Skip to main content

The SMOSAR algorithm

Introduction to the fundamentals of the SMOSAR system and its features.



SMOSAR stands for Soil MOisture retrieval from multi-temporal SAR data.

The SMOSAR algorithm is the core of an in-house developed system of CNR-IREA, which has the goal of retrieving the soil water content from multi-temporal radar imagery at 1 km scale.

SMOSAR in brief

    Soil moisture (SM) is an Essential Climate Variable that plays an important role in the exchange of water, energy and biochemical fluxes between the Earth's surface and the atmosphere. Soil moisture is expressed as volumetric soil water content, which is defined as volume of the water devided by the total volume of a soil sample [/]

The concept of SM retrieval from radar observation arises from the fact that the dielectric properties of soil ‐ which act as a proxy for SM ‐ affect the backscattering properties of soil surfaces in microwave frequencies. In particular, Spaceborne Synthetic Aperture Radar (SAR) sensors are currently the most suitable systems to retrieve SM at high spatial resolution at spatial scales ranging from local to regional and continental.

As example, SAR observations come from the European Radar Observatory Sentinel-1 (S-1), developed in the framework of the Copernicus programme. S-1 systematically provides C-band SAR imagery from two identical spacecraft, (S-1 A & S-1 B), at high spatial and moderate temporal (6-day exact repeat cycle) resolutions with a sustained observation strategy for the next decades which foresees first the S-1C & S-1D satellites and then the S-1 Next Generation satellites from 2028 onwards.

CNR-IREA has developed a pre-operational product of SM with high spatial resolution (1km) derived from S-1 constellation. The SM product has been validated according to international protocols using measurements of hydro networks distributed between Europe, the United States, Canada and Australia. SM forecasts from S-1 can help in agronomic and hydrological applications, such as, for example, the early identification of drought conditions of the soils or the extension of irrigated areas, the alert of flood events through new generation weather forecasts that work on grids with resolutions of ~ 1 km and efficient management of irrigation resources at the basin scale.

The implemented SM retrieval algorithm exploits a change detection approach requiring SAR time series with a short revisit time and, for this reason, can be referred to as a short-term change detection (STCD) approach. Its rationale is that temporal changes of surface parameters influencing the radar backscatter, apart from SM, (e.g., soil roughness, canopy structure, vegetation water content, etc.) usually take place at longer temporal scales than SSM changes (excluding cultivation practices). Therefore, SAR time series with a sufficiently short repeat cycle are expected to track changes in SM only, since other parameters affecting radar backscatter can be considered constant. This approximation makes robust and relatively simple the retrieval approach and also expedites the processing.

The algorithm can be applied to bare and vegetated surfaces dominated by soil attenuated scattering that enables good radar sensitivity to SSM throughout the growing season. This land cover restriction implies that before the retrieval, a masking process obscuring those areas showing a poor radar sensitivity to SM (i.e., areas dominated by volume scattering) is required. Masking is implemented as a two-step process. The first one consists of using global land cover maps (e.g. GlobCover, CCI LandCover) to mask areas such as forests, urban areas, water bodies, etc. The second masking level exploits an adaptive thresholding method applied to S-1 cross-polarized observations. As a result of the masking process, the SM retrieval algorithm is applied only over those land surfaces dominated by soil attenuated scattering that show an adequate radar sensitivity to SSM.

In summary, the implemented code transforms dense time series of N co-registered S-1 images at 40m pixel size into N-SM maps. The last step in SMOSAR is a low pass filter, with a kernel of W × W pixels (W = 13), applied to the SM maps at 40 m pixel size. The advantage is twofold. First, the uncertainty on the SM retrieved is reduced and, second, the impact of errors due to abrupt changes of vegetation and/or soil roughness, which normally take place at field scales and can be wrongly interpreted as SM changes, are mitigated. The SM product consists of a SM map of 520 m pixel, which corresponds to a spatial resolution of approximately 1 km, and the standard deviation associated with the mean SM value at 1 km resolution is also estimated and delivered as a companion layer.

  • Balenzano A., Mattia F., Satalino G., Lovergine F., Palmisano D. et al., “Sentinel-1 soil moisture at 1 km resolution: a validation study”, Remote Sensing of Environment, Vol. 263, 2021,112554.
  • Satalino G., A. Balenzano, F. Mattia, and M. W. J. Davidson, “C-band SAR Data for Mapping Crops Dominated by Surface or Volume Scattering”, Geoscience and Remote Sensing Letters, Vol. 11, Issue 2, pp. 384-388, Feb. 2014.
  • Balenzano A., F. Mattia, G. Satalino, M. W. J. Davidson, “Dense temporal series of C- and L-band SAR data for soil moisture retrieval over agricultural crops”, IEEE Journal of Selected Topics in Applied Earth Obs. and Remote Sensing (J-STARS), Vol. 4, No. 2, pp. 439-450, June 2011.