Detection and quantification of irrigation water amounts at sub-kilometric scale using Sentinel-1 surface soil moisture

Detailed information about the timing and the amount of water used for irrigation at a high spatial resolution is critical for monitoring and improving agricultural water use efficiency. However, neither statistical surveys nor current remote sensing-based approaches can accommodate this need. Being a natural source of information on the amount of water entering the ground, soil moisture is directly related to irrigation and precipitation. Hence, we present a novel approach based on the TU Wien Sentinel-1 Surface Soil Moisture product to fill this gap, i.e. detect and quantify irrigation water amounts at sub-kilometric scale. Theoretically, irrigation occurring in a specific field should be reflected by a local increase in soil moisture, while surrounding non-irrigated fields exhibit a different behavior.  We harness the spatio-temporal patterns of soil moisture to identify individual irrigation events, and then to estimate irrigation water heights. To retrieve the latter we include formulations of evapotranspiration and drainage as such vertical fluxes have a significant impact on sub-daily soil moisture variations. The proposed approach is evaluated against field scale irrigation data reported by farmers at three sites in Germany with heterogeneous field sizes, crop patterns, irrigation systems and management. 
Our results show that most irrigation events occurring in a field can be detected using soil moisture information at 500 m and 1-3 days resolution, however, lower accuracy is found during the peak of the growing season. The retrieved irrigation water heights increase with the fraction of pixel under irrigation as higher water amounts are expected over largely irrigated pixels. Finally, irrigation estimates are in agreement with reference data, in terms of temporal dynamics as well as spatial patterns, regardless of field-specific characteristics (e.g. crop type, irrigation system). Unlike most approaches based on microwave soil moisture data, the proposed framework does not rely on additional datasets, which are either locally available or do not even exist at (sub-) kilometric resolution. Hence, the proposed approach has the potential to be applied over large regions with varying cropping systems (e.g. national and even continental scale).