Multiple regression analysis was used to delineate hydrologically distinct regions in Kentucky, and to develop regression models for estimating peak discharge for unregulated streams in these regions. The regression models provide estimates of flood quantiles with associated average recurrence intervals of 2, 5, 10, 25, 50, and 100 years. The data base used in the analysis included annual peak discharge records (through water year 1985) at 266 continuous- and partial-record gaging stations in, and adjacent to, Kentucky. Selected drainage basin characteristics upstream of each gaging station were used to develop the regression equations. Flood quantiles at the gaged stations were estimated on the basis of log-Pearson Type III distribution and the methodology recommended by the U.S. Water Resources Council. Seven hydrologic regions were delineated in Kentucky on the basis of analysis of residuals from statewide and regional regression models. Regression models for estimating flood quantiles in the hydrologic regions are based on measurements of contributing drainage area, main channel slope, basin shape index, and main channel sinuosity. The regression coefficients indicated an increase in flood discharge with increasing drainage area and channel slope, and a decrease in discharge with increasing channel sinuosity and basin elongation. Accuracy of the discharge estimates from the regression models as measured by the standard error of the estimate ranged from 21 to 52%. The procedures for estimating flood quantiles vary depending on whether the estimate is for an ungaged site, an ungaged site near a gaged site on the same stream, or a gaged site. Also considered is whether the drainage area crosses hydrologic region boundaries or state lines. The methods apply only to natural flow streams drainage areas < 1 ,000 sq mi.
