Dear SEISAN users,
I would be very grateful if someone could help me out... You don't need to answer all the questions. Any feedback would be deeply appreciated. Thank you, Emanuel A. 2015-03-27 16:55 GMT+00:00 Emanuel Antunes <[email protected]>: > Dear SEISAN community, > > > I once again resort to this collective pool of knowledge to seek advice > and help. I've been trying on my own, with the help of the manual and > experimenting with parameters, to find the answers, but so far I have been > unsuccessful, or the answers are unclear. > If someone could help me out solving these questions I would be highly > grateful, as I've wasted too much time trying to work around them... > I post the questions here in order of priority/urgency: > > Concerning the input file corr.inp: > > 1 - What format should the event list ("SFILE INDEXFILE" parameter) have? > Should it look like the DIRF produced files? Or a simple S-File with all > the events? I have tried both of these without success. Can anyone provide > me with an example? Also, can the list point to S-Files outside the folder? > > 2 - I have a big amount of master events, and a few years of continuous > data to serve as target waveforms where the correlation should be run to > find similar (new, unfound) events. Must I really input every master event > in a separated corr.inp/folder? > > 3 - Is there a way to point to the continuous waveforms and not to the > events? I noticed that CORR gave some segmentation faults, slicing the > waveforms in 12h segments did the trick. Since I have a big amount of data > I would prefer to store all the waveforms in one directory and just point > to that directory, so that I don't have to register these waveforms as > events in a SEISAN database... I am a bit confused here. > What I did in my trials was put the 12h segments in the same folder and > have "S-files" containing just this: > > > L 1 > 20110906_000000_taurus_0302_BHZ_00-12.mseed > 6 > > Line 1 is always required to have at least "L" another another region > nominator, and line 6 simply tells the program which waveform to use. Is > this correct or is there a better way to point to the waveforms? > > 4 - Can we address other folders in this SFILE EVENT /MASTER parameters? > (I am asking this because after much trials I can only input s-files which > are on the same folder from where I am running CORR. > > 5 -When running with CONTINUOUS MODE = 1: There is no output of the > correlation values, I can't find it neither in the console output, nor in > the corr.out or corr.trace. How can I get the correlation value? The only > place I can read the correlation value is in the extracted waveform when it > plots the correlation trace, this is OK for small runs, but for over > hundreds of events it's not practical. > > 6 -When running with CONTINUOUS MODE = 0: It looks like for each phase in > the S-File, and each event S-File, CORR only picks the maximum, and not all > the local maxima above the minimum correlation threshold, am I right? It > seems like it even picks events which have a correlation below the > threshold, so far as it's a maximum. > > 7 - When we use Continuous Extract in CORR, it seems that in the extracted > waveforms I always get a time window equal to that of the master signal > duration + 10 seconds. How can I extend this time window? I don't mind > changing the source code, but can't find where is this hardcoded in the > source code > > 8 - I've got some bizarre results using CONTINUOUS MODE = 0. In one event > which had a waveform running from 00h till 12h I got the following sfile in > corr.out: > > L 1 > 2011-09-06-0000-00S.A0302_002_P > 6 > 2011-09-06-0000-00S.A0302_002_S > 6 > STAT SP IPHASW D HRMM SECON CODA AMPLIT PERI AZIMU VELO AIN AR TRES W > DIS CAZ7 > A0302BZ P 240 > 1.90 > A0302BZ S 1320 > 2.68 > > Which should read 2:40 1.9s for P and 13:20 2.68s for S, correct? Well, P > phase isn't visible (nor any seismic event in that neighbourhood) and S > phase is outside the waveform! Is this a bug, or am I just misinterpreting > the results? Also, correlation trace is around 0.3 0.2 in these cases, so > it seems it's just picking a random maxima bellow the threshold. > > 9 - Finally, I think I don't have to ask this, but does CORR deal with a > master and event waveform with different sample rates? My guess is we have > to make sure everything has the same SPS prior to running the routine. > > > I thank you very much for all the time and effort put in helping solve > this issues. > > Best regards to all, > > > -- > Emanuel Antunes > > Seismology Laboratory > Instituto Superior Técnico (IST) > Universidade de Lisboa (UL) > -- Emanuel Antunes MSc in Physics Engineering Seismology Laboratory Instituto Superior Técnico (IST) Universidade de Lisboa (UL)
_______________________________________________ seisan mailing list [email protected] http://mailman.uib.no/listinfo/seisan
