Hydro Leader: Please tell us about yourself and about your professional background.
John France: I have been a private consultant for about 46 years. Since about 2018, I have worked on my own as an independent consultant through JWF Consulting LLC. My career has been focused on dam engineering, safety, and risk analysis.
Hydro Leader: Please tell us about the failure of Edenville and Sanford Dams.
John France: On May 19, 2020, two dams failed near the city of Midland in central Michigan. The dams were two of four dams owned by Boyce Hydro. All four dams were built in the mid-1920s on the Tittabawassee River. From upstream to downstream, they are Secord and Smallwood Dams, both of which suffered some damage but not failure; Edenville Dam, which experienced the initial failure; and Sanford Dam, which subsequently failed.
Storms brought heavy rainfall to the area on the days before the incident. At Secord, the rainfall between May 17 and May 19 was as much as about 6 inches, and at the other three dams the rainfall was in the range of 3–4 inches. It was a significant but not extreme rainfall—nowhere near approaching the probable maximum precipitation. Over the night of May 17, the reservoirs began to rise, but they were still within their normal operating range. All four dams have gated spillways, and by the morning of May 18, the dam operators began to operate the gates. There were gate operations throughout the day, concluding at Sanford at about 8:00 p.m. The water continued to rise on the night of May 18–19. On the afternoon of May 19, lake levels peaked and began to recede at Secord and Smallwood, the two most upstream dams, but the water continued to rise at Edenville until 5:35 p.m.
At daybreak on May 19, the dam operators noted a significant amount of erosion occurring on the upstream slope of Edenville Dam. In the middle of the day, they began to deploy silt curtains and other materials to try to reduce the erosion. Michigan dam safety personnel were on site in the late morning and early afternoon of May 19.
In the early afternoon of May 19, before 3:00 p.m., the lake was rising by as much as 3 inches per hour, but then the rate of rise began to diminish. The lake level never reached the crest of the embankment. However, at 5:35 p.m., there was a sudden failure of a section of the downstream slope of Edenville Dam, resulting in the dam’s breach and release of water behind it. That water flowed downstream to Sanford, which had two spillway facilities. The gated spillway, as I mentioned earlier, had been opened the day before. Sanford Dam also had a fuse plug spillway that was designed to erode and provide some additional spillway capacity for water if the lake got that high. The water did reach the top of the fuse plug spillway at about 7:20 p.m., and the fuse plug began to erode, adding to the spillway capacity. But the water was rising by as much as 3 inches per minute during this time because of the large inflow from the Edenville failure. The combination of the fuse plug spillway and the gated spillway simply wasn’t enough for that amount of inflow, and at 7:46 p.m., water went over the top of Sanford Dam. Around 8:00 p.m., the embankment started to erode significantly, and there was an overtopping erosion breach of Sanford Dam. That breach sent the storage from Sanford Dam and the failure flood from Edenville down into the town of Midland and other areas. Fortunately, there was no loss of life, because the emergency managers had made a proactive decision late in the evening of May 18 to evacuate people downstream. Ultimately, they evacuated around 10,000 people out of harm’s way. However, two dams and lakes were lost, more than 2,500 homes were damaged, and there was property damage estimated to be in excess of $150 million.
Hydro Leader: Would you tell us more about the structures themselves and how they were constructed?
John France: As I mentioned, these four dams were constructed in the mid-1920s. Each dam was originally constructed with a combination of two or more earth embankment sections and at least one concrete gated spillway with tainter or radial gates for control. And, of course, each one had a powerhouse for hydropower generation. During that era, before the advent of soil mechanics, or what was later called geotechnical engineering, dam designers and builders relied on rule-of-thumb guidelines for how to build embankment dams. The dams were built with local earth materials.
When we looked at the specifications for how the Edenville Dam embankments were supposed to be built, they indicated that the embankments were supposed to be built with soil materials that would be placed in layers and compacted. They were supposed to be constructed with an upstream section made of low-permeability clays and sandy clay material and a downstream section made of more permeable and water-conductive sands and silty sands. They were also supposed to have clay-tile drains placed underneath the downstream sections, extending perpendicular to the dams’ axes from the downstream toes to beneath the crests of the dams.
