Well, anyway, it's just from Environment Canada, a hybrid store number joining us today, discussing her research in the booth.

Just go.

So, I actually I'm going to start by saying that the title is kind of completing Originally, I was a lot of time talking about you have measurements, deformation, changes in motion in eastern Canada.

Very new and interesting work and how it hasn't over.

Some events have superseded that talk about other, like the B C, which I will touch on and and seven kinds of going to prison in Italy.

Theories on that.

In a brief outline, I will finish with, um, uh, work with work on deformation and earthquakes in eastern Canada.

And I will cut on that, but more time, probably on the other again.

Thank you for attending.

And thank you for having me.

There are a lot of fun for me, and then I'm gonna start by talking about what earthquake is, what we know about it.

And then when I move into earthquake forecasting of it and thats we'll talk again a little bit about that.

And I'm not finished with nation Canada.

This, of course, is the Haitian quick is we know that it killed 1000 people.

Damages I like to show to talk about this.

Like when I talk about this quake hazard, because this early, into the intense January 12 not three months later, we had a quick of the almost exact same magnitude.

Magnitude seven in Southern California just felt in San Diego and it killed nobody.

And as a matter fact, most you probably don't even have mostly art respect times probably never even heard of.

Or remember the Eastern earthquake of 2010 in California and the reason that we can prevent it, we can prevent death.

But we have to have do that.

And we have to know a bit more about what those risks are in place courses the quake of 1906 I use this image with my students in, uh, natural passes on how the earth moved.

And of course, we talk a lot about you, and I'd like to point out to them that the great terrific was supposed to occur before this when I was a teenager.

I don't tell them how many years ago that was because they wouldn't be impressed.

I know that I was told this was certainly happen again before the year 2000 that we would see it again.

And of course it happened happened because we aren't really all that good at forecasting recurrence when how often the large events occurred.

Um is easy and maybe even to get a good average on that.

But the variation in that average in a lot of variability.

So we could say the non average earthquake occurs every 100 or 121 150 years.

We may know that the average number we may know that pretty well, but the variability on that 100 years could be as much a plus mine.

One.

Yes, forecasting Earthquakes are predicting that good.

This is the friend I was quick and tsunami of last year 2011.

And again, this is approximately a magnitude nine progressive.

It's still, but it killed 20,000.

We know that the tsunami killed most of those people and caused we want to think about forecast in half.

We know a lot about you won't know about, has a map.

We will use them, but and has a Messer are an exceptional tool.

I feel kind of looking to build a better has exceptional tool for planners.

Engineers for building codes for building police don't fall down, but forecast is different.

Forecast is gonna provide you a probability of an earthquake occurring at a specific location over a fixed period of time forecast.

In very in time, we could have a forecasted a matter of days a week.

Short term forecasting.

We can have forecast that are a matter of years, tens of your spot years or tens of years, 5 to 10 years.

Or we can have four cats which were effectively again.

I have a map which are really a your time frame.

Historically, many attempts have been made to do earthquake forecasting.

However, today we are as close as we ever will be.

And the reason is that we, in my opinion and the reason is that today we collect more information about where and when they occur into small event.

We used to rely on forecasting be done using large man.

We would say that we know how often a magnitude six occurs.

We're gonna look for particular kinds of signals before that magnitude 78 and we are going to then try forecast that very hard to do, and it was extraordinary.

Unsuccessful, because we have very foreign statistics.

On magnitude six, we have very poor statistical magnitude seven or a magnitude eight because they occur so infrequently.

However, we have lots of information about magnitude three of them because they had personal office, they incurred 100,000 more often, then a magnitude five for that, so we can collect a lot of statistics that we can use.

Those statistics tell us something about earthquake forecasting about when the next picture likely example of that again, we could be record today, 500,000 earthquakes per year worldwide and 10,000 in Southern California alone.

Those air small events, not nobody feels.

But I'm gonna say this again that most of those small events are going to be used as a sensor for the larger events that are coming.

We're gonna assume that those small events are telling us something about the underlying seismicity there, acting as a Spencer underlying strips.

