UCSD Jacobs School of Engineering

Masters in Data Science and Engineering

Capstone Project: Post-Fire Debris Flow Likelihood Prediction

• Contributors:
  • Alejandro Hohmann
  • Bhanu Muvva
  • Chunxia Tong
• Digital Object Identifier: https://doi.org/10.6075/J0PG1RZH

Project Abstract

Debris Flows are a distinct type of landslide that suddenly occur without warning. They are fast-moving channels of water and soil that carry large natural objects like boulders and trees, or human-made objects including cars. In the American West, Debris Flows have directly caused death and property damage. Debris Flows often occur after rain events and the burn scars left behind by wildfires increase their likelihood. Given the increasing frequency of extreme weather events, it is critical to predict Debris Flows and take precautionary action before they occur. This project builds upon prior research of predicting Debris Flows using additional geological features and more advanced machine learning techniques. The project also includes an intuitive interface for decision makers to access these probability estimates.

Using this Repository

This repository is organized so that everything in the notebooks folder can be run sequentially. However, the data preparation steps are already latent in this repository and do not need to be executed to use.

• To skip straight to the final output, use the app folder. This folder was copied to EC2 instance for end-user access. It can also be run locally by simply executing the app.ipynb or app.py files.
• To experiment with NN architecture, use the notebooks/03_ML_models folder.

├── app
    ├── data
    ├── model
    ├── app.ipynb
    └── app.py
├── data
├── images
├── Landfire
├── notebooks
    ├── 00_archive
    ├── 01_staley
    ├── 02_data_prep
    ├── 03_ML_models
    └── 04_data_viz
├── ABC Debris - Final Poster.pdf
├── ABC Debris - Final Presentation.pdf
├── ABC Debris - Final Report.pdf
├── LICENSE
├── README.md
├── requirements.txt
└── venv.py

Dependencies managed with virtual environment and requirements.txt file:

• open terminal window and clone the repository (git clone ...url)
• open a terminal window in the repository folder
• to create a virtual environment and install the required packages with necessary dependencies, run the following commands in terminal:

> conda create -n venv_debris_flow python=3.10.10
> conda activate venv_debris_flow
> conda -y install jupyter jupyterlab ipykernel
> python -m ipykernel install --user --name=venv_debris_flow
> # you may need to restart the environment to use it in jupyter lab
> conda deactivate
> conda activate venv_debris_flow
> pip install -r requirements.txt

when you launch a notebook, make sure to select the venv_debris_flow kernel

Prior Work Referenced

• The Staley et al. (2016) model report and data:
  • https://pubs.er.usgs.gov/publication/ofr20161106