Multi-model (dynamical and statistical) high-resolution seasonal climate forecasting system: An application to the southeast US

Construction of the multi-model high-resolution seasonal forecasting system is discussed in this study. A set of dynamical regional climate models and statistical downscaling models are considered for the construction of this system. In the present study, coarsely resolved seasonal integrations of surface air temperature (2 meter height) and precipitation from the global climate model are downscaled to a fine spatial scale of ~20 km for crop growing seasons (March through September), which have been extremely challenging for the southeast United States region. Regional climate models (dynamical downscaling) and statistical downscaling models are applied for the region, covering Florida, Georgia, Alabama, South and North Carolina. Dynamical regional climate models considered are the Florida State University/Center for Ocean-Atmospheric Prediction Studies Nested Regional Spectral Model (FSU/COAPS NRSM), Experimental Climate Prediction Center Regional Spectral Model (ECPC RSM), Regional Climate Model 3 (RegCM3), and Weather Research and Forecasting Model (WRF). We additionally present a set of statistical downscaling models. The simple localization with sophisticated bias-correction, an advanced eigentechnique with multiple regression, nonlinear canonical correlation analysis, and geo-spatial weather generator are bases of the statistical downscaling models developed in this study. We present how much improvement of predictability skill on seasonal anomaly and frequency of subseasonal extremes is expected from this forecasting system.