Sentinel-2 Optical Pipeline

A Sentinel-2 remote sensing pipeline for generating RGB imagery and water-related spectral indices (NDVI, NDWI, MNDWI) from Copernicus Data Space imagery.

A Python pipeline for remote sensing and satellite imagery processing that generates RGB composites and spectral indices (NDVI, NDWI, MNDWI) from Sentinel-2 Level-2A imagery using the Copernicus Data Space Ecosystem (STAC API + S3 access).

This repository provides a reproducible geospatial workflow for downloading, processing, and exporting Sentinel-2 optical data for applications such as:

• satellite remote sensing analysis
• SAR validation and comparison
• flood monitoring and water detection
• vegetation monitoring using NDVI
• optical reference generation for SAR studies

⚠️ This repository does NOT contain Sentinel-2 imagery.
All data must be downloaded directly from the Copernicus Data Space Ecosystem.

Overview

This pipeline automates a typical Sentinel-2 remote sensing workflow:

1. STAC search with cloud filtering and date constraints
2. Band download via S3 API
3. RGB composite generation (TCI or B02/B03/B04)
4. Spectral index computation
   • NDVI (Normalized Difference Vegetation Index)
   • NDWI (McFeeters Water Index)
   • MNDWI (Modified NDWI)
5. Export to GeoTIFF and PNG quicklooks

The workflow is designed for geospatial analysis, satellite data processing, and remote sensing research.

Data Source

• Sentinel-2 Level-2A (Surface Reflectance)
• Accessed via the Copernicus Data Space Ecosystem
• Search performed via STAC API
• Data download via S3 protocol

This repository does not redistribute Sentinel-2 data.

When publishing derived products, follow the Copernicus / Sentinel attribution guidelines required by your journal or institution.

Repository structure

.
├ README.md
├ LICENSE
├ CITATION.cff
├ environment.yml
├ s2_stac_pick_cloudfree.py
├ s2_s3_download_rgb.py
├ s2_make_rgb.py
└ s2_make_indices.py

Required bands

Product	Bands Used
RGB	B02, B03, B04 (10 m)
NDVI	B08 (10 m), B04 (10 m)
NDWI	B03 (10 m), B08 (10 m)
MNDWI	B03 (10 m), B11 (20 m → resampled)
Mask	SCL (20 m → nearest resampled)

All outputs are standardized to 10 m resolution.

Notes

• 20 m layers (B11, SCL) are read on the 10 m reference grid (B03_10m) using on-the-fly resampling.
• Bilinear resampling is used for continuous reflectance bands (e.g., B11).
• Nearest-neighbor resampling is used for categorical layers (SCL).
• Pixels flagged as cloud/shadow/snow/invalid in SCL are masked (set to NaN).
• Indices are exported as float32 GeoTIFF and PNG quicklooks.

Requirements

• Python: 3.10+
• Conda: Miniconda/Anaconda recommended (use conda-forge)
• Environments:
• s2api: STAC search + S3 download
• s2viz: processing + visualization (GDAL/Rasterio stack)

All dependencies are pinned in environment_*.yml (single source of truth).

Setup

Create environment:

conda env create -f environment.yml conda activate s2viz

Create a .env file:

CDSE_S3_ACCESS_KEY=your_key CDSE_S3_SECRET_KEY=your_secret

Workflow

1) Scene selection

python s2_stac_pick_cloudfree.py

2) Download required bands

python s2_s3_download_rgb.py

3) Generate RGB composite

python s2_make_rgb.py

4) Compute spectral indices

python s2_make_indices.py

Outputs are written to:

downloads/S2_INDICES/

Intended use: SAR validation (optical reference)

These Sentinel-2 products are intended as optical reference layers to support SAR-based analysis (e.g., MicroSAR / small-SAR satellite studies), for example:

• Qualitative comparison of SAR-derived water masks vs. optical water proxies (NDWI/MNDWI)
• Seasonal context layers for SAR acquisitions (vegetation state via NDVI)
• Cloud-contamination screening using SCL (to avoid misinterpreting optical indices)

Important: spectral indices are not absolute ground truth for water. They can be biased by turbidity, algae blooms, sunglint, terrain shadows, mixed pixels, and adjacency effects.

Citation

If you use this repository in academic work, please cite the software and cite the Sentinel-2 data source separately.

Data citation (guidance)

Sentinel-2 data are provided by the European Union’s Copernicus Programme via the Copernicus Data Space Ecosystem. Follow your journal/institution guidelines for Copernicus/Sentinel attribution.

Software citation (recommended metadata)

If you use this repository in academic work, please cite it as:

Yu, H. (2026). Sentinel-2 Optical Pipeline (Version v0.1.0).
GitHub. https://github.com/Ha-eunYu/s2-optical-pipeline

Citation metadata is also available in CITATION.cff.

• Author: Ha-eun Yu
• Title: Sentinel-2 Optical Pipeline
• Year: 2026
• Version: v0.1.0
• URL: https://github.com/Ha-eunYu/s2-optical-pipeline

License

This project is licensed under the MIT License. See the LICENSE file for details.

---

Keywords

• Sentinel-2
• remote sensing
• satellite imagery
• spectral indices
• NDVI
• NDWI
• MNDWI
• Copernicus
• STAC
• geospatial
• Python
• SAR validation
• flood detection
• water detection
• optical reference
• earth observation