Unpacking the nonlinear relationships and interaction effects between urban environment factors and the urban nighttime heat index

Abstract

In this study, we identified the urban environmental factors affecting the nighttime heat index (NHI), which has policy implications for improving the urban thermal environment using Smart Seoul Urban Data Sensor (S-DoT) data and an interpretable machine learning methodology. S-DoT sensors are ideal for deriving the heat index at a micro-scale in space and time, unlike ground-based observations and satellite-based imagery traditionally used to measure urban temperatures. Moreover, the interpretable machine learning methodology can overcome the nonlinearity argument of the influential factors identified in previous studies. The main results of this study are as follows. First, the Sky View Factor generally has a negative association with the NHI, which means that a higher SVF is associated with a lower NHI. Conversely, the surface roughness commonly has a positive association with the NHI, which signifies that a higher surface roughness results in a higher NHI. Second, the building view factor significantly increases the NHI, and the gross floor area of the building has a positive association with the NHI regardless of its use. Third, although the center of the city has a high temperature, if it has physical environment factors similar to the outskirts of the city, the NHI there may be relatively low compared to the nighttime air temperature. This study is meaningful because it empirically shows the nonlinear relationships and interaction effects between urban environment factors and the NHI and suggests policy implications to improve the urban nighttime thermal environment.