As dam failures and incidents occur both nationally and internationally, there is a pressing need to understand the underlying causes of failure to help minimize such occurrences in the future. Current information on historical dam incidents is sometimes scattered, incomplete, missing, or misleading – making it difficult for owners and practitioners to easily access meaningful information that could assist them with critical design and operational decisions. If lessons learned and best practices are not effectively communicated, there is a possibility that poor practices will be repeated, and preventable incidents will not be averted.
Presented within DamFailures.org are links to individual case studies and 'lessons learned' pages that summarize historical dam incidents and failures and the valuable information gleaned from them. Each page contains background and description, photographs, videos, best practices, and other resources related to the case study or lessons learned being addressed. The content of this website encompasses a range of failure modes, dam types, and dam safety topics, including best practices regarding engineering and design practices, human factors, emergency planning and response, operation and maintenance, and regulatory issues.
Nine case studies and two lessons learned were added in 2024. The site now has over 100 studies, including 66 case studies and 38 lessons learned!
Thank you to all the contributors, the ASDSO Dam Failures and Incidents Committee, and the FEMA National Dam Safety Program, which provides research funding.
Case Studies
Barahona No. 1 Dam (Chile, 1928)
Researcher: Meghan Walter, P.E.
Reviewer: John France, P.E.
On December 1, 1928, an 8.2M earthquake struck central Chile. Barahona Dam Nr. 1, located just over 100 miles from the epicenter, survived the quake then failed catastrophically moments after. Four million tons of copper mine tailings were released downstream, killing 54 people and destroying several bridges and a railroad. At the time of failure, the dam was 65 m (~213 ft) high and 1,885 m (6,184 ft) long. The cause of the failure was seismically induced flow liquefaction, resulting from a decrease in the shear strength of the tailings and large displacements of the sand embankments. The dam was repaired and abandoned after the failure. This was the first record of an earthquake induced collapse of a tailings dam in Chile.
Camará Dam (Brazil, 2004)
Researcher: Paul Risher, P.E.
Reviewers: Juliana Tabet, C.PEng; Abbas Dorostkar, PhD., PE; and Cassandra Wagner, P.G.
Camará Dam is a roller compacted concrete dam in Brazil that was completed in 2002 and failed during the first filling in 2004. The dam was originally designed as an earthen embankment, but the design changed to RCC after the contractor was selected. Irregularities and potential conflicts of interest in the contracting process were allowed by the dam owner. The left abutment was excavated to competent gneissic migmatite rock, and an upstream-downstream zone of poor-quality rock near the dam toe was excavated and backfilled with concrete. Unfortunately, the geology was very complex, and an unforeseen and unpredictable geological feature/fault existed which extended up the left abutment about 16 feet (5 meters) below the exposed rock surface creating a weak plane under the whole left side of the dam. During a large rainstorm, the reservoir rose to a new record level but still at least 20 feet (6 meters) below the spillway crest (67% of capacity). Excessive leakage was observed from the drainage gallery relief drains, which ceased flowing when the drains became plugged with material being washed away from the foundation. This pushed the full reservoir head deeper into the foundation, resulting in delamination of the abutment rock. The dam failed when a portion of the concrete fractured and blew out like a huge concrete plug. The investigation uncovered problems with design, construction, oversight, and monitoring, ultimately leading to the state-level dam owner being held liable for lack of maintenance and construction oversight.
The breach drained the reservoir and inundated the city of Alagoa Grande and its vicinity along the Riachão and Mamanguape Rivers, resulting in five fatalities, 3300 people left homeless, and almost four million dollars in damage at the time.
Cleveland Dam (British Columbia, 2020)
Researcher: William Johnstone, Ph.D., P.E.
Reviewer: Cory Miyamoto, P.E.
Cleveland Dam is a concrete structure located in North Vancouver, British Columbia, Canada. The 300-foot-high dam impounds the Capilano Reservoir, one of three major Metro Vancouver sources of drinking water for the Greater Vancouver region. On Thursday, October 1, 2020, in the early afternoon, the dam’s drum gate unexpectedly started to lower, releasing a torrent of water into the Capilano River. Members of the public located downstream were enjoying recreational activities such as hiking, fly fishing, picnicking, and visiting the local fish hatchery. No public warning was issued, and the population at risk was affected within minutes. There are reports that people had to self-rescue, and at least five individuals were swept downstream. One person was killed, and the body of a second person was never recovered.
