We propose a process-based and object-based analysis of bias in the simulation of the physical climate. This will bring focus to organized structures that are important components of the climate system and will link spatial and temporal scales normally associated with weather to the climate system. We plan to use existing Clouds and the Earth's Radiant Energy System (CERES) cloud-object data and a number of existing integrations of models developed at NASA Goddard Space Flight Center and the National Center for Atmospheric Research. These models have both climate and forecast configurations, and have been run at a wide range of resolutions, with the finest resolution of about 1/8 degree. We choose two geophysical phenomena as a focus of the investigation: tropical precipitation and clouds, and precipitation in the eastern half of the United States. These are important climate features and are not well represented in current climate models. The investigation of tropical precipitation and clouds will be initiated with the CERES cloud-object data. We will extend the object-based approach to Atmospheric Infrared Sounder (AIRS) radiance observations to take advantage of the rich information content and dense sampling patterns of AIRS. This will be important in the study of eastern United States precipitation. We, also, plan to investigate empirical parameterizations that are anchored to cloud-object information. The purpose of this is to provide insight into role of resolved-scale meteorology versus physical parameterization in model-calculated climate sensitivity. We strive to isolate causal relationships between model formulation and model performance, with the goal of improvement of the robustness of climate predictions, which is inline with the NASA roadmap objectives in particular with "the improvement of the robustness of climate predictions and their use in adaptation to climate change".
