Surface soil moisture retrieval for drought monitoring
L-band microwave radiometry measures the dielectric contrast between wet and dry soil to retrieve volumetric moisture in the top 5 cm, globally. SMOS and SMAP are the dedicated missions; Sentinel-1 SAR fills the resolution gap they cannot close.
Sensors
- SMAP L-band Radiometer (NASA): Passive radiometer at 1.41 GHz (L-band). Native spatial resolution approximately 36–40 km after antenna footprint averaging. Global revisit every 2–3 days. Retrieves volumetric soil moisture to a target accuracy of 0.04 m³/m³ (1-sigma) over low-to-moderate vegetation. Radiometer failed in 2015; active radar no longer operational, but radiometer continues.
- SMOS MIRAS (ESA): Synthetic aperture interferometric radiometer at 1.4 GHz. Spatial resolution approximately 35–50 km depending on incidence angle. Global revisit roughly 3 days. First dedicated L-band soil moisture mission, launched 2009. Susceptible to radio-frequency interference (RFI) from terrestrial transmitters, particularly over Europe and Asia.
- Sentinel-1 SAR (ESA, C-band): C-band (5.405 GHz) synthetic aperture radar providing backscatter sensitive to surface roughness, soil moisture and vegetation water content. Spatial resolution 10–20 m in Interferometric Wide Swath mode. Revisit 6–12 days per satellite over most land areas. Does not penetrate canopies as deeply as L-band; change-detection approaches are needed to isolate moisture signal from roughness.
- ASCAT (Metop, EUMETSAT): C-band scatterometer at 5.255 GHz on Metop-A/B/C. Provides a relative soil moisture index (0–100%) rather than volumetric m³/m³. Spatial sampling 25–50 km. Daily to sub-daily global coverage. Long archive from 2006 enables climatological drought baselines.
Why 1.4 GHz and not something more convenient
Water has a dielectric constant near 80 at microwave frequencies; dry mineral soil sits around 3–5. That contrast is the entire basis of L-band soil moisture retrieval. At 1.4 GHz the emitted thermal microwave radiation originates from roughly the top 5 cm of soil, a depth that matters agronomically because it governs germination, evaporation and the early stress signal that precedes visible crop damage by days to weeks.
Higher frequencies, say C-band at 5 GHz or X-band at 10 GHz, see a shallower layer and are far more sensitive to surface roughness and the water held in plant leaves rather than the soil beneath. L-band is also the only frequency window where passive radiometry can penetrate moderate vegetation canopies with a tractable correction. The 1.400–1.427 GHz allocation is a protected radio-astronomy band, which is why it was chosen for SMOS and SMAP. That protection is imperfect in practice.
The radio-frequency interference problem nobody advertises
Despite international protection, L-band is contaminated by RFI from radar systems, television satellites uplinks and other terrestrial transmitters across large parts of Europe, the Middle East, South and East Asia. Both SMOS and SMAP carry onboard RFI detection and filtering algorithms, but flagged pixels are simply discarded. Over densely populated or militarily active regions this can remove 20–40% of retrievals in a single overpass, creating systematic data gaps in exactly the areas where agricultural drought monitoring is most economically consequential.
The practical consequence for an analyst is that SMOS/SMAP drought products over, say, northern India or the North China Plain carry more uncertainty than the headline accuracy figure suggests. Validation studies published in Remote Sensing and related journals consistently show that RFI-flagged retrievals, when they survive filtering, have root-mean-square errors roughly double those over clean-spectrum regions. Users should always inspect the RFI quality flag layer before drawing conclusions.
Vegetation optical depth: the correction that makes or breaks a retrieval
Vegetation attenuates and emits microwave radiation. The standard tau-omega model, used in both SMAP and SMOS retrieval algorithms, corrects for this by estimating vegetation optical depth (VOD) from ancillary data, typically MODIS-derived vegetation indices or a climatological look-up table. Over sparse grassland or bare soil the correction is small and the retrieval is reliable. Over maize or soy canopies at peak biomass, VOD can exceed 0.6 and the correction dominates the uncertainty budget.
Dense forest is effectively opaque at L-band. Retrievals in tropical forest regions are physically meaningless and are masked in operational products. This is not a failure of the mission; it is an honest limit of the physics. The SMAP Level-2 passive product documentation from NASA explicitly flags high-VOD pixels as having degraded accuracy. Any drought monitoring system that ignores this masking will produce spurious wetness anomalies under forest canopies.
Coarse pixels over complex landscapes: what 36 km actually means
A single SMAP or SMOS pixel covers roughly 1,000–1,600 km² of land surface. In a heterogeneous agricultural landscape with irrigated fields, rain-fed plots, fallow land and urban edges all within that footprint, the retrieved value is a weighted spatial average that may not correspond to any single field's actual moisture state. A farmer's 10-hectare plot is invisible.
Two approaches address this. The first is downscaling: using the spatial detail in Sentinel-1 C-band backscatter, or in thermal infrared evapotranspiration proxies, to disaggregate the coarse L-band signal to 1 km or finer. The SMAP Enhanced product already applies a Backus-Gilbert interpolation to sharpen the native footprint to approximately 9 km, but this recovers spatial detail only where the signal-to-noise ratio permits. The second approach is to use Sentinel-1 directly as a soil moisture sensor, accepting that the C-band retrieval is shallower and roughness-contaminated but gaining 10–20 m spatial resolution. Neither method is universally superior; the right choice depends on canopy density, terrain and the decision the client is actually trying to make.
