SAR Flood Detection Application

A web application for detecting surface water and flooding events using free Sentinel-1 SAR (Synthetic Aperture Radar) data from Google Earth Engine.

🎯 Overview

This application enables users to:

• Draw an Area of Interest (AOI) anywhere on Earth
• Automatically fetch the latest Sentinel-1 SAR imagery
• Detect surface water using adaptive Otsu thresholding
• Visualize results interactively with dual basemap options
• Download detected water polygons as GeoJSON

🏗️ Architecture

┌─────────────────────────┐
│ Frontend (GitHub Pages) │
│  - React + Leaflet      │
│  - AOI drawing          │
│  - Basemap toggle       │
└─────────────┬───────────┘
              │ GeoJSON AOI + params
              ▼
┌─────────────────────────┐
│ Backend (Cloud Run)     │
│  - FastAPI              │
│  - Google Earth Engine  │
│  - Sentinel-1 SAR       │
│  - Water detection      │
└─────────────┬───────────┘
              │ GeoJSON water polygons
              ▼
┌─────────────────────────┐
│ Visualization           │
│  - Water overlay        │
│  - Metadata display     │
│  - GeoJSON download     │
└─────────────────────────┘

SAR.Flood.Detection.-.Brave.2026-01-28.13-37-21.mp4

🚀 Demo

Live Demo

🚀 Features

• Adaptive Water Detection: Otsu thresholding automatically adjusts to scene characteristics
• Location Search: Search and navigate to any location worldwide
• GeoJSON Export: Download detected water polygons for further analysis
• Free Data: Uses Copernicus Sentinel-1 SAR via Google Earth Engine

🛠️ Technology Stack

Backend

• FastAPI: Modern Python web framework
• Google Earth Engine: SAR imagery access and processing
• Pydantic: Request/response validation
• Google Cloud Run: Serverless deployment

Frontend

• React 18: UI framework
• Leaflet: Interactive mapping
• Leaflet Draw: AOI drawing tools
• Turf.js: Geospatial calculations
• Axios: API communication

📋 Prerequisites

Backend

• Python 3.10+
• Google Cloud Platform account
• Google Earth Engine service account
• gcloud CLI (for deployment)

Frontend

• Node.js 16+
• npm or yarn

🔧 Setup Instructions

1. Google Earth Engine Service Account

Follow the detailed guide in docs/GEE_SETUP.md to:

1. Create a GCP project
2. Enable Earth Engine API
3. Create a service account
4. Download credentials JSON
5. Set up for local development and Cloud Run

2. Backend Setup

cd backend

# Create virtual environment
python -m venv venv
source venv/bin/activate  # On Windows: venv\Scripts\activate

# Install dependencies
pip install -r requirements.txt

# Create .env file
cp .env.example .env

# Edit .env and set:
# GEE_SERVICE_ACCOUNT_PATH=./credentials/gee-service-account.json

# Place your GEE service account JSON in backend/credentials/

# Run locally
python run_local.py

Backend will be available at http://localhost:8000

3. Frontend Setup

cd frontend

# Install dependencies
npm install

# Run development server
npm start

Frontend will be available at http://localhost:3000

🧪 Testing with Sample AOIs

The application includes 2 pre-loaded test locations:

1. Venice Lagoon, Italy - Permanent water baseline for algorithm validation
2. Lake Mead, Nevada/Arizona - Large reservoir for drought monitoring

Select these from the sidebar to quickly test the application.

🚀 Deployment

Backend to Google Cloud Run

# From project root
./deploy.sh

# Or manually:
cd backend
gcloud run deploy sar-flood-api \
  --source . \
  --region us-central1 \
  --platform managed \
  --timeout 300s \
  --memory 1Gi \
  --allow-unauthenticated \
  --set-secrets GEE_SERVICE_ACCOUNT=gee-service-account:latest

Frontend to GitHub Pages

cd frontend

# Update .env.production with your Cloud Run URL
# REACT_APP_API_URL=https://your-cloud-run-url

# Deploy
npm run deploy

Or use the GitHub Actions workflow (see .github/workflows/deploy-frontend.yml)

📊 Water Detection Algorithm

1. Data Ingestion: Fetch most recent Sentinel-1 GRD imagery (last 30 days)
2. Preprocessing:
   • Radiometric calibration
   • Terrain correction (SRTM)
   • dB conversion
   • Lee speckle filtering
3. Feature Derivation:
   • VV and VH polarizations
   • VV-VH difference
   • Texture (local std dev)
   • Slope
4. Adaptive Detection:
   • Otsu thresholding (or manual override)
   • Polarization rules
   • Terrain masking
   • Texture filtering
5. Refinement:
   • Morphological operations
   • Small object removal
   • Geometry simplification
6. Vectorization: Convert to GeoJSON polygons

🎨 Advanced Parameters

For fine-tuning detection (accessible via collapsible panel):

• End Date: Select a specific date to search for imagery on or before that date (default: latest available)
• Max Slope (degrees): Exclude steep terrain unlikely to retain water (default: 5°)
• Min Area (pixels): Filter out small noise artifacts (default: 100 pixels)

The algorithm uses adaptive Otsu thresholding which automatically determines the optimal VV backscatter threshold based on the image histogram.

⚠️ Limitations

Technical Constraints

• AOI Size: Maximum 50×50 km (2500 km²) for real-time processing
• Temporal: Uses most recent acquisition within specified date range
• Orbit: ASCENDING pass only for consistency
• Resolution: 10m Sentinel-1 ground resolution
• Processing Time: 15-30 seconds depending on AOI size

Algorithm Limitations

• No-Water Scenes: Otsu thresholding may incorrectly identify "water" pixels in completely dry areas (unimodal histograms with no actual water present)
• Frozen Water Bodies: Ice and snow on rivers/lakes appear similar to land in SAR imagery, causing frozen water to be missed or misclassified

Environmental Conditions

• High Winds: Wind-roughened water surfaces increase backscatter, reducing detection accuracy or causing water to appear as land
• Vegetation: Dense floating vegetation may mask water presence

🔮 Future Enhancements

• Temporal change detection (before/after flooding)
• Multi-temporal stacking for stability
• Sentinel-2 optical cross-validation
• Permanent water vs flood water classification
• Time-series animation

📄 License

MIT License - Free for educational and commercial use

🙏 Acknowledgments

• Copernicus Sentinel-1 mission for free SAR data
• Google Earth Engine for cloud-based processing
• OpenStreetMap and Esri for basemap tiles

📧 Contact

For questions, issues, or collaborations, please open an issue on GitHub.

---

Note: This is a demonstration project. For operational flood monitoring, consider additional validation and temporal analysis.