4 ms·

Possibly a silly question, but... I hear the term 'overdue' applied particularly to earthquakes, rather than any other naturally occurring event. Does this mea

Possibly a silly question, but... I hear the term 'overdue' applied particularly to earthquakes, rather than any other naturally occurring event. Does this mean 'overdue' in the same sense that rolling a six is 'overdue' if you've not seen one after six rolls?

991219921 10y ago
No, It means that the tension between tectonic plates steadily builds up, so an earthquake is more likely.
- marcosscriven 10y ago
  I see - the article though seems to talk about 'surface-breaking' earthquakes occurring every 100 years in that region, and that we're now at 159 years without one. This idea of a time expectation doesn't seem to match your explanation (which seems more intuitive to me).
  - rgossiaux 10y ago
    They're the same, really -- those quakes happen every 100 years because that's how long it takes for that much pressure to build up. 159 years without one means there's already going to be quite a lot of pressure built up that's going to have to give way sooner or later.
    - sevenless 10y ago
      We know that fluid injections for fracking can cause earthquakes. How about injecting some extremely viscous fluid deep underground within the plate boundaries? So it can release the shear force over months or years rather than seconds. Something like pitch (https://en.wikipedia.org/wiki/Pitch_drop_experiment https://en.wikipedia.org/wiki/Pitch_drop_experiment) Of course, the major problem with trying quake mitigation strategies might be the legal responsibility if you cause a quake.
      - cossatot 10y ago
        Aside from the very real liability issue, the idea of lubricating faults is interesting. From a practical standpoint it would be hard to inject something that remains viscous enough at the relevant temperatures. Earthquakes nucleate at about 10 km depth in the San Andreas, which is about 350 degrees C. So aside from the difficulty of drilling in these conditions, you would have to have a fluid that maintains its viscosity from surface temperatures to 350 C.
        sevenless 10y ago
        Given fracking injections are a lot shallower than 10km, I'm not sure it would have to be injected that deep. But it's a good point, that temperature will be a factor. Getting a high-melting-point fluid down there might be done by dissolving or suspending it in water that is allowed to boil off, or having a heated injection pipeline. Silicone and silicate lubricants are viable at those temperatures. Might not be viscous enough, though. A low melting phosphate glass might work.
gradys 10y ago
My understanding is that the probability and expected magnitude of an earthquake actually increases over time as tectonic tension builds. So no, not like dice.
- marcosscriven 10y ago
  I meant in the sense of probability, not literally like dice :)
  - tlrobinson 10y ago
    Well yes, pretty much any naturally occurring process involves some probability...
  - maldeh 10y ago
    It essentially comes down to whether you assume that the probability of an earthquake every subsequent year is independent of the result in previous years. The notion of tectonic pressure build-up suggests that that assumption isn't true. The author has a reasonable idea in applying survival analysis to estimate the probability without making any assumptions about the underlying model.
  - pc86 10y ago
    I think it was clear what you meant. If you roll a dice 10 times and don't get a six, the probability of rolling a six is still 1/6 for the next roll. However, as the time since the last earthquake increases, not only does the probability of one increase (unlike dice which has a constant probability) but also the likely magnitude of it increases as well. The equivalent with dice would be you roll a non-six 10 times in a row, and instead of 1/6 probability of rolling a six, you now have a 1/3 probability of rolling a six and a 1/100 probability of setting off a nuclear explosion.
    - marcosscriven 10y ago
      Thanks for the helpful response. I still don't think I got quite the question across that I meant though. It wasn't so much the probability model (independent or otherwise) I intended to question (hence my very poor choice of the die rolling analogy), but the choice of describing an event as 'overdue'. At the risk of introducing another down-voted analogy, presumably hurricanes have a probability of occurring based on things like ocean temperature, time of year etc., but I don't think I've ever heard the phrase 'the east coast of the US is overdue a hurricane'.
avar 10y ago
It's like knowing that 1/2 dice rolls yields an even number, and after rolling twice in a row and getting odds making a big deal of the supposed statistical anomaly, because we got two odd numbers in a row we're really overdue for an even number, don't you think? When volcanos/earthquakes etc. are said to occur every N years, that's averaged over geological periods. An event that occurs on average every 100 years over tens or hundreds of thousands of years could be occurring once every 10 years some millennium, or once every 250, 500 or 1000 years for some time span.
- batbomb 10y ago
  Please read up on binomial and Gaussian distributions.
  - cossatot 10y ago
    Avar has a good point in general, although I'm not sure it's relevant to the recent San Andreas earthquake history. Over geological time, the actual probability distributions are unstable, because the boundary conditions causing the events change as plate configuration, climate, etc. changes.
cossatot 10y ago
With dice, each roll is independent of the others. With earthquakes, the probability of an earthquake occurring in any given year is (almost always) dependent on how long it has been since the last event. That is represented by the hazard function in the post (lambda t). I say 'almost always' because for a few distributions the hazard rate is constant with time. An interesting thing about the hazard function with earthquakes is that it isn't always monotonically increasing; I've done this analysis on some other faults that show a very high hazard in the years to decades after an earthquake, and then the hazard drops down and doesn't recover until the far right tail of the distribution. This reflects temporal clustering of earthquakes, which is something we're gradually becoming aware of.
sethammons 10y ago
I would compare it to filling water balloons and timing when they pop. Our current water balloon should have popped a while ago, but the tension/pressure is still building.

