New Look at Regional Flood‐Frequency Relations for Arid Lands

In the southwestern United States, flood‐frequency relations for streams that drain small arid basins are difficult to estimate, largely because of the extreme temporal and spatial variability of floods, many years of no flow, and short periods of systematic records of annual peaks. A new method is proposed that combines records for several streamflow‐gaging stations, as in the station‐year approach, and produces regional flood‐frequency relations using an iterative regression technique. This technique eliminates the need to extrapolate the flood‐frequency relation to the flood probability of interest. The resulting multiparameter regional flood‐frequency relation is based on all the available annual peak‐flow data. The method was applied to a group of records from 42 gaging stations in Nevada with many years of no flow and with many poorly defined flood‐frequency relations. One‐ and two‐parameter models were developed in which much of the variance in peak discharge is explained by drainage area. The log‐Pearson type III and Weibull probability distributions were used in the models. Part of the error is directly assessed using randomly selected subsamples of the annual peak discharges.