Good evening, all.
Bottom-line short response: Thank you, I have managed to use "dataselect" to
create a BUD archive that, it appears, SEISAN can reliably use. So far everything is
working with the BUD archive.
And I cannot thank you enough for your incredibly quick & responsive support. I can
only say: "Wow."
Spent all weekend servicing stations; now migrating and backing up the data.
Hope to get back to processing very soon, then automating.
Longer response:
Going in, it appeared that trying to create a BUD or SEISCOMP archive
had more opportunities for making mistakes than did using SEISAN or
miniseed continuous databases. Looking back, it is easier to create a
BUD archive than it is to make a SEISAN or other continuous database
work. From my limited experience, an archive is definitely the way to
go.
I am on a metered connection, so I have not tried downloading the most
current version of SEISAN programs and documentation, but the version I
downloaded a couple of months ago includes directions to use "msrouter"
to create a BUD or SEISCOMP database. As far as I can tell, "msrouter"
is no longer available: use "dataselect" instead.
The success that I am having with the BUD archive--in contrast with
the many difficulties I had trying to use continuous miniseed or SEISAN
waveform files-- suggest that maybe support for the non-archive waveform
files should be deprecated.
All the best,
Tim
Tim Sickbert
Devon Labs Coordinator
Boone Pickens School of Geology
105 Noble Research Center
Oklahoma State University
Stillwater OK 74078
(405) 744-6358
[email protected]
-----Original Message-----
From: Jens Havskov [mailto:[email protected]] On Behalf Of Jens Havskov
Sent: Friday, October 09, 2015 13:31
To: Sickbert, Tim <[email protected]>
Cc: [email protected]
Subject: Re: [seisan] Problems with continuous databases
tim
On Fri, 9 Oct 2015, Sickbert, Tim wrote:
Good morning.
OK. I am having many, many issues working with SEISAN. In spite of all
the issues, I am very pleased with what SEISAN enables me to do when
it behaves well, so I don't want to give up on it. I would very much
like to figure out how to make it work well for me.
Background:
This is not my day job. I have financial support for equipment and
operational expenses, plus occasional advice, but otherwise I am
working solo.
I started deploying ten Guralp CMG-6TDs just over a year ago. No
telemetry, I have been periodically downloading data from the onboard
digitizer/logger (essentially a Guralp CD24) as 24-hour both GCF and
miniseed files. I am also using data available from IRIS DMC for
nearby stations on networks OK, GS, and ZD. I was initially processing
with Seismic Handler (SHM), but have been unable to reliably extract
amplitudes or magnitudes, so I am transitioning to SEISAN.
Having 24-hour continuous digital waveform files in miniseed format,
I spent a few weeks trying to modify and recompile SEISAN to be able
to handle 24-hour files. I was unable to make that work.
if enough memory it should be possibel, but not very usefull since too much
data would be displayed.
Then I found "mscut" that allowed me to split the 24-hour miniseed
files into 1-hour files. Good. I have been able to "dirf" the 1-hour
miniseed files into "filenr.lis" and "autoreg" the list of miniseed
files into
<seismo>/REA/<database>/<year>/<month>/<dd>-<hhmm>-<ss>L.S<yyyymm>
parameter files and, running "mulplt" in the local directory
containing the continuous waveform files, process them with some limited
success. More on the limits below.
then you are only looking at one station or one channel at a time, not so
useful. if you do for all channels, you can use a seisan continous data base,
see training document
>
Having hit significant limits processing miniseed files, I searched
further and found "seisei" that merged the miniseed files into SEISAN
(<yyyy>-<mm>-<dd>-<hhmm>-<ss>S.<database>_###) files, the structure of
which I do not yet understand. Let me know if it matters very much.
it does.
if you want to use a seisan continous data base, there will be s-files for
each waveform files.
So I would run
"seisei", then "dirf <yyyy>-mm>-<dd>" to create "filenr.lis", and "autoreg"
the files into the REA database, and then "mulplt" to pick phase
onsets and amplitudes. It sometimes works, but with limitations
similar to but not exactly the same as running "mulplt" against a database of
miniseed files.
In the following, I THINK that what I write is generally but not
specifically true for processing both a database of miniseed files and
a database of SEISAN files, unless otherwise noted. "Generally" in
that when processing either database, "mulplt" will plot more than the
requested length of time (55 or 60 minutes); "not specifically" in
that with a database of miniseed files "mulplt" may display three
hours of data, while with a database of SEISAN files "mulplt" may display five
hours of data.
not sure i follow
It may be worthwhile to note here that the continuous waveform files
downloaded from the Guralp instruments are not always continuous over
a period of an hour. Typically, one or two of the miniseed files has a
discontinuity. Even so, the gap is short and the data is nearly complete.
With this text I am also sending screen captures from PQL II that
illustrate the breaks in the data.
