Understanding rainfall spatial variability in southeast USA at different timescales

This study seeks to understand the spatial variability of monthly and daily rainfall in Alabama, Georgia, and Florida, USA. Monthly spatial statistics are needed to improve downscaling from climate models producing seasonal rainfall forecasts, and spatial correlation of daily rainfall is needed to inform spatial weather generators used in climate risk analysis. We first determined the historical record length that is stationary followed by an analysis of the monthly spatial characteristics of rainfall variables. Rainfall data from 523 weather stations (National Climate Data Center) were obtained for the period 1915-2004 and divided into 15-year subsets for comparisons. Differences in rainfall were found between the most recent 15-year period and all others occurring during the 90-year period of record. Thus only data from 1990 to 2004 (208 weather stations) were used to avoid the detected changes in climate in the region. Correlation, covariance and variance matrices of daily and monthly rainfall amounts were calculated at monthly steps. The same statistics were also computed for frequency of rainfall events and monthly number of rainy days. Results show different spatial patterns at different temporal scales; two spatial patterns were well established. A widely spread correlation in a northeast - southwest direction was found around weather stations during the frontal rainy season, and a concentric short distance-decay in correlations existed around weather stations during the summer convective season. Spatial correlations among daily rainfall amounts are needed for spatial weather generators in a storm-by-storm basis while monthly spatial statistics are needed to ensure the validity of downscaled data from numerical seasonal rainfall forecasts. Copyright (c) 2006 Royal Meteorological Society