Can We Predict Earthquakes?

Hey, welcome to sign Stuff, a production of iHeartRadio. I'm hoor Chimp. And today we are answering the question can we predict earthquakes? What are ways humans have tried to get a heads up on when the ground is gonna shake? Do animals have a sixth sense about it? Or can AI see things in the data that we cannot. We're gonna sit down with two earthquake scientists who are going to explain that we can actually predict earthquakes, just not

So this interview is maybe an hour or so, probably a pretty high likelihood, and in California even probably quite a high likelihood.

So at the end we're going to find out if there was an earthquake during our interview, and spoiler alert, there was. Okay, let me take a step back. I went over there to talk to my friend doctor Alan Husker, the professor at the Seismological Laboratory at Celtech and the manager of the Southern California Seismic Network, and he introduced me to doctor Elizabeth Cochrane, a researched geophysicist with the US Geological Survey. So here's my visit to the hub

Now here we go.

Hello, how are you good to see? Thank you for letting me crash your meeting? No problem, fun well, thank you doctor Husker, Thank you Doctor Cochran for joining and talking with me here today.

Happy to be here.

Yeah, thanks for having us.

Can you please tell us would you do what your every day like?

So, typically most of my days are spent looking at seismic wiggles, so trying to understand what happens when an earthquake occurs and the effects on people and things around us.

I spend a lot of time managing so budgets, people, making sure equipment gets out to where it's supposed to be in the field. And then I'll have a grad student who's looking at moonquakes.

Oh wow, okay. I started by asking them to give me a rundown. The way people have tried to predict earthquakes starting with animals.

So people think that animals can predict earthquake. It's a very common. Most people do actually this, well, a lot does that learn. So some people just think that their dog or whatever has extra senses that humans don't have, and so they can sense when something's going to happen before it happens, because of better vision or hearing or

So it's kind of like your pet acts odd a lot, and then if an earthquake happens after that, you attribute the odd behavior that day, do them knowing an earthquake is happening, when in fact, you know five days out of seven they're doing something odd. So there is a really nice video from one earthquake in Seattle in two thousand and one that happens to be focused on a dog that's sleeping on the floor, and.

The dog gets up, runs.

Away, and then you sort of clearly see that building shaking, But it turns out that in fact, if you actually look very closely, the building does start shaking when the dog gets up the dog noticed the first seismic wave to arrive, and you know, because they're laying on the floor, they felt that vibration, were startled by it and got up, and then the stronger wave arrived. Then that's when people noticed and started moving around.

Yeah, according to our experts, there's no convincing evidence dogs can predict earthquakes, but it has to stop people from trying other animals.

There was one experiment I heard about that happened in China where they put electrodes on goats nipples, Oh, if they could predict earthquakes or whatever.

Barently nothing came out of There was.

Also a different experiment that was done similar thing, where someone kept tanks of these eels. And because one of the ideas are some sort of electrical signal as the rocks are breaking, and you get some electrical signals that animals that are sensitive to that would be able to detect it, it didn't work.

Now, to be fair, scientists have tried more serious ways to predict earthquakes. For example, they've looked to see if the ground expands or swells right before an earthquake, or if there's an extra amount of radon gas that comes up from the ground right before an earthquake hits, but none of these studies have found anything that can be useful. Scientists have even looked to the sky for signs that predict earthquakes, specifically a layer of our atmosphere called the ionosphere.

So people have looked for other things after really really big earthquakes, like magnitude eight point five and above. It's so big that it can sometimes see in a signal in the ionosphere. So there are all sorts of different agnotometers and other equipment on satellites around the Earth all the time. So they tried to look for changes in the iosphere that what happened beforehand, and it hasn't worked either. I would say I'm on the skeptical side.

Okay, the I in the sphered doesn't quite work either. How about using AI?

Yeah, so I think most of the studies that have suggested that earthquakes might be predictable using sort of deep learning AI techniques are applied to laboratory data, so those I think potentially are successful. Boothquakes are quite a bit different than earthquakes.

In the lab.

The ones in the lab are more predictable.

Because they're more predictable.

The other thing that I've seen with some of these AI techniques is they'll say that they get ninety percent prediction rate, which is great, but ninety percent of the earthquakes don't matter.

The ones that matter are the big ones.

So there was a kind of a prediction conference in China where if they can predict an earthquake using AI. So they said that all their predictions were doing really well. There was a nineteen six point four that actually occurred during the conference that none of them predicted.

Okay, so apparently dogs can't predict earthquakes, and neither can goats or eels, or studying the ground or analyzing the gases emitted underground or measuring the Earth's ionosphere, and not even AI can predict when the next earthquake is going to happen. So does this mean it's impossible. When we come back, we're going to ask our experts why it's so hard to predict earthquakes and how in some cases

So an earthquake is a rapids breaking.

Or slip along a fracture on the Earth. So basically, stresses build up in the earth and they're released really rapidly during an earthquake.

Now, according to our experts, there are two things that make earthquakes so hard to predict. The first is that it all happens miles under the ground.

