Hydrogen sulphide detection above wastewater treatment and industrial lagoons
Thermal infrared sounders can detect large H2S plumes from industrial accidents, but the detection floor sits far above routine facility emissions. This page explains the retrieval physics, current limits, and what next-generation hyperspectral sensors may change.
Sensors
- IASI (MetOp-A/B/C): Infrared Atmospheric Sounding Interferometer covering 645–2760 cm⁻¹ at 0.5 cm⁻¹ apodised resolution. Footprint approximately 12 km diameter at nadir. Twice-daily global coverage per satellite. Published H2S retrievals exploit the absorption band near 1270 cm⁻¹; detection threshold in accident-scale studies is on the order of several hundred tonnes in a single overpass column.
- CrIS (Suomi-NPP, NOAA-20): Cross-track Infrared Sounder with spectral coverage including the mid-wave infrared window. Footprint 14 km at nadir, twice-daily revisit. Sensitivity to H2S is comparable to IASI; no peer-reviewed routine H2S product exists yet, but the spectral range overlaps the H2S absorption feature and the instrument is a plausible second source for large-event confirmation.
- EMIT (ISS): Earth Surface Mineral Dust Source Investigation imaging spectrometer on the International Space Station. Covers 380–2500 nm at 7.4 nm spectral sampling and approximately 60 m spatial resolution. H2S has no strong absorption in this range; EMIT is included here as a proxy tool only, potentially mapping lagoon surface conditions and co-located methane plumes rather than H2S directly.
- Next-generation thermal hyperspectral imagers (planned): Instruments such as SBG-TIR (NASA Surface Biology and Geology thermal component) and ESA's LSTM are designed for higher spatial resolution thermal infrared imaging. Neither is operational as of mid-2025. The expectation in the published literature is that finer spatial resolution in the thermal infrared could lower detection thresholds for facility-scale point sources, but quantified H2S detection limits for these systems are not yet peer-reviewed.
Why H2S is hard to see from space
Hydrogen sulphide absorbs thermal infrared radiation in a band centred near 1270 cm⁻¹, overlapping a window region used by sounders like IASI and CrIS. In principle, a sounder measuring upwelling radiance can detect the spectral fingerprint of H2S in the atmospheric column above a source. In practice, several factors conspire against routine monitoring.
The primary problem is the detection floor. IASI retrievals published after the 2019 Rouen chemical plant fire demonstrated that the instrument could identify an H2S plume from an accident releasing an unusually large, concentrated mass of gas. The detection threshold implied by that work is far above what a functioning wastewater treatment plant or produced-water lagoon emits under normal conditions. A large municipal treatment plant might emit tens of kilograms of H2S per hour; the accident-scale events detectable by IASI involve column enhancements orders of magnitude larger. The gap is not a calibration problem. It is a signal-to-noise reality.
A second complication is spectral interference. Water vapour, CO2, and ozone all absorb in overlapping regions. Retrieving H2S requires careful forward-model fitting, and the solution is non-unique when the true H2S column is small. Thermal contrast between the surface and the overlying atmosphere also matters: at night or in humid, isothermal conditions, sensitivity collapses regardless of instrument quality.
What the IASI record actually shows
The most cited demonstration of satellite H2S detection used IASI data from MetOp to track a plume from the Lubrizol chemical plant fire in Rouen, France, in September 2019. The event released an unusually large H2S mass alongside SO2 and other species, and the IASI retrieval produced a spatially coherent plume consistent with trajectory modelling. That result is genuinely significant: it proved the spectral fingerprint is accessible from orbit under the right conditions.
What it did not prove is routine detectability. The Rouen event was an outlier in emission magnitude. Published sensitivity analyses suggest IASI can detect H2S total column enhancements of roughly 0.1–1 ppbv averaged over the instrument footprint, depending on thermal contrast and retrieval assumptions. Converting that to a source rate requires knowledge of mixing layer height and wind speed, both of which carry their own uncertainties. The honest summary is that IASI gives you accident-scale forensics, not a compliance monitoring tool for routine facility emissions.
The retrieval chain, step by step
A thermal infrared H2S retrieval starts with level-1 radiance spectra. An optimal estimation or similar inversion algorithm fits a forward radiative transfer model to the observed spectrum, with H2S column as one of several retrieved state variables. The forward model requires temperature and humidity profiles, surface emissivity, and the spectroscopic line parameters for H2S (primarily from the HITRAN database). The retrieval yields a total column in molecules per square centimetre, with an averaging kernel that describes which altitudes actually contribute to the sensitivity.
For boundary-layer sources like lagoons, the averaging kernel is the critical weakness. IASI is most sensitive to H2S in the mid-troposphere; near-surface sensitivity is limited because the thermal contrast between the warm ground and the cool gas layer is small during the day and the instrument footprint averages over a 12 km circle. A plume that is highly concentrated over a 10-hectare lagoon is diluted by the surrounding clean-air column within the footprint before it reaches the detector. This is why spatial resolution matters so much for the next generation of instruments.
Where the detection floor might fall for next-generation sensors
Planned thermal hyperspectral missions aim for spatial resolutions in the 30–60 m range in the thermal infrared, compared to IASI's 12 km footprint. If an instrument with that resolution also covers the 1270 cm⁻¹ H2S band with adequate spectral sampling and noise-equivalent temperature difference, the geometric dilution problem largely disappears for a point source confined to a lagoon or treatment pond.
