TRACE THE
FAILURE
SPILLWAY FAILURE

The dam held.
188,000 people were ordered out.

On February 7, 2017, part of Oroville's main spillway tore away while releases were being raised toward 52,500 cfs, about a third of the flow the chute carried in 1997. The forensic team traced it to water pressure beneath thin, cracked concrete over the drains. Five days later, erosion below the emergency spillway forced an evacuation.

Follow the evidence
Illustrative AI image of a damaged concrete spillway chute running full, with water blasting out through a torn section into a raw brown gully; not a photograph of Oroville.
Illustrative AI image from the episode; not a photograph or recording of the Oroville spillway in February 2017.
THE COMPANION FILMOroville Dam: What Really Broke in 2017 Watch episode 8:43

Why the Oroville spillway failed

Oroville Dam itself never broke. The 2017 emergency began on its gated main spillway, a concrete chute about 179 feet wide and just over 3,000 feet long. On the morning of February 7, part of the chute’s floor tore away while operators were raising releases from about 42,500 toward 52,500 cubic feet per second. In 1997 the same chute had carried about 160,000 cfs without failing. Independent Forensic Team, printed pp. 9, 18, 25 and 34.

The Independent Forensic Team concluded that water pressure beneath the slab most likely lifted a weak section into the flow. Water leaked into the drains below the chute through cracks and joints, many of them right above the drain pipes, where the concrete was 7 inches thick or less. The ground underneath was weathered rock and soil, poorly cleaned before the concrete was placed. Once a slab was gone, fast water eroded that foundation and ripped out the slabs around it. IFT, printed pp. 19, 22 and 29 to 38.

The evacuation came from a second route. With the main spillway damaged, the lake rose over the separate emergency spillway for the first time, and erosion on the bare hillside below it advanced toward its concrete wall. On February 12 about 188,000 people downstream were ordered to evacuate. IFT, printed p. 26.

Watch the episode

The film rebuilds the chute in 3D, from the bang on the spillway to the lake rising over the emergency spillway, and cuts the slab open at its true thickness to show where the water got in.

Watch “Oroville Dam: What Really Broke in 2017” on YouTube.

The bang on the spillway

At about 10:10 a.m. on February 7, two state water department workers servicing a siren beside the chute heard what sounded like an explosion. Then they saw spray coming off the spillway. When operators shut the gates to look, a large section of the chute’s concrete was gone. IFT, printed pp. 25 and 30.

Storm water was still pouring into the lake, so operators had to release it somewhere. After the first inspection they resumed using the damaged spillway while they managed the rising reservoir. IFT, printed pp. 25 and 26.

The spillway choice

Both routes carried risk. More water down the damaged chute could tear away more of it and threaten the hillside and the structures nearby. Holding water back would push the lake over the emergency spillway: an ungated concrete ogee weir about 930 feet long and an 800-foot broad-crested weir, both topped at elevation 901, with natural, unprotected hillside below. IFT, printed pp. 8 to 9 and 40.

Water first crossed the emergency crest between about 7:00 and 8:00 a.m. on February 11. Erosion went deeper than expected, and by the afternoon of February 12 it was advancing rapidly upslope toward the crest structure. Operators began opening the main spillway’s gates further at 3:35 p.m., and the evacuation order for about 188,000 residents followed at 3:44 p.m. As the lake fell, water stopped crossing the emergency crest that evening. IFT, printed p. 26.

The order became a warning on February 14, and that warning was lifted about five weeks after the original order. Officials feared the erosion would undermine the emergency spillway’s crest; it never did. IFT, printed p. 26.

Beneath the concrete

Investigators asked what had changed since 2006, the last year the chute carried more water. Part of the answer was in the repair records. Documented patches date from 1977, 1985, 1997, 2009 and 2013, but they were shallow, did not address the causes, and typically began to deteriorate within a year or two. IFT, printed pp. 19 and 34.

The drain pipes under the chute were meant to carry groundwater away, but they protruded up into the slab. The design called for at least 15 inches of concrete; above the pipes there were 7 inches or less. Cracks appeared above the drains within a month of the concrete being placed, and by 2017 almost every drain had one. Whenever the chute ran, water leaked through those cracks and open joints into the drains. IFT, printed pp. 11 to 12, 19 and 21 to 23.

