Newer
Older
This repository contains codes and datafiles that were generated within the
scope of our study "Ground-Motion Modeling as an Image Processing Task:
Introducing a Neural Network Based, Fully Data-Driven,and Nonergodic Approach"
(Lilienkamp et al. 2022)
Core functions are stored in the file <code>unetgmm/unetGMMlib.py</code>
Some precalculated datafiles are stored in <code>unetgmm/data/predefined</code>
## Prerequisites
This code was developed under Ubuntu 18.04
[1] Clone this repository:
<code> git clone https://git.gfz-potsdam.de/lilienka/unetgmm.git </code>
[2] Create virtual environment:
<code> python3.8 -m venv ~/py38_unetgmm </code>
[3] Activate virtual environment:
<code> source ~/py38_unetgmm/bin/activate </code>
[4] Upgrade pip:
<code> pip3 install --upgrade pip </code>
[5] Install required packages:
To reproduce the results from our study in the Kanto basin presented in
Lilienkamp et al. (2022) conduct the following steps:
[1] Download Bahrampouri et al. (2021) strong motion database from:
https://www.designsafe-ci.org/data/browser/public/designsafe.storage.published/PRJ-2547
and store files in unetgmm/data/strong_motion/
[2] Download the "depth to seismic bedrock (V.3.2 ESRI shapefile “Subsurface
Structure” (layer 30)
(The Headquarters for Earthquake Research Promotion, 2021)
from https://www.j-shis.bosai.go.jp/map/?lang=en
[3] Sample "depth to seismic bedrock" at regular grid sites given in the file:
unetgmm/predefined/regular_grid.csv;
Normalize sampled values between 0 and 1 and store the result as a numpy
array of shape (572x572) in:
unetgmm/predefined/bedrock.npy
[4] run <code> python3 unetgmm/processing/process_bahrampouri.py </code>
[5] run <code> python3 unetgmm/modeling/training_kanto.py </code>
[6] run <code> python3 unetgmm/modeling/predict_kanto.py </code>
[7] run <code> python3 unetgmm/modeling/evaluate_kanto.py </code>
Results will be stored in unetgmm/results/figures/
To reproduce the results from our synthetic study presented in
Lilienkamp et al. (2022) conduct the following steps:
[1] run <code> python3 unetgmm/processing/generate_synthetic_dataset.py </code>
[2] run <code> python3 unetgmm/modeling/training_synth.py </code>
[3] run <code> python3 unetgmm/modeling/predict_synth.py </code>
[4] run <code> python3 unetgmm/modeling/evaluate_synth.py </code>
Results will be stored in unetgmm/results/figures/
## References
Bahrampouri, M., A. Rodriguez-Marek, S. Shahi, and H. Dawood (2021).
An updated database for ground motion parameters for KiK-net records,
Earthq. Spectra 37, no. 1, 505–522, doi: 10.1177/ 8755293020952447.
The Headquarters for Earthquake Research Promotion (2021).
Modeling concept of subsurface structures from seismic bedrock to ground
surface, available at https://www.jishin.go.jp/evaluation/strong_motion/underground_model/integration_model_kanto_2021/
(last accessed January 2022) (in Japanese).
Lilienkamp, H., S. von Specht, G. Weatherill, G. Caire, and F. Cotton (2022). Ground-Motion
Modeling as an Image Processing Task: Introducing a Neural Network Based, Fully Data-
Driven, and Nonergodic Approach. Bulletin of the Seismological Society of America 112,
no. 3, 1565–1582. doi:10.1785/0120220008.