Paper Conference

Proceedings of Building Simulation 2021: 17th Conference of IBPSA


Spatio-temporal data analysis for development of microclimate prediction models

Tageui Hong, Yeonsook Heo
Korea University, Korea, Republic of (South Korea)

Abstract: Microclimate prediction models have been developed to consider the effects of urban variables on urban microclimate. However, existing studies have not fully exploited the spatio-temporal microclimate data and focused on either spatial or temporal aspects of microclimate conditions. In this study, we analyze the characteristics of full spatio-temporal data of 246 weather stations in Seoul, Korea through the widely used multiple linear regression and Gaussian process regression. We created a set of datasets with different levels of spatialtemporal variability and evaluated the suitability of the two methods and the characteristics of the microclimate data. The statistical analysis results indicate that the accuracy of predicting the urban heat island (UHI) intensity depends on a level of variability contained in the spatio-temporal data and the two methods cannot fully explain the effect of meteorological and urban variables on the UHI phenomena. The results suggest the need to develop an appropriate modelling methodology that can accurately capture full variability in the spatial-temporal data of microclimate conditions.
Keywords: Urban microclimate, Urban characteristics, Urban heat island, Spatio-temporal data
Pages: 878 - 885