Will the Three Gorges Dam experience a partial failure before the end of September?

Last updated: July 26, 2020

Metaculus question

Will there be a partial or complete failure of the Three Gorges Dam before the end of September?

Resolution will be by credible media reports, for positive resolution the reports must indicate that a failure occured before midnight (local time) on 2020/09/30.

Failure as a result of deliberate human action with the intent to cause it, for example sabotage or terrorist activity, triggers ambiguous resolution.

GJOpen question


Current Prediction

Spreadsheet

July 26th: 4%


Base Rate

In the last 20 years, there have been at least 45 dam failures giving a rate of 2.25 dam failures per year. To calculate the probability of a randomly selected dam failing, I need to know the total number of dams. If I limit the tally to only “large dams,” which are dams that are greater than 49 feet tall, there are around 40,000. So, the probability of a randomly selected large dam failing in a year is 2.25/40,000 or 0.006%.

Another way to estimate the base rate is to look at how long the Three Gorges Dam has stood so far. The dam started generating electricity in 2003, but the final section of the dam wasn’t completed until 2006. So the dam has stood for between 17 and 14 years depending on when you start counting. Using the rule of succession, that means there is a (14+1)/(14+2) to (17+1)/(17+2) chance of not failing this year, or a 5.55% to 6.25% chance of failure this year. This is a very pessimistic way of estimating the chance of failure because it ignores the evidence that other dams are typically very reliable, but it does give an upper bound. This might be a reasonable estimate if I thought the Three Gorges Dam was very poorly constructed, and the previously calculated 0.006% couldn’t be used as a prior.

Another way of estimating the base rate would be to look at the rate of failure of dams similar to the Three Gorges Dam. The Yangtze River area has a rather complicated network of dams to control flooding. At least one has already been intentionally destroyed. It would be interesting to see how often this happens, but I haven’t researched this yet.

So the bounds on my base rate are 0.006% to 6.25%.

I have a hard time thinking about really tiny probabilities, so I’m going to use 0.5% for my calculation.

Evidence For and Against

Similar rumors circulated in 2019

Rumors of an impending failure also circulated last year around this time. The dam did not fail. An anonymous post on 5ch says that rumors have been circulating for 10 years. There is an entire wikipedia entry on these rumors.

Why are rumors so common? I’m not sure, but it might have something to do with how unpopular the project has been since its inception. Construction displaced millions of people, caused widespread ecological damage, and destroyed historical and cultural sites. Are these rumors politically motivated? Or are they legitimate, and whistle-blowers are being silenced by the Chinese government? It’s hard to tell the difference between these scenarios, so I don’t update much on this.

Satellite images

Satellite images have been circulating on Twitter showing the dam deforming.

Satellite images supposedly showing the dam deformed

I think this Twitter thread does a pretty good job of debunking these claims. The warping is more likely a side effect of how the images are stitched together and displayed on Google Maps. Not to mention these images also circulated in 2019. To be honest, this doesn’t pass the sniff test in my opinion. I’m no civil engineer, but I don’t think concrete bends like this.

Weird bendy concrete

That said, the operators of the dam have admitted that the dam has deformed on the order of millimeters. This type of movement isn’t visible from satellite images (or at least not the Google images being passed around). Again, I’m not a civil engineer, so I’m not sure how worried to be about this, but the operators claim it’s within operating limits. Overall I don’t update too much on this.

Record rainfall

I think this is really what matters. The 2020 flooding is being called the worst since 1998. It’s bad enough that at least one dam had to be destroyed with explosives to reduce flooding upstream of the dam. Another dam failed when part of it “slipped off a slope” (?!).

Another article says it’s the most precipitation in 80 years, so this is on the order of a once per century event.

The downpours have intensified since last week, causing dozens of Yangtze-basin waterways to post record-high water levels, while more than 400 had exceeded warning levels, Vice Minister of Emergency Management Zheng Guoguang said on Monday.

