Optimizing Post-Earthquake Decision-Making with GEOAI: Identification and Classification of Damaged Buildings in the El Haouz Earthquake, Morocco

Abstract. Post-earthquake reconstruction is complex and must strictly comply with current regulations. Authorities immediately began planning rapid reconstruction of residential buildings to provide shelter to those who lost their homes. Moreover, it must be accelerated to minimize impacts on affected communities. Hence, the idea of drafting a roadmap to leverage geospatial artificial intelligence (GeoAI) and Geographic Information Systems (GIS) to identify and classify buildings collapsed by the El Haouz earthquake. The data used are satellite/drone imagery, orthophotos, and GIS-based geo-risk studies for the affected douars. The chosen study area is part of the commune of Tizi N'Test, within the province of Taroudant, Morocco, where a significant portion of the douars were severely affected by this earthquake. In this paper, we outlined a strategy based on the use of a GeoAI solution composed of an XGBoost machine-learning model and a YOLOv9 deep learning model. The results showed that both the XGBoost and YOLOv9 models achieved high overall accuracy of 97% and 96%, respectively, on validation data. This work brings significant value to the field of post-earthquake management by making the identification of reconstruction sites more efficient and automated.