The foundation was weak too. Specified preparation was relaxed, and in some areas half or more of the foundation was not properly cleaned of weathered rock and soil. The steel anchors holding the slab down reached about five feet into that ground. IFT, printed pp. 36 to 38.

On February 7 the flow was being raised. The IFT’s most likely sequence is that faster water drove more water through the cracks than the drains could carry, pressure built beneath a weak section, and its upstream edge lifted into the flow, where the pressure beneath it jumped and the section tore away. IFT, printed pp. 29 to 33.

The first movement is still disputed

Nobody saw the first slab fail, and the flow washed the evidence away. The IFT could not establish the exact location or size of the first failure. It favored uplift as the first motion while acknowledging that minor local settlement could have created an offset into the flow.

The later FERC After Action Panel considered prior settlement an equally plausible mechanism: long-term internal erosion could have undermined a local section, so that it sagged first and then let pressure in beneath it. Both accounts end with water under the concrete. They differ over what happened just before. IFT, printed pp. 30 to 33; FERC panel, pp. 13 to 17.

Warning signs from the first year

High drain flows were seen the first time the spillway was used, in 1969. A report called them “mystifying but probably not dangerous as the chute is anchored” and said an inquiry to the designers would be in order. The IFT found no record that anyone made one. IFT, printed p. 22.

Many regular inspections viewed the chute only from the gate deck at the top, where just its upper, flatter section can be seen. Over the decades the cracking and the drain flows came to be accepted as normal. A 2014 failure-mode review essentially identified the failure that began in 2017, then ruled it out as unlikely to release the reservoir, mainly because the rock was judged generally non-erodible. IFT, printed pp. 50 to 51.

The emergency spillway had a similar blind spot. A 1962 geology report warned that the weathered rock on that side “will of course be subject to relatively accelerated erosion; where this is critical, the rock should be protected.” A 2005 state memo called the emergency spillway “a safe and stable structure founded on bedrock that will not erode.” IFT, printed pp. 53 to 55.

What had to change

The IFT found no single root cause. It described a long-term failure of the state’s water department, its regulators and general industry practice to recognize the spillways’ design and construction weaknesses, poor bedrock and deteriorated chute over almost 50 years, and said the incident “cannot reasonably be ‘blamed’ mainly on any one individual, group, or organization.” IFT, printed pp. S-1 and 47.

It found no record that anyone had ever reviewed the design, construction and performance of the chute slab in detail, and it recommended periodic comprehensive reviews of exactly that, because physical inspections alone are not enough. After Oroville, California’s dam safety regulator and FERC required owners to review their concrete chute spillways. IFT, printed pp. 52 and 75 to 77.

The main spillway was repaired enough by November 1, 2017 to work as a flood-control outlet that winter and was rebuilt to its final design in 2018. The emergency spillway gained a concrete buttress, an underground secant pile wall and a hillside splashpad to stop erosion from working uphill. When the rebuilt spillway was first used, on April 2, 2019, state water department experts and federal and state regulators agreed it performed as designed. California Water Commission, 2019 annual review, p. 5.

At Oroville the dam held, while both of its routes for letting water out came close to failing.

For another structure that failed in a way its builders had not planned for, read why the Tacoma Narrows Bridge collapsed. The mechanisms differ.

Primary sources and imagery

Independent Forensic Team, Oroville Dam Spillway Incident final report, January 5, 2018, printed pp. S-1, 8 to 12, 18 to 38, 40, 47 and 50 to 77. Report. FERC After Action Panel report, November 23, 2018, pp. 13 to 17. Report. California Department of Water Resources, Oroville Spillways FAQ. California Water Commission, 2019 Annual Review of the State Water Project, p. 5. Review.

The image at the top is an illustrative AI image from the episode, not a photograph of Oroville. The film’s DWR photographs and figures are real and credited on screen; its 3D shots and slab cutaways are simplified reconstructions built from the forensic report, and the slab-lifting sequence shows the investigators’ most likely explanation, which nobody observed. The narration is an AI voice.