Building a drought signal from moisture anomalies
A single moisture retrieval is not a drought indicator. Drought is a departure from climatological expectation. The operational approach is to compute a soil moisture anomaly: the difference between today's retrieval and the long-term percentile for the same day of year, derived from the SMOS archive (2010-present) or the SMAP archive (2015-present). ASCAT extends the usable climatology back to 2006 using its relative surface soil moisture index.
Percentile-based anomaly products, such as the SMAP Soil Moisture Active Passive Drought Monitoring product distributed through NASA Earthdata, express current conditions as a percentile rank. A value below the 20th percentile sustained over several weeks is a recognised early-warning threshold used by food security agencies. The latency matters here: SMAP Level-3 products are typically available within 24 hours of acquisition, which is fast enough to inform weekly agricultural bulletins but not real-time irrigation decisions.
Satellize runs anomaly compositing and threshold alerting on open L-band and C-band data streams. The crop-estimation work for the Kingdom of Tonga demonstrated how moisture anomalies integrated with vegetation indices can separate stress-driven yield reduction from purely phenological variation, a methodology transferable to any smallholder agricultural context.
Honest limits and what to do about them
Summarising the constraints plainly: L-band radiometry sees only the top 5 cm, misses root-zone moisture entirely without a land-surface model, loses accuracy under dense canopies, is blind in heavy RFI environments, and cannot resolve fields smaller than several kilometres without downscaling that introduces its own assumptions. C-band SAR resolves fields but conflates moisture with roughness and sees an even shallower soil layer. No single sensor solves all of this.
The productive framing for a government or agribusiness client is sensor fusion: SMAP or SMOS anomalies for regional drought classification, Sentinel-1 change detection for field-scale moisture dynamics, and ASCAT for long-baseline climatology. Where in-situ validation networks exist, even a sparse one, they dramatically reduce the uncertainty of satellite-derived anomalies. Commissioning a soil moisture monitoring system without budgeting for at least minimal ground truth is a common and expensive mistake.
Typical figures
| Native spatial resolution (L-band passive) | 36–40 km (SMAP); 35–50 km (SMOS, angle-dependent) |
| Enhanced / downscaled resolution | ~9 km (SMAP Enhanced product); 1 km with Sentinel-1 fusion |
| Sentinel-1 SAR resolution | 10–20 m (IW mode, C-band) |
| Revisit period | 2–3 days (SMAP); ~3 days (SMOS); 6–12 days per satellite (Sentinel-1) |
| Sensing frequency | 1.41 GHz L-band (SMAP, SMOS); 5.405 GHz C-band (Sentinel-1); 5.255 GHz C-band (ASCAT) |
| Sensing depth | ~0–5 cm (L-band); ~0–2 cm (C-band) |
| Target retrieval accuracy (L-band) | 0.04 m³/m³ volumetric (1-sigma) over low-to-moderate vegetation, per SMAP mission specification |
| Product latency | ~24 hours (SMAP Level-3); 3–5 days (SMOS NRT); ~1 hour (ASCAT NRT via EUMETSAT) |
| Archive depth | SMOS from 2010; SMAP from April 2015; ASCAT from 2006; Sentinel-1 from 2014 |
| Coverage | Global land (SMAP, SMOS); global land excluding dense forest and high-RFI zones |
Analytics Satellize can run
| Regional drought anomaly map | Percentile-rank anomaly computed against SMAP or SMOS climatology for same day-of-year; thresholded at 20th and 10th percentile | Weekly GIS layer (GeoTIFF or GeoJSON) with percentile rank and anomaly class per pixel |
| Field-scale moisture change detection | Sentinel-1 C-band backscatter change relative to dry-season baseline; roughness normalisation using multi-temporal stack | Bi-weekly raster at 20 m resolution flagging significant moisture increase or decline per agricultural parcel |
| Drought early-warning alert feed | Rolling 30-day anomaly persistence check against SMAP Level-3; alert triggered when anomaly below 20th percentile for 3 consecutive overpasses | Automated alert (API push or email) with affected area polygon and severity class |
| Long-baseline climatological drought index | ASCAT relative soil moisture index (2006-present) harmonised with SMAP anomalies to extend the record; z-score standardisation | Annual report with trend analysis and return-period estimates for drought severity classes |
| Downscaled soil moisture at 1 km | Disaggregation of SMAP 36 km retrievals using Sentinel-1 backscatter and MODIS land surface temperature as spatial predictors; published DisPATCh or similar regression approach | On-demand GeoTIFF at 1 km, with uncertainty layer derived from residual variance of the downscaling model |
| Crop stress attribution report | Integration of soil moisture anomaly with NDVI anomaly (Sentinel-2 or Landsat) to separate water-stress-driven greenness decline from other causes | Seasonal bulletin identifying zones where moisture deficit is the primary driver of vegetation stress, with confidence classification |
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.