Evaluation of the CMIP5 Decadal Hindcasts in the State of California

Abstract

EVALUATION OF THE CMIP5 DECADAL HINDCASTS IN THE STATE OF CALIFORNIA by Colin J. McKellar This study investigated the ability of the new Coupled Model Inter-comparison Project phase 5 (CMIP5) decadal hindcasts to predict the observed decadal variability for maximum temperature (Tmax) and minimum temperature (Tmin) in California over two historical periods in the 20th century and one future period in the 21st century. Annual and seasonal California temperature trends were computed by averaging 54 United States Historical Climate Network version 2 temperature observations from 1960-1990 and 1980-2010. Modeled California temperatures were reconstructed with bi-linear interpolation from the CMIP5 decadal hindcasts and 20th century experiments. The individual model ensemble averages (MEA) and mean model ensemble averages (MMEA) were then compared to the observations during 1960-1990 and 1980-2010. The decadal hindcasts displayed a similar overall skill as the 20th century experiments in predicting the observed annual and seasonal temperature trends during both historical periods. However, the predictive skill for individual models showed that the decadal hindcasts systematically improved the MEA predictions and that certain models, such as the MRI-CGCM3, outperformed the MMEA in each experiment. Also the higher performing models, such as the MRI-CGCM3, provided better future Tmax and Tmin trend predictions. Future predictions show increasing annual and seasonal temperature trends that indicate a longer growing season by the year 2035.