Identifying predictors for regional climate in the Apalachicola-Chattahoochee-Flint (ACF) and Alabama-Coosa-Tallapoosa (ACT) River Basins

Abstract:

Statistical climate models of a regional scale are useful for understanding local processes affecting climate and to improving seasonal forecasting. We explore local historical weather station precipitation and temperature measurements from the United States Historical Climatology Network (1895-2012) for 21 stations in the study area. Teleconnections considered as predictors include: Southern Oscillation Index (SOI), Atlantic Multidecadal Oscillation (AMO), Pacific Decadal Oscillation (PDO), Tropical North Atlantic (TNA), Pacific North American pattern (PNA), North Atlantic Oscillation (NAO), Arctic Oscillation (AO), and Bermuda High Index (BHI). Wilcoxon rank sum testing revealed a strong effect of ENSO on precipitation in the southern part of the study area and a weaker or opposite effect in middle to northern parts of the study area. Warm AMO and warm PDO phases greatly reduced the effect of El Nino rainfall in the study area. Rainfall, temperature and climate indices were examined using nonlinear time series analysis including singular spectrum analysis, convergent cross mapping and granger causality. Aside form seasonality we did not identify strong nonlinear effects. Seasonality was found to be an important factor for selecting lagged predictors in a multivariate time series regression model. Time series clustering was used to reduce the study area to four regions.