Key points are not available for this paper at this time.
Extreme hydro-meteorological impact events are difficult to predict in space and time as they frequently result from localised, high-intensity convective precipitation events. Societal impacts can occur when extreme precipitation events interact with multiple other geomorpholocial, hydrological and societal predisposing and preparatory factors. Due to limitations in spatial and temporal resolution, it is assumed that climate models likely underestimate the magnitude and frequency of future extreme precipitation events (Slingo et al., 2022).In the context of disaster risk reduction, it is important to understand the relationships between the multiple driving factors of geomorphic high impact events. Knowing when and where potential adverse consequences are likely to occur and under which conditions can support the design and provision of risk reduction measures (e.g., impact-based forecasts and warnings). Moreover, impact models can inform on likely changes in the frequency of extreme events under future climate regimes.We address this problem by developing a data-driven machine-learning model aimed at predicting the likelihood of past and future weather extremes that cause societal impacts. Using a risk framework as a conceptual underpinning, a stratified space-time modelling approach is implemented, sampling from combined landslide, debris-flow and rock-fall damage inventories across Austria and South Tyrol (Italy) spanning the period 2005-2022. Building on previous method developments (Steger et al., 2023), multiple meteorological indicators available at different spatial scales, including a sub-model used to predict the likelihood of deep convective events, are combined with morphometric, geological, hydrological, land cover data as well as data on potentially exposed assets to train a hierarchical generalised additive mixed model (GAMM) on the basis of slope units. The modelling results are evaluated through multiple perspectives using variable importance assessment, spatial and temporal cross-validation procedures as well as qualitative plausibility checks.We present first model results, showing the importance of simultaneously considering spatio-temporal variations in hazard components as well as exposure data to predict localised impact events. Further strengths, opportunities and limitations of the approach are discussed. The research leading to these results has received funding from Interreg Alpine Space Program 2021-27 under the project number ASP0100101, How to adapt to changing weather eXtremes and associated compound and cascading RISKs in the context of Climate Change (X-RISK-CC).ReferencesSlingo, J., Bates, P., Bauer, P. et al. Ambitious partnership needed for reliable climate prediction. Nat. Clim. Chang. 12, 499503 (2022). https://doi.org/10.1038/s41558-022-01384-8. Steger, S., Moreno, M., Crespi, A., Zellner, P., Gariano, S.L., Brunetti, M., Melillo, M., Peruccacci, S., Marra, F., Kohrs, R., Goetz, J., Mair, V. Pittore, M. Deciphering seasonal effects of triggering and preparatory precipitation for improved shallow landslide prediction using generalized additive mixed models. Natural Hazards and Earth System Sciences. 23, 14831506 (2023). https://doi.org/10.5194/nhess-23-1483-2023.
Spiekermann et al. (Fri,) studied this question.