Alpine Fault Anisotropy Inversion, Whataroa, New Zealand

<p>This project aims to perform anisotropy inversion of bedrock along the Alpine Fault using Distributed Acoustic Sensing (DAS) data collected near Whataroa, South Island, New Zealand. The inversion methodology involves forward modeling of seismic wavefields in anisotropic media, automated picking of qP- and qS-wave arrivals, and comparison with manually picked arrivals obtained from field measurements. Results indicate significant damage to the rock mass within the fault zone compared to intact rock conditions.</p>

<p>This repository contains files and scripts necessary for performing anisotropy inversion using Distributed Acoustic Sensing (DAS) data.</p>

<h2>Data Files:</h2>

<ul>
<li><code>dasXYZ.csv</code>: Locations of DAS channels in local coordinates.</li>
<li><code>ibedrock_small.npy</code>: Binary geological model stored as a NumPy array.</li>
<li><code>srcXYZ.csv</code>: Source locations in local coordinates.</li>
<li><code>wellGeom.csv</code>: Well geometry represented by consecutive points in local coordinates.</li>
<li><code>Massiot_data.txt</code>: Fracture set data from Massiot et al. (2018).</li>
</ul>

<h3>Manually Picked Arrival Times:</h3>

<ul>
<li><strong>SpickedRealData/</strong>: Directory containing manually picked S-wave first-break data in <code>.fbp</code> format.</li>
<li><strong>PpickedRealData/</strong>: Directory containing manually picked P-wave first-break data in <code>.fbp</code> format.</li>
</ul>

<h2>Neural Network and Processing Scripts:</h2>

<ul>
<li><code>network_parts.py</code>: Defines the structure of the neural network used for inversion.</li>
<li><code>train_model.pth</code>: Pre-trained PyTorch model weights.</li>
<li><code>Helpers.py</code>: Helper functions for applying the trained neural network and post-processing to generate automatic picks.</li>
</ul>

<h2>Inversion Prototypes:</h2>

<ul>
<li><code>job-TI.py</code>: Inversion script for transversely isotropic media.</li>
<li><code>job-orth.py</code>: Inversion script for orthotropic media.</li>
<li><code>job-mono.py</code>: Inversion script for monoclinic anisotropy type 1.</li>
<li><code>job-mono-type2.py</code>: Inversion script for monoclinic anisotropy type 2.</li>
</ul>

<h2>Fracture Modeling:</h2>

<ul>
<li><code>fracturemodeling.py</code>: Script for modeling fractures.</li>
</ul>

<h2>Usage</h2>

<p>Use scripts provided (<code>network_parts.py</code> and <code>Helpers.py</code>) with the pretrained neural network (<code>train_model.pth</code>) to perform automatic picking of seismic arrivals and anisotropy inversion.</p>

<p>Ensure dependencies such as PyTorch, NumPy, and any other necessary libraries are installed before execution.</p>