🌊 HydroSAR: Satellite-Based Flood Intelligence Platform

HydroSAR is a production-grade geospatial intelligence platform that detects flood inundation through cloud cover using Sentinel-1 Synthetic Aperture Radar (SAR) data.

By combining Google Earth Engine with a FastAPI async backend and a modern Streamlit dashboard, HydroSAR delivers near real-time flood impact analytics for disaster response and planning.

---

🚀 Key Features

• SAR-Based Flood Detection: Uses Sentinel-1 VH backscatter for cloud-penetrating flood mapping
• Dynamic Otsu Thresholding: Automatically determines optimal flood threshold using statistical variance separation
• Terrain-Aware Filtering: Reduces false positives using slope-based correction from DEM data
• Land Cover Impact Analysis: Intersects flood extent with ESA WorldCover to quantify damage by class
• Async Job Processing: Non-blocking FastAPI job system for long-running Earth Engine tasks
• Interactive Map UI: Before/After comparison using Folium split map visualization
• Smart Caching Ready Architecture: Designed to support cached geospatial queries

---

🛠️ Tech Stack

Layer	Technology
Cloud GIS	Google Earth Engine
Backend	FastAPI, Uvicorn
Frontend	Streamlit, Folium
Geospatial	GeoPandas, Shapely
Data Sources	Sentinel-1 SAR, ESA WorldCover, SRTM DEM

---

🏗️ System Architecture

graph TD
    A[Streamlit UI] --> B[FastAPI Backend]
    B --> C[Job Manager]
    C --> D[Google Earth Engine Pipeline]
    D --> E[SAR Processing + Otsu Threshold]
    D --> F[Land Cover Analysis]
    E --> G[Flood Mask]
    F --> H[Impact Statistics]
    G --> I[Response Payload]
    H --> I
    I --> A

Loading

---

⚙️ Installation

1. Clone Repository

git clone https://github.com/sanatladkat/hydrosar.git
cd hydrosar 

2. Create Environment

conda env create -f environment.yml
conda activate hydrosar

3. Install Dependencies

pip install -r requirements.txt

4. Authenticate Google Earth Engine

earthengine authenticate

---

▶️ Run the Project

Start Backend

uvicorn backend.main:app --reload

Start Frontend

streamlit run app.py

---

📊 Output Example

• Total Flooded Area (ha)
• Cropland Impacted
• Urban Infrastructure Affected
• Primary Damage Zone

---

🧠 Core Methodology

SAR Flood Detection

HydroSAR uses Sentinel-1 VH backscatter differences between pre-event and post-event scenes to detect inundation.

Dynamic Thresholding (Otsu Method)

Automatically separates flooded vs non-flooded pixels by maximizing inter-class variance in radar backscatter.

Terrain Correction

Slope-based filtering reduces false positives caused by radar shadow and terrain distortion.

---

🌍 Use Cases

• Disaster response and flood mapping
• Agricultural damage assessment
• Urban flood risk analysis
• Insurance and risk modeling
• Government emergency planning