Wild Land

GeoNet is trialling an automated earthquake locating application that will provide faster reports of earthquakes in New Zealand territory.

The present GeoNet website provides information on many of the earthquakes that occur within New Zealand territory. The process is partially automated, gathering information provided by seismic instruments and calculating earthquake strength and location. However, to ensure that known and unexpected errors do not corrupt the data, the final calculation of an earthquake’s parameters is made by a seismologist or a trained earthquake analyst.

In the case of an earthquake that is felt and therefore of immediate interest to the public, the process can take twenty minutes or more. If the event occurs after hours, a duty seismologist will be paged and alerted to the event. The seismologist then has to log on to GeoNet’s network to examine the data and determine what will be published on GeoNet’s website.

In the case of smaller events, these are left for processing during normal working hours. There is no clear policy as to what will be published as soon as possible. Significant events are swiftly dealt with, but smaller earthquakes which may or may not be felt are often left for processing during “normal hours” and most are never published on the website.

With the on-going earthquake sequence in Canterbury, this has sometimes led to frustration over the length of time that people have to wait to get details of a notable earthquake. Some events outside Canterbury have been left until the next business day causing residents outside Canterbury to wonder why shallow magnitude 3 events in Canterbury are published throughout the weekend, while a shallow magnitude 4 in or near the North Island isn’t.

The answer is that GeoNet isn’t staffed 24 hours a day and events in Canterbury have a higher profile than slightly larger events elsewhere in the country at the moment. If an arbitrary rule such as publishing all events over magnitude 3.5 at a depth of 40 km or less were to be made, the on-call seismologists would be run ragged at times. For example, GeoNet calculated the locations of more than nineteen thousand earthquakes last year ”“ about one every 27 minutes on average. Nevertheless, the frustration remains.

To address this, GeoNet has introduced an automated system that will automatically calculate earthquake parameters and publish them on a website. The system will begin calculating quake data as soon as it arrives from a seismometer, refining its calculations as information is sent in from more instruments. The SeisComP3 software should be able to calculate its first location within a minute and information about magnitude, depth and location will be published on the website, with the system updating the report as more information comes to hand. The “quality” flag for this initial calculation will be set to “caution.” As more data arrives, the software recalculates the earthquake solution until it is able to set the quality flag to “good.” A manual review by a seismologist will follow at a later stage.

In the case of a smaller earthquake, a final solution should be quickly established from data supplied by as many as half a dozen instruments. Larger events will be recorded by many more instruments and the final calculation may take longer. Even so, an approximate location and magnitude should be available to the public within a few minutes.

Completely automating the publication of earthquake information has its risks. There will be conditions that haven’t been allowed for in the software, and errors in calculation are possible. Situations such as multiple earthquakes in the same location over a short period of time, or twin events where two earthquakes occur almost simultaneously but at some distance from each other have given seismologists headaches in the past, and may produce some interesting results in an automated system. The 7^th magnitude Darfield earthquake of September 2010 was later discovered to be a multiple event, and one of the Dusky Sound aftershocks in July 2009 struck within a few minutes of a magnitude 5 quake in the South Taranaki Bight. The complex activity involved in both events was not immediately apparent, and it will be interesting to see how SeisComP3 analyses this sort of information.

This is a significant change for GeoNet, which will run its existing system in parallel for some months, while it fine-tunes GeoNet Rapid. Initial testing was favourable, and the trial system is now available to the public.

But note that GeoNet is still the authoritative website for accurate earthquake information in New Zealand. The trial site will be enhanced in the coming months as the performance of the two systems is compared, bugs are ironed out, and the software tuned.

The trial site for GeoNet Rapid, which can be found by clicking on this link, is running in what is known as “beta mode.” In other words, this is a trial site available for use by the public with a view to enhancing the product from observations and feedback. It is running on infrastructure that is light (only one web server) when compared with GeoNet’s redundant server architecture and, if subject to sudden load, there may be crashes or failures. Even so, it will be interesting to see how accurate it is for earthquakes recorded in the near-future.

There will be discrepancies between earthquakes located by GeoNet and those located by GeoNet Rapid even after GeoNet Rapid has refined its calculation. The present system uses a one-dimensional velocity model whereas Rapid uses a three-dimensional model. The new system uses vastly more information and should give an improved result.

There are some known issues with SeisComP3, the software being used for GeoNet Rapid. The calculated location of earthquakes with a reported depth of 0 km should be treated with caution.

A new version of SeisComP3 has been released and will be uploaded to the trial site in the near future.

In the meantime, should a large event occur, the duty seismologist’s priority will be to update the GeoNet website, as this is still the official source. In such an event, the trial of GeoNet Rapid will take a lower priority.

[Compiled from a GeoNet news release]

This entry was posted on Tuesday, March 6th, 2012 at 8:13 pm and is filed under Earthquakes - Observations. You can follow any responses to this entry through the RSS 2.0 feed. You can skip to the end and leave a response. Pinging is currently not allowed.