| Despite a number of benefits for seasonal streamflow forecasts, macroscale hydrologic models (MHMs) remain underutilized by the operational community, due partly to misalignments between experimental and operational methodologies. This research addresses this problem through a series of conceptual frameworks that leverage unexploited strengths of MHMs and are demonstrated in the western United States. A hybrid dynamical-statistical approach is developed and tested in the 14 major watersheds of California's Central Valley drainage. The approach employs gridded precipitation and snow water equivalent (SWE) simulated by the Variable Infiltration Capacity (VIC) MHM as predictors in equations generated via the principal components regression methodology of the Natural Resources Conservation Service (NRCS). Results offer improvement over forecasts issued by California's Department of Water Resources, with particular benefits in watersheds poorly sampled by observations.;The approach is then modified to capitalize on its ability to identify locations with strong predictive power for network design applications. The modified approach is applied toward the expansion of the NRCS SNOTEL network in 24 western U.S. basins using two forecasting scenarios—one assuming the conventional predictors of SWE and precipitation, and one considering soil moisture as an additional predictor. Results indicate that, for most basins, substantial skill improvements are only possible when soil moisture is considered. Furthermore, locations identified as optimal for soil moisture sensors are primarily found at low- to mid-elevations, in contrast to the higher elevations typically occupied by SNOTEL stations.;Finally, the significance of groundwater for seasonal streamflow forecasts is assessed by evaluating its contribution to interannual streamflow anomalies in the 29 Colorado River sub-basins. Terrestrial water storage changes are simulated by a version of VIC modified to include an underlying aquifer. These estimates are evaluated with satellite data and basin-scale water balances derived from observations. Simulated groundwater storage changes are then compared to those derived via baseflow recession analysis. Statistical analyses reveal little relationship between groundwater and streamflow anomalies, suggesting that operational forecasts are likely not degraded by the omission of groundwater conditions for the Colorado River. Viewed collectively, the research provides conceptual and operational contributions toward the adaptation of MHMs to a forecasting environment. |