We know that the drains were built, because we have photographic documentation from construction, and some of the drains were still exposed at the dam. But as to whether the Edenville Dam embankments were built with that upstream-downstream configuration, the information is a little contradictory. We think it’s likely, but we can’t be certain. However, we are relatively confident that the compaction did not happen. We don’t see any evidence of compaction equipment in the construction photos that we’ve been able to find. In addition, over the years there have been a number of exploratory test borings drilled in the embankments, and when penetration tests were done, those sands and silty sands were found to have low blow counts— some less than five blows per foot—which indicates that the sands are loose, and in some places very loose. We even have photographs that show material being dumped in place in some sections of the embankments. So it appears that the embankment included some loose and very loose sands and silty sands, and we believe that was a primary factor in the mechanism of failure of Edenville Dam.
Hydro Leader: Why was Edenville Dam’s FERC license revoked in 2018, and was it related to the failure incident?
John France: We’re still looking into that, so I don’t have a definitive answer for you. But as we understand it right now, the reason was not related to the physical mechanism of the failure. The dam had a couple of owners from the late 1990s up to the time of the failure. In the early 2000s, it was owned by a Canadian organization. In the mid- 2000s, Boyce Hydro acquired the dams. For years, FERC’s principal concern was that the spillways were too small to safely convey the probable maximum flood without water overtopping the embankments. Our understanding at this point—and we’re still doing interviews, so this may change—is that in 2018, FERC revoked the owner’s hydropower license over the spillway capacity issue, not over a geotechnical embankment instability issue, which was the ultimate cause of the failure.
Hydro Leader: After the incident, how did the formation of the IFT come about? Is it normal practice to establish an IFT after dam failures, or are they formed only after certain incidents?
John France: That process has evolved over time. In recent years, it has become FERC’s practice to require an independent forensic investigation after major incidents and failures at dams it regulates. You may recall the Oroville Dam spillway incident that happened in California in 2017. Although Oroville Dam itself did not fail, a failure of its service spillway resulted in the evacuation of almost 200,000 people. In that case, FERC required the owner of the dam, the California Department of Water Resources (DWR), to engage an IFT, of which I was actually the leader. We were engaged and paid by the DWR.
In the case of Edenville and Sanford Dams, FERC again directed the owner to engage an IFT. Our names were submitted to FERC by the owner and were approved, but we were not able to reach a contractual agreement with the owner, and after about 3 months, FERC chose to engage us directly. In both cases, however, our rules of engagement as an IFT are that we do our work independent of the organization that has engaged or contracted us, particularly because both the owner and FERC are part of the human factors investigation. Our interim report was not reviewed by FERC; the owner; the Michigan Department of Environment, Great Lakes, and Energy (EGLE); or any of the other parties associated with the dam before it was issued. That will also be the case with our final report. They have had no input into the way the investigation is done; it’s done fully independently.
Hydro Leader: Please tell us about the IFT’s investigation of the Edenville and Sanford incident.
John France: The investigation is ongoing. We issued the interim report because we had reached conclusions concerning the physical mechanisms of what happened, but we are still evaluating the hydrology and hydraulics of the flood itself. We are also still evaluating the human factors: the various judgments and decisions made and the actions taken or not taken throughout the history of the project. Early in the investigation, we interviewed eyewitnesses of the incident. Now we are talking to individuals who may have been involved in the history of the project—engineers; operators; former employees of Boyce Hydro; and dam regulators from FERC and EGLE.
A video of the Edenville Dam failure taken by a local resident proved to be particularly valuable in investigating this event. We spoke to the resident and to some of his neighbors who also witnessed the event. The video and photos they supplied were instrumental in forming our understanding of this failure. Two representatives from the state regulator were on site during the event, but they were involved with the crews to try to reduce the upstream slope erosion, so they were not situated where the failure happened. Without the video, I don’t know whether we would have reached the conclusion we did regarding the failure mechanism. The video shows that this section of the embankment really did not show any significant distress up until about 5:00 p.m. About 5:00, there was a visible settlement or depression of the section of the crest at the location where the ultimate failure happened. At that point, the residents, who had been watching the dam from upstream, walked around the end of the dam to the downstream side to continue watching. Moments before the failure, they noticed some water start to flow down the downstream slope, and one of the individuals started recording a video on his phone. The video shows that within about 10 seconds, a significant section of the downstream
portion of the embankment over a length of 40–80 feet actually failed, effectively flowed out of the downstream side of the embankment, and was deposited downstream of the dam. It appears that a remnant of the embankment on the upstream side held in place for about another 10 seconds before it ultimately gave way, and then the water in the reservoir started to flow through the breach, which widened over a couple of hours.
