Lag-time characteristics in small watersheds in the United States

Abstract

Time is an important element in hydrologic design. Most hydrologic models require a watershed characteristic that reflects the timing of runoff. The time parameter used in this study was lag time, defined as the time from the centroid of rainfall excess to the centroid of direct runoff. Lag times were evaluated from rainfall-runoff data in over 40,000 events in 116 small watersheds in the United States. The watersheds ranged from 0.243 to 3490 acres, with periods of rainfall-runoff records from 3 to 58 years. Rainfall-runoff event characteristics were used to determine a unique value of lag time for each storm. A tendency towards a constant value of lag time for the "bigger" storms was observed, "bigger" meaning higher values of either previous 48-hour rainfall, average effective rainfall intensity, average runoff intensity, or peak flow. The variable peak flow best showed this tendency; higher peak flow was associated with constant lag time in over 90% of the watersheds. Several hydrologic relationships involving lag time previously described in the literature were not verified in this study. Watershed characteristics were evaluated as "predictors" of lag time within a given watershed. The geomorphic variables used were area, length, width, slope, and storage coefficient (Curve Number). All variables were significant in explaining the variation of lag time by the regression analysis. The watersheds were divided into groups to try to explain the variation of lag time between watersheds. Management practices, geographical region, and the tendency toward constant value of lag time for the "bigger" storms had significant effects in the regression analyses, whereas land use and hydrologic behavior did not. When only the watersheds with the tendency described above were used, no groupings significantly improved the regression equations. Rainfall-runoff data should be used to compute lag time directly, especially for the bigger storms. If data are insufficient, regression predictions can be improved by grouping watersheds by regions and management practices. Width, slope and Snat are the best variables for prediction of lag time. The multiple linear regression model developed in this study had a higher coefficient of determination than other models in the literature.