Evaluation of a Distributed Streamflow Forecast Model at Multiple Watershed Scales

Traditionally, streamflow forecasts have been made at large watershed scales (> 500 km²).  However, demand for flood forecast information at smaller and smaller scales is increasing as floods become more common and communities more vulnerable.  Unfortunately, streamflow information on scales less than 500 km² is limited, primarily due to infrastructure and maintenance costs required to keep these networks functioning properly.  Also, changes occurring due to land use and climate are making prediction in ungauged basins highly uncertain, posing significant problems for forecasting.  The purpose of this work is to evaluate the skill of the National Weather Service (NWS) Hydrology Laboratory Research Distributed Hydrologic Model (HL-RDHM), the spatially distributed version of the Sacramento Soil Moisture Accounting (SAC-SMA), at various watershed scales to better understand how it might perform on small ungauged basins.  We evaluate the performance of the model at two City of Ames and 11 Iowa Flood Information System (IFIS) continuous stage sensors sites placed throughout the Squaw Creek, IA study area, a 528 km² watershed.  Rating curves constructed using LiDAR data were field verified.  Using calibrated parameter adjustments found in a prior study, the model is run on a 4km spatial scale at a 1 hour time step from 2007 to mid-2016.  Preliminary model analysis indicates good accuracy for streamflow timing, but errors in the magnitude for several locations. In particular, skill declines significantly at watersheds less than 50 km² in size.