The catalogue

As we’re sure you are all aware, GeoNet has a seismograph network that allows us to locate earthquakes in Aotearoa-New Zealand; more than 22,000 in 2021 Those locations, close to 700,000 in total, are stored in a catalogue that is available for everyone to access. A brief history of earthquake location and the catalogue is available in a FAQ document.

We are going to work with a small subset of the catalogue, earthquakes beneath Lake Taupō since 1980, provided they were no more than 30 km deep. Any area will do, but we’ve chosen this one because it’s a beautiful spot, and also to help illustrate the value of understanding catalogue meta-data. More on that later.

We used the Quake Search application on our website to get the list of earthquakes, just over 18,000 of them. If you’ve never given Quake Search a go, try it out, it’s very easy to use.

The key information about each earthquake in the catalogue is its location, depth, magnitude, and when it occurred. Here’s a map to give you an idea of where our 18,000 earthquakes occurred. Each symbol is one earthquake.

The catalogue also contains a lot of information about how each earthquake was located. This is what we call the meta-data for each earthquake. We are going to examine a couple items of meta-data that highlight how the earthquakes were located, the different location programs and location methods used, together with the model of the speed of earthquake waves through the Earth (velocity model).

Changes in location program and method

Since 1980, three computer programs have been used to locate earthquakes in Aotearoa-New Zealand. First, a program called LOCAL created by DSIR Geophysics Division in New Zealand, then one called CUSP which was adapted from a program used in California, and lastly in 2012 SeisComP3 (SC3) which we continue to use today. The LOCAL program used a location method also called LOCAL; the CUSP program used a location method called GROPE (a bit of an odd name, but that’s what it was called!); and SC3 has two different location methods, LOCSAT and NonLinLoc. Each of these programs achieved the same things; they located earthquakes and determined their magnitude, but they did it using different methods and earth models.

If you think this is starting to sound quite complex, and maybe a little confusing, then you aren’t wrong! Hopefully a graphical view will help. Below we show the date of each of our 18,000 earthquakes on a background that shows the location program used, and on the left we label the location method and on the right side the velocity model, which we’ll cover soon. There are a couple of important events that we add, the start of GeoNet in 2001, and when our National Geohazards Monitoring Centre (NGMC) went live, in December 2018.

So, what can we learn from this?

First, and perhaps the most surprising, is that until the early-1980s, a few earthquakes in the Lake Taupō area were still located by “felt reports”. In 1983, there were two swarms of felt earthquakes in the northern Lake Taupō area. The seismograph network in the central North Island at the time wasn’t good enough to record and locate every earthquake that was felt by local people, so the felt reports themselves were used to estimate, very roughly, where an earthquake originated. This was the method commonly used in the “old days” before instrument recording of earthquakes became common.

The next thing to note is that when the CUSP location program was introduced in 1987, there looks like an increase in the number of earthquakes located in the Lake Taupō area. As the number of local seismographs didn’t change that suddenly, this increase suggests there must have been some other changes. What we think is the reason is that digital recording of earthquakes began in Aotearoa-New Zealand. This made a big difference as it’s quicker and easier to locate earthquakes with digital data, so more time was available to locate the numerous smaller earthquakes that couldn’t be located when seismologists had to collect all the necessary information from “squiggles” written on paper charts.

In 2001, when GeoNet started, nothing initially changed in how earthquakes were located, because the GeoNet team was heavily focused on building up the seismograph network to provide the coverage and quality we now have.

The big change finally came in 2012 when started to use “GeoNet Rapid”, which is our implementation of SeisComP3 (SC3), which gives us automatic, near-real earthquake detection and location capability. Due to the complex geology in Aotearoa-New Zealand, we configured SC3 to have two different location methods, NonLinLoc and LOCSAT. These operated in quite a different way and typically didn’t give exactly the same location; usually the depth differed.

At the time there was no clear procedure stating when each location method should be used and where, and NonLinLoc was used a lot more than LOCSAT.

When our NGMC went live, in December 2018, LOCSAT became the location method of choice, and since then virtually no earthquakes have used the NonLinLoc method. LOCSAT uses a very simple velocity model which provides earthquake locations quickly and efficiently.

And velocity model too

Locating an earthquake with a computer program relies on a velocity model that describes how the speed of earthquake waves varies with location and depth below the surface of the Earth. The velocity model used to locate an earthquake can affect the estimated location, and particularly the depth. Each of the location methods have used a different velocity model.

The LOCAL program used a model called “nz1d”. This assumed that seismic velocity only varied with depth below the surface and was the same everywhere in Aotearoa-New Zealand. But in some places, including the Lake Taupō area, the velocity was quite different from the model! There, the velocity of rocks in the top 10-15 km of the Earth are quite a bit less than many other parts of the country.

With the CUSP program, the old “nz1d” model was updated to “nz1dr”. The “r” stands for regional, as the model included several sub-models for specific regions of the country so that the velocity better matched reality, good for the Lake Taupō area!

The velocity models available with SC3, “iasp91” and “nz3drx” are very different. “iasp91” is pretty similar to the old “nz1d” and is used by the LOCSAT method. While “nz3drx” is what we call a three-dimensional model, where velocity varies with location as well as with depth, and is used by the NonLinLoc method.

For the first six years we used SC3, the use of the NonLinLoc method meant that the “nz3drx” model was most commonly used to locate our earthquakes, but since NGMC came online, LOCSAT has become the dominant location method, meaning that the “iasp91” model the most frequently used.

So what?

While this is all really interesting, or maybe just to some of us, why does it matter?

If you are just interested in a pretty good estimate of where the earthquake you just felt originated, then none of this matters! Whether GeoNet used the SC3 program with the LOCSAT method and “iasp91” velocity model to locate the earthquake or some other location program, method, and velocity model, it won’t make that much difference. But if you need to analyse changes in earthquake locations or rates of activity in a particular area over a period of years, but especially decades, then it’s essential that you think about this stuff and consider what impact it might have on your work.

The Lake Taupo data set we’ve been using is a great practical example of why you might need to understand this stuff. In the last couple of months, we’ve seen an upsurge in the number of earthquakes GeoNet has located in the area. GeoNet’s Volcano Monitoring Group (VMG) considers earthquake activity as part of their assessment of any unrest at our volcanoes. A question they will surely have been asking recently is “how does the recent activity compare to past activity?”. To answer that question, you really need to understand the catalogue meta-data and consider how it might affect how many earthquakes are in the catalogue, where they are located and their magnitudes.

The last notable swarm of earthquakes under the lake was in 2019, and the earthquake location procedures were the same as now. But what about comparing the current activity with previous swarms under the same part of the lake? If those swarms were before the NGMC started in December 2018, then the VMG should at least take a look at the meta-data. If nothing else, the earthquakes located using felt report information shouldn’t be used for a modern comparison as we know those locations won’t be very reliable.

Most users of earthquake location data know that changes in the number and location of stations in the seismograph network can influence how many earthquakes are located and where they are located. We hope that you can now see that any changes in the technical aspects of how the earthquakes were located – location program, velocity model, etc. can be important too. This is why we make sure we include meta-data with the earthquake location and magnitude information.

Meta-data are really important to all of GeoNet’s data sets, and we are planning to talk about it some more in future blogs. Bye for now!

If you missed our first blog, on the Volcanic Alert Level data set, you can find it here. If you are a returning reader, thank you. We must be doing something right!

We welcome your feedback, and if there are any GeoNet data topics you’d really like us to talk about, please let us know.

Contact: info@geonet.org.nz