An Empirical Reevaluation of Baseflow Recession Analysis

Baseflow recession analysis is a widely used hydrological tool which relies on readily available discharge measurements to estimate otherwise obscure watershed-scale properties and to predict streamflow during periods of drought. Traditional methods rely on the simplifying assumptions of outflow from a single Dupuit-Boussinesq aquifer, which predicts the slope of the recession curve relating streamflow (Q) to change in streamflow (dQ) to decrease from a value of around 3 during early period recession to around 1 to 1.5 during late period recession. However, many recent studies have shown that the slopes of individual events are often much greater than 3, and generally greater than the slope of the recession point cloud including all events. Also, the prediction of recession slopes decreasing with time (i.e., a concave recession curve) has not been empirically tested. We analyze the recession curves of over 1900 streams for a period of 10 to 95 years from across the continental US. For all streams in this study, we find that the average slope for individual events is significantly greater than that of the point cloud, and generally greater than 3. Further, the recession slopes of individual events decrease with time in fewer than 15% of cases, and instead actually increase with time (i.e., a convex recession curve) in over 80% of cases, contradicting predictions of traditional recession curve analysis. We explore alternative models for describing recession behavior, highlighting recent work on parallel aquifers as offering the simplest model which accommodates these empirical results. Both our empirical results and the parallel aquifers conceptual model imply greater streamflow stability during periods of severe drought than previously predicted.