Monitoring the dynamics of surface water fraction from MODIS time series in a Mediterranean environment.

•We estimated surface water fraction dynamics with Random Forest and Cubist methods.•MODIS BRDF data, derived water indices, and a DEM was used as input.•Training and validation were performed with water masks from two Landsat tiles.•Models trained on both tiles outperformed those trained on a single tile.•Our approach is promising for monitoring surface water at large spatial scales.