How to tell ghost quakes from real ones

Back in the days of the Canterbury earthquake sequence, Cantabrians had to wait up to half an hour to learn the magnitude of a quake. These days, GeoNet Rapid tells us within a couple of minutes, or at least gives us a good estimate. But this rapid report can occasionally produce ghost quakes, alerts for quakes that didn’t really happen. So in the immediate aftermath of a quake report, how do you tell a real quake from a ghost quake?


Geonet has a network of seismographs that keeps track of what’s going on seismically in our shaky little country. You can get a rough idea of where a quake is located by looking at the snapshot of the whole seismograph network. This the snapshot for the morning of 4 September 2010, 80 minutes after the quake that started it all for us here in Christchurch:

If you look closely, you can see that the Canterbury seismographs show the quake before the others throughout the country.

More recently (this morning), a 5.2 quake near Culverden resulted in a ghost quake measuring 6.3 near Opotiki:

Immediately after the Opotiki quake alert, the seismograph snapshot looked like this:

There’s just a single line for a single large quake sending seismic waves up and down the country. Again, the leading edge of the curve is centred on South Island sensors. That’s the first hint that the other quake reported might be a ghost quake.

For the second hint, we look at the individual seismographs. The seismograph nearest a quake picks it up before those far away, and the one that is near a quake shows it differently than a drum further away does. The trace on a seismograph for a nearby quake tends to be a bit cleaner than one from faraway. Nearby sensors may show part of the event as a red rectangle. This is where the sensor is indicating full scale deflection of the measurement at the displayed sensitivity range – the needle is literally off the chart. Although in this case there isn’t a literal needle, as these are electronic sensors.

This is the 5.2 quake that occurred near Culverden this morning (12th of February 2017), shown on the nearest drum:


Now here’s the trace from a drum a fair bit further away, and much closer to the supposed 6.3 Opotiki quake, the Urewera drum near the Bay of Plenty:

A bit messier, yeah? This is the Culverden quake we see on the Urewera trace shown above, even though the Urewera sensor is over 600 km away. There’s no big local quake showing up on the Urewera trace (although there might be a small 3.2 that occurred around the same time as the Culverden one that could be partly the reason the ghost quake got spawned; at time of writing, that’s not yet clear).

Drums on GeoNet are updated every five minutes, so looking at these drums is a good way of quickly determining whether a quake is a real one or a ghost one. Of course, once the duty officer has a look at the data, the information on the quake will be updated and its size and location will be revised. You can see this process by looking at the event summary for individual quakes. This is for a quake that was picked up by the automated system, analysed and revised by that system and then checked by the duty officer:


At 1 minute and 49 seconds, the GeoNet Rapid system published an initial magnitude and location when the data was received from the equipment closest to the location. Over the next ten minutes, the automated system continued to update magnitude and location as more information was received from more distant sensors. Then, at 12 minutes and 31 seconds after the quake, the duty officer posted their review of the information the system gathered.

This is what those statuses mean:

Equipment further from the location picks up the seismic waves as they travel away from the epicentre. The sensors are sharing their data and the system matches movements at each sensor to identify which are ‘the same’ quake, which will slow up slightly later and smaller on a more distant sensor when compared to one closer to the epicentre. This automatic matching is not perfect. If the system mismatches a quake with a smaller, later shake on another sensor, it might extrapolate that into a larger quake in the opposite direction, and a ghost quake is born. Once reviewed by the duty seismologist, these quakes are removed from the actual quakes list.

See this story of a ghost quake that showed up in Central Otago after a quake west of the Macquarie Islands.

See Dilemma of deep, distant earthquakes for more information about why ghost quakes occur.

See GeoNet Rapid is here! for a good overview of how GeoNet Rapid works.