This work has been done by using open data from INGV. The aim is to process seismic data from SoS Enatos region.
Data available at: https://eida.ingv.it/en/getdata/ for SoS Enatos region it is necessary to search for SENA.
What has been done now is the q-Transform of data.
The approach to stablish a "SNR" of the signal is the ratio between a STA (short time average) and LTA (long time average) and had promissing results unstil now. With this in hand, a "cutoff" is set and we are able to get the times of the events that we want.
After this, the q-Transform is created, using GWPy. The optimal value of the parameters are still being tested before we proceed with the dimensional reduction techniques that we want to use so we can have coherent data.
A study of microseisms was perfomed by making use of BLRMS (Band-Limited Root Mean Square). With this, it was possible to observe that diferent bands will have diferent responses.
CEEMDAN (Complete Ensemble Empirical Mode Decomposition with Adaptive Noise) was used in a file in order as a commissioning and benchmark of this type of algorithm for future pespectives.
First, a study about how to work on seismic data was done by using one of the seismic bulletins from INGV. At this step, .qml files were used and we could undestand what instances are necessary to perform a useful workflow (specially when it comes to spectral analysis). After this, we have tried to obtain data from the Netherlands station (available at knmi.nl) in order to have other region to compare, but all of this was without suscess. Another files downloaded at this preliminar studies of seismic data was the Einstein Telescope data (also from INGV). This marked an important step, as it was at this point that we began working with .mseed files and established a standardized workflow for seismic data processing.
All of this work is at the /old folder and has been archived just for a report of what was previously done.
The data from the SoS Enatos region was downloaded month by month strating at 00:00:00 of the 1st say, and ending at 23:59:59 pof the last day of the respective month. To keep an standart of the data, it started at 21/06/2021 and ended at 31/12/2025 (DD/MM/YYYY). Since it starts at the very begining of the measurements and ends at the last complete year at the beging of this work.
A preliminar study of complete month was done first in order to debug and to optimize the codes. For this, the folder "/SENA-mseed" was used, containing a folder of the complete month of jan/2024 (explained at the info.txt file). The files at "/sena-jan24-raw-plots" folders were the first ploted data of amplitude versus time. At "/sena-jan24-filtered-test" we've made a cutting on microseismic data. Finally, at "/sena-jan24-filtered-test" we've obtained the spectogram of each day.
With all of this in mind, we could perform a PSD analysis of each month of the year searching for patterns. Some of our results are at the folder "/sena-mseed-3d_plots".
With an workflow stablished, we have proceded to complete year analysis. The data is at the /SENA-files and it is shorted by year at the respective folders for each year. the "/teste" folder contains the data from june/2021 and has been used for the first tests and debugs.
Note from May: now these folders are located at /FFT_PSD_Tests.
We have obtained a way to develop the calculations of the FFT and PSD of the data. An atempt of paralelizing was done, with openmp, with promissing results. These results are stored at the /FFT_PSD_Tests folder. Some of the paralelizing steps are at the folder /Parallel folder.
We also processed the velocity of seismic data in order to obtain a physical meaning and to have a better understanding of data so the classification could be done in a better way. We had some issues with the removing of the station response, so there is a little debug program at the /Station folder, so this specific test can be done quickly.
Optmizing the computaional time of the analisys of our microseismic data has been an important step, since we want to implement the q-transform and use dimensional reduction techniques (such as t-SNE).
Study of the periodicity of seismic events.
Classification of seismic events trough band-limiting.