Introduction

This software repository contains the scripts necessary to reproduce the results from the article "Self-Supervised Coherence-Based Denoising on Cryoseismological Distributed Acoustic Sensing Data." The preprint of the article is available on authorea: DOI: 10.22541/au.172779667.76811452/v2

Abstract:

One major challenge in cryoseismology is that signals of interest are often buried within the high noise level emitted by a multitude of environmental processes. Events of interest potentially stay unnoticed and remain unanalyzed, particularly because conventional sensors cannot monitor an entire glacier. However, with Distributed Acoustic Sensing (DAS), we can observe seismicity over multiple kilometers. DAS systems turn common fiber-optic cables into seismic arrays that measure strain rate data, enabling researchers to acquire seismic data in hard-to-access areas with high spatial and temporal resolution. We deployed a DAS system on Rhonegletscher, Switzerland, using a 9 km long fiber-optic cable that covered the entire glacier, from its accumulation to its ablation zone, recording seismicity for one month. The highly active and dynamic cryospheric environment, in combination with poor coupling, resulted in DAS data characterized by a low Signal-to-Noise Ratio (SNR) compared to classical point sensors. Our objective is to effectively denoise this dataset. We use a self-supervised J -invariant U-net autoencoder capable of separating incoherent environmental noise from temporally and spatially coherent signals of interest (e.g., stick-slip or crevasse signals). The method shows enhanced inter-channel coherence, increased SNR, and significantly improved visibility of the icequakes. Further, we compare different training data types varying in recording position, wavefield component, and waveform diversity. Our approach has the potential to enhance the detection capabilities of events of interest in cryoseismological DAS data, hence to improve the understanding of processes within Alpine glaciers.

Setup

1. Download the code using the terminal

   gh repo clone JohannaZitt/MAIN_DAS_denoising
   

   Alternatively, download the code via the GitHub user interface:

   Code -> Download ZIP
   

   and decompress the folder.

2. Install and activate the environment environment.yml via conda:

   conda env create -f environment.yml
   conda activate env_main_das_denoising
   

3. Download the data folder , unzip it, and replace the empty data folder in the Jupyter project with the downloaded one.

Optional: 4. To simplify code execution without the need to denoise the real-world data, the denoised data from experiment 03_accumulation_horizontal is provided: . Download the folder, unzip it and place it under the directory experiments/03_accumulation_horizontal.

Description of single .py files

calculating_cc.py

Calculates the local waveform coherence and cross-correlation between DAS data and co-located seismometer data for each experiment, as detailed in Section 4.1.1 and Section 4.3. The results are saved in a .csv file named cc_evaluation_id.csv within the respective experiment folder.

Requirements: To execute this script for all experiments, the raw DAS data must be denoised beforehand. For experiment 03_accumulation_horizontal, you can download the denoised data (Optional Step 4. in Setup).

denoise_synthetic_DAS.py

Denoises the synthetic data located in the folders data/synthetic_DAS/from_DAS and data/synthetic_DAS/from_seis, as detailed in Section 3.4. The denoised data is saved in the directory: experiments/experiment_name/denoised_synthetic_DAS.

denoising_DAS.py

Denoises the raw real-world data located in the folder data/raw_DAS, as outlined in Section 3.4. The denoised data is stored in the directory experiments/experiment_name/denoisedDAS.

generate_DAS_preprocessed_training_data.py

The initial waveforms are generated from real-world DAS data sections for model fine-tuning as detailed in Section 3.3.

generate_seismometer_preprocessed_training_data.py

The initial waveforms are generated from seismometer data for model training as detailed in Section 3.2.

generate_synthetic_test_data.py

Generates synthetic test data from seismometer data and sections of DAS data. This generated test data is utilized in Figure 3 and Figure S2.

helper_functions.py

Includes several functions that are used repeatedly: e.g., bandpass filter and computation of local waveform coherence.

main_training.py

Primary training is conducted on the initial waveforms generated using seismometer data.

models.py

Class for the model structure and training data generator.

plotting_fig3.py

Generates Figure 3: Raw data sections, denoised data sections and wiggle comparison for synthetically generated DAS data derived from seismometer data.

Requirements: Denoise synthetically generated DAS data obtained from seismometer data with denoiser 03_accumulation_horizontal or download denoised DAS data sections (Optional Step 4. in Setup).

plotting_fig4.py

Generates Figure 4: Raw and denoised real-world DAS data sections with local waveform coherence and wiggle comparison.

Requirements: Denoise real-world DAS data obtained with denoiser 03_accumulation_horizontal or download the denoised DAS data sections (Optional Step 4. in Setup).

plotting_fig5.py

Generates Figure 5: Frequency content of denoised and raw DAS records

Requirements: Denoise real-world DAS data obtained with denoiser 03_accumulation_horizontal or download the denoised DAS data sections (Optional Step 4. in Setup).

plotting_fig6.py

Generates Figure 6: Average local waveform coherence and mean cross-correlation between DAS data and co-located seismometer data. The calculated cc values are saved in experiments/experiment_name/cc_evaluation.csv for each experiment under its respective name.

plotting_figS1.py

Generates Figure S1: Representative initial waveforms.

plotting_figS2.py

Generates Figure S2: Raw data sections, denoised data sections and wiggle comparison for synthetically generated DAS data derived from real-world DAS data.

Requirements: Denoise synthetically generated DAS data obtained from DAS data with denoiser 03_accumulation_horizontal or download denoised DAS data sections (Optional Step 4. in Setup).

plotting_figS3-S5.py

Generates Figure S3, Figure S4, and Figure S5: Waveform section plots for three exemplary DAS data sections displaying icequakes.

Requirements: Denoise real-world DAS data obtained with denoiser 03_accumulation_horizontal or download the denoised DAS data sections (Optional Step 4. in Setup).

plotting_fig6.py

Generates Figure S6: STA/LTA trigger of raw and denoised DAS data section.

Requirements: Denoise real-world DAS data obtained with denoiser 03_accumulation_horizontal or download the denoised DAS data sections (Optional Step 4. in Setup).

retraining.py

Fine-tuning the models with real-world DAS data sections as described in Section 3.3.

Help

If you need assistance, have any questions, or have suggestions for improvements, feel free to contact me via email.

References

Parts of the code are built upon the software provided by van den Endet et al. [1].

[1] van den Ende, M., Lior, I., Ampuero, J.-P., Sladen, A., Ferrari, A. ve Richard, C. (2021, 3 Mart). A Self-Supervised Deep Learning Approach for Blind Denoising and Waveform Coherence Enhancement in Distributed Acoustic Sensing data. figshare. doi:10.6084/m9.figshare.14152277.v1