Analysis of urban heat island effect using k-means clustering

The characteristics of the urban heat island effect were often examined using urban heat island intensity (UHII). We collected daily mean, maximum and minimum temperature data from the Kaifeng Weather Station in the Kaifeng City and Huibei Station in the surrounding rural area for 2004–2008. Daily UHIIs calculated from daily mean, maximum and minimum temperature were regarded as components of vectors in a 3-D space and they clustered using the k-means algorithm. Clustering results showed that there was usually a slight UHI for most days, while sometimes nocturnal UHI effect was significant. There were very few cases with extremely obvious UHI effect and cooling effect was unignorable. The clustering results confirmed the graphical interpretations. The clustering method presented in this paper provides an alternative way for analysis of UHI effect.