MQZ issue with broadband vs. accelerometer comparison

[elidana]

Jesse Hutchinson (University of Canterbury) has noted an issue with MQZ seismic data when comparing data from the broadband and strong motion sensors installed at the station.

His method calculates the difference of the natural log values of PGAStrong from PGABroadband instruments, for all M>5 events recorded by the station.

A plot he provided is showing a consistent discrepancy for all events recorded between ~2014 and ~2020.

Further investigation to understand the cause are needed

[salichon]

the graph is an indication though it does not help to troubleshoot really.

I encourage to have a look at the raw data (say using daily streams) as it should be done prior to any processing of some sort :) - especially when it is made with big batch (though tedious i know)

Would be helpful to have from Jesse:

1. A detailed step by step procedure of the processing of the data in order to frame the discrepancies and the actual source(s) -Or at least the earliest steps from :querying to deconvolving and using the data.
2. A list of the M5 earthqaukes would help to identify the most relevant periods

[salichon]

Summary of the PSA process above from what i understood

Input:

• M5 event waveforms (time window length?)
• Waveforms from high quality station (national sites)
• Use in priority strong motion HN (200HZ) (then strong BN, and then weak HH 100Hz)

Process:

• Remove instruments response
• Compute Pseudo Spectral Accelerations spectra (PSA),
  From:
  1 - Strong motion
  2 - Weak motion
  --> PSA methods for each data type to be described briefly

Tools and Sources:

• Used Python packages to perform this computation
• Query mechanisms ?

[salichon]

MQZ strong motion periods (delta/fdsnwebservice: https://service.geonet.org.nz/fdsnws/station/1/query?network=NZ&station=MQZ&level=channel&format=text&channel=?NZ)

• NZ|MQZ|20| HNZ/BNZ |Accelerometer|427336.117780|1.000000|m/s**2| 2009-08-06T00:00:01|2016-02-22T20:00:00
• NZ|MQZ|20|HNZ/BNZ|Accelerometer|427336.117780|1.000000|m/s**2| 2016-02-24T01:02:00|2019-06-17T05:25:00
• NZ|MQZ|20|HNZ/BNZ|Accelerometer|427818.982500|1.000000|m/s**2| 2019-06-17T23:20:00| NOW

Periods of identified Strong Motion default in quality:
2016.054 till 2017.218 ( !!2017.219 is --> when Pb is solved on the ticket (to be detailed) )

Today:
Raw archived data --> MQZ strong has "Stepped traces" ("long period Steps every couple of hours") as below :

Does this affect Computation on the event traces ?
Are the data actually valid (over one single event for instance?)
Any other issues identified with this installation over that time ?
Can we explain the HHvsStrong observations with this ?

---

With regards to the comparison with the broadband HH It looks like the mount of the HH is old stable and performing since 2009:

• MQZ|10|HHZ| Broadband Seismometer|838860800.000000|1.000000|m/s|100.000000|2009-08-06T04:00:00|

[SquirrelKnight]

The procedure is as follows:
All stations for which an event arrival was detected and within a certain distance defined by a magnitude-distance scale are used to download waveforms. The time window extends from 10 seconds prior to the P arrival to the S arrival + the ds595 value * 1.2. The ds595 value has been modified from the original equations by Afshari and Stewart (2016) to include an additional term to capture more of the waveform coda for larger magnitude events.

After the raw mseed files are downloaded, they are then evaluated with a ground-motion classifier to determine whether the event meets certain quality criteria. High quality events are then pre-processed for intensity measures. In pre-processing, I first demean the data and pad the data for five seconds on either end. I then remove instrument sensitivity and differentiate any velocity data into acceleration. The data are then rotated to ZNE components and trimmed back to their original window length. Finally, the data are low-pass filtered at a corner frequency of 0.05 Hz.

The pre-processed data is used to compute intensity measures (PGA, PGV, pSAs, etc.) for horizontal, vertical, rotd100, and rotd50 components.

