SO2 emissions monitoring at smelters and cement plants
UV backscatter satellites detect sulphur dioxide columns above smelters and cement kilns independently of any reported figure. TROPOMI resolves facility-scale plumes daily, and wind-field integration converts those columns into source-rate estimates regulators can act on.
Sensors
- Sentinel-5P TROPOMI: The primary workhorse for facility-scale SO2. Nadir pixel footprint of 3.5 × 5.5 km (upgraded from 7 × 3.5 km in August 2019). Daily global coverage. Offline L2 SO2 total vertical column product available within roughly 3 hours of overpass; reprocessed offline product within about 5 days. Detection threshold near 1 Dobson Unit (DU) under clear-sky conditions; sensitivity degrades sharply under cloud fractions above ~0.3.
- Aura OMI: NASA's Ozone Monitoring Instrument, operational since 2004, provides the longest continuous UV-backscatter SO2 record. Nadir pixel 13 × 24 km, daily global revisit. Coarser than TROPOMI but invaluable for decadal trend analysis. Row-anomaly artefacts from 2008 onward reduce effective swath by roughly 25%.
- GOME-2 (MetOp-A, -B, -C): EUMETSAT's Global Ozone Monitoring Experiment-2 series. Nadir pixel 40 × 80 km in standard mode, 20 × 40 km in narrow-swath mode. Lower spatial resolution than TROPOMI limits it to large industrial districts or volcanic sources, but three satellites in succession extend the archive back to 2007 and provide morning-overpass geometry complementary to TROPOMI's early afternoon pass.
- TROPOMI offline L2 SO2 product (ESA/Copernicus): The processed data product rather than the raw instrument. Delivered as NetCDF-4 files via the Copernicus Data Space Ecosystem. Includes total vertical column, planetary boundary layer column, and stratospheric correction flags. The PBL column product is specifically designed for low-altitude industrial sources and is more sensitive to near-surface emissions than the total column under certain aerosol loading conditions.
Why UV backscatter works where stack inspectors cannot
Sulphur dioxide absorbs strongly in the ultraviolet, particularly around 310 nm, where its absorption cross-section is well characterised in laboratory measurements. Satellites carrying UV spectrometers measure the backscattered solar radiance and compare it against a reference spectrum free of SO2 absorption. The differential optical absorption spectroscopy (DOAS) technique isolates the SO2 signal from ozone, aerosol and surface reflectance contributions. The result is a total vertical column density, expressed in Dobson Units, integrated through the atmospheric column above each pixel.
The critical point for enforcement is independence. A satellite does not need access to a facility, does not rely on reported operational hours, and cannot be switched off during an inspection visit. The measurement geometry is fixed by orbital mechanics. For a smelter that shuts its scrubbers at night and reports clean daytime figures, the daily TROPOMI overpass, arriving at roughly 13:30 local solar time, captures whatever the stack is actually emitting at that moment. That is not a complete audit, but it is a consistent, timestamped, third-party observation.
What a 1 DU detection floor actually means in practice
TROPOMI's published detection limit of approximately 1 Dobson Unit sounds abstract. One DU corresponds to a column of roughly 2.69 × 10^16 molecules per square centimetre. For a pixel 3.5 km wide, that translates to a detectable emission rate that varies with wind speed and plume dilution, but published studies using TROPOMI data have attributed SO2 to individual large copper and nickel smelters with annual emissions above roughly 50 kilotonnes per year. Smaller facilities, or those in regions with frequent cloud cover, will fall below detection.
Cement kilns are a harder target than non-ferrous smelters. A large clinker kiln burning high-sulphur fuel might emit 5,000 to 15,000 tonnes of SO2 per year, which is detectable in aggregated monthly composites but rarely in a single overpass. A primary copper smelter such as those documented in Zambia's Copperbelt, by contrast, can emit hundreds of kilotonnes annually, producing plumes that saturate the TROPOMI SO2 signal across multiple pixels on individual days. The method is most powerful at the top of the industrial emission distribution.
Converting a column map into a source rate
A vertical column measurement tells you how much SO2 is overhead at a given location and time. To estimate how much the source is emitting, you need to know how fast the wind is carrying the plume downwind. The standard approach, sometimes called the cross-sectional flux method, integrates the SO2 column across a transect perpendicular to the plume axis and multiplies by the wind speed at plume height. Reanalysis wind fields from ERA5 (ECMWF) or MERRA-2 (NASA) are typically used when no local radiosonde data are available.
The method introduces meaningful uncertainty. ERA5 horizontal resolution is 31 km, which is coarser than a TROPOMI pixel in its upgraded configuration. In flat terrain with steady synoptic flow, the resulting source-rate estimate may carry an uncertainty of 30 to 50 percent. In complex terrain, that figure is worse. Mountain valleys channel and accelerate flow in ways that reanalysis grids cannot resolve. A smelter in a narrow Andean valley, for example, might have a local wind speed that differs from the ERA5 grid cell value by a factor of two or more. Results from such sites should be treated as order-of-magnitude estimates rather than precise inventories.
