Developing An Automated Method To Estimate Reservoir Sedimentation At ~30,000 Reservoirs Across The United States

The majority of dams in the United States were constructed between 1940 and 1980. Based on design estimates, many reservoirs were designed to accommodate 100-years’ worth of sediment input from rivers. Therefore, sedimentation is becoming an increasingly important issue facing our nation’s reservoir assets. The rates of sedimentation vary due to natural stochastic events, such as floods or wildfire. However, in many cases, the emplacement of upstream dams is the dominant control on downstream sediment availability. Our work seeks to automate the calculation for reservoir sediment yield, accounting for the emplacement of upstream dams and their sediment-trapping efficiency (following Minear and Kondolf, 2009).

We acquired reservoir survey data for 535 sites across the United States from Bureau of Reclamation, US Army Corps of Engineers, and publicly available databases. We used the National Inventory of Dams (NID) and the National Hydrography Dataset Plus High Resolution (NHDPlus HR Beta Version) to locate over 80,000 dams on high-resolution flow lines. We developed a Python code to: (1) clean the NID dataset and remove duplicate dams, dikes, and off-stream storage facilities; (2) snap the cleaned-NID dataset to NHD flowlines for the entire nation; and, (3) link dams along flowlines. We identified approximately 30,000 dams upstream from our 535 study sites. We then utilize Matlab to order dams from upstream to downstream and calculate the sediment-contributing drainage area at our sites through time, accounting for upstream dams and trap efficiency. We use the time-weighted sediment-contributing drainage area to calculate sediment-yield rates at our sites. Using the original drainage area under-estimates the sediment yield (m3km-2yr-1) by an average of 25% at our sites. This indicates a significant portion of the upstream-drainage area is disconnected by dams, and the sediment in the study reservoirs is therefore sourced from a smaller contributing-drainage area. Accurately calculating sediment yield based on the sediment-contributing watershed is critical for future projections of sedimentation. Our preliminary data indicate that the cumulative capacity at the 535 study sites will be reduced by approximately 11% of total design storage by year 2050. Our next steps include modeling the sedimentation of the approximately 30,000 upstream NID-registered dams based on basin sediment-production rates inferred from site data. We intend to make our codes publicly available and plan to convert Maltab code to Python to ensure open-access. Because we have performed dam ordering for the entire nation, we can seamlessly add new survey sites or other users can utilize our code to calculate sediment yield at their own sites.