Machine Learning Classification of Fire Types Using MODIS Satellite Data
This project develops a comprehensive machine learning system to classify fire types in India using MODIS satellite data from 2021-2023. The solution includes data preprocessing, feature engineering, model training, evaluation, and deployment through an interactive web application.
- Primary: Develop accurate fire type classification using MODIS thermal and geographic features
- Secondary: Create deployable web application for real-time fire type prediction
- Impact: Support environmental monitoring and disaster management initiatives
Source: NASA FIRMS (Fire Information Resource Management System)
- Coverage: India, 2021-2023 (3 years)
- Satellites: Terra & Aqua MODIS sensors
- Resolution: 1 km spatial resolution
- Size: 500,000+ fire detection records
- Format: 3 CSV files (annual datasets)
- Geographic: latitude, longitude coordinates
- Thermal: brightness, bright_t31, frp (Fire Radiative Power)
- Sensor: scan, track, confidence levels (0-100%)
- Temporal: acquisition date, time, day/night flag
- Metadata: satellite, instrument type
- Target: fire type classification (MODIS/VIIRS)
- Data Integration: Merged multi-year datasets with validation
- Quality Assurance: Missing value analysis, duplicate detection, outlier treatment
- Feature Engineering: Temporal extraction (hour, month, season), categorical encoding
- Data Standardization: StandardScaler normalization for model consistency
- Class Balancing: SMOTE implementation for imbalanced dataset handling
Implemented and evaluated multiple classification algorithms:
- Logistic Regression: Linear baseline model with good interpretability
- Decision Tree: Non-linear decision boundaries with feature importance
- Random Forest: Ensemble method achieving 99.9%+ accuracy (Selected Model)
- K-Nearest Neighbors: Instance-based learning approach
- Performance Metrics: Accuracy, Precision, Recall, F1-Score
- Cross-Validation: Robust model validation with confusion matrices
- Feature Importance: Analysis of key predictive features
- Model Comparison: Comprehensive performance benchmarking
- Best Model: Random Forest Classifier
- Accuracy: 97.77%+
- Precision/Recall: High performance across all fire type classes
- Feature Importance: Thermal features (brightness, FRP) most predictive
- Temporal Patterns: Clear seasonal fire detection trends
- Geographic Distribution: Regional clustering across Indian subcontinent
- Confidence Levels: Bimodal distribution indicating detection certainty
- Class Distribution: Significant imbalance requiring SMOTE correction
- Fire-Themed Color Schemes: Custom palettes for consistent branding
- Interactive Geographic Maps: Folium-based visualization with 5000+ fire points
- Statistical Distributions: Histograms, box plots, correlation heatmaps
- Temporal Analysis: Monthly trends, seasonal patterns, hourly distributions
- Model Performance: Confusion matrices, accuracy comparisons, feature importance
- Clickable Maps: Detailed fire information popups
- Multi-Layer Visualization: Satellite imagery, street maps, terrain views
- Real-Time Updates: Dynamic filtering and zoom capabilities
- Professional Styling: Publication-ready plots with enhanced aesthetics
- User Interface: Professional fire-themed design with gradient backgrounds
- Input Features: Interactive forms for all model parameters
- Real-Time Prediction: Instant fire type classification with confidence scores
- Responsive Design: Mobile-friendly interface with custom CSS styling
- Parameter Validation: Min/max constraints with error handling
- Loading Animations: User experience enhancements
- Color-Coded Results: Visual fire type classification output
- Detailed Descriptions: Comprehensive fire type explanations
- Professional Footer: Developer attribution and contact links
india-fire-type-classifier-modis/
├── 📁 data/ # Raw and processed datasets
├── 📓 Classification_of_Fire_Types_in_India_Using_MODIS_Satellite_Data.ipynb # Main analysis notebook
├── 🐍 app.py # Streamlit web application
├── 💾 models/ # Trained models and scalers
├── 📊 visualizations/ # Generated plots and maps
├── 📄 README.md # Project documentation
- Python 3.8+
- Jupyter Notebook
- Git
- Core: pandas, numpy, scikit-learn
- Visualization: matplotlib, seaborn, folium
- ML: imblearn, joblib
- Web App: streamlit
- Utilities: datetime, warnings
- Clone repository and navigate to project directory
- Install required dependencies
- Open Jupyter notebook for analysis
- Run Streamlit app for web interface
- Wildfire Detection: Early warning systems for forest fires
- Agricultural Monitoring: Crop burning detection and analysis
- Urban Planning: Heat island effect and urban fire risk assessment
- Emergency Response: Rapid fire type classification for resource allocation
- Risk Assessment: Historical fire pattern analysis for prevention
- Policy Support: Data-driven environmental policy recommendations
- Climate Studies: Fire pattern correlation with weather data
- Ecological Research: Impact assessment on biodiversity
- Remote Sensing: Advanced satellite data processing techniques
- End-to-End Pipeline: Complete ML workflow from raw data to deployment
- Advanced Preprocessing: Comprehensive data cleaning and feature engineering
- Model Optimization: Systematic algorithm comparison and selection
- Production Ready: Scalable and maintainable code architecture
- Interactive Deployment: User-friendly web application interface
- Geographic Intelligence: Spatial analysis with interactive mapping
- Custom Visualizations: Fire-themed design with professional aesthetics
- Real-World Impact: Practical application for environmental monitoring
- Deep Learning: CNN/RNN implementation for enhanced accuracy
- Time Series Forecasting: Predictive fire occurrence modeling
- API Development: RESTful services for system integration
- Cloud Deployment: Scalable AWS/Azure infrastructure
- Real-Time Processing: Live satellite data stream integration
- Mobile Application: Cross-platform mobile app development
- Advanced Analytics: Multi-temporal analysis and trend prediction
- Integration Capabilities: Weather data fusion for enhanced predictions
- NASA FIRMS Portal
- MODIS Fire Product Documentation
- LP DAAC MODIS Products
- Scikit-learn Documentation
- Streamlit Documentation
@misc{modis_fire_classification_india_2025,
title={MODIS Fire Type Classification for India (2021-2023):
Machine Learning Approach for Satellite-Based Fire Detection},
author={Arshdeep Yadav},
year={2025},
url={https://github.com/arshdeepyadavofficial/india-fire-type-classifier-modis},
note={Machine Learning classification system using NASA MODIS satellite data}
}Contributions are welcome! Please follow these steps:
- Fork the repository
- Create feature branch (
git checkout -b feature/AmazingFeature) - Commit changes (
git commit -m 'Add AmazingFeature') - Push to branch (
git push origin feature/AmazingFeature) - Open Pull Request
MIT License - see LICENSE file for details.
- NASA FIRMS for comprehensive satellite fire data access
- MODIS Science Team for advanced fire detection algorithms
- Open Source Community for machine learning tools and libraries
- Streamlit Team for intuitive web application framework
Developer: Arshdeep Yadav
GitHub: arshdeepyadavofficial
LinkedIn: Arshdeep Yadav
Email: Available through GitHub profile
⭐ Star this repository if you found it helpful! ⭐
Made with ❤️ for environmental monitoring and disaster management
Empowering data-driven decisions for a safer, sustainable future