Artificial intelligence to classify the cooling effect of tree-shade in buildings’ façade: a case study in Brazil

Abstract

Urban heat islands, exacerbated by climate change, have become a pressing issue as summer temperatures rise. This study uses data mining techniques to classify the thermal impact of tree shade on building façades in the urban area of a tropical city. Our objective was to develop models to assist stakeholders and policymakers in forecasting the thermal impact of solar orientation and tree shade on building façades in the urban areas of a tropical city. Minimum and maximum infrared surface temperatures were registered in health clinics’ façades in Teresina, Brazil. Random forest methodology was applied to develop classifying models. This technique, known for its robust classification and prediction of categorical variables, offers a significant advantage over other modeling methods. Key input variables included façade infrared surface temperature, solar orientation, environmental temperature, relative humidity, and the extent of tree shade. Critical attributes were identified as solar orientation (North, South, East, and West), tree shade, and façade temperature (maximum and minimum). Two tree-ensemble models were selected for an accuracy rate of 88% and Kappa (κ) = 0.86. The models indicate that tree-ensemble methods can accurately classify and predict the thermal impact of tree shade on building façades. Additionally, the method effectively identified and ranked the factors influencing thermal impact, providing users with reliable predictive capabilities.