After an investigation by the dam owner (Metro Vancouver), the primary contributing factor was deemed to be human error related to the programming of the control system for the spillway gate at the dam. Three workers were fired, a public education campaign was started, and an interim warning system was installed. A free public emergency mobile app for smartphones that can receive alerts from Metro Vancouver was also implemented. Public education is ongoing, and tests of the interim warning system are performed on a regular basis. The design of a long-term public warning system is now underway and is planned to be installed in 2025.
Edenville Dam (Michigan, 2020)
Researcher: Irfan A. Alvi, P.E.
Reviewer: Paul Risher, P.E.
Edenville Dam experienced a static liquefaction instability failure of its downstream slope on May 19, 2020. The resulting breach flow reached and overtopped the downstream Sanford Dam about 2-3 hours later, resulting in the failure of that dam as well. Fortunately, there was no loss of life because a cautious decision had been made to evacuate about 10,000 people about 18 hours before it was clear that either dam was going to fail, resulting in an estimated 10 to 20 lives saved.
The primary reason why Edenville Dam was vulnerable to static liquefaction failure was that it was not built in accordance with the construction plans and specifications, resulting in the downstream slope being too steep in the location of the failure and the embankment soils being loose due to receiving little or no compaction during construction. These geotechnical vulnerabilities and their implications were not recognized during decades of inspections and evaluations of the dam.
The geotechnical vulnerabilities actualized as instability failure when the impounded lake rose to a record high level as a result of operational decisions and of portions of the watershed having become frozen due to overnight freezing temperatures just prior to heavy rain, with the frozen ground resulting in disproportionately high runoff. This watershed behavior was unforeseen, but was not unprecedented.
Ka Loko Dam (Hawaii, 2006)
Researcher: David W. Sykora, Ph.D., P.E., BC.GE
Reviewer: Irfan Alvi, P.E.
On March 14, 2006, Ka Loko Dam experienced an unexpected, catastrophic, and massive breach. Ka Loko Dam was a century-old, 40-foot-high hydraulic fill dam located on the island of Kauai, Hawaii. Owing to a period of heavy but not record-setting precipitation, Ka Loko Dam was overtopped. However, the rain, substantial inflows, and overtopping ceased days prior to the breach.
Although several experts concluded that the dam failed by overtopping, there is substantial evidence to suggest that overtopping was not likely the primary cause or a contributing factor to the failure. Substantial evidence does exist to demonstrate that the likely primary mode of failure was base sliding in a grey saprolite layer beneath the downstream slope of the dam that formed (weathered in place) from concentrated seepage in late-stage volcanic deposits since the construction of the dam in the 1890s. Based on geologic and geochemical analysis, there is evidence of the rapid formation of halloysite clay at key locations in the embankment and foundation. From a stability perspective, the rapid formation of halloysite clay in the form of small, spherical particles created a fragile, sensitive material that became increasingly more susceptible to a progressive sliding failure. The in-place weathering of volcanic deposits into halloysite (clay) has been documented as the cause of landslides in semi-tropical places like Hong Kong, but the Ka Loko Dam failure may be the first known case where in-place weathering caused the catastrophic failure of a major dam. The combination of the presence of late-stage volcanic deposits in the foundation and the substantial duration of dam operation created this dangerous condition. There is a much lower likelihood of in-place weathering for more competent volcanic deposits because of the difference in chemical composition.
Meadow Pond Dam (New Hampshire, 1996)
Researcher: R. Lee Wooten, P.E.
Reviewer: James Gallagher, P.E.
Lynda Sinclair died on March 13, 1996, at the age of 48. She drowned trying to escape the flood torrent caused by the failure of Meadow Pond Dam in her town of Alton, New Hampshire. However, the series of circumstances that led to her death started with the human failures of her upstream neighbor and those he hired to build the dam.
Schaeffer Dam (Colorado, 1921)
Researcher: Dylan Hoehn
Reviewer: Clint Brown and Mark Perry
As part of conquering the arid west, water storage has been paramount for settlers in southern Colorado. To help the settlers and farmers along Beaver Creek the Shaeffer Dam and Reservoir was constructed. With the water supplied by Schaeffer Dam the community of Penrose grew to be the largest fruit growing district in the State of Colorado.