So most of the earthquakes happen sort of five to ten miles deep.

Oh, in subduction zones.

It's much deeper because it's between where the two plates are rubbing against each other, deep underneath the earth, so it might be like fifty miles.

Here's an interesting fact. Earthquakes happen at very specific depth depending on how the rocks are rubbing against each other. Earthquakes don't happen deeper because the rock gets hot, which makes it flow instead of crack, and they don't happen above about a mile or so down because the rock there is too loose or crumply. But the ones that do happen still happen about ten to fifteen miles under the ground, which makes them hard to study.

Well.

It's difficult to predict earthquakes is we can't see, like there's.

Rock in the way, so a nice analogy.

It's quite difficult even to predict what the weather's going to be like tomorrow, and in that case, there's all kinds of information about the pressures in the atmosphere, the temperatures across a whole three D volume, and even then our weather forecasts are not exact.

Right.

In this case, we're trying to do the same thing, except buried under miles of rock, where we don't have any or many measurements of exactly what the temperatures are, exactly what the pressures are.

We're kind of blind. Yeah, So the first reason earths are hard to predict is that they happen ten to fifteen miles and sometimes fifty miles under ten to fifteen miles or fifty miles of rock. We don't have a direct view or even a way to directly measure what's happening down there. Scientists have to infer what's happening from what we can see and hear on the surface from up here and from satellites and kind of guess what's

Have drilled into faults.

For example, there was a hole that was drilled across the San Andreas fault perpendicularly through the fault, and if not that deep, so it's down I think it was two kilometers three kilometers, and from that one hole, we have so much more information about exactly what those materials look like, what the stresses are in that one hole, right,

All right.

The second reason earthquakes are hard to predict is that earthquakes happen basically without warning. Remember I said that earthquakes happen because two rock plates are pressing on each other, and they're pressing really hard, and at some point something gives and the plates slip past each other or one goes under the other. Well, that give is really hard to know when it's going to happen. It's kind of like when you're standing on ice and you're pressing down

An even simpler example is it take a stick and you start to bend it. Can you even exactly when that stick is going to break before it snaps in half? And so this is kind of the same thing except in the earth.

Okay, this is something you can try at home. Rub a stick or a wood pencil, or better yet, a single stick of uncooked dried spaghetti and start bending it as you put more bend into it. You can feel the stresses building up in the stick or pencil, and at some point you can't predict it's going to snap. The same thing happens with these tectonic plates. They press on each other, the stress builds up, and at some point they crack and slip, and that's an earthquake. Now,

About ten percent of the time is a four shock, but it's most of the time there isn't.

Although we've gotten really good at detecting really small earthquakes, we still don't necessarily see a lot of four shocks. In addition, there's nothing different about a foreshock. It doesn't look different. There's nothing that we can when an earthquake happens. We can't be like, oh, this is a foreshock.

You can't take the difference between a four shock to a big one and just a regular, smaller earthquake, right, Okay, So that's why earthquakes are hard to predict. They happen ten to fifteens, sometimes fifty miles under solid rock. It's a process that happens, usually without warning, just from the physics of it. Now, again, does that mean it's impossible to predict an earthquake. Well, it turns out there are two ways in which we sort of can predict earthquakes.

All right, So a fault basically is stuck for most of the time, right, And the reason it's stuck is because the forces pushing the fault together are higher than the forces trying to move it sideways. And so if you change how much fluids are in the fault, you change if you put a bunch of water in, you're effectively reducing that stress that's holding the two sides.

Of the fault together.

I see, you're making it slippery, or.

You're making it easier for the stresses to be larger than the stress is holding.

The fault together.

So it's like if I was pushing down on the table super hard, it makes it really hard for me to move my hand sideways.

But if I.

Put a bunch of water in here, it's pushing back on my hand and making it so it's then easier for my hand to move. So one idea is that, well, if we could trigger an earthquake, uh huh, then we would actually be able to study in detail what happens.

Yes, that's right. Scientists have actually considered drilling holes and pumping a lot of water into a fault in order to start an earthquake.

So this is not just a of idea that we came up with today.

Right, There was actually a meeting.

A few years ago of folks working on earthquake physics where we actually talked about whether it would be possible to have an earthquake experiment where we actually would create an earthquake.

It just seems like a terrible idea.

Most of the workshop consisted of people talking about where we would do this because it had to be away from people, right.

I think ideally we would want to have like a ranch of magnitudes, Like we're talking about a three or four, but if we could create a six or seven, that'd be great. Like, there's plenty of earthquakes that happened really far away from more people are and like nobody feels them, so like it'd be like one of those like.

Well wait, this is just an idea, right, they haven't actually done it, have they.

Well, yes, I know, we've started lots of little ones. Oh yeah, well, I mean the geo thermal regions and stuff. They're doing it all the time.

And there was a purposeful experiment a KTB borehole in Germany. So this was think in the nineties where a deep borehole was drilled and then they injected fluids and they triggered a number of earthquakes, small earthquakes. These were kind of magnitude one to two.

