The primary goal of this project is to gain a better understanding of the local and remote sources of tropical biases in climate models through an analysis of local and remote dynamical interactions. The PIs will attempt this by diagnosing both local coupled interactions and remote interactions among 8 geographically localized tropical areas (four in the tropical Pacific, two each in the Indian and Atlantic basins) in observational datasets, in the NCEP/CFS, NCAR/CCSM3, and all available IPCC model simulations, and through additional model integrations of the PIs. The project will have substantial diagnostic and modeling phases. The plan is to estimate local and remote coupling matrices from observational data and all available climate model simulations, and to perform extensive intercomparisons among them. In the modeling phase of the project, they will attempt to reproduce the results obtained in the diagnostic phase with additional integrations of the atmospheric components of the NCAR (CAM3) and NCEP (GFS) models, but now coupled to a mixed layer ocean with simple parameterized Ekman dynamics.