Performance Evaluation of a Developed Hybrid AOGCM Model under Climate Change

The most comprehensive tools for producing climatic scenarios and consequently hydrological predictions are atmospheric-ocean general circulation models (AOGCMs). This paper evaluates the performance of those models. Possible scenarios of temperature and rainfall in the base period (1971-2000) for the Aidoghmoush basin located in East Azerbaijan are estimated with seven AOGCMs. A hydrological model is first calibrated for the basin and then monthly time series of temperature and rainfall resulting from AOGCMs in the base period are input to the hydrological model. The mean observed runoff (MOR) weighting method is employed to assess the effectiveness of each climate model to produce runoff. An appropriate probability distribution is chosen and relevant statistical parameters extracted and compared with observed runoff statistical parameters. Results show that the hybrid and Hadley Centre Coupled Model, version 3 (HadCM3) models with respective correlation coefficients of 97% [root mean square error (RMSE) = 2.09 m(3)/s, mean absolute error (MAE) = 1.51 m(3)/s, and Nash Sutcliffe efficiency (NSE) = 0.89] for the hybrid model and 90% (RMSE = 3.22 m(3)/s, MAE = 2.58 m(3)/s, and NSE = 0.87) for the HadCM3 model can best simulate runoff. Comparison of the results of the probability distributions and the transition probability matrix from the AOGCM models with observed runoff shows that the hybrid and HadCM3 models with respective correlation coefficients of 91% (RMSE = 0.07 m(3)/s, MAE = 0.004 m(3)/s, and NSE = 0.80) and 86% (RMSE = 0.14 m(3)/s, MAE = 0.04 m(3)/s, and NSE = 0.78) yield reliable outputs. (C) 2016 American Society of Civil Engineers.