Identifying streamgage networks for maximizing the effectiveness of regional water balance modeling

[1] One approach to regional water balance modeling is to constrain rainfall-runoff models with a synthetic regionalized hydrologic response. For example, the Large Basin Runoff Model (LBRM), a cornerstone of hydrologic forecasting in the Laurentian Great Lakes basin, was calibrated to a synthetic discharge record resulting from a drainage area ratio method (ARM) for extrapolating beyond gaged areas. A challenge of such approaches is the declining availability of observations for development of synthetic records. To advance efficient use of the declining gage network in the context of regional water balance modeling, we present results from an assessment of ARM. All possible combinations of ‘‘most-downstream’’ gages were used to simulate runoff at the gaged outlet of Michigan’s Clinton River watershed in order to determine the influence of gages’ drainage area and other physical characteristics on model skill. For nearly all gage combinations, ARM simulations resulted in good model skill. However, the gages’ catchment area relative to that of the outlet’s catchment is not an unquestionable predictor of model performance. Results indicate that combinations representing less than 30% of the total catchment area (less than 10% in some cases) can provide very good discharge simulations, but that similarity of the gaged catchments’ developed and cultivated area, stream density, and permeability relative to the outlet’s catchment is also important for successful simulations. Recognition of thresholds on the relationship between the number of gages and their relative value in simulating flow over large area provides an opportunity for improving historical records for regional hydrologic modeling.