ALPROS - Automatic Landscape Prediction & Rescue Optimiztion System

ALPROS is a comprehensive landslide risk prediction and monitoring system designed for Kerala's vulnerable regions. It combines classical statistical analysis with state-of-the-art machine learning to provide accurate, real-time landslide hazard assessments across micro-scale geographic zones.

System Components

ALPROS features three integrated modules:

1. CLI Risk Analyzer (landslide.py) - Traditional rule-based assessment
2. ML Risk Predictor (train_model_with_soil.py & landslide_model.pkl) - XGBoost-based classification
3. Web Dashboard (app.py) - Real-time micro-zone grid visualization

Features

Real-time Analysis

• Live Weather Integration: Fetches current precipitation via OpenWeatherMap API
• Elevation & Slope Calculation: Uses Open Elevation API for precise topographic analysis
• Soil Factor Classification: Distinguishes between rock, laterite, clay, and weathered soil

AI-Powered Risk Assessment

• Machine Learning Model: XGBoost classifier trained on 3,000+ synthetic realistic scenarios
• 5-Factor Risk Evaluation:
  • 24-hour rainfall accumulation
  • 72-hour rainfall accumulation
  • Terrain slope angle (0-1 normalized scale)
  • Digital elevation data
  • Soil composition factor (0.1-0.9 scale)
• Probability-based Scoring: Returns landslide probability (0.0-1.0)

Micro-Zone Grid System

• Divided Coverage: Geographic area segmented into 5×5 micro-zones
• Color-coded Risk Levels:
  • GREEN: Low risk (probability < 0.4)
  • YELLOW: Moderate risk (0.4 ≤ probability < 0.7)
  • RED: High risk (probability ≥ 0.7)

Region-Specific Analysis

• Predefined Regions:
  • Kavalappara
  • Munnar
  • Wayanad
  • Malappuram
• Each region is defined by latitude/longitude bounding boxes and linked to the nearest OpenWeatherMap city for rainfall data.

APIs

• Grid Risk API: /grid_risk?region=<region> - Returns risk assessment for all micro-zones in the specified region.
• Regions API: /regions - Lists all available regions with their display names and associated cities.

Model Training

• Synthetic Data Generation: Simulates realistic rainfall, slope, elevation, and soil conditions.
• Logical Landslide Condition: Incorporates domain knowledge to define landslide-prone scenarios.
• XGBoost Classifier: Trained on 3,000 samples with a test accuracy of ~90%.
• Feature Importance:
  • Rainfall (72-hour) and slope are the most critical factors influencing landslide risk.

Caching for Performance

• Elevation Cache: Stores elevation data for lat/lon coordinates to reduce API calls.
• Rainfall Cache: Caches rainfall data for cities to optimize OpenWeatherMap API usage.
• Soil Factor Cache: Caches soil classification data retrieved from Bhuvan WMS.

Installation

1. Clone the repository:

   git clone https://github.com/your-repo/alpros.git
   cd alpros

2. Install dependencies:

   pip install -r requirements.txt

3. Add your OpenWeather

Map

API key:

• Replace YOUR_OPENWEATHER_API_KEY in app.py with your actual API key.

4. Run the Flask app:

   python app.py

5. Access the web dashboard at http://127.0.0.1:5000/dashboard.

Usage

Train the Model

To retrain the model, run:

python train_model_with_soil.py

This will generate a new [landslide_model.pkl]

API Endpoints

• Home: [http://127.0.0.1:5000/]
• Regions: [http://127.0.0.1:5000/regions]
• Grid Risk: [http://127.0.0.1:5000/grid_risk?region=<region>]

Future Enhancements

• Add support for additional regions.
• Integrate more advanced soil and geological data sources.
• Improve real-time data handling for rainfall and elevation.