Spectral Use Case Validation - Flood 🌊 and Fire 🔥

🌍 Explore the Interactive Maps 🛰️

This repository supports the paper’s LR vs. SR spectral use case validation, comparing native Sentinel-2 imagery against physics-constrained latent diffusion SR outputs for water (flood) and burn-severity (fire) detection. It provides reproducible code, data download helpers, and links to the accompanying walkthroughs on the project site usecases.opensr.eu.

Run the LR vs SR validations locally

The steps below mirror the instructions shown in the notebooks; follow them once to enable both workflows.

1. Install dependencies: create a virtual environment and install requirements.

   python -m venv .venv
   source .venv/bin/activate
   pip install -r requirements.txt

2. Fetch data: download the prepared Sentinel-2 stacks (native, interpolated, and SR products) with the helper script. It populates data_fire/ and data_flood/ under each workflow directory.

   ./fetch_data.sh

   Update ZIP_URL in the script to point at your storage location if you host the archive elsewhere.
3. Open the walkthroughs: the rendered notebooks are available on the project site at https://usecases.opensr.eu/notebooks; they reference the same commands below for local execution.

Running the workflows

The workflows are organized by hazard; each step is a numbered script. Run them from their directory after data download.

Flood (Valencia 2024)

cd flood_workflow
python 01_after_create_SR_flood.py       # visualize SR vs baseline inputs
python 02_create_flood_mask.py           # compute spectral mask
python 03_calc_mndwi.py                  # derive MNDWI index
python 04_create_thresh_mask.py          # apply thresholds
python 05_histograms_and_mndwi_thresh.py # explore histogram-based cuts
python 06_compute_metrics.py             # evaluate detection metrics

Fire (Palisades 2025)

cd fire_workflow
python 01a_before_create_SR_fire.py      # visualize native inputs
python 01b_after_create_SR_fire.py       # visualize SR outputs
python 02_create_fire_mask.py            # compute spectral mask
python 03_calc_dnbr.py                   # derive dNBR index
python 04_create_thresh_mask.py          # apply thresholds
python 05_histograms_and_dnbr_thresh.py  # explore histogram-based cuts
python 06_compute_metrics.py             # evaluate detection metrics

Notes

• Both workflows assume GPU access for LDSR-S2 super-resolution (torch.cuda.is_available() must be true).
• Replace paths or thresholds in the scripts as needed to test additional events or alternative SR baselines.