Translating monthly temperature from regional to local scale in the southeastern United States

This study investigates a statistically based scheme to translate climate information from the regional to local scale. Spatially averaged monthly maximum and minimum temperatures were used in a series of regression equations to predict similar variables at 62 stations across the southeastern United States. The spatial average temperature proved to be a good predictor of temperatures at most stations across the region. Model performance varied according to variable, season, and location. The highest model errors were associated with maximum temperature in summer, the lowest with minimum temperature in summer. While no general spatial patterns of errors were found, anomalously high errors appeared at individual stations. Since such anomalies often were associated with a variety of data discontinuities, the models provide a means of identifying whether individual stations are representative of the regional temperature time series. The regression models could also provide greater spatial resolution for climate impact scenarios as their error is usually far lower than the range of projected regional climate change produced by general circulation models.