Author(s): Rita Pongracz; Istvan Bogardi; Lucien Duckstein
Linked Author(s):
Keywords: Drought; Fuzzy rule-based model; Palmer Index; ENSO; Macrocirculation
Abstract: Results of a fuzzy rule-based methodology is presented to forecast drought indices under the continental climate of the U. S. Great Plains from large scale climatic information, namely atmospheric circulation patterns (CP) and the El Nino/Southern Oscillation (ENSO) phenomena. Great Plains droughts have been shown to be strongly related to certain types of persistent meteorological situations, mostly largescale CPs and ENSO. In the present application, input variables of the fuzzy rulebased model are the monthly relative frequencies of daily CP types (6 premises) and the monthly values of SOI with four different lag periods (4 premises). The response or output variable is the Palmer Drought Index, PMDI observed in the climate divisions of Nebraska. The entire data set 1946-1994 containing 10 (=6+4) variables and 245 (=49 years·6 months) observations are split into two parts: training set and validation set. The training set is used to learn the fuzzy rules while the validation set is used to check how the independently constructed rules reproduce the observed PMDI values. The results show that the fuzzy rule-based model reproduces the statistical properties of observed drought index values. Specifically, for all the climate divisions investigated, the empirical and the estimated frequency distribution functions are not significantly different.
Year: 1999