Geospatial Analysis and Machine Learning Framework for Urban Heat Island Intensity Prediction: Natural Gradient Boosting and Deep Neural Network Regressors with Multisource Remote Sensing Data

Nhat-Duc Hoang, Quoc-Lam Nguyen

The increasing severity of the urban heat island (UHI) effect is a consequence of rapid urban expansion and global climate change. The urban center of Da Nang, Vietnam, is currently experiencing severe UHI effects combined with increasingly frequent heatwaves. This study employs advanced machine learning techniques—including natural gradient boosting machine and deep neural network—to model the spatial variation in UHI intensity. The explanatory variables include topographical features, distances to coastlines and rivers, land cover types, built-up density, greenspace density, bareland density, waterbody density, and distance to wetlands. Experimental results show that the machine learning models successfully explain 90% of the variation in UHI intensity. To identify the primary factors influencing UHI intensity, Shapley additive explanations are utilized. Additionally, a neural network-based cellular automata model is implemented to project future land cover changes. The proposed framework is then employed to forecast UHI intensity in Da Nang’s urban center in 2040. Based on the prediction results, the area with extremely high UHI intensity is expected to increase by 3.7%. The area with high UHI intensity is projected to rise by 4.6%, while the area with medium UHI intensity is anticipated to expand by 12.6%. Notably, it is forecasted that the areas with extremely low and low UHI intensity are forecasted to decrease by 3.9% and 40.8%, respectively. The findings from this study can be useful to assist urban planners in establishing effective mitigation strategies for reducing the impact of UHI effects.