This study proposes a machine learning model for estimating snowpack bulk density ($\rho_s$) from snow depth ($HS$) and other variables that can be measured or derived from the date and location of $HS$ measurements. This repository contains:

• Source code for our paper (DOI will be shared after paper acceptance).
• Instructions for setup and usage.

The dataset used in this study comes from thre sources:

• SNOTEL Dataset - was downloaded using metloom.
• Global Seasonal Snow Classification.
• Maine Snow Survey Data.

This project uses Conda for environment management. However, you can use any environment management tool of your choice. For example, you can manage Python versions with pyenv and create a virtual environment using venv. Go to Step 3 if you wish not to use Conda.

If you don’t have Conda installed, download Anaconda or MiniConda. See Installing Conda to know which installer is right for you.

conda env create --file environment.yml
conda activate ml_density_env 

For those not using Conda, you can set up a virtual environment with venv:

python -m venv ml_density_env
source ml_density_env/bin/activate # activate the virtual environment

Note: the Python version will be the default version on your PC. You can manage Python version using pyenv.

mkdir ml_density
cd ml_density

git clone https://github.com/cryogars/density-models.git
cd density-models

Ensure everything is set up correctly:

python --version  # Should return 3.11.5
pip list  # Displays installed packages

To verify the models are working correctly, you can run the tests:

pytest -v # Run all tests

pip install .

If you wish to modify the source code, install in development mode:

pip install -e .

Note: This project uses conda to only manage the Python version and install Jupyter. All package dependencies are installed via pip. This approach makes the installation process more consistent whether you're using conda or another environment management tool (e.g., pyenv + venv). All dependencies are specified in requirements.txt

Create the data folder and download:

1. SNOTEL Data: link.
2. Global Seasonal Snow Classification on NSIDC: NSIDC link. For this project, download SnowClass_NA_300m_10.0arcsec_2021_v01.0.nc.
3. Main Snow Survey Data: link.

If you would to reproduce our results, you can do so by running the following experiments:

cd experiments
./optimize_ml_models.sh

After running the experiments, you can deactivate the conda environment by running the command below:

conda deactivate

To completely remove the environment, run:

conda env remove --name ml_density_env

The authors would like to thank:

1. USDA NRCS for providing the SNOTEL data
2. M3Works for their metloom package, which we used to download the SNOTEL data.
3. Maine Geological Survey and the United States Geological Survey for providing the Maine Snow Survey data.
4. The creators of the srtm.py Python package for their open-source tool, which we used to obtain the SRTM elevation data.
5. U.S. Army CRREL for the funding (BAA W913E520C0017).

For any questions or issues, please open an issue or reach out to [email protected].