Inverse Streamflow Routing: Toward Runoff-Based Calibration of Distributed Hydrologic Models

Spatially-distributed hydrologic models use a set of spatially-varying soil and vegetation parameters to predict the partitioning of precipitation into infiltration, evapotranspiration, and runoff. While computational resources now make application of such models routine, their calibration remains a challenge. Usually, hydrologic models are calibrated by routing runoff downstream and comparing the aggregated runoff to discharge measurements at stream gauges. This aggregated calibration has drawbacks: information about spatially-varying runoff is lost when the runoff is routed through the river network to gauge locations. For practical purposes, parameter estimates remain semi-lumped, resulting in discontinuities between adjacent basins, and the hydrologic model must be run many times, resulting in excessively long computation times. These problems could be mitigated by calibrating to spatially-distributed runoff, were it available. We build on a previously-proposed runoff estimation method called inverse streamflow routing (ISR) to estimate spatially-distributed runoff from streamflow measurements. We show the effect of calibrating the VIC hydrologic model to distributed runoff estimates vs. calibration to streamflow observations in a test case in the Tuolumne River basin. We note that streamflow estimates from the Surface Water and Ocean Topography (SWOT) mission, now planned for launch in early 2022, will allow ISR to be applied in more places with better accuracy due to the large number of "virtual gauges" SWOT discharge estimates will provide.