Development of a Pedestrian-Feeling Temperature Prediction (PeFT) Model during the Heatwave using Artificial Intelligence

The urban heat island is intensified due to the continuous temperature increase caused by climate change, leading to consistent property and casualty damages in urban areas. Urban areas are more vulnerable to heatwave events than rural areas. In order to reduce the damage caused by heatwaves in urban areas, it is necessary to analyze and predict meter-scale pedestrian-feeling temperature in street canyon surrounded by high-rise buildings. This study developed a pedestrian-feeling temperature prediction (PeFT) model using the data obtained from the building-block 3-dimensional urban meteorological experiment (BBMEX) campaign, conducted at urban center in Seoul, Korea during heatwave period. The thermal comfort types (TCTs) were defined to reflect the comfortability of pedestrians in urban areas. The TCTs and meteorological variables observed by the Korea Meteorological Administration were used as input data for the PeFT model. The temperatures at surface, 0.5m, 1.5m, and 2.5m high, observed by BBMEX campaign, were trained as target data for the model. Four machine learning techniques (generalized linear model, random forest, support vector machine and automatic machine learning) and four types of input data sets were tested. The optimal PeFT model was constructed by considering the root mean square error and determination coefficient (R²). The model was applied to produce the gridded temperature at four levels in the same area during another period. The PeFT model showed a potential to produce the 3-dimentional temperature distribution with a horizontal resolution of less than 5m within 3m height using the operational air temperature.

Keywords: meter-scale, pedestrian-feeling temperature, Pedestrian-Feeling Temperature prediction (PeFT) model, street canyon, Thermal Comfort Type (TCT)

Acknowledgements: This research was supported by the Basic Science Research Program through the National Research Foundation of Korea (NRF , 2021R1I1A2052562) funded by the National Institute of Meteorological Sciences (NIMS).