maps into emission flux
A column map is not an emission rate. To convert VCD fields into tonnes per day, analysts integrate SO2 mass across a cross-wind transect through the plume and multiply by the wind speed at plume altitude. This approach, known as the cross-sectional flux method, requires coincident wind-field data. ERA5 reanalysis winds at 0.25-degree resolution are the standard public-domain source; for volcanic applications, plume altitude must be estimated independently (from ash retrieval, SO2 centre-of-mass fitting, or VAAC reports) before selecting the correct wind level.
A second method, inverse Gaussian plume fitting, fits a dispersion model to the observed column field and retrieves source strength and effective stack height simultaneously. It handles cases where the plume has dispersed beyond a clean transect. Both methods carry uncertainties of roughly 30–50 percent under typical conditions, dominated by wind-field error and the altitude ambiguity described above. Published studies on large smelter complexes in Zambia, Russia and Chile have demonstrated consistent flux retrievals from OMI and TROPOMI data, validating the approach against ground-based DOAS networks where those exist.
Volcanic hazard versus industrial compliance: different problems, same physics
For volcanic hazard, the priority is speed and plume tracking. Civil aviation authorities and volcanic ash advisory centres need to know whether an eruption has injected SO2 into the stratosphere, at what altitude, and where the plume is heading. TROPOMI's daily global pass and the offline L2 product (available within a few hours of overpass) are operationally useful for this, though geostationary sensors such as SEVIRI over Europe and Africa provide faster updates for large events.
Industrial compliance monitoring is a different analytical problem. The question is not where a plume is going but whether a facility's annual SO2 budget is consistent with its reported emissions. That requires multi-year time series, careful screening for cloud cover and unfavourable geometry, and statistical aggregation across many overpasses. TROPOMI's archive runs from late 2017; OMI extends the record to 2004. Together they cover the period when most national SO2 trading schemes and industrial permit regimes came into force, making independent satellite-based verification of reported emissions genuinely feasible for large point sources.
Practical limits buyers should understand before commissioning analysis
Cloud cover is the dominant data-loss mechanism. A single overpass covers a given point once per day, and any cloud fraction above roughly 0.3 in the retrieval scene degrades or invalidates the SO2 column. In persistently cloudy regions, such as the maritime tropics, useful cloud-free overpasses may occur on fewer than half of days. Temporal averaging recovers sensitivity for persistent sources but cannot recover a missed eruption.
Pixel size sets a hard limit on source attribution. At 3.5 km, a plume from a facility must be detectable at that spatial scale. In an industrial cluster where multiple stacks sit within a few kilometres of each other, TROPOMI cannot apportion emissions to individual sources. Higher-resolution airborne or ground-based DOAS is needed for that. TROPOMI is best understood as a fleet-level or facility-group monitor, not a stack-level instrument.
Typical figures
| Nadir pixel size (SO2 product) | 3.5 × 5.5 km (since August 2019 reprocessing; original 7 × 3.5 km) |
| Revisit | Daily global coverage (single polar orbit, 14 orbits per day) |
| L2 offline product latency | Typically 2–3 hours after overpass |
| SO2 retrieval spectral window | 312–326 nm (UV, DOAS fitting) |
| Column height assumptions (operational) | Three box profiles: 1 km (boundary layer), 7 km (mid-troposphere), 25 km (stratosphere) |
| Detection limit (favourable conditions) | ~0.5 DU over dark ocean; 1–2 DU over bright land or high solar zenith angle |
| Archive depth | Sentinel-5P: November 2017–present; Aura OMI heritage: October 2004–present |
| Flux retrieval uncertainty (published) | Approximately 30–50% under typical wind-field and altitude-ambiguity conditions |
| Delivery formats (ESA L2) | NetCDF-4 via Copernicus Data Space Ecosystem; HDF5 via older SciHub archive |
| Cloud screening threshold (standard) | Scene cloud radiance fraction < 0.5 for SO2 retrieval; stricter thresholds applied in post-processing |
Analytics Satellize can run
| Daily SO2 plume extent map | TROPOMI offline L2 column product, altitude-appropriate air-mass factor selection, cloud-screened and regridded | GeoTIFF or GeoJSON layer, delivered within 6 hours of overpass, suitable for GIS ingestion or VAAC briefing |
| Point-source emission flux time series | Cross-sectional flux integration using ERA5 wind fields at retrieved plume altitude; multi-overpass averaging for low-signal facilities | Monthly CSV of daily flux estimates (tonnes SO2 per day) with uncertainty bounds, per named facility |
| Annual facility SO2 budget vs. reported permit | Multi-year OMI and TROPOMI record aggregation, Gaussian plume fitting, comparison against national emissions registry data | Annual compliance summary report with statistical confidence intervals and data-gap accounting |
| Eruption injection altitude and stratospheric SO2 mass | SO2 centre-of-mass altitude fitting across three TROPOMI L2 products; radiative-transfer sensitivity analysis | Event report within 24 hours of eruption detection, including estimated SO2 burden in teragrams and plume trajectory |
| Industrial cluster attribution | Wind-rotated sector analysis separating upwind background from facility-specific downwind enhancement at 3.5 km pixel scale | Facility-group SO2 attribution layer with caveats on co-located source ambiguity |
| Long-term trend and anomaly detection | Seasonal decomposition of multi-year VCD time series; change-point detection against baseline period | Automated alert when 30-day rolling SO2 burden exceeds two standard deviations above historical baseline for a defined region |
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.