Sentinel-1 Synthetic Aperture Radar Time Series for Irrigation Mapping

Mapping and understanding of irrigated areas is crucial to optimize water management and improve agricultural productivity in order to meet the growing demand for food products. Irrigation can double productivity per unit of land compared with rainfed agriculture. This study proposes a scientific approach based on radar data and machine learning model for irrigation detection at plot scale. Our study site covers the entire region of Centre-Val de Loire in France. Using a dense Sentinel1 Synthetic Aperture Radar (SAR) time series to train a Random Forest algorithm, our model achieves an average overall accuracy of 83.3%. This approach effectively addresses the challenge of mapping irrigated areas, and provides valuable information for the implementation of water-efficient agricultural practices.