A Quick Guide to Predicting Earthquakes

It is simultaneously incredibly easy and outright impossible to predict earthquakes. It all depends on how much detail you expect in your predictions.

Sichuan Quake, China. Credit: Lam Yik Fei/Ghetty Images

Geophysicists can predict where earthquakes will happen, the characteristics of earthquakes, and even their probability of occurrence, but they cannot use science to predict exact epicentres and times of earthquakes. Yet every so often, someone gets into the news cycle huffing and puffing about their new trick for predicting earthquakes. I used to roll my eyes, sigh, and possibly even sulk when this happened, but eventually, I developed a new technique for responding. I started making my own predictions.


Every year, I make the same prediction about earthquakes for the upcoming twelve months:

  • 1 or 2 massive magnitude 8 or larger earthquakes along a subduction zone.
  • Between 10 and 20 mid-sized magnitude 7 to 7.9 earthquakes, all along plate boundaries.
  • Between 100 and 200 moderate-sized magnitude 6 to 6.9 earthquakes, mostly along plate boundaries.
  • Between 1,000 and 2,000 small magnitude 5 to 5.9 earthquakes, mostly along plate boundaries.

It doesn't make any difference to me what day I'm asked on; I'll always make the same prediction. Picking June 1st as the arbitrary beginning of the year as it's slightly less weird than picking today's date as the start/end, here's the breakdown of earthquakes for last year compared to my prediction:

Every earthquake M5 or greater from June 1, 2013 to June 1, 2014.


The numbers matched my predictions perfectly! Aren't you impressed?

It'd be pretty dull going through the archives year-by-year in that level of detail, so here's a summary for the past many years of actual earthquakes versus my predicted ranges for earthquakes of different magnitudes:

Historic earthquakes [blue] compared to predictions [red; min/max bounds in grey].


For the most part, the actual recorded earthquakes fell inside my parameters. Sure, there are a few exceptions:

  • From 2012 to 2013, one of the M7.0 earthquakes in Peru was near but not at the fault, roughly 800 kilometers away.
  • In 2011 to 2012, the M9.0 Honshu earthquake in Japan produced a whack of aftershocks, including at least 4 moderate M7.0 to 7.9 earthquakes, 61 strong M6.0 to 6.9, and 608 M5.0 to 5.9 earthquakes. I could compensate by adding a rider to my predictions following the rule of thumb for predicting aftershocks to an M9+ earthquake, but that's too much effort.
  • Years with three major M8+ earthquakes are paired with years with no M8+ earthquakes when years start/end on June 1st. These exceptions are highly sensitive to what specific date is selected as the beginning/end of the year.

Throughout the years, my predictions are mostly right with only minor deviations. It's actually so close that if I tweak the start/end date for the year, I can present the data to be right almost all of the time.

Christchurch earthquake, New Zealand. Credit: Getty Images


All my predictions are is using statistics and vagueness to encompass a wide range centered near the average number of earthquakes per year. I'm actually catering to our fondness for nice, round numbers, both by rounding the center-point of average-earthquakes, and broadening the range to effectively be just 1x to 2x. My accuracy would be more impressive if I abandoned the pretty numbers, picked the actual average as my center-point, and used the statistical distribution to create a range, but the round numbers I use are pretty, work well enough, and are easy to remember.

Did you notice that each order of magnitude decrease in earthquakes is paired with an order of magnitude increase in frequency? That's an eerie sort of symmetry, except that it isn't because magnitude-frequency scales are designed to behave exactly in that manner. In nature, higher impact events happen less frequently and lower impact events happen more frequently. It's a quirk to be grateful for, but also the defining characteristic of logarithmic classification schemes for disaster magnitudes. If earthquake A has 10x the energy of earthquake B, then earthquakes of magnitude A will happen 1/10th as often as earthquakes of magnitude B.

Which all gets down to this: despite being right, my predictions are so vague that they're utterly useless and completely unimpressive.

1906 San Francisco earthquake, California. Credit: Hulton Archive/Ghetty Images



Yet, this is pretty much what happens every time someone grabs the spotlight to start ranting about their weird new trick for predicting the ground-shaking apocalypse. A fear-mongering quake-predictor will say something ominous, like, "We will have an earthquake tomorrow afternoon!" without specifying magnitude, location, or time. Then an earthquake happens somewhere in the world, the hoax-perpetuator puffs in pride, and I gnash my teeth in frustration. With 1,000 to 2,000 small earthquakes a year, of course a prediction that vague will pay off!

Let's put this into action: This article is going live at 11:00 am EDT on July 12th, 2014. I predict we'll have at least one earthquake of magnitude 4 or greater in the next twelve hours, before 11:00 pm EDT. More than that, I predict we'll have at least one magnitude 5 or greater earthquake in the next 24 hours, and at least one magnitude 6 or greater earthquake in the next week.

Anyone else want to chime in with their predictions? We'll reconvene next Saturday to distribute science .gifs to all successful forecasters.

Reply5 replies
Leave a reply