They're acting as a stress sensor.

They're telling us where stresses are increasing or decreasing.

It's not a simple problem is a complicated list, but it's the first order of zeros order effect, you can assume that says misty rate changes.

Changes in the number of events we have are a reaction to changes.

As a result, we're gonna use those to suggest where upcoming events might happen, because those also react the same stresses.

It's just that it takes so again.

This is a forecast map we made over 10 years ago.

Now it was the forecast earthquakes from 4010.

That's the one on the left and, um, again grid again.

It's like it's the forecast.

Raising magnitude five from 4010 and all those red and orange and yellow lots of locations where are expected to happen over the next 10 years.

And you look at that map over plotted on it all.

The earthquake off magnitude five graders have happened in the 10 years following 2000 and you will also notice that there are only two number seven and eight there.

39 place on that plot, 39 races and magnitude five, and all of them are on or near the national air.

We specified, which was, incidentally, 10 kilometers, so within 10 kilometers they all fall within one of those That's not bad.

There's a fair number of false positive, the problem with anyone.

And yet with any respect for casting technique, this or any other is the issue of false.

Okay, you're always gonna forecast some earthquakes that never and you can see that we do have a relatively high false positive right here, about 20%.

But you'll also notice that again we only have two misses.

But in addition, there's a very large part of that map which we say there aren't gonna be.

And that is a benefit to a method like this, because it can tell you where you need to put your resources in the next 5 to 10.

Where should you be upgrading death?

Where should you be reinforcing your schoolroom?

OK, where should you be doing those kinds of things over the next intermediate with intermediate term over the next intermediate term?

And again we think that this was reasonably successful.

If you look at it from that perspective on the right, you have a had a map of the reason for this up again.

What engineers used to build a building with probability exceeding a certain acceleration due to gravity.

And again you'll learn that there are lots of colored areas on.

This is what we use for building codes.

This means that sometime in the next 50 years, there's a 10% chance that Los Angeles good experience and acceleration a jump that's about 80% of 10% chance that in 50 years that's what I would tell you that I So the real issue there is.

What they use it for is used to build buildings that would fall down when that eight, when that 80% of grabbing holds up its building code record.

So this is again that's saying that outline what we've done here and to remind you that what we want to do here is what we are doing here is assuming again.

It's that first line that's most important, that small earthquakes are acting in the centers for the large events.

Acting's a center for the stress changes that occur prior to the large events.

So again, to make this map, we only use earthquakes of magnitude three and above.

Okay, so again, we used there are approximately 25,000 or so that occurred in the last three that we have recorded, I should say, in the last 30 years in California, Magnitude three in a bus.

That's what we use for to create that.

Yet Christie published this prediction back here, too.

Today we submitted it in two dozen.

Didn't actually make press into 2002 huh?

Commentary on publication In today's world, the last 12 years have changed.

So, um, I have learned a lot over the last 10 15 years.

Um, for California, I would change almost nothing.

It turns out you have experimented e I have experimented with other separate using magnitude to about you don't gain much.

So I looked at the information that you get from using different, uh, different magnitude.

Different scale, different sizes.

Way we've been these down into a grid, for example, about 10 plumbers to help.

It turns out there are looking at other no nothing.

I've changed up nothing anyway in California.

What I have learned is that the first probably I have learned, although looking in other areas like Canada, is that is that there are a couple of things that matter a lot here.

One is that you need to be collecting data pretty completely If you have holes in your data, that's it.

Strictly a database estimation technique.

So if you have holes in your data, you won't do well.

For example, the reason we have to Mrs number seven and eight is because we don't collect data in the so the networks are sparks with Finn on the coast, and that means we didn't collect.

We don't we haven't collected enough data.

Allow us to properly look att those areas so that something has to be done if you can actually published.

This really isn't about the public.

You would include in an error estimation associated with what your network coverage looks like.

Second, I have learned that in some places like eastern Canada, which arm which arm or slow moving than California, which which tectonic rates are much slower, the plate motions must lower.

The stress rates are much lower that you have to you larger, larger been larger areas and larger times that to really see anything.

