Physical and Human Factors in a Spillway Failure: A Case Study on Clear Branch Dam

Clear Branch Dam is a 128-foot high hazard embankment dam on the north face of Mount Hood, built by the Soil Conservation Service in 1969 to supply irrigation water to the Upper Hood River Valley. In June 2017, inspectors found a two-inch vertical offset in one 25-foot section of its concrete auxiliary spillway chute. The repair a year later exposed erosion channels and voids beneath the floor and walls. This paper traces the chain of physical and human factors, beginning with design decisions in the 1960s, that produced that offset half a century later, and reads them in the light of the Oroville spillway incident.

The dam

Clear Branch Dam sits on the Middle Fork Hood River in the Mount Hood National Forest. The Soil Conservation Service, now the Natural Resources Conservation Service, built it in 1969 with funding under Section 14 of the Watershed Protection and Flood Prevention Act, Public Law 566. The Middle Fork Irrigation District owns and operates it. The embankment is 1,440 feet long and impounds 3,565 acre-feet. Two concrete chutes serve as the service and auxiliary spillways, with capacities of 8,600 and 1,640 cubic feet per second; the auxiliary chute is 321 feet long and 15 feet wide.

The dam was classified significant hazard when designed. NRCS raised it to high hazard in March 2015, and the Oregon Water Resources Department and the Federal Energy Regulatory Commission agree. FERC is the dam safety regulator, a role it acquired when hydropower was added downstream in 1986.

After Oroville in 2017, FERC opened a spillway initiative reviewing similar structures across its regions. Dams with chute spillways founded on soil or rock, and dams with unlined spillways, were subject to a Focused Spillway Assessment. Clear Branch qualified: its chutes are founded on a terminal moraine of interbedded, poorly to well graded layers of high permeability.

Geology

The dam crosses a steep-walled, U-shaped valley of alpine glaciation. Bedrock is assumed to be Cascade andesite, overlain by pyroclastic, glacial, colluvial, alluvial, and lacustrine deposits. The embankment sits on a terminal moraine ridge 1,000 feet long, deposited in the most recent glaciation; glacial deposits reach depths over 280 feet. During the last recession the ancestral Clear Branch channel incised the moraine through the right abutment. The moraine and underlying drift are interbedded silts, sands, gravels, cobbles, and boulders, with the well graded zones highly permeable. The spillway inlet structures are founded on engineered embankment; the chutes are founded on the moraine.

What the record shows

The site's permeability was known before the dam was built, and the design chose not to address it.

The 2017 spillway failure

FERC's Focused Spillway Assessment was under way when the offset was found. FERC wrote to the district that urgent action was needed to investigate the stability of slab and foundation, and to determine the depth, length, and width of any voids, the material under the slab, and the condition of its base.

The district, NRCS, and FERC worked step by step: monitor, collect data, identify causes, then repair. A manifold system was installed to capture seepage and sediment from between the moved sections, and in October 2017 the seepage there was about 0.5 gallons per minute. Total seepage measured at all the weirs runs 0.7 to 4.2 cubic feet per second and tracks the pool elevation, with pressures rising and falling as the reservoir does. The sediment caught in the manifold was 5 to 7 percent gravel, 85 to 90 percent fine to coarse sand, and 3 to 5 percent silt and clay. By September 11, 2018, the trap had collected 132 pounds of it.

FERC required a drawdown in November 2017 to reduce pressure on the chute until repair. In December the district hand-dug to expose the 12-inch corrugated metal lateral drain about 15 feet upstream of the affected section and scoped it with a borehole camera: iron bacteria were clogging it, and chemical analysis showed high manganese and iron. In May 2018 video of the five bench drains showed flow limited by blocked perforations rather than by bacteria in the filter rock. The drains were jetted clean in July and appeared to function afterward, though total flow at the weir did not change much.

NRCS completed its independent investigation in August 2018. The committee found the drainage system did not lower the phreatic surface enough to relieve hydrostatic pressure under the chute. If a flow event coincided with high pressure, more displacement was expected, and displacement could let water beneath the floor, erode the subgrade, and remove parts of the chute. Iron bacteria were clogging the perforated pipe and probably the drain aggregate.

