Investigation the Robustness of Standard Classification Methods for Defining Urban Heat Islands

Abstract

In the process of studying the spatiotemporal cause mechanism of urban heat island (UHI) effects, the classification method used will directly affect the robustness of urban surface heat classification. Applying five commonly used standard classification methods, we divided Beijing's urban surface temperatures in the summer of 2020 into five levels. We then compared the reliability of the five classification methods in resolving 12-period data and the seasonal average temperature in UHI patches, based on two indicators: UHI area and UHI intensity. The actual land-use composition of the UHI patches obtained with traditional methods was applied to confirm our results. The mean-standard deviation method and natural breaks (Jenks) method were more robust with regard to UHI classification and 12-period data reliability. For the UHI area index, the mean-standard deviation method produced the smallest total area of UHI patches for summer days and nights. For the UHI intensity index, the quantile method, mean-standard deviation method, and natural breaks (Jenks) method were associated with smaller errors. Considering the composition of land-use types in UHI patches, the mean-standard deviation method, and natural breaks (Jenks) method were more rigorous. Thus, our research results provide guidance for method selection when classifying UHI.

Keywords

• Moderate resolution imaging spectroradiometer (MODIS) land surface temperature (LST)
• robustness
• standard classification method
• urban heat island (UHI) effect
• urban thermal grades