An assessment of CMIP5 climate models for the ocean temperature and salinity

Abstract

As a great reservoir of thermal energy, ocean plays a crucial role in determining climate sensitivity to an external forcing. A change in ocean heat storage can cause a change in global mean surface temperature, a major indicator of global warming. Although the CMIP5 (fifth phase of Coupled Model Intercomparison Project) climate models have been widely used to understand and project a climate change, an assessment for the ocean simulations has been limited to upper ocean including sea surface temperature and surface heat fluxes or to the changes in those variables. The purpose of this study is to make an assessment of the CMIP5 models on the ocean climatology with temperature and salinity. As a preliminary result, the median of 37 CMIP5 models, compared with a reanalysis data for the 30years of 1975 to 2004, showed a relatively large warm bias greater than 2 degrees in the Atlantic Ocean below 500 m and in the Indian Ocean above 500 m. These warm biases are well correlated with a higher salinity in those regions. Further analysis results will be presented and our results may provide information on the CMIP5 models systematic bias and simulation uncertainty.warming. Although the CMIP5 (fifth phase of Coupled Model Intercomparison Project) climate models have been widely used to understand and project a climate change, an assessment for the ocean simulations has been limited to upper ocean including sea surface temperature and surface heat fluxes or to the changes in those variables. The purpose of this study is to make an assessment of the CMIP5 models on the ocean climatology with temperature and salinity. As a preliminary result, the median of 37 CMIP5 models, compared with a reanalysis data for the 30years of 1975 to 2004, showed a relatively large warm bias greater than 2 degrees in the Atlantic Ocean below 500 m and in the Indian Ocean above 500 m. These warm biases are well correlated with a higher salinity in those regions. Further analysis results will be presented and our results may provide information on the CMIP5 models systematic bias and simulation uncertainty.