Short Period Sensors Improve SeisComP Performance (LARGE N!!!!)

Yesterday was an epic day for automated event detection throughout Central America.

Over the years, Angel and I have installed SeisComP’s at nearly every geophysical institute from Southern Mexico to the Colombian border. We spend a lot of time with clients tuning their SeisComPs to ensure that in the moment of truth, performance is top notch. The moment of truth struck yesterday when a Mw 6.8 earthquake (later revised down to Mw 6.6) sent network saturating waves from its epicenter south of Panama City through Central and South America:

hub1 300x207 Short Period Sensors Improve SeisComP Performance (LARGE N!!!!)

Click to enlarge


We were excited to see that all SeisComP systems nailed the event location and magnitude, providing the region with solutions before the USGS and Geofon manually reviewed solutions were published/ emailed:

Institution | Magnitude (Mw) | First Location Time | Automatic or Manual
OSOP’s Principal Server | 6.7 | 1 minute 24 seconds!!!!!!!! | Automatic
University of Panama | 6.9 | 1 minute 51 seconds!!!!!!!! | Automatic
Panama Canal | 7.0 | 3 minutes 27 seconds | Automatic

Other Clients:
Client 1 | 6.6 | 3 minutes 38 seconds | Automatic
Client 2 | 6.8 | 3 minutes 27 seconds | Automatic
Client 3 | 6.7 | 3 minutes 34 seconds | Automatic
Client 4 | 6.6 | 3 minutes 45 seconds | Automatic

International Organizations:
GFZ | 6.6 | n/a | Manual
USGS | 6.6 | n/a | Manual
CSEM-EMSC | 6.7 | n/a | Manual

Notice the amazing first location timing at OSOP and UPA compared to the ACP and Clients 1 –> 4. Why the difference? Station density. Plain and simple. For the price of a single broad band installation the Panamanian government contracted OSOP to install a dense network of 75+ short period sensors to complement the broad band network. This has dramatically improved first location times and has reduced the magnitude of completion for the event catalog. The ACP’s system has full access to the Panama broad band network but has not yet integrated the new network of short period sensors.

OK, back to the electronic’s lab…hasta Friday icon wink Short Period Sensors Improve SeisComP Performance (LARGE N!!!!)

–Branden

4 comment(s) on “Short Period Sensors Improve SeisComP Performance (LARGE N!!!!)

  1. Hi, Branden,what the equipment do you use for shot period stations?Mechanical shot period sensor is comparable with broadband in price. Alex

  2. Hi Branden,Nice shot, for sure. It’s always cool to see how thing work in real-life.I have a couple of questions and a comment.I suppose you report the first location for OSOP and UNA, right ? Or is it the first reasonably good location ?For the other clients, I’m surprised that you don’t have any location before 3 minutes after the quakes. Do they use much farther networks ? It specially surprising for Panama Canal solution, which have logically stations as close to the epicenter as the OSOP and UNA system. I though SeisComP3 was able to output a first location with as low as 6 phases, and I though for this quake you would have 6 arrivals on broadbands in Panama much sooner than 3 minutes, no ? And for the comment, I’m pretty sure that the three agencies listed have had in internal some automatic location much earlier, but they weren’t made public. Regarding CSEM, they don’t do location themselves, they collect metadata (arrivals, location, magnitudes) from member agencies and then eventually relocate. So the first location they have is tied to the the first location they collect from a member agencies (mostly USGS, GFZ or a couple of other, if you look at their data contribution page for this quake). Finally, I’m agree with you that for automatic systems, you need just to point arrival times, thus not necessarily from broadband. But a question : do you use the short-periods for magnitude calculations ?

  3. Jean-Marie:

    Yeah, the real-life part is why we love doing this. In small Latin American countries like Panama, when big earthquakes strike the President commonly talks directly to the heads of the seismic networks like Dr. Eduardo Camacho at the UPA. This can make for quite a bit of excitement (and stress).

    We have a software that reports the first location, regardless of whether it comes from SeisComP, Earthworm or Earlybird. As you know, for big events, SeisComP can have dozens of re-locations as the P-wave spreads out and more and more stations trigger. So receiving a new update for every re-location is undesirable. But in order to remain informed about important changes in the event location, depth and magnitude, we allow end-users to define conditions such as: if the location changes by a user-defined X kilometers from initial location, or by Y kilometers in depth or by Z units of magnitude, then the system should send out a new automated notice to recipients. This can be done via email, sms, Twitter, Facebook, fax, phone, …

    It is not that they have much farther networks, just sparser networks. And this event was some 200+ kms outside the confines of the local networks. Surprisingly though, even the initial locations were very good (this is not always true for weird geometries).

    ACP does have stations as close the to epicenter, but limited to the broad bands of which there are only 9 in Panama. They have not yet integrated the short-period sensors into their system. You definitely had 6 arrivals on the stations in Panama and the region in under 3 minutes, I guess the systems took their time locating :)

    Thanks for the point on the CSEM. I never knew they were aggregators of member agency solutions. They usually agree very well with GFZ.

    And yes, from the automated system’s perspective the P’s are what matter most, regardless of whether they come from broad bands or short periods. Deploying short periods are, essentially, an easy, inexpensive way to feed more P’s into an automated detection system, improving density and distribution. Later on the academic-grade science can be done with the broad bands.

    Yup, we use short periods for magnitude calculations. Just like the broad bands, our short-period sensors have 144 dB of dynamic range, send real-time continuous data, good timing (some GPS, others NTP), and they all have response files that we roll into dataless seed volumes. We let SeisComP decide which stations to use for which magnitude calculations whether it be MLv (which saturates between 6 and 7), Mwp, Mw(Mwp), mB, Mw(mB), mb, … Surprisingly, despite the size of this event, the short periods reported very reasonable magnitudes (6.6–>6.8).

    Saludos!

    Branden

Leave a Reply

Your email address will not be published. Required fields are marked *

*

You may use these HTML tags and attributes: <a href="" title=""> <abbr title=""> <acronym title=""> <b> <blockquote cite=""> <cite> <code> <del datetime=""> <em> <i> <q cite=""> <strike> <strong>

Current day month ye@r *