SAR DERIVED PRODUCTS FOR AGRICULTURE

Climate change, anthropogenic pressure and the excessive exploitation of natural resources threaten the sustainability of the food and agriculture sector. To cope with this threat, transformative change in agriculture and food systems are required worldwide. An effective response is promoting sustainable agriculture. It requires efficient and parsimonious use of water, soil resources and energy. For instance, improved water governance is a crucial challenge in agriculture (FAO, 2020). This is particularly urgent in arid and semi-arid regions. Updated information on the extent of irrigated areas and related soil moisture conditions can be valuable input for basin-scale approaches to irrigation planning. Another important measure to reduce the footprint of agriculture is promoting conservation agriculture. It requires the adoption of minimum tillage practices, which increase soil fertility while reducing greenhouse gases emission and soil erosion (Troccoli et al., 2015). An integrated and flexible Earth Observation system, designed to exploit synergies between active and passive observing systems, can provide a broad range of surface parameters at high temporal/spatial resolution and global scale, thus supporting a transition to green agriculture. Synthetic Aperture Radar (SAR) systems, such as Sentinel-1 (S-1), have already demonstrated a great ability to integrate with optical systems, such as Sentinel-2 (S-2), to ensure spatial and temporal continuity for monitoring agricultural areas and improving farming. They also offer complementarity to optical systems. Indeed, SAR data are sensitive to the geometric structure and water content of the crops and the underlying soils. This complementarity is enhanced by using multi-frequency and multi-polarization SAR sensors. The objective of the study is to present a surface soil moisture (SSM) product at a field scale derived from S-1 and S-2 data. This is an evolution of an S-1 SSM product at 1 km validated by Balenzano et al. (2021). The SSM retrieval exploits a change detection approach that requires SAR observations with a short revisit and, for this reason, it is referred to as a short-term change detection (STCD) approach. The strength of the algorithm is its conceptual simplicity and robustness. The evolution to "field scale" consists of integrating S-2 Normalized Difference Vegetation Index (NDVI) to mask abrupt changes of the vegetation and/or soil roughness that may affect SSM estimates at the "field scale". If NDVI is not available (e.g. cloud cover), then the ratio of VH/VV is adopted as a proxy (Palmisano et al., 2020). Besides SSM, two added-value products stemming from the methodology developed to retrieve SSM are illustrated. They consist of maps of fields irrigated and undergoing tillage changes. The rationale for using SSM maps to identify irrigated fields is the correlation between SSM and irrigation water. Therefore, an irrigated field will show an SSM level higher than those non-irrigated, at least for a certain time span, which may range from a few days to hours. The segmentation of the irrigated/nonirrigated fields amounts to a two classes classification problem. It is tackled both in the space and time domain. For the tillage change identification, the algorithm exploits S-2 & S-1 data to first segment agricultural surfaces into vegetated and bare (or sparsely vegetated). Then, multiscale temporal change detection is applied to cross-polarized S-1 backscatter coefficient to single out local changes of soil roughness. Those are likely related to tillage practices (Satalino et al., 2018). The paper gives examples of the S-1 & S-2 SSM product over various sites in Europe and provides an outlook about the irrigation extent and tillage change monitoring at a regional scale.