ABSTRACT

Streamflow-characteristic estimation methods for unregulated rivers and streams of Tennessee were developed by the U.S. Geological Survey in cooperation with the Tennessee Department of Environment and Conservation. Streamflow estimates are provided for 1,224 stream sites. Streamflow characteristics include the 7-consecutive-day, 10-year recurrence-interval low flow, the 30-consecutive-day, 5-year recurrence-interval low flow, the mean annual and mean summer flows, and the 99.5-, 99-, 98-, 95-, 90-, 80-, 70-, 60-, 50-, 40-, 30-, 20-, and 10-percent flow durations. Estimation methods include regional regression (RRE) equations and the region-of-influence (ROI) method. Both methods use zero-flow probability screening to estimate zero-flow quantiles. A low flow and flow duration (LFFD) computer program (TDECv301) performs zero-flow screening and calculation of nonzero-streamflow characteristics using the RRE equations and ROI method and provides quality measures including the 90-percent prediction interval and equivalent years of record. The U.S. Geological Survey StreamStats geographic information system automates the calculation of basin characteristics and streamflow characteristics. In addition, basin characteristics can be manually input to the stand-alone version of the computer program (TDECv301) to calculate streamflow characteristics in Tennessee.

The RRE equations were computed using multivariable regression analysis. The two regions used for this study, the western part of the State (West) and the central and eastern part of the State (Central+East), are separated by the Tennessee River as it flows south to north from Hardin County to Stewart County. The West region uses data from 124 of the 1,224 streamflow sites, and the Central+East region uses data from 893 of the 1,224 streamflow sites. The study area also includes parts of the adjacent States of Georgia, North Carolina, Virginia, Alabama, Kentucky, and Mississippi. Total drainage area, a geology factor, a climate factor, and two soil factors were used as explanatory variables in the RRE equations. Average deleted-residual prediction errors for the West-region RRE equations were 18 and 123 percent for the 10-percent flow duration and 7-consecutive-day, 10-year recurrence-interval low flow, respectively. Average deleted-residual prediction errors for RRE equations were 21 and 89 percent for the same respective flow quantiles in the Central+East region.

The ROI method calculates unique multivariable regression equations for a site of interest using the flow and basin characteristics of 45 similar streamflow-data sites selected from the same region. These 45 sites are selected using a metric that measures similarity between the site of interest and the streamflow-data sites based on total drainage area, geology factor, and climate factor. The ROI method estimates streamflow characteristics using total drainage area, geology factor, and a soil factor as explanatory variables.

Average deleted-residual prediction errors for the West-region ROI equations were 18 and 125 percent for the 10-percent duration and 7-consecutive-day, 10-year recurrence-interval low flow, respectively. Average deleted-residual prediction errors for ROI equations were 20 and 85 percent for the same respective flow quantiles in the Central+East region. In general, when compared to the RRE equations, the ROI method performs similarly in the West region and reduces streamflow-characteristic prediction errors by an average of about 7 percent in the Central+East region of Tennessee.

First posted December 28, 2009