Some dam failures stem from long-term use, while other incidents are tied to earthquakes, floods, and other events. How are failures distributed over the life of a dam? And does a long period of satisfactory performance mean there will be no significant incident over the remainder of its life?
Dam safety professionals must be ever vigilant in their efforts to assure the safety of dams and other water retention or control structures under their charge; whether owner, regulator or consultant, none can be complacent when it comes to dam safety.
And yet, all too often complacency creeps in when a dam has had a lengthy history of apparent successful operation. How many times have we heard, or used the words, "The dam's been OK for 50 years. Why are you worried about it now?"
During many Potential Failure Mode Analysis sessions conducted by the Federal Energy Regulatory Commission, this reasoning came up as a way to lower the categorization of a potential failure mode.
We seem to forget that dams are subject to many of the affects of aging and exposure that we are all subject to. Some failures are the result of long-term use. In humans, it may result in carpal tunnel syndrome; in dams, it may result in the failure of a gate hoist mechanism.
Some incidents are simply due to a longer exposure to a potentially hazardous situation.
Diseases like cancer may come on rapidly or progress over a long period of time. Similarly, piping can be a rapidly developing situation on first filling like at Teton Dam or a longer process such as at Kantale Dam in Sri Lanka that failed from piping after 117 years of, apparently, successful operation.
This article attempts to address four questions:
- How are dam incidents distributed over the life of a dam and in particular, does a significant period of apparently satisfactory performance indicate that a dam will have no significant incidents over the remainder of its life?
- Is there a difference in long-term performance among the different types of dams?
- Are there particular potential failure modes that contribute significantly to safety incidents over time?
- Is there a difference in the types of delayed incidents depending on the year the dam was constructed?
To address the questions stated above, the author developed a database of dam failures and safety related incidents that at the present time includes over 4000 individual dam failures and incidents from 84 countries. The age of the dam when the failure or incident occurred could be determined for 1158 of these incidents.
How are dam incidents distributed over the life of a dam and, in particular, does a significant period of apparently satisfactory performance indicate that a dam will have no significant incidents over the remainder of its life?
It is generally assumed that the initial years of a dam's life are the most dangerous and the data bears out this assumption.
About 31 percent of the dam safety incidents analyzed for this paper occurred during construction or the first five years of a dam's life (see Figure 1). Among dam types, there was a statistically significant variation in certain types of dams with 18 percent of gravity dams and 29 percent of arch dams experiencing incidents within the first five years, while 42 percent of both earthen dams and rockfill dams suffered incidents during construction or within the first five years.
The high percentage of dam safety incidents occurring within the first five years of operation points out the importance of thoroughly examining a potential dam site, making sure that the dam's design accounts for site-specific conditions that could result in the initiation and development of a potential failure mode, constructing the dam carefully in order to minimize the potential for a failure mode to initiate, and implementing a focused surveillance and monitoring program to examine how the dam is behaving.
The second half of the first question was examined by considering only the data for those dams where the incident occurred after five years of operation.
The data for dams that suffered a safety incident after the first five years of operation was plotted as an exceednce graph that shows the percentage of reported dam incidents that occurred beyond a given age. According to a database of dam failures and safety-related incidents developed by the author of this paper, 49 percent of all dam incidents occurred at age 50 or beyond as shown in Figure 2. The available data does not show a major difference in the number of incidents that occurred in any five-year period.
The two five-year periods with the highest number of incidents after the first five years of operation were six to ten years of age with 72 incidents and 11 to 15 years of age with 61 incidents. However, the third highest was 81 to 85 years of age with 55 incidents.
Certainly the number of dams that suffer a failure or serious incident is a relatively small percentage of the total dams constructed. However, data indicates that a significant percentage of dams that suffer a delayed safety related incident do so at age 50 or beyond.
Therefore, we should not be less vigilant as a dam ages nor be surprised if a dam safety incident develops as a dam ages. In fact, some failure modes may become more likely with age.
The average date of construction of dams in the incident database used for this paper, where the age at the time of incident could be determined, is 1933, 77 years ago.
However, this average date of construction is skewed by a few very old dams, the oldest being constructed in 1550, while 451 (39 percent) of the dams were constructed after 1945 and are therefore, at most, about 60 years old.
Is there a difference in long-term performance among different types of dams?
When the data for incidents that occurred after the first five years of operation for each type of dam is plotted as an exceedance graph, the graph shows some distinct variations in the longer-term performance of dams (see Figure 3).
Earthfill dams have a higher than average failure rate during the first 50 years of operation with only 31 percent of incidents occurring at age 50 or beyond, as compared to nearly 50 percent for all types of dams combined. In contrast, gravity dams exhibit better than average performance for the first 70 years of operation compared to all types of dams. Rockfill and arch dams have somewhat worse performance than average over the initial years of operation but, in general, trend closer to the average than either earthfill or gravity dams.
The difference in performance between earthfill and gravity dams may be partially explained when the failure modes for each type are examined. The principal failure modes for earthfill dams are: Internal erosion/seepage/piping, 110 incidents; overtopping, 105 incidents; and structural issues, 47 incidents.