Yeah, okay, So the idea would be to create an earthquake in order to get more data and understand them better. But we also talked about starting an earthquake to make it more predictable.

Another thing along those lines, if you wanted to say we can't predict earthquakes, but we could generate one. Let's say, if we think we could generate a magnitude seven or eight on the San Andreas. Essentially we would kind of speed up the clock for when it would happen.

What our experts are saying here is if we can induce earthquakes, and it looks like we can, then maybe for a region like Los Angeles where we know a big earthquake is coming, the best thing to do is to get it.

We're with and so we could, you know, evacuate LA for a while, pump it as hard as we could, and see if we can create a giant earthquake on the San Andreas fault, uh huh, and then we've alleviated it now for a while, and so people can move back into La and like feel safer.

In other words, we could potentially schedule earthquakes. I mean, it doesn't get more predictable than that. Of course, there are a few problems with this idea.

The problem with that is that every large earthquake is then followed with many, many aftershocks. So and typically the largest aftershock is about only one magnitude unit less than the main shock. So if we trigger an eight, we're then going to have a magnitude seven aftershock. Plus you know many six's, hundreds of five, you know, thousands of fours,

Yeah, let's think about that one a little more, all right. And then the other way in which earthquakes can be predicted is if you consider the timescale of your prediction. Remember the analogy of bending a stick and not knowing when it's going to break, Well, that's not quite true. You do have some sense of when it's going to snap.

You have a general idea of on that timescale. It might not happen instantaneously, but there's a within a few seconds you can make it happen.

Right, Meaning like if I start bending it stick, I sort of know within a few stens of seconds when it's going to snap exactly.

And so like we're actually really really accurate in predicting earthquakes, but it's on the Earth timescale, right, So the Earth, the Earth is you know, billions of years old, and can say there's going to be an earthquake, you know, at the San Andreas in the next hundred years, and you're like, oh, that's really good, right, But no, it's not useful for humans.

It is useful to forecasts in that way because that's how we develop building codes, right. Our buildings last for hundreds of years sometimes, and so we develop these earthquake forecasts so that we know how to build the buildings to withstand the likely events that those buildings will see.

So we may not be able to predict earthquakes down to the minute or hour or even month or year, but through some amazing science by awesome scientists, we can somehow see beneath miles of rock and have a sense of when big earthquakes can occur, which maybe is enough. It can give us guidance about building codes and put people unnoticed to be prepared. And speaking of being prepared, let's find out if there was an earthquake during our interview.

Let's do it.

So where are you taking?

Just right? Monitor? Yeah?

What is it?

This is data coming from the Southern California sized network. This is a monitor that we're collecting all the four hundred stations.

So we're obviously not showing all of the different centers, but they're all across Southern California and listening to all the little events happening on the faults nearby.

But most of this is just noise. But this line right here, this flat line, that's an earthquake. Whow So at about twelve to ten there was an earthquake. It's probably an aftershock of the ridge gross earthquake.

Yeah, or there's a little swarm of events happening right here.

Oh yeah, there too.

Okay, So at twelve ten pm Pacific time, there was an earthing and we were talking, Yes, how was this earthquake?

It was been tiny, maybe like a magnitude one or two, and.

It's telling you, hey, there was something yep, okay. The last question I asked doctor Husker and doctor Cochrane whether it would matter if we could predict earthquakes.

You know, there's also questions of sort of if we had predictions, what would we do with it? If I knew half an hour before a magnitude eight on the San Andreas, what.

Would we do right?

Right?

So you couldn't evacuate Los Angeles in half an hour? Right? It takes half an hour to get ten miles down the road on any given day, just two to traffic.

So I asked them what would you do if you could predict earthquakes at different timescales? Like what if you knew an earthquake was likely to happen in the next ten years, or the next month, or in the next five minutes.

I mean ten years would be helpful if we knew which fault and how big, right, because that would be a specific scenario that we could plan for. A lot of our water comes across the San Andreas fault, A lot of our electricity are you know, so you could imagine if you had ten years and you knew that this pipe was the one that was going to be broken, you would spend that ten years hardening that pipe. You know, all of those supply lines cross the San Andres.

A month is still really long, but you could start to prepare. From the governmental point of view, from a fireman FEMA, these kind of things like you, okay, let's get water, large tanks of water everywhere. If it was of magnitude eight long San Andres, is it going to break? The I five and I ten those are the major arteries that leave Los Angeles, so people won't be able

Warning of five to ten to thirty minutes could be very useful. So we know that there's a number of buildings, for example, in Los Angeles that have not been fully retrofit to with stand earthquakes. These are older buildings, they were built well before we knew the full earthquake hazard, and so you could evacuate those buildings.

Oh, I see, it might give you some time, right, all right, So to recap can we predict earthquakes? Well, they're unpredictable. They happen deep in the earth where you can't see them, and they happen because of rocks pressing on each other and slipping without warning. Despite that, science can observe the general movement of tectonic plates and build models that can give us some estimates of when they're