Machine Learning Modeling of Outdoor Thermal Comfort (UTCI) in Southwestern Nigeria

Rapid urbanization and climate change are intensifying outdoor thermal stress in tropical cities, posing significant risks to public health and urban livability. This study develops and evaluates a machine learning framework for modeling the Universal Thermal Climate Index (UTCI) in Southwestern Nigeria, a rapidly growing and climate-vulnerable region.

A long-term hourly meteorological dataset derived from ERA5 reanalysis was utilized, incorporating key climatic variables such as air temperature, relative humidity, wind speed, and solar radiation.

Multiple machine learning models, including Artificial Neural Networks (ANN), Random Forest, and Extreme Gradient Boosting (XGBoost), were implemented and compared against a baseline linear regression model. Model performance was evaluated using Root Mean Square Error (RMSE), Mean Absolute Error (MAE), and the coefficient of determination (R²).

The results show that the ANN achieved the highest modeling accuracy, while ensemble models (Random Forest and XGBoost) also demonstrated strong and consistent performance, effectively capturing complex non-linear atmospheric interactions.

Feature importance analysis identified air temperature and solar radiation as the dominant drivers of thermal comfort variability, followed by relative humidity and wind speed. However, UTCI was derived from the same meteorological variables used as model inputs; therefore, the machine learning models are interpreted as approximating the UTCI functional relationship rather than predicting an independent observational target.

This study is among the first to apply machine learning-based UTCI modeling in a tropical African urban context. The findings provide a robust, data-driven framework for high-resolution thermal comfort modeling and offer valuable insights for climate-sensitive urban planning, heat-health early warning systems, and climate adaptation strategies.