As concerns about water availability increase, there is a growing interest in evaluating potential climate impacts to water users. Watershed modeling plays a central role in these discussions by producing estimates for available water (e.g., streamflow, streamflow), which can be used to inform water management strategies. Current techniques for modeling streamflow and water allocations vary, from process-based simulations that take advantage of domain-informed approaches for hydrologic cycling to data-driven models that leverage the underlying statistical properties of the historical record. In this talk, we will summarize some of the current data-driven techniques to simulate streamflow and assess allocations to different users. We will then describe how we are building on these techniques through: 1) the application of Hidden Markov Models to simulate streamflow under both historic and possible future climate scenarios and 2) the implementation of machine learning techniques (e.g., neural networks) to learn water management strategy for the allocation of water across users during different scenarios. We demonstrate the utility of these methods in the Colorado River Basin of Texas, including how we are modifying these off-the-shelf techniques to be sensitive to water management nuances (e.g., explicit representation of the spatiotemporal correlations and resource constraints). We will also note important socio-hydrological nuances that could influence the adoption or application of these methods, especially given the use of water across multiple sectors and scales. Such systems thinking will be critical for helping to drive further insights into different approaches for watershed modeling in support of robust and resilient water management under a changing climate.