Targeted Calibrations to Optimize the NWS Research Distributed Hydrologic Model for Runoff Risk Advisory Forecasts

In collaboration with various federal and state agencies, universities, and private industry representatives in Great Lakes states, the North Central River Forecast Center (NCRFC) has developed Runoff Risk tools to serve as guidance for timing of nutrient application on soils. Originally, the Runoff Risk Advisory Forecast (RRAF) was run primarily for Wisconsin and used the Sacramento Soil Moisture Accounting (SAC-SMA) lumped hydrology model. The NCRFC has expanded the RRAF to cover all states bordering the Great Lakes and transitioned from the lumped hydrology model to the Hydrology Laboratory–Research Distributed Hydrologic Model (HL-RDHM) to improve spatial and temporal resolution of the RRAF guidance product. They are currently running the HL-RDHM using the Sacramento Soil Moisture Accounting model with heat transfer and enhanced evapotranspiration (SAC-HTET) and the baseline a priori parameter values provided by the NWS Hydrology Laboratory. The goal of this research project is to perform targeted calibrations to optimize the SAC-HTET a priori parameters in the distributed model in support of the RRAF decision support tool. The targeted calibrations in this project have focused on using scalar multipliers to adjust eleven a priori SAC-HTET parameters within HL-RDHM. The scalar multipliers were calculated using the SAC-SMA lumped model parameters previously calibrated by forecasters at NCRFC and the average of the HL-RDHM a priori parameters over the watershed. This process of calibration forces the average of the distributed parameter over the watershed in the HL-RDHM to match the calibrated value for that same watershed in the SAC-SMA lumped model. Auto-calibration techniques packaged within HL-RDHM have also been used to mathematically optimize the scalar multipliers using observed streamflow. This project compares several calibration techniques and highlights the strengths and weaknesses of these approaches using selected watersheds in Minnesota and Wisconsin. Case study watersheds were selected based on availability of an edge-of-field (EOF) flow measurement site within the watershed, a USGS stream gage near its outlet and a lumped model calibration that was successful in improving forecasts. The study discusses the impacts of watershed versus sub-watershed scale in the targeted calibration process and highlights difficulties in efficient parameter calibration across the multi-state region of the operational RRAF decision support tool.