Droughts continue to be a major natural hazard and mounting evidence of global warming confronts society with a pressing question: Will climate change aggravate the risk of drought at local scale? Are current infrastructure and their operation enough to mitigate the damage of future drought, or do we need in-advance infrastructure expansion for future drought preparedness? To address these questions, this study presents a decision support framework based on a coupled simulation and optimization model. A quasi-physically based watershed model is established for Frenchman Creek Basin (FCB), where groundwater based irrigation plays a significant role in agriculture production and local hydrological cycle. The model, revised from the Soil and Water Assessment Tool (SWAT), simulates the dynamic response of aquifer and base flow to groundwater pumping. The physical model is used to train a statistical surrogate model, which predicts the watershed responses under future climate conditions. The statistical model replaces the complex physical model in the simulation-optimization framework, which makes the models computationally tractable. Decisions for drought preparedness include traditional short-term tactical measures (eg facility operation) and long-term or in-advance strategic measures, which usually require capital investment. A scenario based three-stage stochastic optimization model assesses the roles of strategic measures and tactical measures in drought preparedness and mitigation. Considering uncertainties involved in different climate prediction horizons, the model results show the relative roles of mid-and long-term investments and the complementary relationships between wait-and-see decisions and here-and-now decisions on infrastructure expansion and irrigation system operations. Infrastructure expansion is preferred for the long-term plan than the mid-term plan, ie, larger investment is proposed in 2040s than the current, due to a larger likelihood of drought in 2090s than 2040s.