Quantifying Risk, consider Fukushima earthquake predictions

Can be harder than it seems. I found this on “The Statistics Forum” of ASA and Chance

http://statisticsforum.wordpress.com/2011/03/23/quantifying-uncertainty-is-never-easy/

Many of questioned why the Fukushima reactor was built to withstand only 8.3 in that part of the world known for seismic activity. It turns out they had a seemingly well understood theory and 300 years of data.

Here is the article from the NYT

http://www.nytimes.com/2011/03/22/science/22predict.html?_r=2&ref=science

but note

This was not the first time scientists have underestimated the ferocity of an earthquake fault. Many were also caught by surprise by the magnitude-9.1 quake in 2004 off Sumatra, which set off tsunamis radiating across the Indian Ocean, killing more than 200,000 people.

Sometimes, scientists are blindsided by earthquakes because they occur along undiscovered faults…

In California, for instance, scientists have cataloged 1,400 faults, yet for smaller earthquakes — magnitude 6.7 or less — about one in three still occur on previously unknown faults.

In other words we have good reason to believe our knowledge is incomplete. Hubris is often called Overconfidence bias, http://en.wikipedia.org/wiki/Overconfidence_effect , where we substantially overestimate our certainty. Subjective confidence intervale are not very reliable.

The wikipedia restates a well known study where weather forecasters have no overconfidence but clinical psychologists do. We often make fun of weather forecasters, but their repeated experience with forecasting and prompt outcome feedback means that they are well calibrated. If they say 70% chance of rain, it will rain 70% of the time.

Historic data may not be enough. Although, perhaps had they looked at the data for predictability they would have determined that the earchquake process is neither stable nor fuly understood.

Advertisements

About Bill Nichols

PhD in Physics from Carnegie Mellon University I'm a software team coach and instructor with the TSP Team at the Software Engineering Institute
This entry was posted in metacognition, Models, risk, uncertainty. Bookmark the permalink.

6 Responses to Quantifying Risk, consider Fukushima earthquake predictions

  1. Bill,

    An ex-neighbor here in colorado is an earthquake guy at UC Davis, https://www.geology.ucdavis.edu/faculty/rundle.html it’s be interesting to hear from him on this topic.

    • Bill Nichols says:

      Yes, it would be interesting to hear directly from an expert. It wouldn’t surprise me much to learn that the press reported a single opinion or otherwise glossed the story.

      My non-expert observation is that there does not appear to be a stable “earthquake process” in the Deming sense. Although we can compute objective probabilities with history and models the probability distribution does not appear to be fully reliable., That is, there seem to be special rather than common causes, e.g. undiscovered faults for small quakes.

      Although I worked in the industry, I was never involved in the probabilistic safety analysis. I have some familiarity with how the engineers model this, but I’m not the domain expert. All we know for certain is the quake and tsunami exceeded their design specifications.

  2. Bill,

    The jury is not out yet but, although it is possible that hubris may have a place in the current events, I kind of remember hearing that the Japanese knew that the fault that released was “stuck” and accumulating energy.

    That being said they are known for being very good at planning for seismic events and the actual difference between 8.3 and 9.0 is about 2:1 given the logarithmic Richter scale. It is 2.7:1 when you consider what they expected.

    What is really amazing to me is the compounding of events that, although factored in, had a much worse impact than expected. My take is that they did their homework but still got walloped.

    • Bill Nichols says:

      Patrick,

      Thanks for your comment.

      I have to wonder if their emergency response protocol began with “because the the tsunami s contained by the sea walls…”.

      I suspect in coming years this will be studied extensively. But notice we never hold Congressional hearings (perhaps Parliamentary interrogations of the Ministry, I’m not as familiar with Canada) when everything works out as expected?

  3. Suresh Kayamboo says:

    Hi Bill,

    I am not a physicist but a Program Manager in the IT industry. I live in Japan for the past 10 years. What happened in Fukushima is lack of Risk management. When a severe earthquake hit Niigata a couple of years ago, one of the Nuclear Power Reactors maintained by the same company TEPCO automatically shut down. They are still not able to re-start the reactors after 2 years. And a 9.1 earthquake is not new. It has happened decades ago in Chile before Sumatra. Fukushima reactor may be state of the art than Chernobyl when it was built 50 years ago. But what lacked is they did not revisit their risk management plan (RMP) and failed to re-assess the probability and severity matrix learning lessons from Chile and Sumatra knowing that Fukushima lies on a faulty line that is prone to earthquakes. Onagawa nuclear power plant which sits few tens of kilometers above Fukushima has made arrangements and predicted a Tsunami wave of 15 meters. That is why they could escape from the tsunami whereas Fukushima overlooked the height of a Tsunami ending in a disaster.

    This just reminds me the basics of Risk Management that anticipating all the risks is not possible. But during the course of progress, risk assessment should be done and RMP modified periodically with lessons learned. But as you have quoted in another forum, cost and schedule are mostly dictated by the management leaving little room to apply best practices always.

    Regards,
    Suresh Kayamboo

    • Bill Nichols says:

      Suresh,

      Thanks for your comments and insights. I have several friends and colleagues in Japan, some of whom are justifiably upset with the conditions and ongoing response.

      You touch on several topics l plan expand on in the future. One is risk management under conditions where the risk cannot be directly quantified. A second is ongoing tracking and management of risks. Yet a third is how the project/program manager relies on subject matter experts. Finally, though we say “hindsight is 20-20, we see the past through a refracting lens.

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s