A data science and machine learning project focused on Great Lakes ice forecasting using NOAA NetCDF (.nc) datasets. Includes full data processing pipelines, U-Net forecasting, visualization tools, and a static GitHub Pages dashboard.
Built by students at Grand Valley State University who live, study, and work near the Great Lakes.

• Marcos Sanson – Machine Learning Model Development
  Processed NetCDF data, built the forecasting dataset, developed and trained the U-Net model, and generated all HD visualizations and February predictions.

• Elijah Morgan – Data Reading & Analysis
  Handled ingestion of raw NetCDF files, verified dataset structure and metadata, and organized the initial file system used across the project.

• Darren Fife – Data Reading & Documentation
  Evaluated and interpreted the GLSEA dataset, confirmed the continuity of observations, and contributed to the end-to-end data preparation workflow.

• Diego de Jong – Frontend & Dashboard Development
  Built the GitHub Pages site, designed the layout and visual presentation of predictions, and created the project demo video.

• End-to-end pipeline for Great Lakes ice concentration forecasting
• NetCDF ingestion and visualization tools for NOAA GLSEA data
• Daily visualization generation with stable color scales and custom land shading
• Machine learning dataset builder using 7-day input windows
• U-Net model training for next-day ice forecasting
• Automated February prediction generator with versioned output folders and GIFs
• Static dashboard website hosted through GitHub Pages
• Jupyter notebooks for exploration and development
• HD visualization script with grid overlays for targeted ice-clearing operations
• Clear, modular project structure suitable for future extensions

MISpaceHackathon/
├── assets/
│   └── css/
│       └── style.css
│
├── data/
│   ├── raw/                 # (optional; primary raw data lives outside repo)
│   ├── processed/           # ML-ready arrays (X.npy, y.npy)
│   └── external/
│
├── notebooks/
│   └── 01_data_exploration.ipynb
│
├── src/
│   ├── daily_visualizations/         # Jan11–Jan31 PNGs
│   │
│   ├── data_processing/
│   │   ├── load_and_visualize.py
│   │   ├── inspect_nc.py
│   │   ├── downsample_data.py
│   │   ├── processor.py
│   │   └── nc_visualizer_outputs/
│   │
│   ├── models/
│   │   ├── predict_unet.py
│   │   ├── train_unet.py
│   │   └── checkpoints/
│   │
│   ├── utils/
│   │   └── gif_utils.py (optional)
│   │
│   └── predictions_ver_*/            # Feb predictions + GIFs
│
├── index.html
├── data-science.html
└── README.md

Live site

git clone https://github.com/ElijahLeeMorgan/MISpaceHackathon.git
cd MISpaceHackathon

Open index.html or run:

python -m http.server 8000

Visit: http://localhost:8000

python -m venv venv
source venv/bin/activate        # Windows: venv\Scripts\activate
pip install -r requirements.txt

Open the website or load index.html.

pip install xarray netCDF4 pandas numpy matplotlib

import xarray as xr
import matplotlib.pyplot as plt

ds = xr.open_dataset('weather_data.nc')
print(ds)

temp = ds['temperature']
temp.mean(dim=['lat', 'lon']).plot()
plt.title('Average Temperature Over Time')
plt.show()

NetCDF (Network Common Data Form) files are widely used in climate and weather science. They store large multi-dimensional arrays such as temperature, ice concentration, wind speed, and more.

pip install xarray netCDF4 numpy pandas matplotlib cartopy

The forecasting dataset converts daily 1024×1024 GLSEA maps into:

• Inputs: 7 consecutive days
• Label: next-day ice map

Resulting shapes:

X: (N, 7, 1024, 1024)
y: (N, 1024, 1024)

python src/models/train_unet.py

• Trains for 5 epochs
• Saves checkpoints to src/models/checkpoints/

python src/models/predict_unet.py

Outputs:

predictions_ver_1/
  Feb01.png
  Feb02.png
  Feb03.png
  Feb04.png
  forecast.gif

Enhancements include:

• Clamp to 0–6 range
• Stabilization smoothing
• Light-blue land shading
• HD rendering and GIF generation

notebooks/01_data_exploration.ipynb demonstrates:

• NetCDF inspection
• Visualization of Great Lakes patterns
• Data cleaning
• Early ML experiments

• NOAA GLSEA (Great Lakes Surface Environmental Analysis)
• NOAA CoastWatch / GLERL
• NCEP / HRRR / GFS weather model fields (optional future extensions)

1. Fork
2. Create a feature branch
3. Implement changes
4. Submit a pull request

Possible contribution areas:

• Better visualization
• Additional preprocessing
• New ML architectures
• More interactive dashboard features

MIT License — see LICENSE.

• Marcos Sanson
• Elijah Lee Morgan
• Diego De Jong
• Darren Fife

• xarray and NetCDF communities
• NOAA, NASA, ECMWF for open data
• Python data science community
• PyTorch and scikit-learn teams
• GitHub Pages for hosting

Built with ❤️ by students at Grand Valley State University, inspired by the Great Lakes we live beside and study around.