The volume of water reaching reservoirs during the April-July growing season is critical to meeting water demands for agriculture and other human demands. However, our ability to forecast seasonal water supplies is hindered by extreme and changing snowpack. In this research, we investigate how current water supply forecasts will be impacted by a future with less and earlier snowmelt and what can be done to improve those forecasts. Our analysis over 30+ years shows that statistical regression models are generally more skillful than more complex, conceptual models. However, our results suggest that statistical models are less skillful in low snowpack (i.e. snow drought) years than the conceptual models. Results show a 10-20% loss of skill by the second half of the 21st century, and mid-elevations (1000-1700 meters) most vulnerable to loss of snowpack. As a second part of the project, we evaluate mitigation strategies to buffer loss of forecast skill through the introduction of supplemental observations: 1) basin-wide snowpack measurements) and 2) soil moisture station observations. Remotely sensed snowpack information supplements limited field stations, which buffers loss of skill by an average of 40% before 2050. The role of snow in Sierra Nevada declines by late century, however, there will be potential for the inclusion of soil moisture observations to track the water that was previously stored in snowpack. Together our results suggest that declines in reservoir forecasting skill can be buffered by new observations and that investment in distributed snowpack remote sensing and new ways to track soil moisture storage should be done in concert with improving forecast models. Given the reliance on water supply forecasting in the Sierra Nevada for water allocations, hydropower, and drought management, targeted investments in new observation technologies should be initiated to give sufficient time to test and integrate into public forecasting tools.