“Since June, average precipitation in the Yangtze river basin has been the highest since 1961,” he told a news briefing in Beijing.

Could record rainfall cause the dam to fail?

One article from 2010 says the dam can handle up to 98,800 cubic meters per second. Wikipedia says the spillway can handle 116,000 cubic meters per second. A graph from Wikipedia shows an average peak flow rate at that part of the river of 30,000 cubic meters per second.

Flow rate

This means the dam has enough headroom for a 3x or maybe 4x flow rate event.

The current flow rate is reported here by the Yangtze River Hydrological Bureau of Water Conservancy Committee. Right now the flow rate is way below the maximum. Granted, it’s unclear how reliable these numbers are. If they were bad, it’s plausible the chinese government would doctor them.

Current flow rate

You can also see the current water level at 158 meters. Although this is above the flood control limit (145 meters), it is below the collapse level, which is 175 to 185 meters according to this article.

These numbers are reassuring, but they only show that at this moment in time the dam is not at high risk of collapsing. How much water can we expect in the next weeks and months? This update from July 23rd expects heavy rainfall in the Yangtze Basin over the next week. Below is an image of expected rainfall (millimeters per hour).

Heavy rainfall expected

So many areas around the river are expected to receive on the order of 1+ inches of rainfall per hour. That sounds like a lot.

A WSJ article from July 25th says the last surge produced a max flow rate that is 72% of what the dam can handle, and another surge is expected.

The gi­ant dam is ca­pa­ble of han­dling in­flows of as much as 22.1 mil­lion gal­lons (83.7 mil­lion liters) per sec­ond to pro­tect the lower reaches. The last flood hit the dam with a peak in­flow of 16 mil­lion gal­lons a sec­ond last week, rais­ing con­cerns about the dam’s strength and safety.

I believe this deserves a hefty update. Not only might this be a once in a century event, it has already caused other dams to fail and the worst might be ahead of us.

Update: 10:1

Downstream cities have not been evacuated (I think)

There are large urban populations downstream of the dam. If the government truly thought it was going to fail, I would expect them to try and evacuate. Maybe there’s some sort of Chernobyl effect going on here, and government officials are delusional about the reliability of the dam. But in situations where it’s hard to evaluate something myself, or I don’t have access to all the data, I think it’s informative to watch how people who do have the data act.

Downstream cities

Yichang is the largest city directly downstream of the dam. It has an urban population of 1.7 million, and has already suffered severe flooding. One simulation predicts the complete devastation of this city if the dam were to fail. I can’t find a source saying if it has been evacuated or not, but cell phone videos were posted to social media in late June suggesting some people were still in the city.

Wuhan is hundreds of miles downstream of the dam, but the simulation still predicts 5 meters (16 feet) of floodwater. On July 23rd this article reports that operations at Honda, Bridgestone, Isuzu, and Mazda factories have not been affected.

Honda Motor operates three plants in Wuhan. The facilities were responsible for half the 1.37 million vehicles the Japanese automaker manufactured in China in fiscal 2019.

“At this point, there are no reports of any impact, including at dealerships,” said a Honda representative. But in the case of severe flooding, the blow to Honda’s Chinese business would be significant.

Japanese tire maker Bridgestone operates an automotive seat pad plant in Wuhan. As of Wednesday afternoon, “there has been no impact on facilities and the supply chain for material has not ceased,” said a Bridgestone spokesperson.

No damage has been reported at Isuzu Motors’ truck and engine factories in Chongquing, a metropolis on the Yangtze River. The same goes for Mazda Motor’s facility in the city of Nanjing, also located along the river.

The whole situation is pretty chaotic, and the most up to date primary sources are all in Chinese making this very hard to research. So I could be wrong, but so far I haven’t heard of any evacuations caused by the Three Gorges Dam. There is a list here of major events that have happened in the various provinces. I take this to be a slight update against the idea of a failure. If I had access to more up to date information in English, this would be a stronger update.

Update: 9:10