Gravity dam incidents include 101 related to flood loadings, often related to spillways, and 115 related to structural issues, including concrete deterioration and the structural performance of appurtenant structures such as spillway gates and gate operators. Also, there are 27 incidents of piping in the foundations of gravity dams. Of the 115 structural issues, 41 were related to spillway gates.
Seepage is a potential failure mode that is always active, assuming the water level is at least as high as the critical pathway, and it may also be exacerbated by the higher water levels during floods or a favored pathway may be opened due to seismic induced cracking.
Therefore, internal erosion and piping are potential failure modes that must always be considered for all embankment dams and in the foundations of all dams. On the other hand flood loadings, especially extreme floods, are relatively rare events and the probability of occurrence during early years of a dam's operation is relatively small. Similarly, degradation processes are time dependent and actions such as ASR or freeze-thaw may negatively impact the structure only after an extended period of time.
Degradation is also a significant component in incidents related to electrical/ mechanical equipment. Degradation of electrical/mechanical equipment may be due to simple wear and tear from use or a lack of preventative maintenance.
The shape of the curves in Figure 3 may also be influenced by the large numbers of dams that were built in the last 40 to 50 years.
Are there particular failure modes that contribute significantly to safety incidents over time?
In order to answer this question, the data was sorted by age at the time of the incident and the failure modes grouped into three categories; flood related, structural, and seepage/piping.
Failure modes related to reservoir slides, other reservoir events and other failure modes not related to these three categories made up insignificant percentages of the identified incidents and were dropped from the data. All failure modes related to flooding whether caused by an extreme flood, spillway plugging, failure of a gate to operate when needed, etc., were placed in the flooding group. Structural events included incidents such as sliding, embankment slope instability, earthquake damage, concrete degradation, and gate failures. Seepage/piping included all internal erosion, piping and seepage related failure including those through the embankment, through the foundation, and those into or along outlet works. The data was analyzed for all incidents in the database and for those where the dam had survived the first five years of operation.
Only 65 percent of all seepage incidents occur after five years of operation, whereas 74 percent of flood-related incidents and 78 percent of structural incidents occur after the first five years. It is interesting to note that after 80 years of operation, all three failure modes have nearly the same percentage of incidents that will occur in future years.
Data indicates that after the first five years of operation, there is some difference in the percentage of failures occurring among the different failure modes with structural related failures occurring at a lower rate. However, after 45 years of operation, there is little distinction to be made in the percentage of incidents yet to occur from either flood, seepage or structural failure modes.
At the extreme end of the data, dams that experience a failure or incident after 100 years of operation, there were 45 incidents where both the age and failure mode were identified.
Of the 45 incidents, 22 were related to flooding, 13 to structural issues, and ten to seepage.
This information indicates that, although seismic and flood loadings are generally considered remote events and failures related to overtopping or damage from earthquakes can be easily rationalized as remote incidents from extreme events, we cannot forget about piping and other seepage related incidents when dealing with embankment dams and in the foundations of all dams.
Seepage related incidents are the most common modes of failure in the early years of a dam's life and continue to be an important potential failure mode over the longterm.
Is there a difference in the types of delayed incidents depending on the year the dam was constructed?
The average construction year of dams in the database, where the age at the time of the incident could be determined, is 1933. This is prior to the development of current theories of rock mechanics, filters and drains and the application of modern geotechnical engineering theory into the design on many dams. The incident database shows peaks in number of incidents in certain construction eras due in part to the number of dams of a particular type built in the particular era.
The smaller number of incidents in the later years may be due to better design but may also be due to a smaller number of dams being constructed since the building boom of the 1960s and 1970s, or the shorter period of time that these later dams have been exposed to processes that take time to develop.
We examined the relationship between seepage related incidents in earthfill dams in greater detail by plotting the percent of seepage/piping incidents that occurred beyond a given age by the era in which the dam was constructed. The dams were grouped by those constructed before 1900, those constructed from 1900 to 1950, and those constructed between 1950 and 1975. The data includes 65 incidents for both the 1900-1950 group and the 1950-1975 group and 26 dams in the pre-1900 group.
To account for the difference in exposure between newer and older dams, the data for incidents that occurred in the first five years of operation and those that occurred beyond 45 years of operation was removed to provide a picture of how dams have fared in the five to 45 years of age range.
The data indicates that there is no real difference in the percentage of incidents that occur at advanced ages for earth dams constructed between 1900 and 1950 and those constructed after 1950. Dams constructed before 1900 actually show that once past the initial five years of operation the percentage of incidents that occur at later ages is greater than for newer dams, especially in the five to ten years of age range. After ten years of operation the older dams continue to have a greater percentage of the total incidents occurring at a later age than either of the more recent groups.
Pat Regan, PE, is regional engineer for the Division of Dam Safety and Inspection, Federal Energy Regulatory Commission, Portland, Oregon.
This article has been evaluated and edited in accordance with reviews conducted by two or more professionals who have relevant expertise. These peer reviewers judge manuscripts for technical accuracy, usefulness, and overall importance within the hydroelectric industry.