Regression models of ecological streamflow characteristics in the Cumberland and Tennessee River Valleys

Abstract

Predictive equations were developed using stepbackward regression for 19 ecologically relevant streamflow characteristics grouped in five major classes (magnitude, ratio, frequency, variability, and date) for use in the Tennessee and Cumberland River watersheds. Basin characteristics explain 50 percent or more of the variation for 10 of the 19 equations. Independent variables identified through stepbackward regression were statistically significant in 81 of 304 coefficients tested across 19 models (⬚ < 0.0001) and represent four major groups: climate, physical landscape features, regional indicators, and land use. The most influential variables for determining hydrologic response were in the land-use and climate groups: daily temperature range, percent agricultural land use, and monthly mean precipitation. These three variables were major explanatory factors in 17, 15, and 13 models, respectively. The equations and independent datasets were used to explore the broad relation between basin properties and streamflow and its implications for the study of ecological flow requirements. Key results include a high degree of hydrologic variability among least disturbed Blue Ridge streams, similar hydrologic behavior for watersheds with widely varying degrees of forest cover, and distinct hydrologic profiles for streams in different geographic regions.