The committee's central conclusion was about the original design. It had not incorporated the seepage control features recommended in the 1961 and 1964 geologic reports, following the consultant's study instead. Those recommendations included nothing to protect the chute from hydrostatic pressure, whether from abutment seepage or from discharge out of the gravel blanket drain. No drainage was installed beneath or beside the chute, and the concrete was not designed to resist uplift. Nothing provided for controlled discharge of seepage after it left the blanket drain at the toe; it ran uncontrolled across the downstream face of the left abutment. Construction addressed seepage through the glacial sediments neither with an upstream blanket nor with a grout curtain, though the seepage study itself allowed for 2 to 6 cubic feet per second. These decisions left the dam at risk of piping, internal erosion, and surface erosion, and every seepage issue from first filling in 1968 to the chute failure in 2017 stems from that deficiency.

The repair, and what it exposed

In the fall of 2018 NRCS and the district agreed on an investigative repair: remove and replace two chute sections totalling 50 feet, examine conditions below and beside the spillway, replace the drains parallel to it and beneath the new sections, and commission a new seepage study. After dewatering wells and access pads were in, the contractor saw-cut and lifted the lower section in eight pieces, leaving the subbase undisturbed for inspection, with inspectors from all three organizations watching. A two-stage granular filter and drainage system went in below and on both sides of the chute, with 240 linear feet of new lateral drains outletting to the plunge pool, and inspection and cleanout ports on both systems. With FERC's concurrence the upper section followed. New weir boxes with staff gauges were installed at the outlets of the new laterals and the existing bench drains.

Under the lower section, inspectors found erosion channels and voids up to three inches deep and red bacterial deposits up to an inch thick. The voids ran under the walls and slab and had been cut by water flowing at the interface of subgrade and concrete, with finer particles moving downslope. There was no evidence of piping into the dam. Several large andesite boulders came out from beneath and beside the spillway, one of them directly under an existing keyway joint. The north lateral drain did not extend upslope past the area of movement; the south drain was intact and full of iron bacteria. The slab itself did not match the as-built drawings: 8 to 9 inches thick where 11 was specified, with 2 to 3 inches of rebar cover instead of 3.

Under the upper section the channels ran up to six inches deep, with voids reaching farther toward the center of the chute beneath the north wall, so subsurface flow appeared to enter under the walls and run beneath the slab. Soils at the upper end showed iron staining and roots but no bacterial deposits: brown-grey gravel and cobbles in good rock-to-rock contact, the spaces filled with coarse and medium sand, interpreted as very dense glacial drift.

The seepage under the removed slabs was not piping through the dam. It flowed along the interface between subbase and slab, mobilizing fines and opening voids. The two-stage filter and collection system is expected to control it, though biofouling may return over time. After the repair the district monitored the new weirs and reservoir level and received FERC's approval for a controlled refill.

Physical factors

The geology below and beside the dam is heterogeneous and anisotropic. The terminal moraine forming the left abutment is highly stratified, with sand and gravel layers that give it high horizontal permeability, so groundwater follows preferential paths and concentrates at discharge points. Two high permeability zones in the valley floor downstream of the left abutment, at elevations 2,850 to 2,880 and 2,896 to 2,936, are hydraulically connected to the reservoir and flow harder when it is high.

The dam has a documented history of seepage through the embankment, including where the chute lifted. Drains have been added repeatedly over its life. In 2018 the drains parallel to the auxiliary spillway were flowing very slowly, with iron staining well above the water line, evidence that they had once carried much more. Every other seepage drain was discharging significantly. The slow flow in these lines showed they were compromised, probably by biofouling of the media or the conduit.

The investigation committee concluded that excessive hydrostatic pressure beneath one chute section produced an uplift force greater than the section could resist, mainly because seepage moved through the moraine beneath the chute and the drainage could not lower the phreatic surface enough to relieve it.

