Abstract: Along with the rapid urbanization, the expansion of population and explosion of urban impervious surface, the phenomenon of urban heat island is becoming more and more prominent and influences social and economic development seriously in negative ways. Based on Landsat 8_OLI image data in 2014, a monowindow algorithm was implemented for the retrieval of land surface temperature in Guangzhou. The retrieval temperature, together with GS+/GIS spatial analysis technology, was used for the analysis of the hierarchical characteristics of thermal intensity and its spatial differential characteristics. The results show that the area size of heat island intensity levels completely followed the normal distribution with a fitting degree of R²=0.971 and the dominant pattern effect of Ⅴ and Ⅵ intensity levels. The centers of heat island intensity level were observed in forms of “single center”, “single center surrounded by multiple subcenters”, and “multiple center”, which followed certain orders of succession. The centers of the rank of cold island intensity were also observed in forms of contiguous cold island center, single cold island center and abnormal cold island center with inconspicuous order of succession. The results of the variable function show that the heat island intensity had high spatial predictability, and obvious regional differences as well as probable hierarchical structure. The results of the study may be helpful for the understanding of the effects of urbanization and regional ecological environment, and the promotion of the sustainable development of the city.