BKZ seismic station polarity issue

[elidana]

A user of our seismic data notified us about the following issue, so opening this as it might be useful for others

copying here the user request, and will add some initial info as a first comment

Dear station operator， I am a seismology researcher in China. I recently processed LHZ data from the NZ network in New Zealand for the period 2008-2012 and found that the waveform records from several stations had polarity flips. The waveform data and instrument response files we used were downloaded from IRIS. In the measurement of the relative traveltime of the 10 mHz surface wave, it is found that all the measured values of the problem station have a half-period (50 s) deviation for several years. Since our measurement method is based on a cross-correlation technique, it is presumed that the cause is flipped polarity of the waveform recording. Afterwards, we used the P wave of the teleseismic (Wenchuan M7.9 earthquake in China on May 12, 2008, and the Tokyo M9.1 earthquake in Japan on March 11, 2011) to conduct a test. Compared with other nearby normal stations (IU SNZO, NZ RPZ), the seismic records of station NZ BKZ on the island of New Zealand have opposite polarities. For details, please refer to the attachment. I recently plan to publish an article about station clock error and polarity reversal, and I hope the station operator can tell me the reason for the polarity reversal of the waveform record of this station. I will thank you in the acknowledgments of the article.

Earthquakes and station locations picked for this example

The first arrival P waves recorded by the three stations, raw data is unprocessed except for filtering (0.01Hz-0.02Hz). The record in the middle shows a clear polarity reversal

[elidana]

Unfortunately the reverse polarity on some of the GeoNet broadband stations is a known issue (see https://github.com/GeoNet/help/issues/36 for another example).

When known, reversed polarities are tracked into our metadata repository delta (https://doi.org/10.21420/0VY2-C144).

The info can be found in the following table: https://github.com/GeoNet/delta/blob/main/install/streams.csv and when Reversed value is set to true, it means that we are aware on some issue at the sensor/digitizer level where we have been able to pinpoint the root cause.

This is not a comprehensive list, and we are aware that there are other station with polarity inversion issue that is not tracked there, and we have been working (though not actively right now) on how we can present and made this information available to end users, when known.

A good explanation of what (in same cases) can be the root cause can be found in this comment: https://github.com/GeoNet/help/issues/36#issuecomment-283538808

For the specific request about BKZ, we can confirm that there is a known polarity inversion issue for the broadband sensor installed in the period 2004-2013 (below an extract of the table available on delta).

Station	Location	Band	Source	Sampling Rate	Axial	Reversed	Triggered	Start Date	End Date
BKZ	10	V	H	0.1	false	true	false	2004-02-11T02:00:00Z	2013-10-09T00:00:00Z
BKZ	10	V	H	0.1	false	false	false	2013-10-09T00:00:01Z	2013-12-31T00:00:00Z
BKZ	10	L	H	1	false	true	false	2004-02-11T02:00:00Z	2013-10-09T00:00:00Z
BKZ	10	L	H	1	false	false	false	2013-10-09T00:00:01Z	9999-01-01T00:00:00Z
BKZ	10	H	H	100	false	true	false	2004-02-11T02:00:00Z	2013-10-09T00:00:00Z
BKZ	10	H	H	100	false	false	false	2013-10-09T00:00:01Z	9999-01-01T00:00:00Z

The issue was resolved in 2013 after a digitiser change, so I am assuming (but will need some better investigation, I just did a very quick search) that this might have been caused by a cabling misconfiguration, as the sensor remained the same across this time period?

I have also asked some help to our superstar seismologist in house, @jristau, and he's comment is reported below

BKZ is one of the stations that I flagged as having a polarity reversal from 2004 until mid-2013. In my notes I have 9 October 2013 as a date that the logger was changed at BKZ and the next time I was able to check the polarity in 2014 it was correct. So, when the logger was changed must be when the polarity issue was fixed. I should add that I was looking at our broadband sites (HHZ, HHN, HHE) while they have LH (long-period) in the document. I don’t know if that’s the same thing and I didn’t know we even had strictly long-period instruments operating anymore in 2008/2009.

hope this helps!

[shun-zz]

Hi, elidana. Thank you for your reply. I compared my test results with the list you provided. Most of the problem stations are consistent. But one station, QRZ, showed polarity reversal in my detection(My detection is only for the LHZ component). But in the list you gave it is normal. I used the teleseismic P waves of >M7 earthquakes for a test, and found that the polarity reversal did occur at this station.

In my detection results, the first period of polarity reversal at QRZ station: 232 days before 2008, and the second period: from the 290th day in 2008 to the 220th day in 2009. These timings coincided well with the timing of instrument changes at station.

I hope the station operator can verify this.

[elidana]

Dear @shun-zz ,

Thanks for providing these further evidences and confirming the already existing information! That's super helpful! :)

I have missed another important files that contain polarities information in delta, apologies for that! This is https://github.com/GeoNet/delta/blob/main/install/polarities.csv

That file is only listing the time periods for which a specific site and channel may have reversed polarity. QRZ is listed there as a station with reversed polarities for time periods that seems to match with your observations

Station	Location	Sublocation	Subsource	Primary	Reversed	Method	Citation	Start Date	End Date
QRZ	10			true	true			2003-08-06T00:00:00Z	2003-10-29T22:00:00Z
QRZ	10			true	true			2003-10-30T21:00:00Z	2007-08-24T09:50:01Z
QRZ	10			true	true			2007-08-24T10:00:01Z	2008-08-24T23:45:00Z
QRZ	10			true	true			2008-08-24T23:55:01Z	2009-08-12T23:00:00Z

We also have confirmations from @jristau analysis for QRZ (John's analysis is also looking at vertical components only), reported below

QRZ: Unknown polarity from 2003-09-25 to 2004-07-15, then reversed from 2004-07-15 to 2009-08-12. Likely reversed the entire time and there were other issues with the station early on which meant I couldn’t check the polarity.

Station	Start Date	End Date	Reversed(*)
QRZ	2003-09-25T19:50:26Z	2004-07-15T04:27:14Z	U
QRZ	2004-07-15T04:27:14Z	2009-08-12T23:00:00Z	R
QRZ	2009-08-12T23:01:01Z	9999-01-01T00:00:00Z	N

Reversed: U=Unknown, R=reversed, N=normal

By looking at the station metadata change, there have been a few dataloggers changes during this time period.

Details can be obtaining using our FDSN station metadata service