Image source: xkcd (accessible here)
When I lived in California, earthquakes were just a part of life. In fact, I've had a hard time adjusting to the idea that I no longer have to worry about a earthquake striking suddenly (the other night my first thought at feeling the subway passing by underground was that an earthquake was upon us!). Today, earthquakes seem to be a hot topic in the blogosphere, with both FiveThirtyEight and Pacific Standard posting about them. The topic of earthquakes is an inherently interesting one: they are one of the few natural disasters that we currently have no warning system for, and large earthquakes are an extremely rare event that cause a lot of fear and apprehension (probably also related to their unpredictability). They also represent an interesting lens through which to consider risk communication (one of my main interests). Thus, I decided to jump on the earthquake bandwagon and summarize the articles below (plus, I got to make a reference to a Taylor Swift song, so how could I not?).
In Nate Silver's blog, FiveThirtyEight, Carl Bialik discusses the latest developments in predicting earthquakes, and all the models (there are approximately 5,000 being considered by the USGS alone) that are trying to accurately make these predictions. Currently, it seems the best we may be able to hope for is a system that warns us when the small probability of an earthquake occurring increases. As a behavioral researcher, this seems like a far worse situation to be entrenched in. People generally do not do well with very small probabilities - they often overweight them and overreact to them. While one seismologist argues that "...Americans have experience processing low probabilities for catastrophic events. We’ve grown accustomed to hearing about heightened awareness of terrorist attacks, and wildfire warnings have become a feature of Californian life," I think this is a dangerous mindset for a potential risk communicator to have. Hearing stories about catastrophes may make us accustomed to hearing about them, but this does not mean we know how to properly process and account for their true risk. In fact, it may be this "accustomization" process (reading major headlines about small probability catastrophes) that makes the general public so bad at accounting for these types of events: the headlines make the events more salient and vivid which leads to a higher predicted probability of the event occurring - i.e., reasoning by availability. And any small amount of risk above zero may be too much risk for the extremely risk averse general population, which makes changes in these small risks ultimately unhelpful in terms of preparation and response.
The post at Pacific Standard is also about earthquake prediction, but takes a much different approach. The article discusses a study conducted by Japanese researchers assessing whether animals can warn us about impending earthquakes. The authors of the study collected survey data from cat and dog owners in Japan who were located at different distances from the epicenter of the March 2011 earthquake. They find that approximately 19% of dog owners and 16% of cat owners reported unusual behavior in the time leading up to the earthquake. This data is suspect given it relies on retrospective accounts of pet behavior (I am a doting cat owner, but I could never tell you if my cats were behaving unusually as I always feel as if they are behaving unusually). And people may be looking for behavior or cues that confirm the hypothesis (see: their pets knew the earthquake was coming).
More interestingly, the researchers also looked at milk production in dairy cows to see if it was affected by the earthquake. They surveyed three separate farm locations located various distances from the epicenter. While the cows located the furthest away from the epicenter showed no effect, cows located closer to it (211 miles away) showed significant decreases in milk yield for four days prior to the earthquake. Of course, it's possible this decline correlates with some other occurrence unrelated to the earthquake, but it's potentially interesting nonetheless.
While the FiveThirtyEight post focuses on a very scientific and objective approach to predicting earthquakes that suggests they may remain predictably unpredictable, the Pacific Standard post suggests there may be other, less conventional avenues through which we can attempt to predict these occurrences. Perhaps the key to predicting earthquakes isn't in the ground but on top of it.
And just for fun, I've included the TSwift video below.
UPDATE 10/15/2014: A few articles came out in the last day saying that Northern California is overdue for a "megaquake" (see here and here and here). Given the unreliable nature of earthquake predictions (see above), it's surprising that any scientist would want to put their name on such a forecast. It also makes me wonder whether earthquake predictions are susceptible to faulty (no pun intended) causal inferences (gambler's fallacy anyone?). In other words, we haven't had a major earthquake in Southern California for quite some time, so we must be "due."
In Nate Silver's blog, FiveThirtyEight, Carl Bialik discusses the latest developments in predicting earthquakes, and all the models (there are approximately 5,000 being considered by the USGS alone) that are trying to accurately make these predictions. Currently, it seems the best we may be able to hope for is a system that warns us when the small probability of an earthquake occurring increases. As a behavioral researcher, this seems like a far worse situation to be entrenched in. People generally do not do well with very small probabilities - they often overweight them and overreact to them. While one seismologist argues that "...Americans have experience processing low probabilities for catastrophic events. We’ve grown accustomed to hearing about heightened awareness of terrorist attacks, and wildfire warnings have become a feature of Californian life," I think this is a dangerous mindset for a potential risk communicator to have. Hearing stories about catastrophes may make us accustomed to hearing about them, but this does not mean we know how to properly process and account for their true risk. In fact, it may be this "accustomization" process (reading major headlines about small probability catastrophes) that makes the general public so bad at accounting for these types of events: the headlines make the events more salient and vivid which leads to a higher predicted probability of the event occurring - i.e., reasoning by availability. And any small amount of risk above zero may be too much risk for the extremely risk averse general population, which makes changes in these small risks ultimately unhelpful in terms of preparation and response.
The post at Pacific Standard is also about earthquake prediction, but takes a much different approach. The article discusses a study conducted by Japanese researchers assessing whether animals can warn us about impending earthquakes. The authors of the study collected survey data from cat and dog owners in Japan who were located at different distances from the epicenter of the March 2011 earthquake. They find that approximately 19% of dog owners and 16% of cat owners reported unusual behavior in the time leading up to the earthquake. This data is suspect given it relies on retrospective accounts of pet behavior (I am a doting cat owner, but I could never tell you if my cats were behaving unusually as I always feel as if they are behaving unusually). And people may be looking for behavior or cues that confirm the hypothesis (see: their pets knew the earthquake was coming).
More interestingly, the researchers also looked at milk production in dairy cows to see if it was affected by the earthquake. They surveyed three separate farm locations located various distances from the epicenter. While the cows located the furthest away from the epicenter showed no effect, cows located closer to it (211 miles away) showed significant decreases in milk yield for four days prior to the earthquake. Of course, it's possible this decline correlates with some other occurrence unrelated to the earthquake, but it's potentially interesting nonetheless.
While the FiveThirtyEight post focuses on a very scientific and objective approach to predicting earthquakes that suggests they may remain predictably unpredictable, the Pacific Standard post suggests there may be other, less conventional avenues through which we can attempt to predict these occurrences. Perhaps the key to predicting earthquakes isn't in the ground but on top of it.
And just for fun, I've included the TSwift video below.
UPDATE 10/15/2014: A few articles came out in the last day saying that Northern California is overdue for a "megaquake" (see here and here and here). Given the unreliable nature of earthquake predictions (see above), it's surprising that any scientist would want to put their name on such a forecast. It also makes me wonder whether earthquake predictions are susceptible to faulty (no pun intended) causal inferences (gambler's fallacy anyone?). In other words, we haven't had a major earthquake in Southern California for quite some time, so we must be "due."
RSS Feed