991219921 10y ago

No, It means that the tension between tectonic plates steadily builds up, so an earthquake is more likely.

marcosscriven 10y ago
I see - the article though seems to talk about 'surface-breaking' earthquakes occurring every 100 years in that region, and that we're now at 159 years without one. This idea of a time expectation doesn't seem to match your explanation (which seems more intuitive to me).
- rgossiaux 10y ago
  They're the same, really -- those quakes happen every 100 years because that's how long it takes for that much pressure to build up. 159 years without one means there's already going to be quite a lot of pressure built up that's going to have to give way sooner or later.
  - sevenless 10y ago
    We know that fluid injections for fracking can cause earthquakes. How about injecting some extremely viscous fluid deep underground within the plate boundaries? So it can release the shear force over months or years rather than seconds. Something like pitch (https://en.wikipedia.org/wiki/Pitch_drop_experiment https://en.wikipedia.org/wiki/Pitch_drop_experiment) Of course, the major problem with trying quake mitigation strategies might be the legal responsibility if you cause a quake.
    - cossatot 10y ago
      Aside from the very real liability issue, the idea of lubricating faults is interesting. From a practical standpoint it would be hard to inject something that remains viscous enough at the relevant temperatures. Earthquakes nucleate at about 10 km depth in the San Andreas, which is about 350 degrees C. So aside from the difficulty of drilling in these conditions, you would have to have a fluid that maintains its viscosity from surface temperatures to 350 C.
      - sevenless 10y ago
        Given fracking injections are a lot shallower than 10km, I'm not sure it would have to be injected that deep. But it's a good point, that temperature will be a factor. Getting a high-melting-point fluid down there might be done by dissolving or suspending it in water that is allowed to boil off, or having a heated injection pipeline. Silicone and silicate lubricants are viable at those temperatures. Might not be viscous enough, though. A low melting phosphate glass might work.

gradys 10y ago

My understanding is that the probability and expected magnitude of an earthquake actually increases over time as tectonic tension builds. So no, not like dice.

marcosscriven 10y ago
I meant in the sense of probability, not literally like dice :)
- tlrobinson 10y ago
  Well yes, pretty much any naturally occurring process involves some probability...
- maldeh 10y ago
  It essentially comes down to whether you assume that the probability of an earthquake every subsequent year is independent of the result in previous years. The notion of tectonic pressure build-up suggests that that assumption isn't true. The author has a reasonable idea in applying survival analysis to estimate the probability without making any assumptions about the underlying model.
- pc86 10y ago
  I think it was clear what you meant. If you roll a dice 10 times and don't get a six, the probability of rolling a six is still 1/6 for the next roll. However, as the time since the last earthquake increases, not only does the probability of one increase (unlike dice which has a constant probability) but also the likely magnitude of it increases as well. The equivalent with dice would be you roll a non-six 10 times in a row, and instead of 1/6 probability of rolling a six, you now have a 1/3 probability of rolling a six and a 1/100 probability of setting off a nuclear explosion.
  - marcosscriven 10y ago
    Thanks for the helpful response. I still don't think I got quite the question across that I meant though. It wasn't so much the probability model (independent or otherwise) I intended to question (hence my very poor choice of the die rolling analogy), but the choice of describing an event as 'overdue'. At the risk of introducing another down-voted analogy, presumably hurricanes have a probability of occurring based on things like ocean temperature, time of year etc., but I don't think I've ever heard the phrase 'the east coast of the US is overdue a hurricane'.

avar 10y ago

It's like knowing that 1/2 dice rolls yields an even number, and after rolling twice in a row and getting odds making a big deal of the supposed statistical anomaly, because we got two odd numbers in a row we're really overdue for an even number, don't you think? When volcanos/earthquakes etc. are said to occur every N years, that's averaged over geological periods. An event that occurs on average every 100 years over tens or hundreds of thousands of years could be occurring once every 10 years some millennium, or once every 250, 500 or 1000 years for some time span.

batbomb 10y ago
Please read up on binomial and Gaussian distributions.
- cossatot 10y ago
  Avar has a good point in general, although I'm not sure it's relevant to the recent San Andreas earthquake history. Over geological time, the actual probability distributions are unstable, because the boundary conditions causing the events change as plate configuration, climate, etc. changes.

cossatot 10y ago

With dice, each roll is independent of the others. With earthquakes, the probability of an earthquake occurring in any given year is (almost always) dependent on how long it has been since the last event. That is represented by the hazard function in the post (lambda t). I say 'almost always' because for a few distributions the hazard rate is constant with time. An interesting thing about the hazard function with earthquakes is that it isn't always monotonically increasing; I've done this analysis on some other faults that show a very high hazard in the years to decades after an earthquake, and then the hazard drops down and doesn't recover until the far right tail of the distribution. This reflects temporal clustering of earthquakes, which is something we're gradually becoming aware of.

sethammons 10y ago

I would compare it to filling water balloons and timing when they pop. Our current water balloon should have popped a while ago, but the tension/pressure is still building.