Process Level Investigation of the Role of Convection and Cloud Parameterization in Tropical Pacific Bias in GFDL Next Generation Global Climate Models

Despite their climatic importance, simulations of tropical Pacific mean climate and variability in state-of-the-art global coupled ocean-atmosphere climate models remains unsatisfactory. During the past few decades, tremendous community efforts have been devoted to understand common problems, especially those associated with models analyzed by the IPCC assessment reports. New theories, hypotheses, and analysis approaches have been proposed. We are at a point when some major progress could happen. At GFDL, recent model development efforts have led to a range of prototype configurations for the next generation GFDL climate model AM4/CM4. These configurations exhibit large differences in tropical Pacific biases, such as the equatorial Pacific cold/dry bias, the double ITCZ, the overly strong trades, and the biases related to MJO. The models include configurations based on previous generation AM3 and HIRAM models, as well as configurations using a new double-plume convection (DPC) scheme under development. Preliminary coupled simulations suggest that DPC produces significant improvements in the tropical Pacific mean climate, ENSO, and MJO in coupled simulations. However, we do not understand at this time the mechanisms by which changes in the convection scheme lead to the improvements, since these improvements result from complex interactions and feedback among convection, clouds, radiation, atmospheric circulation, boundary layer processes, and the underlying ocean processes. Without an in-depth understanding of the mechanisms, new improvement cannot be generalized and transferred to other models.

Given the abundance of theories on tropical Pacific biases in the literature and given that we have various prototype configurations of AM4/CM4, we propose a research project to bring these together in order to strengthen our understanding of the mechanisms by which parameterization details lead to model improvements. We believe that the next generation GFDL climate models could significantly benefit from an integration of the existing theories, hypothesis, analysis approach, as well as new observational data in the area of convection-cloudair- sea interactions into the model development routine. Experiments and analyses will be designed and conducted to connect our model results to theories and conceptual models. Results will be compared with previous analyses to assess their generality for the broader GCM modeling community.