Applying hillslope-storage models to improve low flow estimates with limited streamflow data at a watershed scale

This study presents a framework to evaluate the performance of rainfall-runoff models for the estimation of low flow at sites with limited streamflow data. Estimates of low flow statistics are important for water supply, waste-load allocation, irrigation, hydropower, and ecological and habitat assessment. Paradoxically most rainfall-runoff models focus on flood simulations and use oversimplified representations of baseflow processes resulting in poor performance simulating low flow statistics. Such baseflow models cannot account for variations in topography and hydrogeology that impact baseflow processes and have limited applicability to evaluate land use and climate change impacts on low flow. Both a hillslope-storage Boussinesq model (hsB) and a kinematic wave hillslope-storage model (kw) have shown good results in simulating baseflow in synthetic hillslopes; one major challenge is how to apply these models in real watersheds. In this study hsB and kw are coupled to the Sacramento Soil Moisture Accounting (SAC-SMA) model and tested at two similarly sized watersheds in North Carolina with different watershed slopes. The partitioned kw and hsB models are also compared to the original SAC-SMA model (Sac) and SAC-SMA applied to a partitioned watershed (Sacm). Both 5 years and 1 year of full and reduced ranges of streamflow data are employed for model calibration. All partitioned models improved their estimation of low flow when calibrated to a lower range of streamflows but with kw and hsB performing slightly better at the steeper sloped watershed. The performance of the coupled models with limited streamflow data is encouraging and can potentially improve the estimation of low flow statistics at sites with limited streamflow data.