The published literature on this is cautious. No peer-reviewed study has demonstrated sub-kilometre H2S detection from orbit. The analogy to methane is instructive but imperfect: methane has a stronger absorption feature, a higher atmospheric background concentration that paradoxically aids retrieval, and has benefited from a decade of targeted instrument development. H2S is at an earlier stage. Facility operators and regulators should treat next-generation thermal hyperspectral H2S detection as a credible near-term prospect, not a current capability.
Practical use cases given today's limits
Given the detection floor, the realistic near-term applications for satellite H2S monitoring are three: post-accident attribution, chronic large-source screening, and multi-signal corroboration.
Post-accident attribution uses the IASI archive, which extends back to 2006 for MetOp-A, to reconstruct plume trajectories and estimate emission magnitude after an industrial release. This is forensic work, not real-time alerting, but it has genuine value in regulatory investigations and insurance assessments.
Chronic large-source screening applies to facilities that are genuinely anomalous in scale: very large produced-water impoundments in arid regions, or kraft pulp mills with known high H2S emission rates. These sources may approach or occasionally exceed the IASI detection threshold under favourable thermal contrast conditions. Systematic screening of the IASI time series over known facility locations can flag episodes worth ground investigation.
Multi-signal corroboration pairs H2S retrieval attempts with co-emitted tracers. SO2 from pulp mills, for example, is detectable at lower thresholds by instruments like TROPOMI. A facility that shows elevated SO2 in the TROPOMI column record is a candidate for H2S co-emission; the two retrievals together constrain source attribution better than either alone. Satellize applies this kind of multi-tracer logic across open constellations, with H2S treated as a conditional layer rather than a primary detection channel.
The honest advice to any operator or regulator considering satellite H2S monitoring is to begin with what is demonstrably possible today, build the archive, and plan for a step-change in capability when next-generation thermal hyperspectral data become available.
Limits worth stating plainly
Cloud cover blocks thermal infrared retrievals entirely. IASI and CrIS both require clear-sky or near-clear-sky conditions for valid H2S retrievals. In humid tropical or temperate climates, cloud-free overpasses over a specific facility may occur only a fraction of the time. Combining MetOp-A, B, and C gives up to six overpasses per day globally, but cloud filtering can reduce usable observations to a handful per month at persistently cloudy sites.
Nighttime thermal contrast can actually improve sensitivity at some sites, because the surface remains warm relative to the cooler nocturnal boundary layer. IASI does observe at night. However, boundary layer stability at night also suppresses vertical mixing, which can concentrate H2S near the surface but also makes the retrieval geometry more complex.
There is no validated routine H2S product from any operational satellite programme as of mid-2025. The retrievals described here are research-grade, requiring specialist expertise to run and interpret. Buyers should not expect a turn-key operational product analogous to the TROPOMI methane or SO2 columns.
Typical figures
| Primary sensor footprint | IASI: ~12 km diameter at nadir; CrIS: ~14 km at nadir |
| Revisit (per satellite) | Twice daily (IASI per MetOp satellite); three MetOp satellites give up to 6 overpasses/day globally |
| Spectral band for H2S retrieval | Thermal infrared, ~1270 cm⁻¹ (7.87 µm); IASI spectral resolution 0.5 cm⁻¹ apodised |
| Detection threshold (accident scale) | Column enhancements of order 0.1–1 ppbv averaged over footprint; routine facility emissions below current threshold |
| Cloud sensitivity | Clear-sky or near-clear-sky required; cloudy pixels rejected in retrieval |
| Archive depth (IASI) | MetOp-A from 2006; MetOp-B from 2012; MetOp-C from 2018 |
| Retrieval latency (research-grade) | Hours to days depending on processing pipeline; no operational near-real-time H2S product exists |
| Next-generation spatial resolution (planned) | 30–60 m thermal infrared for SBG-TIR class instruments; H2S detection limits not yet peer-reviewed |
| Delivery format (analytics output) | GeoTIFF column maps, NetCDF retrievals, PDF forensic reports, GIS-ready shapefiles |
Analytics Satellize can run
| Post-accident plume reconstruction | Optimal estimation retrieval on IASI level-1 radiance spectra, combined with HYSPLIT or similar trajectory modelling | PDF forensic report with mapped plume extent, estimated column mass, and trajectory attribution |
| Facility-level H2S episode screening | Systematic time-series extraction of IASI retrievals over known facility coordinates, flagging statistically anomalous column enhancements | Monthly screening report with episode flags and cloud-coverage statistics per facility |
| Multi-tracer co-emission analysis | Correlation of H2S retrieval attempts with co-emitted SO2 (TROPOMI) and thermal anomaly (VIIRS) signals to improve source attribution confidence | GIS layer with co-emission event markers and confidence scores, updated per overpass |
| Archive baseline characterisation | Statistical analysis of IASI archive (2006 to present) over target facility to establish background variability and identify historical anomalies | Historical baseline dataset in NetCDF with annotated anomaly log |
| Thermal contrast assessment for retrieval quality | Automated screening of surface skin temperature versus boundary-layer air temperature from ERA5 reanalysis to predict retrieval sensitivity per overpass | Retrieval quality flag appended to each overpass record; summary statistics in quarterly 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.