Dynamically and statistically downscaled seasonal temperature and precipitation hindcast ensembles for the southeastern USA

摘要

We present results from a 15-year 10-member warm season (March-September) hindcast ensemble of maximum and minimum surface air temperatures and precipitation in southeast USA. The hindcasts are derived from the Florida State University/Center for Ocean-Atmospheric Prediction Studies Global Spectral Model (FSU/COAPS GSM) and downscaled using, both the FSU/COAPS Nested Regional Spectral Model (NRSM) and a statistical downscaling method based on stochastic weather generator techniques. We additionally consider statistical bias correction of the dynamical model output. Basic descriptive statistics indicate that the bias-corrected and statistically downscaled data reduce the FSU/COAPS GSM bias considerably in terms of basic climatology. Statistics describing the daily precipitation process are improved by both downscaling techniques relative to the bias-corrected GSM. Improvement in monthly and seasonal hindcasts relative to FSU/COAPS GSM is spatially and temporally varying. Precipitation hindcasts are generally less skillful than those for temperature, although useful precipitation predictability exists at many locations. Hindcast improvements due to downscaling are greatest over peninsular Florida. The smallest root mean square errors (RMSE) for temperature hindcasts are found in the southern part of the study region during the spring months of March, April and May (MAM) for maximum surface air temperature, and in the summer, June, July and August (JJA), for minimum Surface air temperature. Overall. there is no indication that either downscaling method has a direct advantage over the other. Copyright (c) 2008 Royal Meteorological Society