Model weighting based on mesoscale structures in precipitation and temperature in an ensemble of regional climate models

摘要

We present a weighting scheme specifically designed for regional climate models (RCMs) in that it is based on the model performance in simulating the sub-global climate model (GCM) mesoscale climate signal. The functional form of the weights is based on multiple variables (temperature and precipitation) and metrics (correlation and root mean square error). The weighting scheme is applied to an ensemble of RCM simulations for the European region recently completed as part of the ENSEMBLES project. As a test of the successful implementation of the scheme, the weighting leads to an overall improvement of the performance of the ensemble when measured with the same metrics used in the weighting. The improvement is particularly pronounced over topographically complex regions (e.g. the Alps) in which a larger inter-model range of performance, and thus a more aggressive weighting, is found. When applied to the generation of probabilistic climate change projections, this scheme is designed to be used in conjunction with other RCM weighting metrics developed in the ENSEMBLES project and corresponding weighting schemes for the GCMs driving the regional models.