Datensatz

Zusammenfassung

The utilization of General Circulation Models (GCMs) plays a crucial role in forecasting future climate changes and is heavily relied upon by policymakers in managing responses to human-induced global warming and climate change. To attain a robust global signal and assess uncertainties, GCMs are often combined in Multi-Model Ensembles (MMEs) using various approaches such as the Multi-Model Mean (MMM) or its weighted variants. Recently, Thao et al. (2022) proposed a new model comparison approach that is based on graph cut optimization. This optimization method, originally developed in computer vision for tasks like image segmentation, is used for selecting the best-performing model at each gridpoint for a given variable, resulting in a patchwork of models that maximizes performance while avoiding spatial discontinuities. In contrast to methods like MMM that use global weights, this approach considers the local performance of individual models, resulting in improved global predictions. Here we present a new combination approach of GCMs that utilizes graph cuts. Compared to the univariate method, this approach ensures that the relationships between variables are locally preserved while producing coherent spatial fields. Furthermore, we replace the use of distances between multi-decadal means with statistical distances between multi-decadal distributions, enabling the combined model to represent not only the average behavior (e.g. mean temperature or precipitation) but the entire multivariate distribution, including extreme values that have substantial societal and environmental impacts.Thao, S., Garvik, M., Mariethoz, G., & Vrac, M. (2022). Combining global climate models using graph cuts. Climate Dynamics, February. https://doi.org/10.1007/s00382-022-06213-4