AbstractRapid and informed response is needed to ensure effective water management during droughts, including reliable and immediate data synthesis, near-real-time forecasting, and model-based decision support for water operations. A service-driven approach has been developed previously to couple river modeling and genetic algorithm (GA) optimization services for determining optimal water allocation strategies under daily drought scenarios. However, the computational effort in handling the constraints, which involves executing computationally intensive models repetitively, is a major obstacle to enabling an effective real-time Web application for decision support. The objective of this work is to develop a computationally efficient metamodel approach to reduce the computational burden of the simulation-optimization model. Two types of metamodels are developed: a pretrained metamodel that is built offline before the optimization and an adaptive metamodel that is built and updated during the optimization. The metamodel is a classifier algorithm that evaluates whether a constraint is satisfied, which simplifies the prediction and leverages the metamodel’s role in water management. The metamodel framework was tested for a drought event in the Upper Guadalupe River Basin, Texas, in April 2015 and the performance of the different approaches is compared. A conservative model, which narrows the feasible region by increasing the constraint probability threshold, is needed with the pretrained metamodel to ensure convergence to a feasible near-optimal solution, but not for the adaptive metamodel. The results also show that the adaptive metamodel GA performs best in model accuracy and reduces computation time by 58%, compared with the pretrained metamodel GA, which reduces computation time by 78% but does not reliably obtain the optimal solution. The approach also does not require users to select classifiers, tune parameters, or execute simulation models offline. Therefore, a prototype Web interface is implemented that uses the best-performing adaptive metamodel approach to more efficiently assist decision makers with real-time drought management.