Item

Abstract

Landslide mapping is critically important for providing detailed spatial information on hazard extent ina timely manner that ultimately contributes to the protection of human lives and critical infrastructure.In the context of increasing demands for scalable and automated solutions, Earth Observation (EO) datacoupled with deep learning offer great potential to enhance the speed and accuracy of emergencymapping. This study explores the utility of a deep learning model with the U-Net architecture forautomated landslide mapping using data from optical Sentinel-2 and Synthetic Aperture Radar (SAR)Sentinel-1 satellites. We investigate the effectiveness of various optical (visible, near-infrared, and short-wave infrared) and SAR-derived features (backscatter coefficients, polarimetric features, interferometriccoherence), used both independently and in combination. Additionally, we assess the impact ofincreasing the number of pre-/post-event SAR observations on classification performance. The U-Netmodels are trained and tested using globally distributed and limited reference data (563 uniquepatches). Optical features consisted of one pre-/post-event feature, whereas SAR features had threefor each reference sample. Our analysis shows that the highest classification accuracies are consistentlyachieved using optical features (F1-score of 0.83 with visible, near-, and short-wave infrared bands). Nosubstantial improvements were recorded when SAR features were combined with optical features. Theusage of the most common optical features (visible and near-infrared) shows the lowest accuraciescompared to their combination of short-wave infrared or red-edge bands. Increasing the number ofpre-/post-event SAR features improves the SAR-based accuracies. To promote further advancements inautomated landslide mapping using deep learning, the landslide reference dataset generated in thisstudy is freely available at (https://doi.org/10.5281/zenodo.15284357)