The list of M5 events used for MQZ are as follows (note that this analysis has only been performed between HN and HH channels):
'2016p858069', '2016p858147', '2016p858156', '2016p858160','2016p858279', '2016p858848', '2016p858895', '2016p858951','2016p859051', '2016p859105', '2016p859270', '2016p859378','2016p859625', '2016p859628', '2016p859929', '2016p860053','2016p860215', '2016p860224', '2016p860287', '2016p860567','2016p860729', '2016p861632', '2019p809993', '2019p927023','2020p201743', '2020p475488', '2020p606562'

[salichon]

thanks for this description:

• from the 27 events ID perspectives and the Period of time identified Strong Motion default in quality 2016.054 till 2017.218.
  It looks like 22 events may fall into the questionable period of the SM instrument issue.

• the 5 others '2019p809993', '2019p927023','2020p201743', '2020p475488', '2020p606562' should have better data set with the strong motion

From the Broadband side of the process: MQZ broadband had a fairly stable sensor and datalogger since 2009 to now:
Guralp Systems Ltd.,CMG-3ESPC,T35861,MQZ,10 ,2009-08-06T04:00:00Z,9999-01-01T00:00:00Z
Kinemetrics,Q330/6,3386,Glencoe Station,,2009-08-06T00:00:01Z,9999-01-01T00:00:00Z

so let s focus on the most recent and 5 events first.

[salichon]

MQZ is located in Christchurch

• https://www.geonet.org.nz/earthquake/technical/2019p809993 M5
• https://www.geonet.org.nz/earthquake/technical/2019p927023 M5.4
  are located to distance ~700/800km upper north island
  Strong motion tool does not provide SM data for both events.
  This is legitimate due to signal to noise ratio too low for relevant result afaik
  https://strongmotion.geonet.org.nz/#/event/2019p927023
  https://strongmotion.geonet.org.nz/#/event/2019p809993

---

• https://www.geonet.org.nz/earthquake/technical/2020p475488 M5.9 is located in Fjordland ~ ~400 KM from MQZ
  Strong motion tool does provide SM data https://strongmotion.geonet.org.nz/#/event/2020p475488

• https://www.geonet.org.nz/earthquake/technical/2020p606562 M5.4 is located in Fjordland ~ ~400 KM from MQZ
  Strong motion tool does provide SM data https://strongmotion.geonet.org.nz/#/event/2020p606562

• https://www.geonet.org.nz/earthquake/technical/2020p201743 M5.1 is located in Canterbury about 100km
  Strong motion tool does provide SM data https://strongmotion.geonet.org.nz/#/event/2020p201743

Out of 5, the 3 latest and closest have some reasonable data to compare with for a start

[salichon]

@SquirrelKnight In order to focus on the most relevant data sets:
What are the certain distance criteria for using the data :
"All stations for which an event arrival was detected and within a certain distance defined by a magnitude-distance scale are used to download waveforms." ?
I find that the "certain distance" can be pretty loose for some distant events and a resulting acceleration can be pretty low for M5 > 500km distance
thanks :)

I guess also it would be interesting to get the raw PSA data derived from HH and HN ( as a text file) of the method to be able to compare the values for a least the events: 2020p475488, 2020p606562 , 2020p201743
cheers!

[SquirrelKnight]

Here is an image of the magnitude-distance relationship. As you can see, stations selected for lower magnitude data are selected from considerably smaller distances.

I can certainly provide the raw PSA data. Where would be a good place to upload it?

[salichon]

thanks a lot @SquirrelKnight
"I can certainly provide the raw PSA data. Where would be a good place to upload it?" here on git (as *.text) so that we some psa to compare :)

okay thanks for the distance relationship !
still for having MQZ used in far north 2019p809993 M5 and 2019p927023 M5.4
I reckon that the selection should be tighter the epicenter distance does not seem right.

cheers
j