A single triggering event such as strong precipitation, may result in swarms of shallow landslides that cumulatively pose significant risks to population, infrastructure and the environment. There is large scientific and societal interest in hazard maps that display the susceptibility to shallow landslide occurrence while accounting for spatio-temporal variability of triggers and influencing factors. In this study, we implemented a generic and scalable hazard mapping workflow in accordance with the FAIR principles that relies on a Random Forest Classifier. It is a purely data-driven approach which due to its modular implementation allows a flexible adaptation to different areas of interest and integrated data. It’s application results in individualized and reproducible hazard maps. The validation of these maps as well as the underlying workflow is considered as well. The workflow facilitates dynamic hazard maps of varying temporal scope. This temporal dynamic is achieved through the integration of highly volatile features such as precipitation events. In this way, hazard maps can be created depending on the weather forecast. Geohazard potential changes dramatically in response to volatile triggering conditions making their inclusion into the mapping process a natural extension to the established static hazard mapping features. First results of enriching data-driven landslide hazard mapping with temporal data are promising. The resulting temporally dynamic hazard maps offer a wide range of opportunities to improve landslide risk management.

The 28th IUGG General Assembly (IUGG2023) (Berlin 2023)