Development of regression models to estimate flow duration statistics at ungaged streams in Oklahoma using a regional approach

Multiple-regression analysis was used to develop equations for estimating annual and seasonal flow-duration statistics at ungaged streams in and near Oklahoma that are not substantially affected by human alteration. Ordinary least-squares and left-censored (Tobit) multiple-regression techniques were used to develop equations that relate these statistics, from continuous streamflow data at gaged locations with 10 or more years of record, to physical and climatic basin characteristics. Separate equations were developed to estimate these statistics for stations within similar hydrologic and geologic regions. Use of separate regressions by region substantially improved the accuracy of the estimate for streams in eastern and central Oklahoma when compared with estimating equations developed for the entire State, especially for regressions estimating lower flow duration values. For all regions, the equations were more reliable for estimating higher flow duration values. The accuracy of regressions for estimating flow duration statistics in western Oklahoma was very poor, especially for lower flow duration values. ?? 2009 ASCE.