A further complication is plume age. SO2 oxidises to sulphate in the atmosphere with a lifetime that depends on humidity, temperature and OH radical concentration, typically ranging from a few hours to a couple of days. If the satellite overpass captures a plume that has already travelled several hundred kilometres from the source, the retrieved column underestimates total emission because some fraction has already converted to sulphate and left the SO2 absorption band.
Volcanic degassing as calibration and complication
Volcanic SO2 is not an industrial problem, but it matters operationally for two reasons. First, volcanoes are the best-characterised large SO2 sources on Earth, and published TROPOMI retrievals above Kilauea, Etna, and Ambae have been used to validate the column-to-flux conversion method against ground-based DOAS traverses. That validation literature gives buyers confidence that the same method applied to a nearby smelter is grounded in something real.
Second, volcanically active regions present a genuine attribution problem. A smelter located downwind of a degassing volcano, as is the case for several copper operations in Chile and Indonesia, may have its industrial signal partially masked or inflated by volcanic SO2 advected into the same pixel. Separating the two requires either temporal filtering during quiescent volcanic periods or spatial decomposition using multiple overpass geometries. Neither approach is fully reliable. Buyers operating in volcanically active regions should request explicit treatment of this ambiguity in any analytical deliverable.
Building a verification record regulators can use
A single overpass is an observation. A multi-year time series is evidence. The OMI archive extends to 2004, GOME-2 to 2007, and TROPOMI to May 2018. Combining these instruments, with appropriate cross-calibration corrections, allows an analyst to reconstruct the emission history of a facility across nearly two decades. That kind of record is directly comparable to national emissions inventories submitted under the Convention on Long-range Transboundary Air Pollution (CLRTAP) or regional permitting regimes.
Satellize runs TROPOMI SO2 analytics on open constellation data, applying the cross-sectional flux method with explicit uncertainty bounds and cloud-fraction filtering. The output is a facility-level time series delivered as a GIS layer and a structured report, formatted for direct comparison against permit conditions or self-reported stack data. The workflow is the same one underlying our Overhead column, where we have published satellite-derived SO2 observations for smelter districts in southern Africa. Buyers who want to understand the method before commissioning a full analysis can request a single-facility demonstration run covering the most recent 90-day TROPOMI archive.
Typical figures
| Spatial resolution (TROPOMI, post-Aug 2019) | 3.5 × 5.5 km nadir pixel |
| Spatial resolution (OMI) | 13 × 24 km nadir pixel |
| Revisit frequency | Daily global coverage (TROPOMI ~13:30 local solar time overpass) |
| Spectral range for SO2 retrieval | 305–340 nm UV (DOAS fitting window) |
| Minimum detectable column (single overpass, clear sky) | ~1 Dobson Unit (TROPOMI); ~0.5 DU in aggregated monthly composites |
| Cloud sensitivity threshold | Retrievals flagged unreliable above ~0.3 cloud radiance fraction |
| Product latency (TROPOMI offline L2) | Offline product available within ~3 hours; reprocessed offline within ~5 days |
| Archive depth | TROPOMI from May 2018; OMI from October 2004; GOME-2 from 2007 |
| Source-rate uncertainty (flat terrain, steady wind) | Typically 30–50%; higher in complex terrain |
| Delivery formats | NetCDF-4 (native L2), GeoTIFF column composites, GIS vector layers, structured PDF/CSV reports |
Analytics Satellize can run
| Facility SO2 column time series | TROPOMI offline L2 SO2 product extraction per facility bounding box, cloud-fraction filtering (CF < 0.3), daily overpass aggregation | CSV time series and GIS point layer with daily column values and quality flags, updated on each overpass |
| Source-rate estimate (tonnes SO2 per day) | Cross-sectional flux integration of TROPOMI SO2 columns across downwind transect, ERA5 wind field at plume height, with explicit uncertainty propagation | Structured PDF report per facility per quarter, including uncertainty bounds and wind-field provenance |
| Multi-year emission trend analysis | Cross-calibrated OMI + TROPOMI time series, seasonal decomposition, Mann-Kendall trend test on annual mean columns | Trend report comparing satellite-derived emission trajectory against reported inventory figures, with statistical significance statement |
| Exceedance alert | Threshold trigger on daily TROPOMI SO2 column exceeding user-defined DU value within facility pixel cluster, automated within 6 hours of product availability | Email or API alert with overpass timestamp, column value, cloud fraction, and a GeoTIFF of the overpass SO2 field |
| Plume attribution under volcanic interference | Temporal filtering to volcanically quiescent periods using TROPOMI volcanic SO2 flag; spatial decomposition comparing upwind and downwind column gradients | Annotated analysis note with explicit statement of attribution confidence and periods excluded from the record |
| Permit compliance comparison | Satellite-derived annual source-rate estimate compared against facility's permitted annual SO2 tonnage, formatted for regulatory submission | Compliance summary table with satellite estimate, reported figure, percentage difference, and methodological caveats |
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.