The Selection of Global Climate Models for Regional Impact Studies Should Consider Information from Historical Simulations and Future Projections

Abstract The selection of Global Climate Models (GCMs) based on their ability to represent precipitation patterns of a region is required for hydrological climate change impact studies to address time and computational constraints. Generally, the selection of GCMs is determined based on their abilit...
Ausführliche Beschreibung

Abstract The selection of Global Climate Models (GCMs) based on their ability to represent precipitation patterns of a region is required for hydrological climate change impact studies to address time and computational constraints. Generally, the selection of GCMs is determined based on their ability to reproduce observed climate statistics in historical simulations, assuming they will continue to perform well in the future. However, the performance of GCMs varies over time in ways that are not sensitive to their historical performance, indicating that GCMs’ selection needs to consider historical simulation and future projection information. We propose a framework to account for future GCM projection convergence to and divergence from the ensemble mean, along with historical performance, to select the GCMs that are applicable to a particular regional climate impact study. The framework uses Reliability Ensemble Averaging (REA) with 30 Coupled Model Intercomparison Project-6 (CMIP6) GCMs to select GCMs based on the ensemble mean and variability of projections. We demonstrate the framework using three climate indices (annual maximum precipitation, annual total precipitation, and wet day precipitation intensity) in the Chesapeake Bay watershed of the United States. Our analysis shows that using only the GCM performance during the historical period could result in the selection of GCMs that are extreme outliers due to an inherent underprediction of precipitation extremes by all GCMs and requires an efficient bias correction before selection. There was also no significant correlation between the historical period performance of GCMs and future GCM convergence for more than 95% of the cases in the study region. This highlights the need to consider convergence and divergence information from climate projections when selecting GCMs for practical and computationally intensive applications. The proposed framework can be adapted to any study region and can help identify GCMs for computationally intensive climate change impact studies. Ausführliche Beschreibung