Machine learning based dynamic numerical climate multi-model ensembleweighting for high impact weather affecting the energy infrastructureThe infrastructure for renewable energy production and electrical grid itself is affectedby weather and climate conditions and vulnerable to high impact weather andcascading events. A reliable representation of the meteorological conditions leading tosuch events including their uncertainty is therefore needed for both weather and climatetime scales. Individual numerical weather and climate models exhibit systematicstrengths and weaknesses across scales, and geographic regions, despite thedifferences in model physics and parametrizations. One way to tackle this and avoidrunning single-model ensemble climate simulations are multi-model or poor-manensembles consisting of a set of different climate or weather models combined.Ensemble mixing offers a way to mitigate these weaknesses while providing uncertaintyquantification. Simply ensemble averaging can dilute forecast and climate signals andpenalize outliers and rare extremes. Different approaches have been proposed to tacklethis problem by assigning non-uniform weights to individual model fields andparameters, however, these methods often rely on domain knowledge such as modeldependencies [1,2].Here, we propose a machine learning-based multi-model ensemble model-mixingframework that is domain-agnostic and assigns spatially and temporally dynamicweights, in addition to an error metric. The domain of interest is the Alps, which exhibitchallenging terrain and localized extreme events, e.g. precipitation extremes that aredifficult to capture in conventional climate models. The CERRA reanalysis data at ~5.5km resolution serves as the target grid. We build a multi-model ensemble by combiningdynamically downscaled simulations of 2 m air temperature, precipitation, and windspeed from COSMO-CLM (6 km) and WRF (10 km). Each regional model is driven by twoCMIP6 global climate models (MPI-ESM and EC-Earth) under two scenarios (SSP1-2.6and SSP5-8.5), with an additional historical period used for training. Static informationsuch as orography and seasonal dependencies are considered. We evaluate theensemble’s performance on selected extreme events (e.g., heavy precipitation,windstorms, heatwaves) that can (and did) harm energy infrastructure, such as theEuropean derecho 2022.[1] Christensen, Jh, E Kjellström, F Giorgi, G Lenderink, and M Rummukainen. 2010.‘Weight Assignment in Regional Climate Models’. Climate Research 44 (2–3): 179–94.https://doi.org/10.3354/cr00916.[2] Merrifield, Anna Louise, Lukas Brunner, Ruth Lorenz, Iselin Medhaug, and RetoKnutti. 2020. ‘An Investigation of Weighting Schemes Suitable for Incorporating LargeEnsembles into Multi-Model Ensembles’. Earth System Dynamics 11 (3): 807–34.https://doi.org/10.5194/esd-11-807-2020.