Estimating Soil Moisture with SAR: Why Radar Sees What Optical Can't
Quick Answer: Radar backscatter is sensitive to soil moisture because water dramatically changes soil's dielectric constant. Dry soil (dielectric ~3-5) reflects little radar energy; wet soil (dielectric ~15-30) reflects significantly more, appearing 3-8 dB brighter in SAR images. Sentinel-1 C-band can detect soil moisture changes in bare or sparsely vegetated fields, but dense vegetation canopy obscures the soil signal. The relationship is strongest for VV polarization on bare soil. Operational soil moisture products (like SMAP and Sentinel-1-based retrievals) combine radar with models to provide global estimates.
A farmer in southern Spain once asked me why his SAR images looked completely different after rain. Fields that had been medium gray turned almost white. Roads barely changed. Water bodies stayed black. He'd stumbled onto one of radar's most useful properties — its sensitivity to soil moisture — without knowing the physics behind it.
The Physics: Dielectric Constant
The connection between radar and soil moisture runs through a material property called the dielectric constant (ε). The dielectric constant describes how a material interacts with electromagnetic waves — specifically, how much it polarizes in response to an electric field.
Water has a very high dielectric constant (~80 at microwave frequencies). Dry soil minerals have a low dielectric constant (~3-5). When you add water to soil, the mixture's dielectric constant increases dramatically:
| Volumetric Soil Moisture | Dielectric Constant (approx.) | SAR Response |
|---|---|---|
| 5% (very dry) | ~4 | Low backscatter |
| 15% (moderate) | ~10 | Medium backscatter |
| 25% (moist) | ~18 | High backscatter |
| 35% (wet) | ~25 | Very high backscatter |
| 40%+ (saturated) | ~30+ | Maximum backscatter |
This relationship is roughly linear in the 5-35% moisture range, which conveniently covers the range most relevant to agriculture. A 10% increase in volumetric soil moisture typically produces a 2-4 dB increase in C-band backscatter — easily detectable by Sentinel-1.
Why VV Polarization Works Best for Soil Moisture
Sentinel-1 transmits and receives in two polarizations: VV (vertical transmit, vertical receive) and VH (vertical transmit, horizontal receive).
For bare soil moisture estimation, VV polarization is preferred because:
- VV backscatter from bare soil is dominated by surface scattering, which is directly related to dielectric constant (moisture) and surface roughness
- VH backscatter involves depolarization, which is more influenced by volume scattering (vegetation canopy) and less directly tied to soil moisture
The practical sensitivity: VV backscatter changes by approximately 0.2-0.4 dB per 1% change in volumetric soil moisture on bare, smooth soil. With Sentinel-1's radiometric accuracy of about 1 dB, moisture changes of 3-5% are reliably detectable.
The Surface Roughness Problem
Here's the complication that keeps soil moisture retrieval from being simple. Backscatter depends on two factors simultaneously: moisture and surface roughness.
A rough, dry field can produce the same backscatter as a smooth, wet field. Without knowing the roughness, you can't isolate the moisture signal.
Approaches to handle this:
Multi-temporal change detection: If surface roughness doesn't change between two dates (no plowing, no erosion), the backscatter difference is attributable to moisture change. This works well for monitoring moisture changes over time but can't give absolute moisture values.
Co-polarization ratio: The ratio VV/VH is less sensitive to roughness than either polarization alone, providing a more robust (if less sensitive) moisture indicator.
Model inversion: Physical backscatter models (like the Integral Equation Model or Oh model) relate backscatter to both roughness and moisture. With multi-polarization or multi-angle data, both parameters can be estimated simultaneously.
The Vegetation Problem
Vegetation above the soil attenuates the soil-surface signal and adds its own volume scattering contribution. As crop canopy develops through the growing season:
- Early season (bare soil or seedlings): SAR sees the soil directly. Moisture estimation works well.
- Mid-season (partial canopy): Vegetation partially obscures the soil. Estimation accuracy degrades.
- Full canopy (closed crop): Vegetation dominates the signal. Soil moisture estimation becomes unreliable at C-band.
The transition point depends on crop type and radar wavelength. For C-band over wheat, the soil signal becomes negligible once the crop height exceeds about 30-40 cm (roughly NDVI > 0.6). For corn, the canopy effect is even stronger due to the larger plant structure.
L-band radar (23.5 cm wavelength, like NISAR) penetrates deeper into vegetation canopy, maintaining soil moisture sensitivity at higher biomass levels. This is one of the key motivations for L-band missions.
Operational Soil Moisture Products
Several global soil moisture products are available:
Sentinel-1 Based Retrievals
ESA and various research groups produce soil moisture maps from Sentinel-1 at approximately 1 km resolution. The Copernicus Global Land Service provides a Surface Soil Moisture (SSM) product updated every 1-3 days (depending on Sentinel-1 overpasses).
The approach typically uses change detection relative to dry and wet reference conditions, normalized to produce a relative moisture index (0-100%) rather than absolute volumetric moisture.
SMAP
NASA's Soil Moisture Active Passive mission (launched 2015) provides global soil moisture at ~36 km resolution using an L-band radiometer. While the active radar component failed shortly after launch, the passive radiometer continues to provide the most reliable global soil moisture dataset.
SMAP's coarse resolution limits its utility for field-scale agriculture, but it excels at regional drought monitoring and climate applications.
Combined Products
The most promising approaches combine C-band SAR (high resolution, moisture + roughness sensitivity) with L-band radiometry (coarser resolution but more direct moisture measurement) and physical models. Downscaling SMAP data using Sentinel-1 spatial patterns can produce field-scale moisture estimates with regional-scale accuracy.
Practical Tips from Field Experience
Timing matters: Acquire SAR data early morning (ascending passes in some orbit configurations) when soil moisture reflects recent precipitation rather than afternoon evaporative losses.
Frozen soil looks dry: Frozen water doesn't contribute to the dielectric constant the same way liquid water does. Winter SAR data over frozen ground shows low backscatter regardless of actual water content.
After heavy rain: Very wet soil surfaces can become smooth (rainfall breaks up soil aggregates), reducing roughness-related backscatter while increasing moisture-related backscatter. These opposing effects can partially cancel, making post-rain moisture retrieval less straightforward than expected.
Irrigation detection: The 6-12 day Sentinel-1 revisit can detect irrigation events as sudden backscatter increases in agricultural fields. This has practical value for water resource management — tracking which fields are being irrigated and how frequently.
What This Means for Agriculture
Soil moisture is arguably the most important variable in crop management that can't be easily measured at scale from the ground. Tensiometers and capacitance probes measure individual points; satellite SAR covers entire regions simultaneously.
The current state of the technology: SAR-based soil moisture provides useful relative information (this field is wetter than that field; this week is drier than last week) rather than precise absolute measurements. For agricultural decision-making — when to irrigate, where drought stress is developing, which fields received recent rainfall — relative information is often sufficient.
The future: as L-band missions like NISAR come online and data fusion techniques improve, the accuracy and coverage of satellite soil moisture will continue to advance. The physics is well-understood; the engineering challenge is in separating the moisture signal from the many other factors that influence radar backscatter.
