An efficient method for choosing the best sub-ensemble of climate models for ΔT Projections


Climate scientists produce projections of future climate using Global Climate Models (GCMs), however, for a regional climate study, it is sometimes inconvenient or otherwise impractical to analyze each of the available models individually. Often, scientists will instead perform an ‘ensemble’ study, by averaging the output of all available models. Recent studies have shown, however, that representative sub-ensembles may outperform individual models and full ensembles alike. We propose an efficient method for the selection of representative sub-ensemble members for ΔT projections based on the sub-ensemble’s ability to reproduce an observed climate baseline. We used our method to validate GCM reproduction of long-term baseline temperature averages at Toronto and Montreal, and found that sub-ensembles consistently outperformed individual models and total model ensembles when tested for individual variables and stations. Furthermore, sub-ensembles were more effective at reproducing the baseline average across combinations of variables and stations. Our method, derived and developed in the R programming language for statistical computing, provides a fast, computationally efficient solution to the selection of model members of selective, representative sub-ensembles.

Theoretical and Applied Climatology