Shaeffer Dam was constructed in 1910 in Fremont County, Colorado. The dam was 100-foot-tall, 1,100-foot-long and had a storage capacity of 3,190 acre-ft. The dam had a 3H:1V upstream and 2H:1V downstream slopes. The dam was constructed with a 100-foot-wide spillway with 10 feet of freeboard. The dam also had cut-off walls constructed of concrete and timber.
From June 2-5, 1921, one of the largest rain events in local history occurred in the Arkansas River Valley. While the rain was not a constant and heavy fall, the multiple “cloud bursts” and flooding overwhelmed Beaver Creek and the Schaeffer Dam spillway, resulting in the overtopping and eventual failure of the dam. The failure of the dam drained the entire reservoir in approximately 30 minutes and released a wall of water that “removed every tree, house, and object in its path.” The dam breach occurred on June 5th, and the community was already on alert from the previous two days of flooding. No lives were lost due to the breach. The community of Penrose, Colorado, saw lasting impacts due to the loss of soil and property, the effects of which can still be seen to this day.
Silver Lake Dam (Michigan, 2003)
Researcher: Jim Pawloski
Reviewer: Jonathan Pittman, P.E.
The emergency fuse plug spillway at Silver Lake Dam in Michigan failed on May 14, 2003, resulting in a nearly complete release of the reservoir. The dam was modified in 2002 to increase spillway capacity with the construction of the fuse plug spillway. A fuse plug spillway is an engineered earth embankment section designed to fail sacrificially to prevent failure of the main dam structure. Silver Lake Dam is in the Dead River Basin, a tributary to Lake Superior, in Michigan’s Upper Peninsula. The main stem for the Dead River extends 25 miles and includes impoundments, which are part of a hydropower generation system licensed by the Federal Energy Regulatory Commission (FERC) and owned and operated by a regional utility provider. Silver Lake is the most upstream impoundment in this system, is used for water storage, and does not have power generation facilities.
South Fork Dam (Pennsylvania, 1889) - UPDATED TEXT
Researcher: Michael D. Bennett, P.E.
Reviewers: William Bingham, P.E.; Neil Coleman, P.G.; Christopher Coughenour, Ph.D.; Carrie Davis Todd, Ph.D.; Brian Greene, Ph.D., P.G; Dina Hunt, P.E.; and Andrew Rose, Ph.D., P.E.
The Johnstown Flood of 1889 continues to hold important lessons. The event reminds civil engineers, especially dam and geotechnical engineers, that they must overcome challenging project conditions mainly with technical expertise, not business or managerial judgment. Doing so requires meeting the contemporary standard of care in civil engineering and constantly remembering that failed work by the profession can have awful human impacts. Moreover, the dam breach and flood illustrate for people in all careers that historical events, like current ones, were never preordained and that both legal sanctions and constraints of conscience are necessary to guard against human self-interest.
Lessons Learned
Static liquefaction should be considered as a potential failure mode for dams that have loose sands or silts in their embankments or foundations.
Researcher: John W. France, PE, D.GE, D.WRE, M.ASCE
Acknowledgments: Dr. Tiffany Adams, Dr. Gonzalo Castro, Dr. Kaare Hoeg, Dr. Allen Marr, Mr. Paul Ridlen, Mr. Christopher Saucier, Ms. Jennifer Williams, and Ms. Christina Winckler
Static liquefaction is a phenomenon in which saturated, loose sand or silt loses strength and collapses rapidly under sustained shear loading, generating high pore water pressure in the soil mass and very low strength. The stress-strain behavior is brittle, and the low residual strength is much lower than the static shear stresses, creating a large force imbalance resulting in acceleration, velocity, and flow of the soil mass. The behavior during failure is referred to as flow liquefaction. A lesson to be learned from the May 19, 2020, failure of Edenville Dam in Michigan is that static liquefaction needs to be considered as a potential failure mode for water dams that include loose sands or silts in their embankments or foundations.
The study of past incidents and failures aids in the assessment of existing dams.
Researcher: Mark Baker, P.E.
Reviewer: Bill Fiedler, P.E.
The knowledge of many past dam failures and incidents provides the engineer reviewing the safety/risks of existing dams with a greater ability to identify weaknesses that could initiate, progress and fail a dam. This lesson learned explains how the study of past failures and incidents can support every step of the risk analysis process. If engineers purport to understanding how to keep dams safe, they need to know how they fail.
