Locating earthquakes with Swarm

Hi Shakers,

I made a video on how to locate earthquakes using Swarm. In many areas of the world now there are enough RShakes to be able to locate events. The density you need is 6 to 10 stations in a circle of 50 to 100 kilometers in diameter. I am planning to make a follow-up video to help dial in the magnitude and how to see how good your location is. Here is the link:

Good hunting for little earthquakes.

Saludos, Angel

I’ve tried the method Angel describes so well and it works!!

Here is the Swarm clipboard from a nearby EQ with the annotations as Angel shows in the video:

The “official” GeoNet results for this EQ using 20 stations (https://www.geonet.org.nz/earthquake/technical/2019p901682) were:
-43.61 N 171.55 E 7 km depth M 4.4
The Hypo71 results (using the 4 RSs shown above) were:
-43.64 N 171.55 E 7.18 km depth M 2.4

These results are remarkably close, except for the magnitude, which Angel is working on…

The first time you do this, it’s a bit laborious, but with practice it gets easier.

Hi Tideman,

Thanks watching the video and doing the location. The picture you sent helps me a bit. In the two middle traces the coda is of the to the right and should not have been picked. On the top and bottom trace I would have picked the coda a little to the right of 14.57.50.

That would move the coda magnitude up but I don’t think that you would have reached the GeoNet 4.4. Now that I can see what stations you used I will see if I can guesstimate a correction factor the you can use in the hypo71.config file.

Angel

Angel

You’re right, the position of the coda makes a big difference to the magnitude. I guess one needs a bit of experience to get that right.
For me, the choice of the time for C1 is much more difficult to detect than either iP or iS.
The human eye is a very precise instrument in detecting changes such as the arrival of the P waves, but not so good at figuring out when the pattern returns to ambient.

Taking the coda to extremes, the magnitude increases to 3.04 (note there are 5 stations in this one):

This morning we had a decent EQ (mag 4.0) at Oxford, about 50 km away.
We didn’t feel it in town, but the local RSs did as shown below:

The SWARM algorithm for detecting the epicentre got it pretty right compared to the “official” location. The latitudes were the same, but the SWARM longitude was 5 km west.

The magnitude from the SWARM algorithm was 2.11 compared with the official value of 4.

2 Likes

Lovely result TideMan. My guess as to the magnitude difference is that the coda 2 picks are too soon, i.e. they should be closer to when the wavetrain becomes indistinguishable from background (may require you to scroll right on your timelines a bit).

We had a 4.3 nearby this morning. I was in the bathroom at the time and it quite unsettled me, though it was short and sharp. Took me back to the days of 2010 and 2011 when we were getting mag 5’s every week or so…
Here are the Swarm plots from nearby stations (my home machine is the top one):

And here is the result of the Swarm calculation for the epicentre compared with the “official” GeoNet result (based on 53 stations). They are only 4.5 km apart.

However, the magnitude calculated by the Swarm algorithm is 2.71 km compared with the official 4.3, even though I’ve moved the C2 limits way to the right (as suggested by Ian for an earlier event).
The GeoNet depth was 5 km compared with Swarm of 7.1 km.

2 Likes

My suggestion would be to find where the signal disappears into cultural noise (probably using the spectrogram) which is most likely even further to the right of where you’ve picked the stop coda. I’m not sure how well it’ll work given that you were so close to the epicenter, and the duration magnitude is an imperfect method, but it might be worth a try.