Performance of CMIP5 wind speed from global climate models for the Bay of Bengal region

Abstract

AbstractRealistic estimates of surface wind speed (WS) are an essential prerequisite for process‐based air–sea interaction studies and numerical modelling needs. In this context, historical and projected WS estimates obtained from global and regional climate models warrant proper assessment and necessary bias corrections before it can be optimally used for rigorous analysis and research needs. Adequate evaluation and bias correction of WS estimates are therefore crucial to understand the extremes. For example, they have direct implications on extreme wind‐wave characteristics that can influence the coastal zones. The present study performed a detailed evaluation of WS obtained from the Coupled Model Intercomparison Project Fifth Phase (CMIP5) products to assess their projections for the Bay of Bengal region. A suite of global climate models (GCMs) is employed to generate the CMIP5 projections under four Representative Concentrative Pathways (RCPs) of 2.6, 4.5, 6.0, and 8.5 and based on differential CO2 emission scenarios. The present study also used the Research Moored Array for African–Asian–Australian Monsoon Analysis and Prediction (RAMA) buoys located in the central Bay of Bengal in order to validate and skill assess the CMIP5 wind products under varying RCPs. Besides, an intercomparison exercise that was performed between RAMA buoys data and merged satellite altimeter data from the French Research Institute for Exploitation of the Sea/Laboratory of Oceanography from Space (IFREMER/CERSAT) provided the necessary confidence to ascertain the quality of CMIP5 WS products. The study signifies that a moderate positive correlation was noticed in the WS comparison between CMIP5 GCM products and the RAMA buoys (maximum correlation of .64), and the correlation factor varied between the suite of models used in CMIP5 experiments. This exercise would provide detailed know‐how on the performance of various GCMs and also provide a basis to select the best performing GCMs for the Bay of Bengal region. Analysis of the upper 10% (90th percentile) showed a maximum underestimation/overestimation of 2.5 and 1.5 m·s−1, respectively, for WS comparison between CMIP5 and RAMA buoys data. Probability density of WS data fitted to Weibull distribution showed an increase of moderate WS (6–8 m·s−1) during the period of study. Although the CMIP5 GCMs are not able to represent the contemporary WS climatology satisfactorily, the models such as HadGEM2‐ES, HadGEM2‐CC, CanESM2, and ACCESS‐1.3 showed the best performance concerning near‐surface WS for the Bay of Bengal region.