Observations:
- I run "mulplt", "cont", <yyyymmddhh>, "55" [or "60"], "1" "o",
<enter>, "o", "t" {where values in quotes are literal entries; values
in angle-braces are variables, e.g., I might enter "20150106" for
<yyyymmdd>}, but meta-keys are also in angle-braces, i.e., <enter> or <return>
indicate pressing a key.
tim@xenoK:/media/tim/OS/scream/xfer_20150117_msd/20150106$ mulplt
Filename, number, filenr.lis (all)
Continuous SEISAN data base: cont
Large SEED volume: conts
Archive: arc
Make a
choice
cont
Give start time, yyyymmddhhmmss
2015010600
Interval in min
60
2015 1 Reading events from base OKSTU 1407
here you should read form many different bases, at least as many as the number
of stations
Plot options: Interactive picking Return
Multi trace plot on screen, def (0)
Multi trace plot on screen (1)
Multi trace plot on screen+laser(2)
Multi trace plot on laser (3)
Continuoues on screen (4)
Continuoues on screen + laser (5)
Continuoues on laser (6)
Stop (q)
The resulting "mulplt" screen is captured in
"mulplt_2015010600_60min_but_3hr.png". I tried a 55-minute interval
and "mulplt" displayed four hours of data for the HHZ component of
station QUOK, i.e., similar results. I tried a 15-minute interval and
mulplt displayed 15 minutes of data for all the A* stations, CROK, and
OK031; 1 hour for KNG1 and two components of QUOK; and 2 hours for
QUOK HHZ. In the attached screen capture, also note the data dropout
on KNG1 and QUOK at the beginning of hour 2: this is a ubiquitous
problem. The attached screen capture "mulplt_2015010600_60min_zoom_missing_first.png"
illustrates that "mulplt"
has some kind of issue with displaying the beginning of the data. In
contrast, the screen capture "PQL_II_2015010600_60min.png"
demonstrates that all the data is available in the *_HH? miniseed continuous
waveform files.
I approximately located the largest event in the period:
date hrmn sec lat long depth no m rms damp
erln erlt erdp
15 1 6 021 51.96 3624.72N 97 19.5W 6.5 17 3 0.13 0.000
1.00 2
2015 1 6 0021 52.0 L 36.412 -97.325 6.5 OSU 11
0.1
OLD: 1 6 0 0 0.0
L
Return to continue
The Oklahoma Geological Survey catalog location is:
15805 2015/01/06 00:21:52.06 36.41689 -97.32362 4.923
1 0.6 2.2 0.51 NOBLE OGS 2.6 ML OGS
The OGS catalog and approximate "mulplt" event times and locations are
sufficiently close to each other to demonstrate that "mulplt" in this
case--where "mulplt" does not properly display the data at the
beginning of the period--the data is still properly placed in time.
Because I had trouble with miniseed continuous databases, I tried
migrating the data to a SEISAN continuous database using "seisei" to
merge the files into <yyyy>-<mm>-<dd>-<ss>S.<database>_### files,
"dirf" the files into "filenr.lis" and "autoreg" the files into REA,
then "mulplt" "cont" for analysis. The results are similar, as
illustrated by, for example,
"mulplt_2015010602_55min_got_7hr_SEISEI_database.png".
it is not clear if you have made a proper seisan continous data base.
please see the example in the test data. if not a proper data base, anything
can happen.
> Further issues:
- Even when "mulplt" properly shows the period I selected from
the continuous database, when I register an event and then
subsequently process it with "eev" "po" (to get back to "mulplt"), I
sometimes get the appropriate period from the original "mulplt" zoomed
window; other times I get the entire period from the original
"mulplt". That is, registering an event in "mulplt" ("p") sometimes
saves the ~3-minute waveform files that I expect, and other times saves the
entire 1-hour file.
- I ran "mulplt":
tim@xenoK:/media/tim/OS/scream/xfer_20150117_msd/20150106$ mulplt
Filename, number, filenr.lis (all)
Continuous SEISAN data base: cont
Large SEED volume: conts
Archive: arc
Make a
choice
cont
Give start time, yyyymmddhhmmss
2015010605
Interval in min
55
2015 1 Reading events from base OKSTU 1424
seems like you have put all channels form all station in one file. it would not
work.
Plot options: Interactive picking Return
Multi trace plot on screen, def (0)
Multi trace plot on screen (1)
Multi trace plot on screen+laser(2)
Multi trace plot on laser (3)
Continuoues on screen (4)
Continuoues on screen + laser (5)
Continuoues on laser (6)
Stop (q)
and ended up with the display shown by
"mulplt_2015010605_misalignment" and
"mulplt_2015010605_misalignment_zoomed". Ignore all the missing data
for the
moment: the data are displaced in time. Note especially "A001 HHE"
where the onset of the signal at about "5h06" preceeds the onset of
the signal on
"A001 HHZ". I encounter this sort of misalignment rather regularly.
Occasionally, I can get around it by selecting a different start time
and interval for "mulplt", and executing "f" to move forward in time,
but this does not reliably align the data in time.
It may be helpful to you to have the continuous waveform data files
that I am using, but they are too large to try to send via e-mail. I
would be happy to make them available by sharing over cloud storage.
------------------------
Start time
201501060200
Interval
55
Displays about 4 hours with several data dropouts < screen print >
PQL II shows
pql 2015-01-06-02*_?H?
i still think the best way to work with you data is a BUD or SEICOMP archicve
of 24 h miniseed files. did you try that. else a seisan continous data base
with one waveform file per hour per stations with corresponding s-files.
jens
Tim Sickbert
Devon Lab Coordinator
Boone Pickens School of Geology
105 Noble Research Center
Oklahoma State University
Stillwater OK 74078
(405) 744-6358
[email protected]
_______________________________________________
seisan mailing list
[email protected]
http://mailman.uib.no/listinfo/seisan
