Subsurface Warm Biases in the Tropical Atlantic and Their Attributions to the Role of Wind Forcing and Ocean Vertical Mixing

Abstract

Abstract Realistic ocean subsurface simulations of thermal structure and variation are critically important to success in climate prediction and projection; currently, substantial systematic subsurface biases still exist in the state-of-the-art ocean and climate models. In this paper, subsurface biases in the tropical Atlantic Ocean (TA) are investigated by analyzing simulations from the Ocean Model Intercomparison Project (OMIP) and conducting ocean-only experiments that are based on the Parallel Ocean Program, version 2 (POP2). The subsurface biases are prominent in almost all OMIP simulations, characterized by two warm-bias patches off the equator. By conducting two groups of POP2-based ocean-only experiments, two potential origins of the biases are explored, including uncertainties in wind forcing and vertical mixing parameterization, respectively. It is illustrated that the warm bias near 10°N can be slightly reduced by modulating the prescribed wind field, and the warm biases over the entire basin are significantly reduced by reducing background diffusivity in the ocean interior in ways to match observations. By conducting a heat-budget analysis, it is found that the improved subsurface simulations are attributed to the enhanced cooling effect by constraining the vertical mixing diffusivity in terms of the observational estimate, implying that overestimation of vertical mixing is primarily responsible for the subsurface warm biases in the TA. Since the climate simulation is very sensitive to the vertical mixing parameterization, more accurate representations of ocean vertical mixing are clearly needed in ocean and climate models. Significance Statement The purpose of our study is to analyze the characteristics of subsurface temperature biases in the tropical Atlantic Ocean and to investigate the causes for the biases. This is important because subsurface biases greatly reduce the reliability of models in climate prediction and projection. It is found that significant subsurface warm biases arise in 100–150 m over the entire tropical Atlantic basin and the biases are mainly attributed to overestimated ocean vertical mixing. Our work highlights that subsurface ocean simulations are highly sensitive to vertical mixing parameterization, and further research is necessary for more accurate representations of ocean vertical mixing in ocean and climate modeling.