SARWater

The project

In December 2024, ESA launched a call for proposals under the SAR – Fundamental Scientific Developments theme, part of the Earth Observation Science for Society component of the Future EO1 Segment 2 Programme. This initiative, aligned with the Sentinel Users Preparation (SUP) program, aimed to foster novel science, innovative methods, and new capabilities to maximize the impact of upcoming missions. SARWATER was selected under the theme “Retrieval of Soil and Vegetation Water Content,” which is expected to benefit significantly from the forthcoming ROSE-L mission. ROSE-L will enhance European radar imaging capabilities beyond those of Sentinel-1, operating with two L-band SAR satellites on a 12-day repeat orbit with a 180° phase difference. This configuration improves revisit frequency and supports interferometric applications. ROSE-L will also align with Sentinel-1, enabling multi-frequency imaging of the same area.

The mission’s nominal imaging mode is ScanSAR dual-polarization (Interferometric Wide Swath Mode, RIWS), with quad-polarization also available. L-band microwaves interact less strongly with vegetation than C-band, allowing penetration through crop canopies that typically obscure surface soil moisture (SSM) signals in C-band (e.g., sugar beet). Thus, ROSE-L will provide additional, accurate information not accessible via Sentinel-1, extending SSM monitoring to vegetated regions. Existing L-band SAR missions — such as ALOS-2/ALOS-4 (JAXA), SAOCOM (CONAE), and NISAR (NASA-ISRO) — offer opportunities to explore synergies with C-band systems.

SARWATER officially began in September 2025.

Project objectives

SARWATER aims to:

• Develop synergistic methods that combine C- and L-band SAR data to retrieve SSM and quantify associated uncertainty at ~250 m resolution.
• Consolidate and assess the performance of single-frequency SAR platforms (C- or L-band) for SSM retrieval and valuate the performance of combined C-/L-band products relative to single-band retrievals.
• Leverage vegetation sensitivity in SAR backscatter to retrieve vegetation water content (VWC), rather than treating it solely as a hindrance.
• Validate results on a large scale and demonstrate their impact on applications, such as estimating root zone soil moisture and determining the extent of irrigated areas.
• Make openly available the project outcomes, ensuring lasting accessibility and scientific reuse

The Short-Term Change Detection Method

The SMOSAR algorithm in C-band (Balenzano et al., 2021) consists of three sequential stages applied to a stack of N co-registered and geocoded Sentinel-1 VV and VH images (pixel resolution: 40 m; spatial resolution: ~100 m). An optional low-pass filter can be applied to reduce uncertainty. The output includes SSM and its standard deviation at ~1 km or ~250 m resolution.

Static and Dynamic Masking

This step identifies and masks vegetation areas where C-band backscatter shows minimal sensitivity to SSM (Satalino et al., 2014).

• Static masking uses land cover maps (e.g., ESA CCI / Global Dynamic Land Cover) to exclude forests, urban areas, and water bodies.
• Dynamic masking filters out extreme backscatter values (e.g., >−3 dB or <−22 dB in VV polarization), which are unlikely to originate from flat soil surfaces.
• VH observations help distinguish volume scattering from soil-attenuated scattering. Areas dominated by volume scattering are masked.
• For high incidence angles, volume scattering dominates, so SMOSAR sets a threshold of 55°. This constraint is relaxed over bare or sparsely vegetated areas.

In L-band, most seasonal crops are expected to exhibit soil-attenuated scattering even at high incidence angles, so dynamic masking may not be necessary. This will be further evaluated in SARWATER. Balenzano et al. (2022) introduced a refinement using Sentinel-2 data and the VH/VV ratio to detect abrupt changes in vegetation and roughness at field scale, improving resolution to ~100 m.

Estimation of Minimum SSM

External data (e.g., SMOS, SMAP, ASCAT) or in situ measurements can be used to estimate the minimum SSM value across the image stack. This value is a key input for the STCD algorithm (Balenzano et al., 2013). The current approach in SMOSAR uses a calibration curve relating S-1 VV observations to low-resolution SSM values.

SSM Retrieval

Using the STCD method, N SSM maps are generated for non-masked areas. The STCD approach is based on a first-order iterative solution of the radiative transfer (RT) equation for co-polarized backscatter. It assumes that SSM changes rapidly, while other surface properties (e.g., roughness, vegetation) evolve slowly. Thus, backscatter changes primarily reflect SSM variations. The ratio of backscatter between successive acquisitions isolates the SSM signal (“alpha approximation”). The maximum likelihood solution for the soil reflection coefficient is derived and inverted in two steps:

1. Invert the relationship between backscatter and dielectric constant.
2. Invert the dielectric constant–SSM relationship using empirical or theoretical models and soil texture data.

While SMOSAR originally used the Small Perturbation Model, any suitable expression for the reflection coefficient — independent of roughness and vegetation — can be adopted, including those from physical optics or multi-layer models. Importantly, SMOSAR’s formulation is immune to time-independent multiplicative corrections in backscatter.

The partners

SARWATER brings together three leading European research institutions:

CNR-IREA (Prime Contractor)

National Research Council, Institute for Electromagnetic Sensing of the Environment, Italy www.irea.cnr.it

CNR-IREA possesses significant expertise in SAR remote sensing, electromagnetic modeling, and agricultural applications. The institute coordinates the project and leads the development and validation of algorithms.

Forschungszentrum Jülich (Subcontractor)

Jülich Research Centre, Institute of Bio- and Geosciences: Agrosphere (IBG-3), Germany www.fz-juelich.de

FZJ provides expertise in land surface modeling, soil moisture networks, and hydrological applications. The team operates the TERENO Rur catchment observatory, a key validation site for the project.

Ghent University (Subcontractor)

Universiteit Gent, Belgium www.ugent.be

UGent brings expertise in evaporation modeling, data assimilation, and surface and root-zone soil moisture and vegetation water stress monitoring through the GLEAM model framework.

Jülich Research Centre, Institute of Bio- and Geosciences: Agrosphere (IBG-3), Germany https://www.fz-juelich.de/en/ibg/ibg-3