Comparing various methods for the regionalization of model parameters in the Sacramento Soil Moisture Accounting Model and Snow Accumulation and Ablation Model

Calibration of river forecast models is a long, laborious process for the over 750 basins in the North Central River Forecast Center’s 285,000 square mile area. Automatic parameter optimization techniques are being tested to speed the calibration process while still providing robust and unbiased parameter values. In order to test the use of the Automatic Parameter Optimization program (OPT3) on the regionalization of calibration parameters, the North Central River Forecast Center (NCRFC) calibrated two river forecast system model components on a headwater basin in the Grand River watershed in Michigan. The two model components calibrated were the Sacramento Soil Moisture Accounting Model (SAC-SMA) and Snow Accumulation and Ablation Model (SNOW-17). The headwater basin was calibrated using traditional manual estimation methods and a three-step automatic calibration scheme with OPT3 (Hogue, 1999). The two parameter sets were then regionalized over hydrologically similar basins in the area using several different methods. A comparative analysis was conducted on a direct regionalization with no further calibration, a manual regionalization with limited further calibration, and three separate OPT3 methods of regionalization. Initial results indicate parameter estimation using an automatic calibration scheme, combined with automated regionalization techniques, have the potential to significantly reduce the amount of time necessary to calibrate model components.