Landslide/mudslide susceptibility is of significance to socioeconomic sustainable development and emergence management. Although remote sensing datasets have been used for landslide/mudslide susceptibility interpretations, the results might be weak owing to the limitations of the single-modal remote sensing dataset. Evolving Earth observation techniques enable the automatic identification of landslide/mudslide susceptibility over a large extent from multimodal remote sensing datasets. This also poses a major challenge in effective organization, representation, and modeling for complex information on landslide/mudslide susceptibility. In this study, we propose a geospatial semantic model to formally represent the interpretation of visual features from optical remote sensing, deformation features from synthetic-aperture radar (SAR) datasets, terrain features from digital elevation models (DEMs), and descriptions by field investigations. First, we applied optical remote sensing image, DEM, and SAR datasets to detect and annotate the features of landslide/mudslide susceptibility. Then, we developed a geospatial ontology to represent these features in a machine-understandable format. Depending on the triple structure of “domain-property-range” and the rules and restriction set by the proposed geospatial ontology, we created a semantic model to conduct semantic query and reasoning for landslide/mudslide susceptibility. The proposed semantic model for landslide/mudslide susceptibility interpretation has been successfully tested in four counties in Yunnan Province, China. We expect this work to be a major contribution to the integration of knowledge from both remote sensing and GIS data, and to deepen the application of semantic web technology in landslide/mudslide susceptibility domains.