Dam Rehabilitation and the Perpetuation of Human Factors across Time and Space

Since 1948 the Soil Conservation Service, now NRCS, has helped local sponsors build more than 11,800 dams. As those dams reach the end of their 50-year design lives, the work has shifted from building to rehabilitation, and with it a different problem. Today's engineers and geologists must navigate the human factors on their own design teams while trying to understand the human factors that shaped the teams of fifty years ago. This paper compares two Oregon dams built in the same year, with similar designs and similar early trouble, whose different geology and post-construction histories now point to very different rehabilitations.

A new era

NRCS assistance built more than 11,800 dams under the Flood Control Act of 1944 and the Watershed Protection and Flood Prevention Act of 1954. These project dams return an estimated $2.2 billion a year in reduced flood and erosion damage, improved habitat, recreation, and water supply for about 47 million people. Construction peaked in the 1960s. Of the 11,800, some 978 were originally high hazard potential; today 2,158 are, mostly because of development downstream. More than 6,200 have passed the end of their design life.

Rehabilitation is a wicked problem, in the sense of a challenge with no clearly identifiable solution, embedded in a fluid and interconnected set of causal influences that make it impossible to know the best path forward. Complexity magnifies human fallibility, and so does time. Rehabilitation layers human factors over time and across space on critical infrastructure. Confronting a predecessor's decisions brings the advantage of hindsight and the burden of new constraints.

Physical and human factors

Making and managing a dam involves physical and human factors that interact. They are unique to each structure and together form a dynamic, site-specific system responsible for both safety and failure.

Rehabilitation usually attends to the physical factors, bringing a dam up to current standards to extend its life. But as Alvi puts it, because physical processes deterministically follow physical laws, with no possibility of physical mistakes, failure of dams, in the sense of not fulfilling human intentions, is ultimately always due to human factors: humans falling short in various ways.

Human factors in dam safety work through three drivers:

• Human fallibility and limitations. Misperception and faulty memory; limited skill, expertise, knowledge, or information; cognitive biases; inaccurate models; unreliable intuition.
• Pressure from non-safety goals. Delivering water, generating power, reducing costs, meeting schedules, building and maintaining relationships, personal and political goals, and the limits of federal programs.
• Complexity. Many interacting components with nonlinearities, feedback loops, and network effects; large effects from small causes; difficulty in modeling, prediction, and control, all of which exacerbate fallibility.

Every dam project is a socio-technical system: the idea that the design and performance of any organizational system is best understood by treating its social and technical aspects as interdependent parts of one complex whole. Coupling physical and human factors this way helps show how social and organizational forces shape technical decisions, on today's teams and on those of the past.

Clear Branch Dam

Clear Branch is a 128-foot high hazard embankment dam on the Middle Fork Hood River, in the Mount Hood National Forest, supplying water to the Upper Hood River Valley under the PL-566 program. The Middle Fork Irrigation District owns and operates it. The embankment is 1,440 feet long and holds 3,565 acre-feet. Built in 1969 with a single concrete chute spillway, it gained a second when hydropower was added off-site in the 1990s and FERC, as the new regulator, required capacity for the probable maximum flood. The new chute became the service spillway and the 1969 chute the auxiliary. Classified significant hazard at design, it was raised to high hazard by NRCS in 2015; Oregon and FERC agree.

Seepage and drainage have troubled the dam since construction, ending in the 2017 failure of the auxiliary spillway, which required removing and replacing a 50-foot section and installing under-drains and laterals. The NRCS investigation found that design decisions to set aside the SCS geologists' recommendations for seepage and drainage led to excessive hydrostatic pressure under the chute. That failure is documented in the authors' 2019 paper.

Three physical factors led to 2017. The geology is complex: a U-shaped glacial valley, an embankment on a moraine ridge 1,000 feet long, glacial deposits over 280 feet deep, interbedded and highly permeable. Seepage is excessive: a 1965 consultant's study contradicted the SCS geologist by concluding minimum foundation treatment was needed, and estimated 2 cfs of loss with 6 possible; first filling in November 1968 produced 3 to 15 cfs along both sides of the auxiliary spillway, and a new seep appeared at the right abutment in 2016, perhaps related to a wildfire upstream. And the drainage was inadequate from the start: despite the study's own estimate, no measures addressed seepage through the glacial sediments of foundation and abutments, which left the dam exposed to piping, internal erosion, and surface erosion.

