Stochastic Modeling of the Effects of Large-Scale Circulation on Daily Weather in the Southeastern U.S.

Abstract

Statistical methodology is devised to model time series of daily weather at individual locations in the southeastern U.S. conditional on patterns in large-scale atmosphere–ocean circulation. In this way, weather information on an appropriate temporal and spatial scale for input to crop–climate models can be generated, consistent with the relationship between circulation and temporally and/or spatially aggregated climate data (an exercise sometimes termed `downscaling'). The Bermuda High, a subtropical Atlantic circulation feature, is found to have the strongest contemporaneous correlation with seasonal mean temperature and total precipitation in the Southeast (in particular, stronger than for the El Niño–Southern Oscillation phenomenon). Stochastic models for time series of daily minimum and maximum temperature and precipitation amount are fitted conditional on an index indicating the average position of the Bermuda High. For precipitation, a multi-site approach involving a statistical technique known as `borrowing strength' is applied, constraining the relationship between daily precipitation and the Bermuda High index to be spatially the same. In winter (the time of greatest correlation), higher daily maximum and minimum temperature means and higher daily probability of occurrence of precipitation are found when there is an easterly shift in the average position of the Bermuda High. Methods for determining aggregative properties of these stochastic models for daily weather (e.g., variance and spatial correlation of seasonal total precipitation) are also described, so that their performance in representing low frequency variations can be readily evaluated.