Upper tropospheric humidity retrieval from infrared sounders
Water vapour in the 200–500 hPa layer controls longwave radiative feedback and marks convective outflow. Thermal infrared sounders retrieve it globally, but optically thick cirrus remains a hard limit.
Sensors
- IASI on MetOp-A/B/C: Infrared Atmospheric Sounding Interferometer covering 645–2760 cm⁻¹ at 0.5 cm⁻¹ apodised spectral resolution. Circular footprint of roughly 12 km diameter at nadir. Twice-daily polar coverage per satellite. The 6.3 µm water vapour band provides weighting functions peaking between 200 and 500 hPa, enabling UTH retrievals with a vertical resolution of approximately 2–3 km in the free troposphere under clear or thin-cloud conditions.
- CrIS on Suomi-NPP and NOAA-20: Cross-track Infrared Sounder with 1305 spectral channels across three bands; the longwave band spans 650–1095 cm⁻¹ and the shortwave band 2155–2550 cm⁻¹. Footprint diameter 14 km at nadir, swath 2200 km. Ascending and descending nodes give two overpasses per day. CrIS retrievals of UTH are operationally produced by NOAA STAR and are used in NWP assimilation.
- SEVIRI on Meteosat Third Generation (MTG): Spinning Enhanced Visible and InfraRed Imager on MTG carries a 6.3 µm water vapour channel at 3 km sub-satellite resolution with a 10-minute full-disc repeat cycle. It does not resolve vertical layers the way a hyperspectral sounder does, but the rapid cadence makes it indispensable for tracking convective outflow and tropical cyclone intensification in near real time.
- AIRS on Aqua: Atmospheric Infrared Sounder with 2378 channels spanning 3.7–15.4 µm. Footprint 13.5 km at nadir. AIRS has operated since 2002, giving a two-decade UTH climate record. The NASA AIRS Version 7 Level 3 product provides gridded UTH at 1° resolution with documented uncertainty estimates, making it a reference dataset for climate trend studies.
Why the upper troposphere matters more than its altitude suggests
The 200–500 hPa layer holds a small fraction of the atmosphere's total water vapour mass, yet it exerts a disproportionate influence on the planet's energy budget. Water vapour at these altitudes absorbs and re-emits outgoing longwave radiation in a part of the spectrum where the surface is otherwise nearly transparent. A modest increase in UTH can offset a significant fraction of the radiative cooling that would otherwise escape to space. Climate models consistently identify this as a positive feedback mechanism, but quantifying it from models alone is circular: you need independent observations.
Convective systems are the primary pump that lifts moisture into the upper troposphere. Deep convection overshoots the tropopause, detrain water vapour into the 200–300 hPa layer, and the resulting moist anomalies can persist for days as they are advected by upper-level winds. This makes UTH a tracer of convective history. In tropical cyclone analysis, the distribution of moist and dry air around the vortex is a recognised predictor of intensification rate, because dry-air intrusion at upper levels promotes evaporative cooling that can disrupt the warm core.
What a 6.3 µm channel actually measures
The 6.3 µm absorption band corresponds to the bending mode of the water vapour molecule. At this wavelength, the atmosphere is optically thick enough that upwelling radiance observed at the satellite originates not from the surface but from a layer whose altitude depends on the local humidity profile. Where the upper troposphere is moist, the effective emission level is high and cold, so the brightness temperature is low. Where it is dry, the emission level descends and brightness temperature rises. This is the fundamental signal.
Hyperspectral sounders such as IASI and CrIS resolve hundreds of individual channels across and adjacent to this band. Each channel has a distinct weighting function, the altitude range from which it draws most of its signal. By combining channels with weighting functions peaking at different pressures, retrieval algorithms can decompose the vertical humidity profile into discrete layers. The practical vertical resolution in the upper troposphere is roughly 2–3 km, set by the overlap and width of the weighting functions rather than by any hard instrument limit.
The retrieval is an inverse problem: given observed radiances, infer the humidity profile that would produce them. Operational approaches include one-dimensional variational assimilation (1D-Var), which iterates a first-guess profile from a numerical weather prediction model toward the observations, and physical-statistical methods that use empirical eigenvectors derived from radiosonde climatologies. Both approaches require an accurate forward radiative transfer model. RTTOV (Radiative Transfer for TOVS) is the standard in European operational centres; CRTM is its US counterpart.
Cloud-clearing: the problem that does not disappear
Optically thick cirrus is the dominant source of retrieval failure. When a cirrus layer is optically thick at 6.3 µm, the satellite sees the cloud top rather than the atmospheric column above it. The retrieved humidity then reflects the cirrus emission temperature, not the true UTH. Thin cirrus is worse in a different way: it partially attenuates the signal without triggering a clear cloud flag, introducing a bias that is difficult to detect without independent validation.
Cloud-clearing algorithms attempt to identify contaminated fields of view and either reject them or correct for the cloud contribution using adjacent clear-sky pixels. AIRS pioneered a two-footprint cloud-clearing method that uses spatial variability in nearby scenes to estimate the clear-sky radiance. Even so, in the tropics where deep convection is most frequent, cloud-affected retrievals can exceed 40% of observations on any given day. This is not a solvable problem with current passive infrared technology. It is a known, documented limit that any honest UTH product must flag explicitly.
Geostationary imagers versus polar hyperspectral sounders: a genuine trade-off
SEVIRI on Meteosat Third Generation images the full disc every 10 minutes, which is invaluable for tracking the rapid evolution of convective systems. But a single broadband 6.3 µm channel cannot separate the contributions of different pressure levels. What you observe is a column-integrated brightness temperature weighted toward whichever layer happens to be most opaque at that moment. For climatology and NWP assimilation, this ambiguity is a significant limitation.
