Diagnosing MJO forecast biases in the NCAR Community Atmosphere Model during the DYNAMO field campaign

This study investigates the MJO forecast biases, especially at the initiation stage, of the Community Atmosphere Model CAM3. We diagnose the dynamic and thermodynamic forecast misfits in a version of CAM3 for the MJO event observed during the DYNAMO field campaign during October 2011. The CAM3 forecasts were initialized from ECMWF Reanalyses fields and the MJO forecast skill was analyzed for daily and climatological SST boundary conditions. Further, CAM forecasts were nudged towards ECMWF Reanalyses fields to obtain nudging increments during the forecast period. These nudging increments were then analyzed to diagnose deficiencies in model physics and thermodynamic structures in the model as observed in the MJO. The forecast experiments show that the initiation of the MJO event is captured well in the runs forced with daily SST. The nudging experiments reveal that the model has too strong a zonal shear during the evolution of the MJO and the heating structure in the model does not emulate the observed top heavy baroclinic heating structure of the MJO. We conclude with a discussion of suggested improvements to the deep convection scheme in the context of tropical convection and the MJO prediction.