On April 18th, Jhonatan Rivera Rivera presented in the NH6.1 session of EGU2024 his scientific contribution “Interferometric Synthetic Aperture Radar to assess the impacts of ground deformation in local, regional and national studies”, which results are part of the SARAI project. In this work, we applied Machine Learning (ML) algorithms to classify ground deformation process, using MOVESAR: a national database based on SAR, integrating geological, geotechnical, hydrological, morphometric, and land cover covariates.
The MOVESAR database compiles MP’s from processed data carried out by the Instituto Geológico y Minero de España (IGME), the Centro Tecnológico de Telecomunicaciones de Cataluña (CTTC), and the European Ground Motion Service (EGMS). Spatially, MOVESAR contains MPs from Arcos, Huelva, Cobre las Cruces, Granada, Lorca, Murcia, Madrid, Barcelona, Valle de Tena, Asturias, Albuñuelas, Rules, Sierra Nevada, La Unión, and Suria. The MPs from EGMS were only used to validate the machine learning algorithms employed (random forest and catboost). The results indicate that noise filtering techniques for MPs (threshold velocity filtering technique “TVF”), class balancing (Cost Sensitive Learning “CSL”), and feature reduction (Feature Importance “FI”) enhance the interpretability, efficiency, and accuracy of the models.