The human factors run alongside. Little in the record suggests the design anticipated seepage emerging high on the downstream left abutment; the decision makers did not heed the SCS geologists and misjudged the risk, and those risk indicators were later normalized, tolerated, and accepted. Cognitive bias may have played a part, since a very similar design was used at Cooper Creek; standardized designs were popular in the federal dam building agencies for the time and cost they saved, especially before computer drafting, GPS survey, and lidar. Economics probably drove the choice of the consultant's study over the geologic reports, and the seepage control that was considered aimed at saving irrigation water rather than at dam safety. Federal funding brings its own requirements, such as periods of performance for spending, which add layers to the socio-technical system.

Complexity has grown. Understanding of Pacific Northwest geology has advanced; detailed mapping and remote sensing give modern geologists insight into depositional history and foundation character, and advances in seismology and geophysics have transformed understanding of regional hazards such as the Cascadia Subduction Zone. In 2011, lidar revealed a new fault system near Clear Branch Dam, now under investigation by the USGS, the Oregon Department of Geology and Mineral Industries, and universities. It is a clear illustration of how perceptions of physical factors change with new technology and information, and of how complexity feeds overconfidence and the underestimation of risk. The dam was also built before the Magnuson-Stevens Act; salmon and bull trout in the reservoir and streams mean rehabilitation must address intake screening, fish passage up and down, and reservoir operation. Three federal agencies now hold jurisdiction or oversight, three more are involved in approvals, and two state agencies and tribes besides.

Cooper Creek Dam

Cooper Creek is an 88-foot homogeneous earthfill dam in Douglas County, planned, designed, and built with SCS assistance under PL-566 for flood control, municipal supply, and recreation. The Sutherlin Water Control District owns and operates it. The embankment is 500 feet long and holds 3,390 acre-feet at normal pool. A six-foot vertical drainage zone just downstream of the centerline controls internal seepage, discharging through a blanket drain under the downstream shell. The principal spillway is an ungated pair of 30-inch reinforced concrete pipes with a drop inlet, discharging into the auxiliary spillway chute on the right abutment; together they pass 3,300 cubic feet per second. It was high hazard by design and remains so.

Within a year of completion, the dam showed minor instability near the toe of the right abutment. In January 1970, with the pool about 50 feet above the outlet works intake, tension cracks appeared in the right abutment perpendicular to the concrete spillway, and inspectors saw excessive displacement and cracking at spillway chute joints. Construction photographs show slides of varying depth on both abutments and beside the spillway. Investigation established that the outlet of the horizontal filter blanket had been covered with soil during final grading, and the increased head eventually eroded through the covering. No definite connection between the spillway movement and the temporary head could be demonstrated, though the two appeared related.

Further exploration in 1970 could not identify the slide plane; the 1978 Phase I inspection suggested movement may have occurred within a zone rather than along a plane. The driving force was not clearly identified either, though saturation of the surficial deposits was considered a factor, and the exploration found permeable zones in the right abutment carrying significant groundwater. Five subsurface drains went in during 1970 and three relief wells in 1988, which lowered pressures. There have been no slides since, and the instruments show little or no movement.

The physical factors resemble Clear Branch closely, and foundation geology is what separates them. Seepage at Cooper Creek, 0.04 to 0.12 cubic feet per second, is a small fraction of Clear Branch's 3 to 15. The gap between seepage and drainage capacity was therefore much smaller and much easier to close. Both dams suffered instability on first filling from inadequate drainage, but the remedies at Cooper Creek have held for fifty years, and those at Clear Branch have not.

The human factors are similar too, with a few differences. Because the seepage volume is smaller, inaccurate models and incomplete information mattered less. Both projects face the same pressures of cost, schedule, and program funding, but Clear Branch is an irrigation supply reservoir, where drought and user demand add more. Better understanding of the Cascadia Subduction Zone affects both, though no new fault has been found at Cooper Creek. Both sit in similar regulatory environments for threatened and endangered species, but FERC and the Forest Service are not involved at Cooper Creek, which removes layers of federal coordination.

The influence of time

Each of the roughly 11,800 project dams is unique in its physical and human factors, and together they represent a culture of engineering design, a snapshot of the federal dam building era. These two case studies show how interconnected those factors are, and how time works on them. Time increases complexity, acting as a catalyst for physical factors to develop and human factors to compound. New individuals, teams, and agencies join a project and add their own factors on top of the old. Hindsight is a privilege, because it shows the outcome of past decisions, and a burden, because it carries the responsibility to recognize and rectify past human factors in a way that helps the project and is transparent to the teams that come next.

