|Van Liew, Michael|
Submitted to: Journal of the American Water Resources Association
Publication Type: Peer Reviewed Journal
Publication Acceptance Date: July 1, 2006
Publication Date: October 1, 2006
Citation: Garbrecht, J.D., Schneider, J.M., Van Liew, M.W. 2006. Monthly runoff predictions based on rainfall forecasts in a small watershed in Oklahoma. Journal of the American Water Resources Association. 42(5):1285-1295. Interpretive Summary: Seasonal precipitation forecasts are made available on a regular basis and offer the opportunity to forecast streamflow. However, the usefulness of such streamflow forecasts for planning and management of reservoir storage has not been established. Here the usefulness was investigated for a small watershed in central Oklahoma. Streamflow response to hypothetical precipitation forecasts was simulated by a computer for wet, average, and dry watershed conditions, and wet, average, and dry forecasts lasting one to three months. The results suggest streamflow forecasts have potential usefulness for reservoir management, but only under certain conditions. Under pronounced wet watershed conditions, the impact of a forecast on streamflow was greatest, while under dry conditions, the majority of forecasted precipitation was absorbed in the soil profile, producing little change in streamflow. Also, forecasts that lasted three months produced a stronger impact on streamflow than one-month forecasts. Thus, the greatest potential for streamflow forecasts for decision support were under wet watershed conditions when the forecast signal was least dampened by soil-storage effects and for precipitation forecasts that persisted over several months. However, the forecasts are expressed in complex mathematical terms and their use in practical applications remains a challenge.
Technical Abstract: It remains an open question whether seasonal probabilistic forecasts of precipitation can be translated into useful streamflow forecasts. The utility of streamflow forecasts was examined for a small watershed in central Oklahoma. Streamflow response to precipitation forecasts was simulated using the hydrologic model SWAT (soil and water assessment tool). Eighteen scenarios were examined, representing combinations of wet, average, and dry antecedent hydrologic conditions, with wet, normal, and dry precipitation forecasts for a single month, and then for a three month period. The results suggest there is potential utility for decision support from precipitation forecasts, but only under certain conditions. Pronounced wet and dry hydrologic antecedent conditions were shown to have greater impact on streamflow volume than precipitation forecasts. As expected, the largest impact of forecasts on streamflow occurred after wet antecedent hydrologic conditions, when the fraction of precipitation contributing to runoff was greatest. Under pronounced dry antecedent conditions, the majority of precipitation was absorbed in the soil profile, producing little immediate impact on streamflow. These nonlinearities in streamflow response to antecedent conditions and forecasted precipitation suggest a highly asymmetric utility function for precipitation forecasts. Persistent three-month forecasts produced cumulative effects with a stronger impact on streamflow than one-month forecasts. Finally, forecasts of streamflow were presented in terms of probability of exceedance curves, providing a convenient way to illustrate projected changes in streamflow in response to precipitation forecasts.