Intercomparison of CMIP5 and CMIP3 simulations of the 20th century maximum and minimum temperatures over India and detection of climatic trends

Climate change impact assessment has become one of the most important subjects of the research community because of the recent increase in frequency of extreme events and changes in the spatiotemporal patterns of climate. This paper analyses the ability of 46 coupled climate models from Coupled Model Intercomparison Project phases 3 and 5 (CMIP5 and CMIP3). The performance of each climate model was assessed based on its skills in simulating the current seasonal cycles (monthly) of both maximum temperature and minimum temperature (T-max, T-min) over India. The performance measures such as coefficient of correlation (Skill_r), root mean square error (Skill_rmse), and the skill in simulating the observed probability density function (Skill_s) are mainly employed for evaluation of the simulated monthly seasonal cycle. A new metric called Skill_All which is an intersection of the above three metrics has been defined for the first time. A notable enhancement of Skill_All for CMIP5 vis-a-vis CMIP3 is observed. Further, three best CMIP5 models each for T-max and T-min were selected. The methodology employed in this study for model assessment is implemented for the first time for India, which establishes a robust foundation for the climate impact assessment study. The seasonal trends in T-max and T-min were analyzed over all the temperature homogenous regions of India for different time slots during the 20th century. Significant trends in T-min can be seen during most of the seasons over the entire Indian region during last four decades. This establishes the signature of climate change over most parts of India