A fuzzy rule-based approach to drought assessment

A methodology for predicting regional droughts from atmospheric pressure patterns is presented. Drought characteristics are strongly related to general circulation patterns (CP). CPs are determined from daily atmospheric pressure data. The link between large-scale CPs and regional scale droughts is modeled using a fuzzy rule-based approach. A fuzzy rule-based model operates on an ''if'' --> ''then'' principle, where ''if'' corresponds to a vector of fuzzy inputs and ''then'' corresponds to some fuzzy consequences. The rules are derived from a so-called training set which includes a daily time series of CP classes and a corresponding monthly sequence of Palmer Drought Severity Indices (PDSI). Split sampling of historical data available for a 35-year time period is used to derive and then to validate the rules. Then, these fuzzy rules may be applied to predict droughts in terms of atmospheric circulation patterns. The occurrence and persistence of CPs are expected to vary under global climate change. Thus the approach may also be useful in estimating the potential impact of climatic change (e.g., 2 x CO2 scenario) on droughts. The methodology is illustrated using drought index data from New Mexico and atmospheric pressure data over the western United States.