Polar hyperspectral sounders resolve the vertical structure but visit any given tropical location only twice per day. For a slow-moving subtropical anticyclone, twice-daily is adequate. For a rapidly intensifying tropical cyclone, the gap between overpasses can span the entire period of maximum intensification. The practical answer is to use both: geostationary imagery for temporal continuity and trend detection, hyperspectral sounders for vertical attribution. MTG's Infrared Sounder (IRS), which will carry 1920 channels in the thermal infrared, is designed to close part of this gap from geostationary orbit, though its operational data stream is not yet fully established.
From brightness temperature to a usable climate or operational product
For climate applications, the quantity of interest is often relative humidity with respect to ice in the 200–500 hPa layer, expressed as a layer-mean value. The GEWEX Water Vapour Assessment has evaluated multiple satellite UTH datasets against radiosonde and GPS radio occultation references. Biases of 5–15% relative humidity are common between products, largely traceable to differences in cloud-clearing, the choice of first-guess profile, and the radiative transfer model used. Users should treat any single UTH product as one realisation within that uncertainty envelope rather than ground truth.
For operational tropical cyclone forecasting, UTH fields from IASI and CrIS are ingested into models such as ECMWF's IFS in near real time, typically with a latency of 2–4 hours from observation to assimilation. The impact on forecast skill in the 24–72 hour range is documented in observing system experiments. The signal is real but modest: UTH from sounders is one input among many, and its value is highest in data-sparse ocean basins where conventional upper-air observations are absent.
Satellize can ingest operational Level 2 UTH retrievals from IASI, CrIS and AIRS alongside SEVIRI water vapour imagery to produce composite diagnostics for specific regions or events.
Validation: what the numbers mean and where they come from
Radiosondes remain the primary validation reference, but they introduce their own biases. Vaisala RS41 sondes, now widely deployed, have a documented dry bias of 2–5% relative humidity in the upper troposphere under certain conditions. Comparisons between satellite UTH and radiosonde humidity must account for the radiosonde's own uncertainty before attributing residuals to the satellite retrieval.
GNSS radio occultation provides a near-independent validation source with no calibration drift, but it retrieves refractivity rather than humidity directly, and separating the humidity and temperature contributions below the tropopause requires an independent temperature profile. The combination of hyperspectral sounder retrievals with radio occultation bending angles in a joint retrieval is an active research area. For now, the honest position is that UTH retrievals are well-characterised in the tropics under clear-sky conditions and poorly characterised under convective cloud, which is precisely where the science question is most interesting.
Typical figures
| Spatial resolution at nadir | IASI: ~12 km footprint diameter; CrIS: ~14 km; AIRS: ~13.5 km; SEVIRI MTG: ~3 km (single channel, no vertical discrimination) |
| Spectral range (water vapour band) | 6.3 µm (≈1595 cm⁻¹ band centre); hyperspectral sounders resolve hundreds of channels across 5.5–8 µm |
| Vertical sensitivity layer | 200–500 hPa (approximately 5.5–12 km altitude); weighting function peak pressure varies by channel |
| Vertical resolution | Approximately 2–3 km in the upper troposphere, set by weighting function overlap |
| Revisit (polar orbiters) | Twice daily per satellite (ascending + descending); three MetOp satellites in constellation extend sampling |
| Revisit (geostationary SEVIRI) | Full-disc every 10 minutes; rapid-scan sector every 2.5 minutes |
| Latency (NWP assimilation) | Typically 2–4 hours from observation to operational Level 2 product |
| Retrieval uncertainty (clear sky) | 5–15% relative humidity bias between products, per GEWEX Water Vapour Assessment |
| Cloud contamination rate (tropics) | Up to 40% of observations affected by optically thick cirrus on any given day |
| Archive depth | AIRS: from September 2002; IASI: from 2007 (MetOp-A); CrIS: from 2012 (Suomi-NPP) |
Analytics Satellize can run
| Layer-mean UTH anomaly maps | 1D-Var retrieval applied to IASI or CrIS Level 1C radiances; anomaly computed against ERA5 climatological baseline | Gridded GIS layer (NetCDF or GeoTIFF) at 0.25° resolution, updated daily with 6-hour latency |
| Convective outflow moisture plume tracking | Lagrangian trajectory analysis driven by ERA5 winds, seeded from moist UTH anomalies identified in SEVIRI water vapour imagery | 72-hour forward trajectory ensemble overlaid on geostationary water vapour animation; PDF report per event |
| Tropical cyclone dry-air intrusion diagnostic | Azimuthal UTH compositing around storm centre using IASI and CrIS swaths; comparison against intensity-change records from IBTrACS | Per-storm summary report with UTH radial profiles at 200 hPa and 400 hPa, flagged for forecast use |
| Cloud-cleared UTH time series for a defined region | AIRS Version 7 Level 3 monthly means with quality flags applied; gap-filled using SEVIRI single-channel brightness temperature regression | Monthly CSV time series with uncertainty bounds; suitable for climate-trend analysis over multi-year windows |
| NWP observation-impact screening | Comparison of IASI/CrIS Level 2 UTH against concurrent NWP first-guess fields to identify systematic departures indicating model bias or retrieval error | Bias monitoring dashboard updated per orbit; alert feed for departures exceeding 10% relative humidity |
| Cirrus contamination frequency climatology | Cloud flag statistics from IASI EUMETSAT Level 2 product aggregated by season and region; cross-referenced against CALIPSO-derived cirrus frequency maps | Annual climatology report with regional maps of retrieval availability, informing sensor selection for specific applications |
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.