Vikctor/wildfire_disaster_management

🔥 Wildfire Prediction API

An advanced ML-powered API for wildfire occurrence prediction and severity assessment using deep learning models.

🌟 Features

• —Wildfire Occurrence Prediction: Binary classification using neural networks
• —Severity Assessment: Multi-class severity prediction (Low/Moderate/High)
• —Weather Integration: Automatic weather data fetching from Open-Meteo API
• —Enhanced Risk Analysis: Risk level interpretation with recommendations
• —Batch Processing: Multiple location predictions in one request
• —Historical & Forecast Data: Supports both historical (ERA5) and forecast weather data

🚀 API Endpoints

Core Predictions

• —POST /predict - Predict using coordinates and date (auto-fetches weather)
• —POST /predict/direct - Direct prediction with provided weather data

Utility Endpoints

• —GET /health - Health check with model validation
• —GET /model/info - Model information and required features
• —GET /api-info - Current API date range information
• —GET /date-info - Date parsing and information

📊 Usage Examples

Coordinate-based Prediction

curl -X POST "https://your-space-name.hf.space/predict" \
     -H "Content-Type: application/json" \
     -d '{
       "lat": 40.7128,
       "lon": -74.0060,
       "date": "2024-07-15"
     }'

Direct Weather Data Prediction

curl -X POST "https://your-space-name.hf.space/predict/direct" \
     -H "Content-Type: application/json" \
     -d '{
       "cloud_cover_mean": 45.0,
       "dewpoint_mean": 18.5,
       "evapotranspiration_total": 3.2,
       "fire_weather_index": 12.5,
       "humidity_min": 25.0,
       "pressure_mean": 1015.3,
       "solar_radiation_mean": 185.7,
       "temp_mean": 28.5,
       "temp_range": 12.3,
       "wind_direction_mean": 225.0,
       "wind_direction_std": 15.2,
       "wind_speed_max": 18.5,
       "daynight_N": 0,
       "lat": 40.7128,
       "lon": -74.0060,
       "month": 7
     }'

📈 Response Format

{
  "occurrence_probability": 0.75,
  "occurrence_class": 1,
  "severity_probabilities": [0.15, 0.35, 0.50],
  "severity_class": 2
}

🎯 Model Information

• —Occurrence Model: Deep neural network for binary wildfire prediction
• —Severity Model: Multi-class neural network for severity assessment
• —Features: 23 engineered features including weather, location, and temporal data
• —Weather Source: Open-Meteo API (ERA5 historical + forecast data)

🌍 Supported Locations

Global coverage with weather data from:

• —Historical: ERA5 data (1940-present)
• —Forecast: Up to 14 days ahead
• —Real-time: Current conditions

⚠️ Important Notes

• —Date Range: Historical data from 1940, forecast up to 14 days
• —Weather API: Automatically switches between historical and forecast APIs
• —Rate Limits: Respects Open-Meteo API rate limits
• —Coordinates: Global coverage (latitude: -90 to 90, longitude: -180 to 180)