Regression models for predicting on‐site runoff from short‐duration connective storms

On‐site runoff resulting from summer convective thunderstorms was studied in the Walnut Gulch Experimental Watershed, using 6− × 12−foot plots at 2 locations, based on 5 location‐years of data from 34 storms. Average runoff increased as precipitation quantity increased, decreased as crown spread of vegetation increased, and decreased as antecedent soil moisture increased. In a stepwise multiple linear regression equation, these independent variables accounted for, respectively, 72, 3, and 0.5% of the prediction variance. Considering regression equations for any one location‐year, storm amount or intensity always was significant, crown spread usually was significant, and antecedent soil moisture rarely was significant. In simple correlations, antecedent soil moisture was never related significantly to runoff. The equations developed appear valid for a set of thunderstorms with at least one‐sixth of maximum 5‐minute intensities exceeding 3.7 inches per hour. (Key words: Hydrology; runoff; thunderstorms)