Optimal selection of representative climate models and statistical downscaling for climate change impact studies: a case study of Rhode Island, USA

Document Type


Date of Original Version



The future climate impact studies rely on future projections obtained from downscaling of Coupled Model Intercomparison Project Phase 5 (CMIP5) climate models. The main challenge is large, yet growing pool of CMIP5 models, posing a high computational cost for analyzing the climate model ensemble. The optimal model selection from the large pool is hence necessary. It is critical that the selected climate models represent the uncertainty range of possible future and have high skill in past performance. In this study, we use a multi-criteria decision process based on the coupling of the envelope and past performance approach to identify potential future climate stresses for RI, USA. The selection is based on projected changes in annual climatic means (precipitation and temperature) followed by a range of projected changes in climatic extremes and past performance among these models. From a pool of 109 models from RCP4.5 and 79 models from RCP8.5, a final subset of 4 models was selected for RCP4.5 and RCP8.5 respectively. The change in annual climatic means for temperature varied + 1.7 to + 3.0 °C in RCP4.5 and + 2.4 to + 5.8 °C in RCP8.5, and the range in climate mean of annual precipitation varied from 0.2 to 12.7% in RCP4.5 and − 4.1 to 15.9% in RCP8.5. The selected climate models are statistically downscaled to produce reliable local-scale climate estimates. Various variants of quantile mapping are studied, and quantile delta mapping is applied to systematically reduce biases and preserve raw GCM signals.

Publication Title, e.g., Journal

Theoretical and Applied Climatology