Urban Heat Island Detection 🏙🌡

A machine learning project for detecting urban heat islands using satellite data classification, focusing on Bucharest, Romania.

🌍 Overview

Urban heat islands occur when urban areas experience higher temperatures than surrounding regions. This project uses XGBoost classification on satellite imagery to predict urbanization levels and analyze heat island effects.

Key Results: 89% accuracy across 5 urbanization classes using multi-temporal satellite data (2018-2024).

✨ Features

• Multi-temporal satellite analysis (Landsat 8/9 + Sentinel-2)
• XGBoost classification with 5 urbanization categories
• Interactive Streamlit web application
• Temporal change detection over 6-year period
• Real-world applications for urban planning

📊 Dataset

• Sources: Landsat 8/9 (thermal), Sentinel-2 (multispectral)
• Area: Bucharest metropolitan region
• Time: 2018 vs 2024 comparison
• Features: NDVI, NDWI, NDBI, LST + temporal changes
• Classes: Water, Vegetation, Moderate/High Urbanization, No change

🚀 Installation

git clone https://github.com/yourusername/Urbanization-assessment.git
cd Urbanization-assessment
pip install -r requirements.txt

Run Web Application

streamlit run home.py

🌐 Web Application

Interactive Streamlit app with 4 main sections:

• Home: Project overview and pipeline
• Data Exploration: Satellite data analysis and visualizations
• Model Details: Architecture, training, and performance metrics
• Conclusions: Results, applications, and future directions

🏛 Applications

• Urban Planning: Heat island mitigation strategies
• Environmental Monitoring: Climate change impact assessment
• Public Health: Heat-related risk identification
• Real Estate: Climate-resilient development planning