Tropospheric humidity retrieval from microwave sounders
Passive microwave sounders near 183 GHz retrieve tropospheric humidity profiles through cloud cover that defeats every optical instrument. This page explains the physics, the key instruments, and what the data honestly can and cannot resolve.
Sensors
- ATMS (Suomi-NPP / NOAA-20): Advanced Technology Microwave Sounder. Channels 18-22 straddle the 183.31 GHz water-vapour absorption line at offsets of ±1, ±1.8, ±3, ±4.5 and ±7 GHz, each sensitive to a different pressure layer from roughly 300 hPa to the lower troposphere. Nadir footprint approximately 15.8 km at 183 GHz channels; swath 2,300 km gives near-daily global coverage per satellite, twice-daily combined.
- MHS (Metop-A / B / C): Microwave Humidity Sounder. Five channels: 89, 157, 183.31±1, 183.31±3 and 190.31 GHz. Nadir footprint 16 km; swath 2,250 km. Three Metop satellites in staggered sun-synchronous orbits provide roughly 3-4 overpasses per day at mid-latitudes, improving sampling for numerical weather prediction assimilation.
- AMSU-B (NOAA-15 / 16 / 17): The predecessor to MHS, with the same five-channel 183 GHz configuration. NOAA-15 remains partially operational; the archive from the late 1990s onward provides a multi-decade climatological record of upper-tropospheric humidity, despite known calibration drift issues that require inter-satellite correction.
- MWHS-2 (FY-3D / FY-3E): China's Microwave Humidity Sounder-2 adds 8 channels on the 183 GHz line plus channels near 118 GHz for temperature context. Footprint approximately 15 km; FY-3E operates in an early-morning orbit that fills a sampling gap left by the NOAA and Metop afternoon/morning pair, making it increasingly important for global NWP.
Why 183 GHz and not something simpler
Water vapour has a strong rotational absorption line centred at 183.31 GHz. Place a radiometer channel close to that centre frequency and the atmosphere becomes opaque high up, around 300-400 hPa, because even modest amounts of moisture absorb the upwelling radiation before it reaches the sensor. Shift the channel further from line centre, say to ±7 GHz, and the atmosphere is more transparent, so the sensor sees deeper, down toward 700-850 hPa. That frequency-dependent opacity is the entire basis of the weighting-function approach: each channel samples a different layer, and inverting the set of brightness temperatures together produces a humidity profile through the column.
The practical advantage over infrared sounders is substantial. Microwave radiation at these frequencies passes through non-precipitating cloud with relatively little scattering, meaning retrievals remain possible in overcast conditions that completely block thermal infrared instruments. The penalty is spatial resolution. ATMS channels near 183 GHz have a nadir footprint of about 15-16 km, compared with sub-kilometre pixels from infrared imagers. For synoptic-scale moisture analysis and numerical weather prediction, that trade is entirely acceptable. For resolving a narrow atmospheric river or a convective boundary, it is a genuine limitation worth stating plainly.
Brightness temperature to humidity: the inversion problem
A sounder measures top-of-atmosphere brightness temperature, TB, in each channel. TB is the integral of the Planck function weighted by the channel's weighting function across all pressure levels, plus surface emission attenuated by the intervening atmosphere. Retrieving the humidity profile from that integral is an ill-posed inverse problem: many atmospheric states can produce the same observed TB. Operational centres handle this through one-dimensional variational (1D-Var) retrieval, which finds the humidity profile that minimises the difference between observed and modelled TB, constrained by a background state from a short-range forecast. The EUMETSAT Satellite Application Facility for Numerical Weather Prediction (NWP SAF) publishes the AAPP and RTTOV software packages that implement this approach for MHS and ATMS.
The retrieved product is typically specific humidity or relative humidity at a set of standard pressure levels, commonly 1000, 925, 850, 700, 500, 400 and 300 hPa, though the vertical resolution is far coarser than the level spacing implies. The degrees of freedom for signal in a five-channel humidity sounder are roughly 2-3, meaning the retrieval can genuinely resolve about two or three independent layers, not seven. Presenting output on a fine pressure grid gives false precision; the honest interpretation is broad-layer moisture content.
What the archive actually covers, and where the gaps are
AMSU-B flew on NOAA-15 from 1998, giving a continuous microwave humidity record now exceeding 25 years when combined with MHS and ATMS. That archive underpins published climatologies of upper-tropospheric humidity, tropical convective moistening, and long-term trends in atmospheric water content. The NOAA CLASS archive and EUMETSAT's Earth Observation Portal both hold level-1 brightness temperatures; NOAA's operational ATMS retrievals are distributed through the Joint Polar Satellite System (JPSS) data portal.
Gaps and caveats deserve equal billing. Over heavy precipitation, ice scattering at 183 GHz depresses TB in ways that mimic dry air, contaminating retrievals inside deep convective cores. Surface emissivity over sea ice and arid land is spectrally variable and poorly constrained, introducing biases in the lowest-troposphere channels. At high latitudes in winter, the surface is cold and the atmospheric signal is weak, degrading retrieval skill. And because each sounder footprint averages 15-16 km of atmosphere, sub-mesoscale moisture structures are simply invisible.
Assimilation versus standalone retrieval: two different use cases
Numerical weather prediction centres, including ECMWF, NOAA/NCEP and the UK Met Office, assimilate ATMS and MHS radiances directly rather than using retrieved humidity profiles. Assimilating TB avoids propagating retrieval errors into the model; the model's own radiative transfer code computes what the sounder should have seen and adjusts the model state to match. This approach has measurably improved 5-day forecast skill in the Southern Hemisphere, where conventional radiosonde coverage is sparse.
