UNDERSTANDING THE EFFECTS OF WEATHER DATA AND SITE CHARACTERISTICS ON URBAN AIR TEMPERATURE IN SINGAPORE USING MACHINE LEARNING

Abstract

With the rapid development of urbanization, many issues have emerged. Being one of the most significant consequences of urbanization, Urban Heat Island (UHI) has driven many studies over the years. In microclimate area, to understand the urban air temperature patterns and settlements, studies have been done to analyze the factors affecting it, using traditional analytical methods. Meanwhile, as machine learning is widely adopted in many statistical studies, microclimate studies can also utilize such methods to explore the correlations between the factors and urban air temperature. In this study, a machine learning approach was adopted to investigate the factors affecting urban air temperature in Singapore, which include weather data and site characteristics at various radiuses of influence. Multiple linear regression lines were derived using gradient descent, a type of machine learning optimization algorithm at each radius of influence. As a result, the correlations between air temperature, weather data and site characteristics in Singapore were presented and discussed. As a result, this study revealed the presence of strong linearity between the factors and air temperature in Singapore at various radiuses of influence. In addition, results also concluded that solar radiation, Sky View Factor (SVF) and grass area are the significant contributing factors to air temperature. This research can be utilized as an assessment and prediction tool for microclimate in Singapore. Furthermore, it can be regarded as a preliminary research of applying machine learning into microclimate area, with which more non-linear relationships can be explored using machine learning algorithms.