The accuracy of ecological flow metrics derived using a physics-based distributed rainfall-runoff model in the Great Plains, USA

Abstract

The development of a hydrologic foundation, essential for advancing our understanding of flow-ecology relationships, was developed using the high-resolution physics-based distributed rainfall–runoff model V flo in a semi-arid region. We compared the accuracy and bias associated with flow metrics that were generated using V flo , gauge data, and drainage area ratios at both a daily and monthly time step in the Canadian River basin, USA. First, we calibrated and applied bias correction to the V flo model to simulate streamflow at ungauged catchment locations. Next, flow metrics were calculated using simulated and observed data from stream gauge locations. We found discharge predictions using V flo were more accurate than drainage area ratios. General correspondence between predicted discharge and the gauge data was apparent; however, flow metrics calculated using the V flo output did not accurately represent flow variability. Results from the V flo model showed systematic discharge over-predictions in the upper basin and isolated over-predictions in the lower basin, likely due to hail events and sparse rainfall data across the large catchment. Goodness-of-fit statistics (Nash–Sutcliffe efficiency, root-mean square error, and the coefficient of variation) indicated the drainage area ratio and V flo were more accurate at a monthly rather than daily time step, even after quantile mapping. This finding limits the number of streamflow metrics available to develop ecological models, but more importantly, the coarser resolution may hinder our understanding of ecological processes that occur at a submonthly time step. Our approach provides a framework for selecting flow metrics that best represent hydrologic patterns across a large semi-arid catchment with the necessary accuracy to address the ecological questions of interest.