Today's climate models have difficulty accurately simulating intraseasonal variability in the tropical atmosphere. For example, the simulated amplitude of the 30-60-day Madden-Julian Oscillation (MJO) is too weak, and its propagation speed too rapid—errors which may result from deficiencies in model physics (e.g. its cumulus parameterization), resolution, mean state, or air-sea coupling. In climate simulations, it is difficult to identify the primary model deficiencies responsible for such shortcomings, since their effects tend to be masked by nonlinear interactions of multiple processes and the resulting compensation of errors.

To address this issue, the USDOE-sponsored project known as CAPT (Climate Change Prediction Program-Atmospheric Radiation Measurement Program Parameterization Testbed) has developed a protocol for running a climate model in weather-forecast mode. We identify incipient model errors (i.e. those developing prior to significant process interaction) by comparing the short-range forecasts of atmospheric variables with highfrequency verification data provided by reanalyses, satellites, and field campaigns such as the Tropical Ocean Global Atmosphere Coupled Ocean-Atmosphere Response Experiment (TOGA-COARE). Analyzing these forecast errors may permit more direct identification of model deficiencies. In this study, we analyze the simulated tropical intraseasonal variability of two U.S. AGCMs: version 3 of the NCAR Community Atmosphere Model (CAM3) and version 2 of the GFDL Atmospheric Model (AM2). After daily initialization of the models with realistic global state variables approximated by the ERA-40 reanalyses during the TOGA-COARE Intensive Observing Period (1 November 1992 to 28 February 1993), we generate a series of short-range weather forecasts, and verify these against available observational data. To assist our analysis of the CAM3, we also employ its single column model which we force with the analysis of TOGA-COARE data. We will present results of our forecast-error analysis for CAM3 and AM2, discussing potential parameterization deficiencies in each model. We also will compare these results to the intraseasonal variability of the models when run in climate mode and for the CAM3 we will contrast the results obtained with alternate parameterizations of convection

This work was performed under the auspices of the U.S. Department of Energy by University of California, Lawrence Livermore National Laboratory under Contract W-7405-Eng-48. UCRL-ABS-223235