Predicting future earthquakes
Protection against the effects of earthquakes matters of course, but if we could predict their impending arrival, that would be even better: we could evacuate people; put rescue services on standby; prepare to turn off gas or electricity supplies if necessary, and so on. But, regrettably, it’s not that easy. And to explain why, Chris Smith was joined by UCL's Joanna Faure-Walker…
Joanna - The problem is it's, as you mentioned, very difficult to actually predict an earthquake. So when you're talking about prediction, you're saying, you know, how big will the earthquake be and exactly where and when is it going to occur? So is it going to be on Tuesday at 5:20 PM and magnitude 8 on this particular fault? We're nowhere near that. The problem is to be able to do that - let's just consider one fault in isolation at the moment. If we wanted to be able to predict when the earthquake would occur, we would need to know how large we expect the earthquake to be, so how much movement on that particular fault do expect that to be? And how long would it take to build up the stress required for that amount of movement? So how fast is the fault moving? We'd also need to know how much of that movement is taken up, what we say seismically, so in earthquakes, versus aseismically, so not in earthquakes. And then we would also need to know when was the last earthquake on that fault so we know where we're starting our timer. Now, all of those things are actually very difficult to measure. We know some of them on some faults, we actually know almost all of them on particular faults, but in most places in the world, most of those elude us. Even if we did know all of those things, it's more complicated than that because faults don't behave on their own. They're not isolated structures, they interact with each other. So the forces that are governing the timing of these earthquakes is just more complicated than at the moment our models can deal with.
Chris - I was going to ask you - do faults talk to each other? in the sense that if you've got one that starts to unzip or move, does that actually have an impact or knock on effect at encouraging another fault to unload itself as well? Does it transfer the stress elsewhere, if you like?
Joanna - It does exactly that. I like your expression there of talking to each other. When an earthquake occurs on a fault, we lose the stress on that fault. It releases the stress there. But it also is causing movement throughout the crust, that's where we're standing on now and that's where these earthquakes occur. So other faults will either be accelerated or you can actually bring delays so you can slow down that rupture. So the problem with this is the idea you'll sometimes hear about recurrence intervals on faults - so the average time interval between earthquakes. But if we've got all these interactions taking place, it's not a regular time interval and so it's much more difficult to predict again.
Chris - It becomes a massive, massive problem doesn't it with so many different dimensions to it. Have new tools helped us though? Because we've got things like GPS that can make very accurate movement assessments of how the land is moving that would previously have been very difficult to do at the sort of scale that we can now do them. Does bringing more data to the table help with these sorts of predictions?
Joanna - It really does. It really helps us understand the fundamental mechanics - so how fast are these faults moving? But the problem is we're looking at timescales here. Now as a geologist, when I talk about short term time periods, I'm talking about hundreds or even maybe thousands of years. So as a geologist, that's a short time period. Now these GPS records that you just referred to can look at years or tens of years, but they're not going back hundreds of thousands of years. Now, if we consider an individual fault, they can move at parts of millimetres or centimetres per year. So about as fast as our fingernails grow. How long would you need to not cut your fingernail before you could have, you know, a metre or even 20 metres of slip in one event? So on an individual fault, it may be hundreds or thousands of years between individual earthquakes. And with these modern methods of analysing the movement, yes, they're giving us a lot of information, but we're looking at such a short time period compared to the length of what we call the seismic cycle.
Chris - Are there any other kinds of markers that might be proxy markers of energy that's being stored in compression in these faults that's slowly building up? Because, you know, I know people say the animals change their behaviour ahead of earthquakes. Now this might just be attaching significance to coincidence might'n it because about 10 years ago there was this lovely paper - unfortunately it came out around the time of April fool's day which meant everyone kind of said they thought it was a joke, but it was a bunch of toads that just disappeared from when they would normally be mating around a site in Italy that then a few days later had a huge earthquake. And the researchers there speculated perhaps there were changes in the Earth's ionosphere, the pattern of how the earth guides radiation round its outer part of space. Are there any markers we might be able to plug into that would point us towards a critical stress building up at certain fault points?
Joanna - There are some ideas and some of them are more promising than others. I mean, the problem with animal behaviour is we are yet to see a statistical study. Something like the toad behaviour, or dogs barking, things like this, if you actually monitored the toads every single day of the year, and you kept doing this year after year, and you actually said that there was statistically significant different behaviour on those dates, then you might be able to start using some of these methods. But as yet, I do not know of any studies where they actually look at these. So you hear people saying things like the toads moved or the dogs barked, and it's like, well, yes, toads move and dogs bark. So I don't think those are particularly promising at the moment. But what we can do, and a lot of the work that I'm involved in, is trying to look at long term rates of deformation. So geological timescales for thousands of years, tens of thousands of years, and we can use markers on the longterm where we look at how rocks have been displaced relative to each other, to try and understand how fast are these faults moving in the long term? And then compare these to things like the GPS records to see where are we in that cycle of stress building up, and is there any helpful information about imminent failure.
Chris - And are we better at doing this these days? Or are we just better at knowing we're no good at it, if you see what I mean.
Joanna - We are better. So we are producing what we call probabilistic seismic hazard approaches. So this is where we use probabilities - so how likely is an event to occur in the next 5, 10, or even a thousand years, depending on what you're interested in, and on those we are making very good progress. It helps us make relative risk decisions. But in terms of actually saying whether we expect an earthquake at 5:00 PM on Tuesday we still have a way to go.