Anomaly Detection of Watershed Variables Related To Sedimentation In Us Reservoirs

Although our nation’s reservoirs provide critical benefits to our communities, sedimentation processes continuously occurring in reservoirs can jeopardize their functions and compromise dam safety. Several reservoirs in the US are approaching the end of their economic design life, and climate change and associated hydrologic uncertainty are introducing additional stressors to these vital systems. The US Army Corps of Engineers (USACE) developed the Reservoir Sedimentation Information (RSI) to enhance reservoir storage information, evaluate reservoir aggradation and life expectancy trends, and improve climate preparedness and resilience. In this study, data from 184 reservoirs in the RSI database were combined with data related to watershed processes that affect erosion and sedimentation such as basin topographic features, upstream reservoir properties, land use/landcover features, and precipitation descriptors. The composite dataset was diagnosed, analyzed, and interpreted to investigate multivariate relationships and identify potentially erroneous data. The analyses included preliminary data filtering, Autonomous Anomaly Detection (AAD), Kolmogorov-Smirnov and Efran (KSE) anomaly detection, and Principal Component Analysis (PCA). PCA results indicated that sedimentation rates and capacity losses had strong relationships with drainage basin size and runoff processes (Curve Number values) while being independent of elevation-related properties. A total of 20 reservoirs had anomalous records detected by the AAD and KSE methods, 5 of which were flagged by both methods, and 6 had more than one record detected. Variables with large Z-scores for anomalous records, likely causing their detection, were related to elevation characteristics (watershed slope, channel slope, and minimum watershed elevation), precipitation trends (median and cumulative monthly precipitation), dam properties (time since dam completion and initial trap efficiency), and the average watershed curve number. The results of the study will be used to refine the RSI database to provide reliable information regarding USACE reservoir conditions and enable the development of indicators related to climate change’s impact on sedimentation.