Abstract Author: Young-Kwon Lim, D. W. Shin, S. Cocke, T. E. LaRow, J. J. O'Brien, and E. P. Chassignet
Abstract Title: Statistical downscaling of NCEP CFS retrospective forecasts for regional climate simulation over the southeast United States
Abstract: The large-scale surface climate simulation for summer has been downscaled to local spatial scale of 0.2°×0.2° (~20 km) over the southeast US region, covering Florida, Georgia, and Alabama. The regionalization from the global model simulation is conducted by statistical methods. Daily surface temperature (Tmean) and precipitation data obtained from the NCEP/CFS retrospective forecasts (2.5° lon.-lat. resolution) are employed for this downscaling. In this study, large scale seasonal integrations starting from consecutively different 10 datas, consisting of 10 ensemble member runs, for each year for the period of 1987 to 2005 are used for downscaling.
The statistical downscaling is conducted based on the clearer separation of prominent local climate signals (e.g., seasonal cycle, dominant intraseasonal or interannual oscillations) over the training period. The statistical information identified from training leads to better prediction of local climate scenario from the large-scale simulations. CSEOF (Cyclostationary EOF) analysis, multiple regression, and stochastic time series generation methods are primarily used for the statistical downscaling under cross-validation framework.
Downscaled temperature and precipitation are compared with the NCEP/CFS large-scale fields and observations. The present talk discusses the skill of downscaling from the NCEP/CFS via a series of evaluations including mean absolute errors, anomaly correlations, frequency of extreme events (heat wave, heavy rainfall, and dry spell), and categorical predictability.