Standalone retrieved profiles serve a different community: climate monitoring, reanalysis validation, and applications where an analyst needs a gridded humidity field without running a full NWP system. ERA5, the ECMWF reanalysis, blends sounder radiances with all other available observations back to 1940 and provides hourly humidity fields at 31 km resolution. For many commercial and government applications, ERA5 is the most practical entry point to sounder-derived humidity, though it carries its own biases and should not be treated as ground truth.
Operational applications: what buyers actually do with this
Tropical cyclone intensity forecasting depends heavily on upper-tropospheric humidity fields from sounders. A dry mid-level environment inhibits intensification; sounder data quantifies that dryness in near-real time over ocean where no other observations exist. Aviation route planning uses sounder-derived humidity to anticipate clear-air turbulence and icing layers. Hydrological forecasting in data-sparse basins uses sounder moisture profiles as boundary conditions for precipitation models.
Satellize processes ATMS and MHS brightness temperatures through published retrieval schemes to produce layer-mean humidity fields and anomaly maps for clients whose operational decisions depend on moisture state over regions without radiosonde networks. The Tonga crop-estimation programme, for instance, required atmospheric correction inputs over a maritime domain where conventional upper-air data is essentially absent. Sounder humidity fields filled that gap directly. For clients wanting to discuss a specific basin or forecast application, the right starting point is a scoping call with our atmospheric team, not a generic data subscription.
Combining sounders with complementary observations
No single instrument type closes all the gaps. GNSS radio occultation provides high-vertical-resolution refractivity profiles that constrain the temperature and moisture structure independently of sounder assumptions, and the two are routinely combined in NWP assimilation. Infrared hyperspectral sounders such as IASI on Metop and CrIS on NOAA-20 offer finer vertical resolution in clear sky, complementing the microwave sounders' cloud-penetrating capability. Geostationary water-vapour imagers at 6.2 and 7.3 µm provide high temporal resolution for tracking moisture advection between polar-orbiter overpasses, though they too are limited to clear and thin-cloud conditions.
The practical message for a buyer is that a humidity monitoring system built on a single sensor type will have predictable blind spots. Microwave sounders are the most weather-independent option for bulk moisture profiling, but they should be treated as one layer of a multi-source analysis rather than a complete solution.
Typical figures
| Nadir footprint (183 GHz channels) | ~15-16 km (ATMS, MHS, MWHS-2) |
| Swath width | 2,250-2,300 km (ATMS and MHS); full global coverage per satellite in ~12 h |
| Revisit (combined ATMS + MHS fleet) | Roughly 4-6 overpasses per day at mid-latitudes; near-continuous at high latitudes |
| Key frequency channels | 183.31 ± 1, ±1.8, ±3, ±4.5, ±7 GHz (humidity); 89 and 157 GHz (window) |
| Vertical layers resolved | ~2-3 independent layers (degrees of freedom for signal); output commonly gridded to 300-1000 hPa |
| Brightness temperature noise (NEDT) | ~0.5-1.0 K at 183 GHz channels (instrument-dependent; ATMS specification ~0.6 K) |
| Latency (operational NWP products) | ~1-3 h from observation to NOAA/EUMETSAT level-2 dissemination |
| Archive depth | AMSU-B from 1998; MHS from 2006; ATMS from 2011; ERA5 reanalysis back to 1940 |
| Primary retrieval limitation | Degraded or failed retrieval inside heavy precipitation (ice scattering); surface emissivity uncertainty over sea ice and desert |
| Delivery formats (operational) | HDF5, NetCDF-4 (NOAA JPSS); BUFR (EUMETSAT NWP dissemination); NetCDF via Copernicus Climate Data Store (ERA5) |
Analytics Satellize can run
| Layer-mean specific humidity anomaly maps | 1D-Var retrieval against ERA5 background; anomaly computed relative to 2000-2020 climatological baseline | Gridded NetCDF layer at 300-500 hPa and 500-700 hPa, updated per overpass, with anomaly magnitude flagged against 1-sigma and 2-sigma thresholds |
| Atmospheric river detection and axis tracking | Integrated vapour transport computed from sounder humidity profiles combined with ERA5 wind fields; IVT threshold 250 kg/m/s applied per published Guan & Waliser methodology | Polygon shapefile of AR corridor with IVT magnitude, updated every 6 h; alert triggered when corridor intersects a defined catchment boundary |
| Tropical cyclone environment moisture assessment | Mid-level (500-700 hPa) relative humidity field extracted within 500 km of NHC or JTWC storm centre; dry-air intrusion index computed per overpass | Per-storm PDF report with humidity curtain cross-sections and time series of mean mid-level RH, issued within 3 h of each overpass |
| Upper-tropospheric humidity trend analysis | Inter-calibrated AMSU-B / MHS / ATMS brightness temperature time series at 183 ± 1 GHz; linear trend and breakpoint detection over user-defined region and period | Annual trend report with uncertainty bounds, seasonal decomposition, and comparison against ERA5 reanalysis for the same region |
| Atmospheric correction input for surface retrievals | Columnar precipitable water and layer humidity from sounder retrieval, ingested as ancillary input to optical or thermal surface-reflectance correction pipelines | Ancillary humidity profile file in CF-convention NetCDF, co-registered to client image footprint, with quality flag per pixel |
| Radiosonde gap-fill for data-sparse regions | Sounder-retrieved humidity profiles at standard pressure levels, bias-corrected against available radiosonde climatology for the region; output formatted to match WMO TEMP message structure | Pseudo-sounding dataset in NetCDF and CSV, covering a defined domain, with per-level retrieval uncertainty estimates |
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.