Classifying Urban Climate Zones (UCZs) Based on Spatial Statistical Analyses

Abstract

The objective of this study is the classification of urban climate zones (UCZs) based on spatial statistical approaches to provide key information for the establishment of thermal environments to improve urban planning. To achieve this, using data from 246 automatic weather stations (AWSs), air temperature maps in the summer of the study area were prepared applying universal kriging interpolation analysis. In addition, 22 preliminary variables to classify UCZs were prepared by a 100 m × 100 m grid. Next, six influential urban spatial variables to classify UCZs were finalized using spatial regression analysis between air temperature and preliminary variables. Finally, the UCZs of the study area were delineated by applying K-mean clustering analysis, and each spatial characteristic of the UCZs was identified. The results found that the accuracy of the air temperature of the study area ranged from ±0.184 °C to ±0.824 °C with a mean 0.501 root mean square predict error (RMSPE). Elevation, normalized difference vegetation index (NDVI), commercial area, average height of buildings, terrain roughness class, building height to road width (H/W) ratio, distance from subway stations, and distance from water spaces were identified as finalized variables to classify UCZs. Finally, a total of 8 types of UCZs were identified and each zone showed a different urban spatial pattern and air temperature range. Based on the spatial statistical analysis results, this study delineated clearer UCZs boundaries by applying influential urban spatial elements that resulted from previous classification studies of UCZs mainly based on pre-determined spatial variables. The methods presented in this study can be effectively applied to other cities to establish urban heat island counter measures that have similar weather observation conditions.