Ideally, the difference between model forcing data used in calibration and operations is minimal. This is accomplished in the MARFC calibration by using combinations of operational precipitation data (gage and radar derived) and reanalysis data archives (analysis of record from the Office of Water Prediction - AORC) combined together to generate basin-averaged precipitation time series. Time series for each basin are compared to forecast group averages to determine consistency through time. Instead of using traditional correction factors, the data sources are interchanged in order to improve this temporal consistency.
We use basin-averaged SWE estimates from our operational archive to help determine if parameter adjustments need to be made to the SNOW-17 model parameters as opposed to runoff parameters in the SAC-SMA model. This archived operational data was created through a forecaster quality control process using the latest observations of snowfall and SWE available at the time of forecasting. These data are useful in understanding the total amount of modeled SWE at the seasonal scale and the correct precipitation type at the storm event scale.
In addition, we use a GIS-based time-area method to generate synthetic unit hydrographs for the channel response function. This method uses a combination of shallow concentrated flow velocities based on NRCS equations and channel velocities from USGS gauges. The channel velocities are based on measurements taken at gage locations in surrounding watersheds and during events that have a return interval of 2 to 5 years.
Greater spatio-temporal consistency in the forcing data and calibrating to manually estimated, basin-averaged SWE estimates helps avoid local overfitting of parameters to bad data. Starting with a basin-wide consistent approach to unit-hydrograph derivation encourages consistent estimation of sub-basin SAC-SMA parameters that control runoff timing. Using these methods to augment the traditional SAC-SMA calibration process, the initial calibration results show substantial average reductions in the variation of monthly biases by 43% and improvement in low-flow simulation biases of 18%. Thus, recalibration benefits are expected not only in flood forecasting but also for partners and the public utilizing forecasted flows for water supply, ecology, and recreation.

