Performance Evaluation of Lake Shelbyville by Stochastic Dynamic Programming

Abstract
The operation policy of Lake Shelbyville, Illinois has been modified several times in the past 20 years due to unusual weather conditions and changes in basin‐wide water resources development. A modified stochastic dynamic programming (SDP) model, which accounts for the unrepeatable agricultural and property damages and improves the accuracy of these damage estimates, is used to evaluate the performance of Lake Shelbyville. The optimal pool levels in the summer months are found to be 2–5 ft lower than the current target level of 599.7 ft. The expected damages in the model increase by 9% when a penalty function is used to force the summer pool to the current target. Moreover, it would take more than one month for Lake Shelbyville to resume the summer pool from the winter drawdown. A transition period of at least two months for lake levels between winter and summer appears warranted.