Both dams were planned, designed, and built in the same period with similar designs, notable because concrete chute spillways on dams over 80 feet are uncommon in NRCS. The geology at both sites is complex, and understanding of it has grown, so the staff now rehabilitating them recognize the projects as more complex than the original teams knew.

Many physical factors are multi-dimensional processes rather than discrete events. They develop over time and often do not appear as problems until long after they begin. They are persistent, evolving components of a dam, not issues to be addressed once and forgotten. Dams are not static objects on the landscape but living resources requiring continuous attention, assessment, and maintenance: mechanisms people use to modify and control the forces within a watershed, and therefore subject to change from both the physical factors that define their morphology and the human factors that define their design, construction, maintenance, and performance.

Seen this way, the potential for failure is the balance between the human factors that contribute to failure, the demand, and those that contribute to safety, the capacity. At Cooper Creek the drainage installed immediately after first filling gave enough capacity that demand never exceeded it. At Clear Branch, repeated attempts have not, and demand still exceeds capacity. In the standard engineering metaphor, the spillway failed in 2017 when demand exceeded capacity.

Conclusions

Though conceptually distinct, the physical and human factors at these two dams are inextricably linked across their lifespans, forming complex socio-technical systems. As dams age, the complexity of managing them increases as those systems evolve. Physical factors drive decisions, and decisions are a product of and susceptible to human factors, which compound over time through changes in personnel, evolving technology and knowledge, and new regulatory and funding requirements.

This paper looks at human factors in technical design decisions and the influence of design teams, but many others shape a project: the decisions of owners, of operation and maintenance staff, and of the people who live, work, and play around a dam. Those deserve further exploration.

It is hard to look back at the decisions of previous generations and understand what drove them. Project files capture physical factors well and human factors poorly, so the record is incomplete. Since failure, in the sense of not fulfilling human intentions, is ultimately always due to human factors, the risk cannot be eliminated. Without human intervention through operation, maintenance, and rehabilitation, dams will fail; and failure is itself a result of human factors. That is the wicked paradox of dam safety. What is possible is to minimize future risk by cultivating a preoccupation with identifying and avoiding failure rather than assuming overconfidence or complacency. Engineers and geologists who understand how these factors evolve and perpetuate will succeed by recognizing the factors affecting their structure, acknowledging their own biases, and balancing demand and capacity as far as they can.

References

1. Morgan, D.F., Ingle, M., and Shinn, C.W. (2019). New Public Leadership: Making a Difference from Where We Sit. Routledge.
2. Durgin, S., Natural Resources Conservation Service. (2019). GeoObserver webinar, April 4, 2019.
3. Alvi, I. (2018). Human factors in the Oroville Dam spillway incident. Webinar, Association of State Dam Safety Officials, August 14, 2018.
4. Alvi, I. (2013). Human factors in dam failures. ASDSO National Dam Safety Conference, Providence, Rhode Island.
5. Alvi, I. (2018). Appendix J: human factors framework and methodology. Independent Forensic Team Report, Oroville Dam Spillway Incident.
6. Leeds University Business School. (2019). Socio-technical systems theory.
7. Walter, M., Gillilan, J., and Markov, J. (2019). Physical and human factors in a spillway failure: a case study on Clear Branch Dam. ASDSO National Dam Safety Conference, Orlando, Florida.
8. Soil Conservation Service (High, R.D., Patterson, P.V., and Scoles, C.C.). (1964). Geologic report of Middle Fork of Hood River watershed, Clear Branch Dam and reservoir site.
9. Natural Resources Conservation Service. (2018). Clear Branch Dam engineering investigation report.
10. Oregon Department of Geology and Mineral Industries (Madin, I.P., Streig, A.R., Burns, W.J., and Ma, L.). (2017). The Mount Hood fault zone: late Quaternary and Holocene fault features newly mapped with high resolution lidar imagery.
11. Soil Conservation Service (High, R.D.). (1966). Final geologic report on Sutherlin Creek watershed, Cooper Creek dam site.
12. Oregon Water Resources Department and U.S. Army Corps of Engineers. (1979). Cooper Creek Dam Phase I inspection report.
13. McMillen-Jacobs Associates. (2020). Cooper Creek Dam evaluation report, revision 2. Prepared for the Sutherlin Water Control District.

The authors thank Craig DeHart, Wade Osbourne, Rick Scott, Joe Gasperi, Carmen Berry, Amod Koirala, Gregg Hudson, Trent Snellings, and Claudia Hoeft for their contributions to this paper and to the written record of both dams.