Chlorine gas plume detection from industrial accidents
IASI and CrIS sounders have retrieved Cl2 and HCl columns above large industrial chlorine releases, but rapid atmospheric dilution means the detection window is narrow and the minimum detectable mass is high. This page sets out what the public record shows, where the physics limits retrieval, and how geostationary infrared sounders could close the timing gap.
Sensors
- IASI (MetOp-A/B/C): Hyperspectral thermal infrared sounder, 12 km circular footprint at nadir, ~8461 spectral channels across 645–2760 cm⁻¹. Retrieves HCl and Cl2 via absorption features near 2800 cm⁻¹ and 550 cm⁻¹ respectively. Each MetOp satellite provides two overpasses per day at any given location; with three satellites the combined revisit is roughly six overpasses per day globally, though local timing is fixed by orbital mechanics and cannot be commanded.
- CrIS (Suomi-NPP / NOAA-20): Cross-track Infrared Sounder, 14 km effective footprint, covering 650–2550 cm⁻¹ in three bands. Spectral resolution up to 0.625 cm⁻¹ in the longwave band. Capable in principle of HCl retrieval; published accident-response retrievals are fewer than for IASI, partly because the IASI spectral database and retrieval heritage are more mature for halogenated species.
- SEVIRI (Meteosat-9/10/11): Spinning Enhanced Visible and InfraRed Imager on geostationary Meteosat. Covers Europe and Africa with a 15-minute repeat cycle and a 3 km pixel at sub-satellite point in thermal infrared channels. SEVIRI lacks the spectral resolution to retrieve Cl2 or HCl columns directly, but its thermal channels can track associated fire or explosion thermal signatures and plume dispersion patterns that provide timing context for sounder retrievals.
- MTG-IRS (Meteosat Third Generation, planned): The Infrared Sounder on MTG, expected to begin operations in the mid-2020s, will provide hyperspectral infrared sounding from geostationary orbit over Europe and Africa with roughly 50 km footprint and sub-hourly repeat. This is the architecture most likely to close the timing gap for accident-response chlorine monitoring, though its sensitivity to low-mass Cl2 releases at that footprint scale remains to be demonstrated operationally.
What the public retrieval record actually shows
The strongest published evidence for satellite-retrieved chlorine from industrial accidents comes from IASI. The 2015 explosion at a chemical warehouse in Tianjin, China, produced an HCl signal detectable in IASI radiances within roughly 24 hours of the event. Researchers at LATMOS and partner institutions used optimal-estimation retrievals against IASI level-1C spectra to derive HCl column amounts above the plume footprint, demonstrating that the technique works at real accident scale rather than just in atmospheric chemistry contexts.
Cl2 itself is harder. Its absorption cross-section in the mid-infrared is weaker than HCl, and it reacts rapidly with moisture and atmospheric constituents, converting partly to HCl and HOCl within minutes to hours depending on humidity and sunlight. In practice, satellite retrievals above accident sites tend to capture the HCl daughter product rather than the parent Cl2 molecule. That distinction matters for emergency responders: the satellite column is a proxy for the release, not a direct measure of the hazardous species still present at ground level.
The detection window is shorter than it looks on a timeline
A large chlorine release from a rail tanker or industrial storage vessel can put hundreds of tonnes into the lower troposphere within minutes. IASI's 12 km footprint means the retrieved column is an average over roughly 113 km². If the plume disperses to fill that footprint before the satellite overpass, the column density may still be detectable. If the release is small or the wind is strong, dilution can drop the column below the instrument's noise floor before any overpass occurs.
Published minimum detectable HCl columns for IASI are on the order of 1–5 × 10¹⁵ molecules cm⁻², depending on thermal contrast between the plume and the surface below. Converting that to a mass threshold is geometry-dependent, but rough estimates from published studies suggest that releases of several hundred tonnes of HCl-equivalent may be needed for confident retrieval, particularly in cloudy or low-thermal-contrast conditions. Smaller accidents, which are far more common, will not appear in IASI data at all.
The orbital timing problem is stark. A polar-orbiting sounder passes a given mid-latitude point roughly twice per day per satellite. With three MetOp satellites, the best-case gap between overpasses is around four hours; the worst case approaches twelve. A plume that disperses to below-detection concentrations in three hours will simply not be seen. This is not a processing failure; it is orbital geometry.
Cloud cover and thermal contrast: the two silent veto conditions
IASI retrieves trace gases by measuring the difference between upwelling thermal radiation and what the atmosphere would emit without the absorbing species. Thick cloud below the plume removes the thermal contrast needed for retrieval. Mid-latitude industrial regions are frequently cloudy, and a major accident in winter over northern Europe or North America could easily coincide with overcast conditions that render the sounder blind to the plume entirely.
Thermal contrast also depends on the temperature difference between the surface and the overlying air. At night, or in conditions of strong temperature inversion, the contrast can be near zero even in clear sky. Some retrievals are actually easier at night over warm surfaces, because the surface radiates strongly and the cooler plume above absorbs against it. The geometry is case-specific and cannot be assumed to be favourable.
What geostationary sounders change, and what they do not
The fundamental appeal of a geostationary hyperspectral sounder for accident monitoring is continuous temporal coverage. MTG-IRS, once operational, would observe the same European scene every 30 to 60 minutes rather than every four to twelve hours. For a chlorine release, that could mean catching the plume while it is still concentrated enough to retrieve, rather than arriving after dilution has erased the signal.