We came to the conclusion that the most plausible explanation is a phenomenon called static liquefaction. In the simplest terms, what happens is that a very loose saturated sand or silty sand can, by a couple of different trigger mechanisms, experience a dramatic and rapid reduction in strength. This results in an overall force imbalance in the soil mass, which creates an acceleration, and the acceleration creates velocity, resulting in rapid flow liquefaction.
We’re confident that the incident at Edenville Dam was not an overtopping failure. There is also another mechanism called internal erosion, which occurs when seepage through an embankment or foundation can actually erode particles of soil out of the embankment, and that can then begin to disintegrate the embankment and cause it to fail. But we don’t see evidence of that in this case. We think that static liquefaction is the most plausible explanation. We labored a lot with that conclusion; it’s a significant conclusion for the water dam engineering profession because that failure mechanism has not historically been looked at much. The professional literature almost entirely says that sands and silty sands will not behave like this except in an earthquake. Our colleagues in the mine tailings dam sector over the past couple of decades have begun to recognize that mine tailings dams have behaved this way, but our colleagues in water dam engineering haven’t. Our conclusion means we have a new failure mode that we need to begin considering when we have loose sands or silty sands in embankment dams.
Hydro Leader: Does this conclusion suggest that static liquefaction is a greater danger for dams than previously believed?
John France: I think what it suggests is that it’s a physical possibility that we as a profession were not necessarily recognizing before. Documented cases of static liquefaction are pretty rare, quite frankly; there are probably only a handful. One of the big reasons we reached this conclusion was the video. I don’t know whether we would have reached it otherwise. That raises the question: During floods in the past, have there been some failures of embankment dams that may have been static liquefaction but were attributed to other causes, such as overtopping or internal erosion, simply because there were no eyewitnesses or videos? It still seems that it takes the combination of an unusual set of circumstances to cause static liquefaction, but this failure has taught us that we need to consider the possibility and understand the required circumstances better.
Hydro Leader: Does that suggest that there might need to be additional monitoring requirements on specific kinds of dams?
John France: I don’t think it’s a matter of monitoring, because the failure was so quick that monitoring would not necessarily have helped a lot. I think it’s more going to be investigation and evaluation. For most of the high-hazard dams in the United States—dams that would likely cause loss of life if they were to fail—we, the dam safety community, know a fair amount about how they were built and what the materials are in them. I think we have a pretty good handle on which dams might have loose sands or loose silty sands like these in them. It’s going to be a matter of doing research on their design and construction, similar to what we did here. Perhaps some investigations will need to be done to evaluate whether the sands in those embankments are loose enough to behave this way. The profession as a whole will need to figure out what guidelines we’re going to use to evaluate this mechanism, because we don’t evaluate it right now. I think we can learn from what our colleagues in the mine tailings dam industry are doing and adapt their research to water dams. Then, hopefully, we’ll start looking at these dams more closely and, when we find problems, take corrective actions before we have failures like this one.
Hydro Leader: What are the next steps for the investigation?
John France: The next steps for the IFT, as I mentioned earlier, are to continue our human factors investigation and our evaluation of the hydrology and hydraulics of this particular flood. We have more interviews to do and more historical documents to go through. We also need to go through all the information that the public provided in response to our press release early in the investigation. One question we’re trying to answer is why the water level in Wixom Lake rose to this historic high level. We need to compile and write our report and then edit it so that all five members of the IFT are comfortable with what it says and are willing to put their names on it. That process takes a while, but we’re all anxious to move it forward as quickly as we can. We decided to issue this interim report now, at the point when we understood the physical mechanisms, in large part because of the significance to the water dam industry of our finding of static liquefaction as the primary physical mechanism of failure.
John France is a private consultant. He can be contacted at johnwfrance.pe@gmail.com.
