Run examples of processing

The case studies used in the manuscript are followings:

id	remove eq	temporal/spectral normalization	dv/v method	reference period
00	yes	no	stretching	2010-2022
01	yes	no	mwcs	2010-2022

You can download the dv/v datasheet from monitoring_stats_uwbackup_2010-2022.tar.gz

In the early development of software tools using TACC FRONTERA, we performed the following case studies for the data between 2002-2020 to investigate the effect of parameters associated with the process flow on the dv/v time history, which are archived as follows.

id	remove eq	temporal/spectral normalization	dv/v method	reference period
02	yes	no	stretching	2010-2020
03	yes	no	mwcs	2010-2020
04	yes	spectral whitening + onebit	stretching	2010-2020
05	yes	spectral whitening + onebit	mwcs	2010-2020
06	yes	no	stretching	2006-2016
07	yes	no	stretching	2007-2010
08	yes	no	stretching	2017-2020
09	no	no	stretching	2010-2020
10	yes	no	robust stuck + stretching	2010-2020
11	yes	no	compute coda Q	2010-2020

Archived dv/v datasheets

We investigated the case studies listed above to check the effect of reference period, normalizations and stacking methods to the dv/v time history. Indeed, it only causes minor differences in dv/v and does not modify the conclusions. However, as we have updated the software tools even after the the jobs done in FRONTERA and extended the study period from 2020 to 2022, we keep those case studies as the archives here.

The archived dv/v datasheets are available in monitoring_stats_TACCbackup.tar.gz The plots of the comparison between the master data sheets and the case study on Frontera can be found in Others/dvvanalysis_onTACC,

How to run the projects

To execute the process flow, we first configure the init file, e.g. Example/init_ex_download.jl, to set the input and output paths as well as the process parameters.

Note

The output directory is specified in the julia scripts as project_outputdir. We need disk space enough to store the raw and intermediate data (CCFs, stacked CorrData).

Then, we run the job using the topo_multi_slurm_run.jl and topo_slurm_multi.slurm. These work flows are summarized in the tutorial of SeisMonitoring.jl.

Note

We performed the casestudy with Slurm Workload Manager in Frontera, while we can also run the job in a workstation such that

#!/bin/bash

timestamp=$(date +PDT-%Y-%m-%d-%H-%M-%S)
NPROCS=16

date > log_$timestamp.txt
julia topo_multi_slurm_run_uwdata.jl  $NPROCS 2>&1 | tee -a log_$timestamp.txt
date >> log_$timestamp.txt

Once you finished the processing of the stacking and the measurement of dv/v, run SeisMonitoring.smstats_read_stretching/mwcs as coded in Utils/run_smstats.jl. This gathers the output of dv/v measurement and dumps to the csv file.

Run the archived casestudy

To conduct the archived casestudy listed in the table above, please initiate projects and run processes as following:

Move to Examples and run sh init_project_all.sh.

Note

You can run from download data to stacking & dv/v measurement in one project at once; However, in this repository we separately process them for the sake of simplicity.

Go to Examples and run/submit jobs from download to stacking (see slurm batches to run jobs).

Note

The batch files need to be manually prepared in each input directory. For the cross-correlation stage, use Utils/make_slurmbatch_parallel.jl to prepare the slurm batch files to parallelize cc process with time chunks. In Frontera, it takes ~1.6 hours to complete all the cross-correlations (13 stations x 3 components x 18 years) using 21 nodes and 37 cores/node. Note that this computational time does not include the waiting time of the job's queue.

Configure the paths and run sh Utils/smstats_seismonitoring.sh to compile the outputs into csv table.

Note

The stacked CorrData has the measurements in its misc (C.misc). run_smstats.jl gathers the measurements from stacked corrdata and output in csv table for post processing.

Development environments

We conducted the case study of the ambient seismic noise processing using TACC FRONTERA and the local workstation (48cores) installed in UW Denolle Lab.

Reference of FRONTERA: Dan Stanzione, John West, R. Todd Evans, Tommy Minyard, Omar Ghattas, and Dhabaleswar K. Panda. 2020. Frontera: The Evolution of Leadership Computing at the National Science Foundation. In Practice and Experience in Advanced Research Computing (PEARC ’20), July 26–30, 2020, Portland, OR, USA. ACM, New York, NY, USA, 11 pages. https://doi.org/10.1145/3311790.3396656

We also used the FASRC Cannon cluster supported by the FAS Division of Science Research Computing Group at Harvard University for the early development of the software tools.

We used the seismic data operated by the High Resolution Seismic Network (HRSN) doi:10.7932/HRSN.