But I learned that the Taconic area makes a difference.

You have to tailor it to the particular physics, but other than that it hasn't know significantly changed much.

What about like Sandra strikes that time generation.

What about that?

Versus subduction?

Like, you know, I guess there's different nuances in terms of having smaller quakes.

So there again in place.

But I had a student who did a lot of work with Taiwanese catalog, for example.

And in Taiwan, you could do quite a good job of this.

Well, two problems with abduction, but I want you to quit a good job because the data is again pretty good.

You don't have to use magnitude threes.

They produce enough events that the smaller the magnitude you go, the more noise you get that will be that good at collecting small, small data, small magnitude it and get more noise dropped on smaller magnitude.

So in Taiwan, you're better off collecting up.

They have enough earthquakes of magnitude.

For that.

You're gonna do this with a magnitude.

For me is one of the things that the other problem productions noisy because a three dimensional, so so the conduction.

Don't look at it.

This is a two dimensional map, right?

And in some ways, like Taiwan, you have really have to do a three dimensional map.

Which student isn't against a job with it in someone like Western Canada.

Our problem is that we don't collect.

We don't have enough networks networks out over the ocean subduction plate.

And as a result, we don't collect data from deep enough yet to do.

There's a lot of errors, a lot of Miss data with death in Western Canada and that one of the problems with deductions on it you're gonna have to do is we're gonna do this on a regular basis of a place like that.

I have to have better that.

Any other questions?

Yeah, well, just maybe in terms of the GPS technology and there's even a cougar can Japanese at all these GPS readings of pressure build up in that seductions?

Your place?

Another staying.

The second biggest area buildup is off the northern Island.

Sapporo.

Is that normal?

Like, how about that technology in terms of forecasting, using GPS technology, actually using GPS technology at the end of the talking you want, uh, maybe a little.

There was a whole different topics.

This is a forecast.

That expression is to pull, I think, um, back in 4 2005 This is the forecast I did back then.

This is actually Excuse me.

You know, points on the, um top.

Matt is noisy again, using some of the techniques I learned I d noises for the bottom one is much less noisy.

And the math we've been using the technique and the massive been using based in Canada ever since.

You could see in general again the Blue circle show you locations of earthquakes that were forget that occurred after the forecast started.

The yellow and orange again are our forecast locations, locations where was forecast and again interestingly again, there's the location of the Ottoman earthquake in 2010 and we do a forecast that correctly as well have a more Children that anybody wants to see it.

But I left out of this talk, but we do forecast that auto quite well.

And we also again produced no false negative except the one and Husbands Day and that one in Lake Huron and Georgian men.

And that one occurred that scenario where again, we just don't collect anything we don't have.

Until very recent years, we haven't had networks up in that area, so we don't have a good record.

There is about it again.

Wouldn't missing about 12 Quick, which again is about the same rate we're getting for California.

2 40 Nothing.

12 again.

This is just a discussion of where we are today, with those mass produced 10 years ago.

And I like to tell people that 10 15 years ago, when when those mats were produced, we were told we went to publish it.

You don't want to do that.

Nobody forecasters done.

And now, today, 12 years later, there are actually agencies and organizations out there that not only take our forecast, but others based forecasts of other forecast techniques used the same idea.

Small earthquakes tell us something about fingers.

They made you the mask differently, but that's a possibility.

They used the same principle.

They're gonna use that seismicity data to make their own four cats and their Intel agencies like the Collaborative for the Study of Earthquake Predictability out in California, which actually evaluate those forecasts one against the other.

Now, enough of the mouth that you can't even put together a group of different forecasts and see which one how they perform relative to each other.

Some people get a contest.

I like to see that as an opportunity to need to figure out which ones are doing better.

And then why that better?

What?

What's different about them that makes them better Halfway improve all.

Not everybody needs that.

Is there a possibility?

Ensemble?

There are two possible ensemble modeling in this.

Wanted to do yet to stack them all up on average amount and tried to sometime like there's actually a model that was produced statistical models of larger over a couple 100 years.