Drought prediction is a major challenge among the hydrological and meteorological community. Although directly related to water deficiency, the meaning of drought varies in different geographical areas and within communities that use water for different purposes. Given that the time scales of the onset and recovery of different types of drought are variable, it is difficult to devise one universal drought index for the assessment and prediction of all kinds of drought. The relationships between meteorological deficits and hydrological deficits vary in time and space, and common precipitation-based indices do not capture this variation. Physically-based hydrology model can play a significant role in this area by simulating the responses of soil moisture, snow water equivalent and runoff on a continuous, gridded domain, across hydro-climatic gradients. We describe the implementation of one such model in a web-based, drought-focused hydrologic monitoring and prediction system for Washington State. The products include current percentiles of soil-moisture and accumulated runoff (at different time scales) as well as more traditional drought indicators such as PDSI. Comparison of the various VIC model-based drought proxies with the operationally-produced indices demonstrate that model-based indices are a potentially viable tool in characterizing different aspects of drought. The system updates in near real time (daily to weekly intervals, depending on the analysis product), but also offers a long term retrospective archive for perspective on current conditions.