Thursday, June 20

Earthquake Prediction - the Fraudulent Business

Me: I can predict the outcome of soccer matches! I’ve been watching the clouds and when they take a certain form, I claim I can predict when a goal is going to be scored — and I’ve been right many times before!
You: That sounds impressive…. But wait, can I see one of your predictions?
Me: Sure! I see that the clouds will become thicker later in the week, so I can confidently predict that a goal will be scored somewhere in the Premier League sometime across the weekend.
You: How can I be sure you’re right about this?
Me: Look, the last time I predicted this, I was right! In fact, I’m right almost 50 percent of the time. That’s incredible, right?
You: I suppose, but what about all the times you were wrong?
Me: Does that matter, I get them right, too, so I must be onto something. So, want to pay me money to make these predictions for you?
How many of you would fall for this? Hopefully not many, but if you look at the internet, you’ll find a number of people trying to say they can predict earthquakes and volcanoes across the planet, yet they are doing almost exactly what I said I could do for the Premier League. It is fraudulent, unscientific and wrong.
Listen, I love the interwebs. It provides more information at your fingertips than almost anything in the history of civilization. However, it is also very easy to manipulate people’s opinions because there is just so much information out there. You type in the words “volcanic eruption” into Google, you get over 3.3 million results. How can you sift through that all if you don’t know where to start — especially when it comes to discerning reliable sources from crackpots. If you’re interested in natural disasters, even a simple search shows you a large population on the web of people who claim (or want to believe) that they can predict when an earthquake or eruption will happen. This is different than stating the probability of such an event — this is what volcanic monitoring organization and geologic surveys do fornatural hazards. For example, after looking at the signs of activity at a volcano (earthquakes, gas emissions, etc.), they say that an eruption may happen in days to weeks. What I’m talking about are these predictions that say that based on some criteria unrelated to the volcano or fault in question, an earthquake or eruption will occur on a specific date (or narrow window). Typically these predictions are vague, rarely saying too specifically where or exactly when the event will happen. Usually these predictions also have multiple windows of opportunity in a given month for their “massive earthquake or eruption.” They are not talking about probability, but rather specific prediction of geologic events.
And that, folks, is not possible. You cannot predict, weeks in advance, specifically when an earthquake or eruption will occur. Let’s put that to rest.
Let’s take a moment and look at some of these predictions. Here is one from Piers Corbyn, a so-called predictor of weather and geologic events based on the activity of the sun. This is the prediction for April 2012. (By the way, he sells his predictions to those foolish enough to believe him.)
"Predictions" of earthquakes and volcanoes in April 2012 made by Pier Corbyn.
What does it say. Well, between Apr. 8 and 10, there is a “very high” threat of an earthquake or eruption in the Pacific ring, maybe in the northern hemisphere. The earthquake is like to be M6.5 or higher. Here is what actually happened:

Actual seismicity on the planet during the Apr. 8 to 10, 2010 "prediction window", via the U.S. Geologic Survey.
But wait, this “window of prediction” was called a confirmation of Corbyn’s method because the M8+ earthquakes off Banda Aceh in Indonesia occurred on … Apr. 11. That is almost the right date, right? It is almost the right hemisphere! Who cares that there were hundreds of earthquakes in the Pacific and Indian rim every day. Who cares that each year, there are over 150 earthquakes over M6.5 or greater, meaning you pick any day of the year, you might have a 1-in-3 chance of picking the right one. Who cares that in your other “windows” nothing happened. This is what we call “cherry picking” the data to fit your prediction, rather than using the data to assess your validity. Make the window big enough and your “prediction” broad enough, you can claim “success” almost every time.
Another example, this time using asteroid passing near Earth. Some people claim that certain close approaches of asteroids will cause earthquakes that can be predicted. This is equivalent to the“supermoon” idea, that if the moon is full and closer to the Earth than normal, then a multitude of massive earthquakes will occur. It was predicted last March and nothing happened. Here are some dire predictions based on asteroids:

Supposed correlations of the approaches of two asteroids and seismicity -- cherry picking data at its worst.

The prediction, made on Apr. 12, of "earthquakes" on Apr. 18.
It claims that this prediction was a “HIT!” because they predicted an earthquake around Apr. 18 because of two asteroids (2012 FC71 and 1996 SK) passing near Earth. Guess what? There were two moderate earthquakes on Apr. 17 — one in the Bonin Islands, one in Chile, both about M6.5 to M6.7. Guess what? There are earthquakes of that size that occur outside these windows as well! Take a close look at the supposed “hits” — they predict earthquakes at closest approach but sometimes they’re early, sometimes late. This, again, is matching the data to your prediction. Guess what? Look at the NASA NEO data for asteroids passing close to Earth and you can see that that asteroids the size of 2012 FC71 (23 to 51 meters across) and 1996 SK (1,600 meters) pass within 0.2 AU of Earth all the time, so unless you want to say all earthquakes are caused by asteroid approaches, this abuse of correlation equally causation is meaningless.
Earthquakes and eruptions are caused by the Earth’s plate motion and mantle dynamics — it is as simple as that. Sure, we see some evidence that tidal forces from the Moon can cause minor variations in volcanic activity at specific volcanoes when they are already erupting, but sunspots, alignments of celestial bodies across the solar systems, approaches of asteroids (or mystery planets) — none of these have shown any scientifically rigorous correlation to lead to invoke that they caused the geologic event in questions.
So, why do people want to believe this nonsense? Part of it might be the desire to feel “smarter” than the experts. Part of it might be the “thrill of disaster” toward which some people are drawn. I think many times the problem is that we’ve let science literacy fall so dramatically in the U.S. (and elsewhere) that people don’t have the basic skills to discern a charlatan from real scientific research. Throw a pile of data and some nifty correlations and that will convince many people that indeed, the pattern is there. If someone disagrees, then it is because they are blind to your insight. You could make the lazy argument that this is like climate change, except that climate change data has been reproduced and peer-reviewed, showing these correlations are real. When it comes to earthquake and eruption prediction, there is no such scientific rigor that can support the claims on these internet wizards.
The real danger here isn’t the predictions, but rather the idea that some people will believe this type of “black magic” rather than scientific information. Instead of being prepared for an earthquake or eruption, they could just wait for the so-called predictors to say when an geologic event might happen. These “earth forecasters” spread easy fear of earthquakes or volcanoes that can cause panic. (See the “Great Rome earthquake of 2011.”) They are the faith healers of the geologic community and should be seen as such.

No comments:

Post a Comment

You can write whatever you want, but technically, the precise mathematical-statistical forecast of earthquakes based on information about the behavior of wild and domestic animals, birds, fish, and individuals available from 1995, with the advent of social networking.

THE STRUCTURE OF INPUT BIG DATA: API applications to social networks