A two-component daily event-based precipitation model previously analyzed only at isolated watersheds is "regionalized" by logically grouping points over Southeast Arizona into a small number of geographic clusters. Based on their grade of membership in successive similarity functions, fuzzy set theory is used to cluster 26 rainfall stations with 19 years of historical data into four groups representing relatively homogeneous probability distributions of rainfall interarrival time and amount per event. Geographic cluster subregions are identified and boundaries are drawn with such ancillary indicators as physiography, vegetation, and seasonal rainfall. Precipitation seasons are determined, theoretical probability distributions are hypothesized, and parameters are estimated by moments for each process in the model. An inverse distance algorithm is used to vary parameters within a cluster based on elevation, latitude, and longitude. An isolated validation of the regional technique is performed at an additional station with records.
展开▼
