Representing Anthropogenic Processes in the National Water Model: A Persistence Approach to Reservoir Operations

Reservoir operation is a complex human-controlled activity that substantially alters surface hydrology. The National Water Model (NWM) version 2.1 includes representation of approximately 5,300 reservoirs to help accurately simulate and forecast streamflow over the entire continental United States and its contributing basins. The current NWM utilizes a level pool scheme which does not represent complex reservoir dynamics and as a result negatively affects the accuracy of the model's simulations and forecasts below large, managed systems. A persistence-based methodology for real-time data assimilation of observed flows downstream of reservoirs is proposed to improve model's skills for the operational NWM version 2.1. A persisted release is the last known valid observation at a reservoir outlet or a downstream gage. A persisted release is used within the NWM short-range forecasts, representing a lead time of 1 to 18 hours, and proved to be more accurate compared to the calculated release by the level pool method. To apply this approach for the medium-range forecast cycles, a weighted average of persisted release and level pool release, called a hybrid model, is considered. A partial autocorrelation function is used to determine how far one may trust an individual persistence model for future forecasts, which generates appropriate weights for the hybrid model. The final reservoir release is then mass balanced to ensure that it does not exceed a reservoir's storage boundary conditions. This methodology is implemented within the operational NWM V2.1 to utilize observed real-time data feeds provided by the U.S. Geological Survey, and the Army Corps of Engineers and is also expected to use reservoir forecasts authorized by River Forecasting Centers. This presentation will discuss the approach to representing these persistence-based reservoir models within the NWM and our results to date.