The trade-off is spatial resolution. MTG-IRS will have a footprint of roughly 50 km, compared to IASI's 12 km. A plume that has not yet spread to fill that footprint will be further diluted in the retrieved column. The net effect on minimum detectable mass is uncertain and will depend on the specific accident scenario. Geostationary sounders improve the timing odds substantially; they do not solve the sensitivity problem.
SEVIRI, already in geostationary orbit, cannot retrieve Cl2 or HCl directly. Its value in accident response is contextual: tracking the thermal signature of an explosion or fire, monitoring cloud cover to predict whether a sounder overpass will be useful, and providing wind-field context for dispersion modelling.
Combining satellite data with dispersion models for emergency response
The practical workflow for accident response does not rely on satellite detection alone. Ground sensors, meteorological observations and atmospheric dispersion models such as HYSPLIT or NAME are typically the primary tools for emergency services. Satellite retrievals serve as an independent check on model outputs: if IASI detects an HCl column downwind of a reported release site, that confirms the release occurred and provides an independent constraint on the total mass emitted.
This mass-estimation function is where satellite data adds clearest value. Emergency responders often do not know how much chlorine was released from a damaged vessel or facility. An IASI-derived HCl column, combined with a dispersion model run backwards from the observation, can bracket the source term. Published post-event analyses of the Tianjin incident demonstrated this approach. The uncertainty is large, typically a factor of two to three in source mass, but that is still more informative than no satellite constraint at all.
Satellize can run retrieval and dispersion-inversion workflows on IASI and CrIS level-1 radiance data for government clients with emergency-response mandates, delivering source-term estimates and plume-trajectory products within hours of a confirmed overpass. The workflow draws on the same optimal-estimation retrieval methods used in the published literature, applied operationally rather than retrospectively.
Honest limits and the cases satellite data cannot cover
Any buyer considering satellite monitoring for chlorine accident response should be clear about what the technology cannot do. It cannot provide real-time detection. It cannot detect small releases below a few hundred tonnes of HCl-equivalent at IASI footprint scale. It cannot see through cloud. It cannot replace ground-based air quality monitoring or emergency response networks for immediate public safety decisions.
What it can do is provide post-event verification, independent source-term estimation, and plume-extent documentation for regulatory, legal and insurance purposes. For a government building a national chemical-incident response capability, satellite data is one layer in a multi-sensor architecture, not the primary sensor. The value proposition is strongest where ground monitoring is sparse, where the release is large, and where the sky is clear at the time of an overpass.
Typical figures
| Spatial footprint (IASI) | 12 km diameter circle at nadir; increases off-nadir |
| Spatial footprint (CrIS) | 14 km effective field of view at nadir |
| Revisit (three MetOp satellites combined) | Up to 6 overpasses per day globally; local gap 4–12 hours depending on latitude |
| Spectral range (IASI) | 645–2760 cm⁻¹ (~3.6–15.5 µm), 8461 channels |
| Key absorption features | HCl: ~2800 cm⁻¹ (R-branch); Cl2: ~550 cm⁻¹ (weaker, lower in atmosphere) |
| Minimum detectable HCl column (IASI, clear sky, good thermal contrast) | ~1–5 × 10¹⁵ molecules cm⁻² (published range; scene-dependent) |
| Indicative minimum source mass for detection | Hundreds of tonnes HCl-equivalent; smaller releases unlikely to be retrieved |
| Cloud sensitivity | Thick cloud below plume renders retrieval impossible; partial cloud degrades sensitivity |
| Archive depth (IASI) | MetOp-A from 2006; MetOp-B from 2012; MetOp-C from 2018 |
| Geostationary sounder outlook (MTG-IRS) | ~50 km footprint, sub-hourly repeat over Europe/Africa; operational mid-2020s |
Analytics Satellize can run
| HCl column retrieval above accident site | Optimal-estimation retrieval against IASI or CrIS level-1C radiances, using published spectroscopic line parameters from HITRAN | Georeferenced column-amount map (molecules cm⁻²) with per-pixel uncertainty, delivered as GeoTIFF and CSV within hours of overpass |
| Source-term mass estimate | Atmospheric dispersion inversion (HYSPLIT or equivalent) constrained by retrieved column; published approach from post-Tianjin literature | PDF report with source-mass range, uncertainty bounds and sensitivity notes; suitable for regulatory or legal use |
| Plume-extent and trajectory reconstruction | Forward dispersion modelling from inferred source term, validated against retrieved column spatial pattern | Animated GIS layer showing modelled plume footprint at hourly intervals post-release |
| Overpass timing alert | Orbital propagation against reported accident location and time; cloud-cover forecast from ECMWF or equivalent | Email or API alert listing next IASI/CrIS overpass time, predicted cloud fraction and estimated detection probability |
| Archive plume search for past incidents | Retrospective screening of IASI level-2 HCl products and level-1C radiances against reported accident dates and locations | Summary report of detections or non-detections with explanatory notes on orbital and meteorological conditions |
| Thermal context layer from SEVIRI | Brightness-temperature anomaly detection in SEVIRI 10.8 µm channel to identify explosion or fire signature and track cloud cover | 15-minute time-series GeoTIFF stack covering 24 hours around reported incident time |
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.