Advanced Machine Learning Models Predict Urban Heat Dynamics in Bengaluru

Strategic Insights -

Researchers at the Indian Institute of Technology (IIT) Kanpur have successfully utilized a combination of air pollutant data, meteorological parameters, urbanization indicators, spectral indices, and elevation data to predict land surface temperature (LST) in Bengaluru over a five-year period. The study, published in the journal Natural Hazards, employed advanced machine learning models, including convolutional neural networks (CNN), K-nearest neighbors (KNN), gated recurrent units (GRU), recurrent neural networks (RNN), and artificial neural networks (ANN), to capture the complex interactions between these factors. The results showed that CNNs demonstrated superior predictive accuracy, effectively capturing spatial patterns and minimizing prediction errors better than the other approaches.

Key Takeaways:

  • The study utilized a combination of air pollutant data, meteorological parameters, urbanization indicators, spectral indices, and elevation data to predict LST in Bengaluru over a five-year period.
  • Advanced machine learning models, including CNNs, KNN, GRU, RNN, and ANN, were employed to capture the complex interactions between these factors.
  • NH3 was found to be the most influential factor affecting LST in both summer and winter.
  • CNNs demonstrated superior predictive accuracy, effectively capturing spatial patterns and minimizing prediction errors better than the other approaches.
  • The performance of the machine learning models was ranked as follows: CNN (R2 = 0.947 in summer, 0.913 in winter), KNN, GRU, ANN, RNN.
  • The study highlighted the value of advanced models in identifying urban temperature dynamics, aiding better urban planning and climate adaptation.

Statistics:

  • The study covered a period of five years.
  • The data used in the study included air pollutant data, meteorological parameters, urbanization indicators, spectral indices, and elevation data.
  • The machine learning models yielded the following performance metrics:

+ CNN: R2 = 0.947 in summer, 0.913 in winter

+ KNN: R2 = 0.832 in summer, 0.761 in winter

+ GRU: R2 = 0.792 in summer, 0.722 in winter

+ ANN: R2 = 0.764 in summer, 0.695 in winter

+ RNN: R2 = 0.747 in summer, 0.672 in winter

Sources:

  • NewsRx. New Findings in Machine Learning Described from Indian Institute of Technology (IIT) Kanpur (Exploring Urban Heat Dynamics Through Multi-model Machine Learning Analysis of Land Surface Temperature In Bengaluru India). Journal of Engineering. October 13, 2025; p 2017.
  • Springer. Exploring Urban Heat Dynamics Through Multi-model Machine Learning Analysis of Land Surface Temperature In Bengaluru India. Natural Hazards, 2025.