PLT - Estimated rates of failure for dams in the United States

We estimate the annual rate of failure for dams in the United States by examining the National Inventory of Dams and a dam failures database containing historical dam failures. Permanent man-made water retaining dams of all purposes are included; landslide dams, test dams, cofferdams, and tailings dams are removed from both datasets. Dams are categorized by type into the following types: concrete, earthfill, masonry, rockfill, timber, and other, as well as by construction era and dam age. We estimate that there has been a total of 5 628 516 dam-years, 2543 dam failures, and thus 0.00045 dam failures per dam-year in the history of dams in the United States (once incomplete data has been imputed). We make examinations of the rate of dam failures per dam-year across each dam type, construction era, and age category, and use the number of dam failures per dam-year as a proxy for estimating annual rate of failure. From this simple approach, we make observations regarding rates of failures of dams of various characteristics. Concrete and earthfill dams both have less than 0.0005 failures per dam-year; masonry and rockfill dams have more than 0.001 failures per dam year and timber dams have more than 0.0035 failures per dam-year. For all dam types, except for earthfill and rockfill, we find that the failures per dam-year are greatest in the first five years after construction with a steady decline in failure rates as dams age beyond 100 years. Earthfill and rockfill dams reach the lowest rates of failures per dam-year when they are between 20-50 years old with a significant increase in failure rate, especially for earthfill dams, after 50 years. We find that dams constructed prior to 1920 and from 1920-1950 have had more failures per dam-year (0.00105 and 0.00041, respectively) than those constructed between 1950-1970 (0.00019); however, dams constructed since 1990 have the greatest rate of recorded failures per dam-year (0.00141) which is suspected to be due to improved modern monitoring programs through the initial years after construction when dams are most likely to fail. We hope these observations assist practitioners doing quantitative risk analysis, by providing estimates of historical rates of failures for dams of various construction types, ages, and eras of construction.