Analyzing Urban Heat Island Effect Using Landsat-8 and Machine Learning: A Case Study of Lucknow, India

This project combines satellite remote sensing data and machine learning technology, taking Lucknow City in India as the research object. By processing Landsat-8 images via Google Earth Engine, it achieves precise monitoring of urban thermal environments and prediction of heat risks. The project code is open-source and highly reproducible, providing decision support for urban planning, public health, and other fields.

With the acceleration of global urbanization, the Urban Heat Island (UHI) effect has become a critical issue affecting livability and residents' health, especially prominent in developing countries like India due to rapid expansion. Traditional ground meteorological station monitoring has limitations such as sparse stations and insufficient coverage, while satellite remote sensing technology can obtain large-scale, high-resolution land surface temperature data at low cost, providing a new solution for thermal environment monitoring.

Core data sources include:

1. Landsat-8 satellite images: Equipped with the Thermal Infrared Sensor (TIRS), providing land surface temperature data at 100m resolution, with historical archives dating back to 2013.
2. Google Earth Engine (GEE): A cloud-based geospatial analysis platform integrating petabyte-scale data and computing capabilities. Technical tools rely on the Python ecosystem: Rasterio (raster data processing), NumPy/Pandas (numerical computation), Scikit-Learn (machine learning), Matplotlib (visualization).

Main analysis methods include:

1. Land Surface Temperature (LST) retrieval: Using Landsat-8 thermal infrared bands, retrieved via the radiative transfer equation, considering parameters such as atmospheric water vapor content and surface emissivity to ensure accuracy.
2. NDVI analysis: Measures vegetation coverage, which is negatively correlated with land surface temperature and is an important means to mitigate heat islands.
3. NDBI analysis: Identifies built-up areas; the low albedo and high heat capacity of building materials form the core of heat islands.
4. Hotspot detection: Uses spatial statistical methods to identify significant hot/cold spot areas and locate priority intervention regions.

The project introduces the Random Forest algorithm to build a heat risk classification model. Input features include land surface temperature, NDVI, NDBI, topographic elevation, and historical temperature trends, outputting heat risk levels for urban areas to support decision-making.

Taking Lucknow City as the research area, key results:

• Minimum temperature: 22.72°C
• Maximum temperature: 33.07°C
• Average temperature: 27.40°C
• Urban heat island severity score: 82.87 (out of 100) The temperature difference within the city exceeds 10 degrees, showing obvious thermal environment imbalance, with high heat stress risks in some areas.

Application Value:

• Urban planning: High-risk areas are prioritized for greening/ventilation corridor planning;
• Public health: Formulate high-temperature early warning and emergency strategies to protect vulnerable groups;
• Policy-making: Quantitative indicators provide scientific basis for evaluating sustainable development policies. Technical Highlights: Clear code structure, explicit dependencies, high reproducibility; uses open-source toolchains to reduce thresholds and costs, facilitating promotion.

This project demonstrates how to use free and open satellite data and cloud computing platforms to build a low-cost and efficient urban environment monitoring system. In the future, with the launch of new satellites like Sentinel-3 and the application of deep learning in remote sensing, monitoring accuracy and timeliness will be further improved; combining with IoT sensor networks, it is expected to achieve a leap from macro monitoring to fine management, supporting the construction of livable and sustainable cities.