Results

Our data compare very well to the 25 validation reaches, with a mean absolute error of 2.84 meters per year and a bias of 0.39 meters per year (Figure 2a). Unfortunately these limited validation sites are not equally distributed around the world and do not represent a very broad range of morphologies (Figure S5). Nevertheless, our validation results suggest that our method and implementation in GEE can capture riverbank erosion with similar accuracy to studies focused on smaller domains.