Municipality-Scale Flood Risk Mapping in Choco, Colombia

Using Sentinel-1 SAR and Ensemble Machine Learning (2015--2025)

Cristian Espinal Maya · Santiago Jimenez Londono

School of Applied Sciences and Engineering, Universidad EAFIT, Medellin, Colombia

·

---

Abstract

The Department of Choco, one of the rainiest and most flood-affected regions on Earth (5,000-13,000 mm/yr), lacks spatially explicit flood risk information at the municipal scale. We present an open-access, reproducible framework that delivers municipality-level flood risk statistics for all 30 municipalities of Choco, Colombia (46,530 km²; ~600,000 inhabitants). Using Sentinel-1 SAR, Google Earth Engine, and an ensemble of Random Forest, XGBoost, and LightGBM, the framework maps flood susceptibility and quantifies population exposure across Choco's predominantly Afro-Colombian and Indigenous communities.

This study extends the Antioquia flood risk framework to Choco's hyper-humid tropical environment, addressing unique challenges including dense tropical forest cover (>85%), extreme precipitation driven by the Choco Low-Level Jet, and limited SAR canopy penetration at C-band.

Repository Structure

.
├── gee_config.py              # Central configuration (Choco-adapted)
├── overleaf/                  # Manuscript (preprint format)
│   ├── main.tex               # Main LaTeX source
│   ├── arxiv.sty              # Preprint style file
│   ├── references.bib         # Bibliography
│   └── figures/               # Figure PDFs
├── scripts/                   # Processing and analysis pipeline
│   ├── 01_sar_water_detection.py
│   ├── 02_jrc_water_analysis.py
│   ├── 03_flood_susceptibility_features.py
│   ├── 04_ml_flood_susceptibility.py
│   ├── 05_population_exposure.py
│   ├── 06_climate_analysis.py
│   ├── 07_visualization.py
│   ├── 08_generate_tables.py
│   └── 09_quality_control.py
└── README.md

Study Area

Feature	Value
Department	Choco, Colombia
Area	46,530 km²
Municipalities	30
Subregions	5 (Atrato, Darien, San Juan, Pacifico Norte, Pacifico Sur)
Population	~600,000
Capital	Quibdo
Major rivers	Atrato (~4,900 m³/s), San Juan (~2,550 m³/s), Baudo
Annual precipitation	5,000 - 13,000+ mm/yr
Climate driver	Choco Low-Level Jet (ChocoJet)

Data Sources

All data are open-access and processed via Google Earth Engine:

• Sentinel-1 GRD (ESA/Copernicus) — 10 m SAR flood detection
• JRC Global Surface Water — 38-year water dynamics
• SRTM DEM v3 — Topographic features (30 m)
• MERIT Hydro — HAND computation (90 m)
• CHIRPS / ERA5-Land — Precipitation and soil moisture
• ESA WorldCover / Sentinel-2 — Land cover and NDVI
• WorldPop — Population density (100 m)
• GADM v4.1 — Administrative boundaries

Pipeline

python scripts/01_sar_water_detection.py       # SAR water detection (~4h on GEE)
python scripts/02_jrc_water_analysis.py         # JRC validation
python scripts/03_flood_susceptibility_features.py  # 18 predictor features (~2h)
python scripts/04_ml_flood_susceptibility.py    # ML training + SHAP (<30min)
python scripts/05_population_exposure.py        # Population exposure
python scripts/06_climate_analysis.py           # ENSO + seasonal analysis
python scripts/07_visualization.py              # Generate figures
python scripts/08_generate_tables.py            # Generate tables
python scripts/09_quality_control.py            # QC checks

Related Work

This study extends the methodology from:

Espinal Maya, C. & Jimenez Londono, S. (2026). Municipality-Scale Flood Risk Mapping in Antioquia, Colombia, Using Sentinel-1 SAR and Ensemble Machine Learning (2015-2025). Available at SSRN. GitHub

Citation

@article{EspinalMaya2026Choco,
  author  = {Espinal Maya, Cristian and Jim\'enez Londo\~no, Santiago},
  title   = {Municipality-Scale Flood Risk Mapping in {Choc\'o}, {Colombia},
             Using {Sentinel-1} {SAR} and Ensemble Machine Learning (2015--2025)},
  year    = {2026},
  note    = {Preprint}
}

License

Source code: MIT License. Manuscript and figures: CC BY 4.0.