| description abstract | The California State Water Project (SWP) along with the Central Valley Project (CVP), under various environmental regulations, manage California’s complex water storage and delivery system. An important part of the water system regulations strictly limit salinity intrusion into the Sacramento–San Joaquin Delta, which is a complex tidal estuary with many factors influencing its salinity. The nonlinear relationship of these factors on salinity make the system operations challenging. Operational models [e.g., California Water Resources Simulation Model (CalSim) and CalSim Lite (CalLite)] are used to provide guidelines to decision makers for efficient planning and management of the system. But these operational models are not designed to directly simulate the salinity. The hydrodynamic and water quality model, California Department of Water Resources (DWR) Delta Simulation Model II (DSM2), is needed to simulate the salinity. Because of a linking problem and longer simulation time of DMS2, it is impractical to use DSM2 directly in operational models. This paper presents the development, improvement, and successful application of an artificial neural network (ANN). The ANN, when fully integrated into CalSim and CalLite, emulates the Delta salinity so that the operational models, when coupled with the ANN, can simulate the salinity management in the Delta. The newly developed and improved ANN reported in this research, when used in the CalSim model, provides more accurate insights on the salinity regime in the Delta, which is conducive to more efficient use of the freshwater in the Delta resulting in the more efficient overall operation of the SWP and CVP. | |