Human factors

The committee found that following the 1965 consultant's study put the dam at risk of piping, internal erosion, and surface erosion. Construction in 1969 incorporated a design that argued against many of the measures the 1964 geologic report had recommended. On first filling, significant seepage emerged from the left abutment at roughly the elevation of the section that moved in 2017. Drains went in beside the chute in 1970. An investigation in 1977 declared design deficiencies including excessive seepage, and bench drains followed in 1979, becoming less effective until rehabilitated in 2002. At no point was anything done about hydrostatic pressure on the underside of the chute, by under-drainage or by redesign for uplift.

Using the framework in Alvi (2013), three drivers account for the human contribution.

• Human fallibility and limitations. Misperception, limited skill and expertise, limited knowledge and information, unreliable intuition, inaccurate models. Every seepage issue from 1968 to 2017 stems from a design that did not incorporate the recommended controls, following the consultant instead of the geologists. There is little evidence the design anticipated seepage emerging high on the downstream left abutment, as it did on first filling. Over time, the volumes, rates, and locations of discharge were normalized and tolerated.
• Pressure from non-safety goals, such as cost and water supply. Minutes of the 1978 West Technical Service Center meeting record that concern for seepage in design appeared more about lost water for beneficial use than about stability, piping, or internal erosion, and that an SCS engineer who believed the choice was a filter blanket or not building at all felt overruled when a consultant was hired for a cheaper option. Economics presumably drove the decision to follow the consultant over the 1961 and 1964 geologic reports.
• Complexity, from many interacting components with nonlinearities, feedback loops, and network effects. Various attempts were made to control seepage, but none addressed the loop between geology, seepage, and hydrostatic pressure under the chute. The project has grown more complex: three federal agencies now hold jurisdiction, responsibility, or oversight, and threatened and endangered species live in the reservoir and streams. Responding to an emergent failure mode means mobilizing large bureaucracies full of feedback loops.

These drivers produced the errors that ended in the 2017 movement. It is hard to look back at the decisions of earlier generations and know what drove them. It is possible to reduce future risk by cultivating a preoccupation with finding and avoiding failure, rather than assuming overconfidence or complacency.

Conclusions

Seepage at Clear Branch Dam is a complex problem that many skilled engineers and geologists have tried to study and control. Dam failures typically involve physical and human factors interacting for years or decades before the event. The auxiliary spillway failed in 2017 through a combination that acted across the structure's 50-year service life: complex site geology, excessive seepage, design decisions that ignored or misjudged risk, and the normalization of risk indicators. A new seepage study is under way using the 2018 geotechnical exploration and what the investigative repair revealed.

The dam is in the planning phase of a comprehensive rehabilitation with NRCS, intended to bring it to current dam safety, regulatory, and environmental standards and extend its life another 50 to 100 years, which means re-evaluating the design and the environmental effects of the structure and its appurtenances, including the drivers of potential failure modes.

Clear Branch offers an unusual chance to study an emergent and complex spillway failure mode, one caught early, monitored closely, and not destroyed by the flood or breach that usually erases the evidence.

References

1. Soil Conservation Service. (1964). Detailed geologic report.
2. Soil Conservation Service. (1961). Preliminary geologic report.
3. Natural Resources Conservation Service. (2018). Engineering investigation report.
4. Soil Conservation Service. (1978). Report of investigation of probable deficiencies.
5. Soil Conservation Service. (1980). Geologic report.
6. Kleinschmidt. (2015). Dam assessment report.
7. Kleinschmidt. (2018). Remedial watershed measures engineering design report.
8. Alvi, I. (2015). Failure of Sella Zerbino secondary dam in Molare, Italy. Association of State Dam Safety Officials.
9. Alvi, I. (2013). Human factors in dam failures. Dam Safety 2013, Association of State Dam Safety Officials, Providence, Rhode Island.
10. Cornforth. (2018). Draft geotechnical investigation report, Clear Branch Dam explorations.

The authors thank Jeff Brown, Craig DeHart, Steve Durgin, Joe Gasperi, Nicholle Kovach, Jim Lyons, Wade Osbourne, Steven Reinsch, Jesse Wilson, Ken Worster, and Kip Yasumiishi for their contributions to this paper and to the written record of Clear Branch Dam.