Statistical evaluation of CMIP5 climate change model simulations for the Ethiopian highlands

This study compares Climate Research Unit version 3 reference observations in the period 1980-2009 with Coupled Model Intercomparison Project Five (CMIP5)-simulated climate patterns and trends over the Ethiopian highlands (7-14N, 36-40E) using the rcp6 scenario. Three metrics are used in a stepwise statistical evaluation: (1) mean annual cycle of area-averaged rainfall (R) and maximum temperature (T-x), (2) spatial pattern of R and T-x over Ethiopia and (3) interannual (1.5-6 years) R and T-x variability. Those models passing the first step by simulating a unimodal annual cycle reaching 28 degrees C in March-May and 9 mm day(-1) in July-August are then evaluated for their ability to represent the cool, wet island' pattern over the highest elevation. Finally, the models' ability to follow observed area-averaged R and T-x anomalies from 1980 to 2009 is analysed by correlation. The stepwise metrics are designed to improve confidence in the use of CMIP5 models for future resource management and development planning across the Ethiopian highlands. The models with optimal performance in the most categories include the HAD, CCSM and GFDL.