A K-Nearest neighbor based stochastic multisite flow and stream temperature generation technique

Hydrologic and climatic uncertainty is increasing in the western United States, and with it the need for models capable of capturing this uncertainty beyond what is seen in the historical record for planning and management purposes. This is especially important for managing water resources on Lake Shasta under water supply and stream temperature constraints. We develop K-nearest neighbor based stochastic simulation methods for daily streamflow and attendant stream temperature at five streams that drain into Lake Shasta. The methods can also generate scenarios conditioned on the larger climate - e.g., extreme wet or extreme dry. The ability of the methods to capture the historical variability of flow and temperature for Lake Shasta is demonstrated. Although, we developed and demonstrated this technique for Lake Shasta, they can be readily applied to any water resource systems.