[MSNoise] MSNoise Digest, Vol 31, Issue 1

Thu Oct 6 20:19:24 UTC 2016

Hi Ashton,

Trying to answer inline below:

On 06/10/2016 19:38, Flinders, Ashton wrote:
> Thanks Esteban,
>
> I have looked through the documentation extensively, as well as the SRL
> paper, and a bit of grepping on the source code (e.g. it took me a while to
> figure out why in the world I couldnt remove instrument responses using a
> RESP or PZ file, since it is listed on the documentation page, until I
> grepped the source code and saw those were not implemented.)
Yeah, that's supposed to pop out if you're using the -vvvv option for 
the commands (DEBUG level). This should be implemeted (if needed = not 
for me as I only have either dataless or inventory, but still)..
>
> 1) Yes. It is more efficient to compute the cross-correlation functions or
> Green’s functions using 30-min chunks. Although you can change this feature
> if you will. Segments are re-build afterwards.
>
> This is only true if the daily segments are always preloaded into memory
> (IO vs CC time). The SRL paper makes it sounds like this is the case, but I
> want to make sure that this is always the case, even for processing many
> stations/days or long segments. I also wasnt sure if these segmented CC
> were used for any statistics (e.g. uncertainty in daily CC values).
They are, the base of MSNoise is a daily job, so traces are preloaded & 
preprocessed, one by one, in RAM before processing. No "moving window 
I/O" is done,. In the future, for LargeN or very high freq cases, I want 
this to be more dynamic, but for the moment it's like this.
>
> 2) You need to define the filters in order to compute the
> cross-correlations and extract the coherent part of the ambient noise for a
> given frequency range. MWCS is used to compute the moving window cross
> spectral analysis after the cross-correlation procedure.
>
> Yes, I understand that (the way its been written) filter values should be
> defined before the CC (which really should only be the whiteneing band or
> preprocessing), however I dont understand why in the web gui the whiten and
> MWCS filters are put within the same group. This make me predefine a MWCS
> filter band, along with my whiten band, even though the MWCS doesnt happed
> for another two processing steps. This leads to confusion, for example;
>
> What happens if I compute the CC's with a defined filter (as written, a
> MWCS and whiten filter defined together) and then want to compute different
> MWCS's later on when I actually get to the MWCS step. Can I just define a
> new filter, with new mwcs_low mwcs_high values? Since the web gui forces me
> to define a filter high low value (whiten, I assume) does it not use these?
> Do I have to go back and recalculate the cross correlations everytime I
> want to add an MWCS filter (which I wouldnt expect I have too)?

The MWCS values don't need to be predefined when computing the CC, but 
the logic we defined is the following:
1/ you compute the CC in a quite narrow band
2/ you compute the MWCS is a (slightly) narrower band

the original idea is thus not:
1/ you compute the CC in a broad band
2/ you compute MWCS in different bands based on the same CC

but I'm completely OK to start a discussion about this, certainly 
because, for example, the lag-time-related parameters are also 
filter-related, and somehow should be linked to the filters iff, like 
now, they are quite narrow-band.

> Similarly, what happens if I want to delete a set of processed MWCS or
> stacks? I cant just delete the folder, since their are still pointers in
> the sql database. Just reprocessing doesnt seem to work either.
Just delete the MWCS & DTT folders, and run "msnoise reset DTT --all" 
and the jobs will be reset.

Cheers,

Thomas
>
> thanks again
>
> -ashton
>
>
>
>
> On Thu, Oct 6, 2016 at 10:17 AM, Esteban Chaves <echfisica at gmail.com> wrote:
>
>> Hi Ashton,
>>
>> As an active user:
>>
>> I highly recommend you to check the documentation: http://msnoise.org/doc/
>> workflow.html#module-msnoise.s03compute_cc <http://msnoise.org/doc/
>> workflow.html#module-msnoise.s03compute_cc>
>> I think the pre-processing and processing steps are well documented in the
>> web site.
>>
>> 1) Yes. It is more efficient to compute the cross-correlation functions or
>> Green’s functions using 30-min chunks. Although you can change this feature
>> if you will. Segments are re-build afterwards.
>>
>> 2) You need to define the filters in order to compute the
>> cross-correlations and extract the coherent part of the ambient noise for a
>> given frequency range. MWCS is used to compute the moving window cross
>> spectral analysis after the cross-correlation procedure.
>>
>> 3) dunno. May be Thomas can give you more info.
>>
>> -Esteban J. Chaves
>>
>>
>>> On Oct 6, 2016, at 5:00 AM, msnoise-request at mailman-as.oma.be wrote:
>>>
>>> Send MSNoise mailing list submissions to
>>>        msnoise at mailman-as.oma.be
>>>
>>> To subscribe or unsubscribe via the World Wide Web, visit
>>>        http://mailman-as.oma.be/mailman/listinfo/msnoise
>>> or, via email, send a message with subject or body 'help' to
>>>        msnoise-request at mailman-as.oma.be
>>>
>>> You can reach the person managing the list at
>>>        msnoise-owner at mailman-as.oma.be
>>>
>>> When replying, please edit your Subject line so it is more specific
>>> than "Re: Contents of MSNoise digest..."
>>>
>>>
>>> Today's Topics:
>>>
>>>    1. new user (Flinders, Ashton)
>>>
>>>
>>> ----------------------------------------------------------------------
>>>
>>> Message: 1
>>> Date: Wed, 5 Oct 2016 13:10:53 -0700
>>> From: "Flinders, Ashton" <aflinders at usgs.gov>
>>> To: <msnoise at mailman-as.oma.be>
>>> Subject: [MSNoise] new user
>>> Message-ID:
>>>        <CAPbWHEO7J_9dyMdKJ3SQkpGy9hG7N1M-COwgzGTpkVmy85ig5g at mail.gmail.
>> com>
>>> Content-Type: text/plain; charset="UTF-8"
>>>
>>> Hi all, long-time obspy user new MSnoise user.
>>>
>>> I just had a couple quick questions I was hoping to get clarification on;
>>>
>>>>  From the SRL article;
>>> "Once all waveforms are loaded in memory, the computation is done by
>>> iteration of the station pairs, on the different components to compute
>> and
>>> the on the different filters for each defined window (30 minutes slices
>> by
>>> default)."
>>>
>>> 1) To be clear, the two time-series from a station-station pair are
>> divided
>>> into segments (30 minutes long, no overlap, by default), and the CC is
>> done
>>> on these segments? Are these segments then rebuilt into the daily CCF? Is
>>> this segmentation purely for efficieny reasons, or is there something
>> else
>>> going on?
>>>
>>> 2) Why does msnoise compute_cc, not work if there is no MCWS filter
>> defined?
>>> 3) Is there an intelligent way to search the mailing list archive?
>>>
>>> Thanks for the help!
>>>
>>> -ashton
>>>
>>>
>>> ------------------------------
>>>
>>> _______________________________________________
>>> MSNoise mailing list
>>> MSNoise at mailman-as.oma.be
>>> http://mailman-as.oma.be/mailman/listinfo/msnoise
>>>
>>>
>>> End of MSNoise Digest, Vol 31, Issue 1
>>> **************************************
>> _______________________________________________
>> MSNoise mailing list
>> MSNoise at mailman-as.oma.be
>> http://mailman-as.oma.be/mailman/listinfo/msnoise
>>
>
>