Characterizing the spatial dynamics of land surface temperature–impervious surface fraction relationship

• Published in: International journal of applied earth observation and geoinformation 2016-03, Vol.45, p.55-65
• Main Authors: Wang, Jiong, Qingming, Zhan, Guo, Huagui, Jin, Zhicheng
• Format: Article
• Language: English
• Subjects: ISF; LST; Relationship; Scale; Spatial dynamics
• ISSN: 1569-8432; 1872-826X
• DOI: 10.1016/j.jag.2015.11.006

•The LST–ISF relationship is spatially dynamic and scales through space.•The dynamics manifests at the phenomenon and formation mechanism levels.•The optimal scale to investigate the relationship in the study area is between 500 and 650m.•Spatial Autocorrelation and Ordinary Least Square regressions unveiled the LST–ISF interaction through scales.•A flexible weight matrix is used to configure the Spatial Autocorrelation regression. The land surface temperature (LST) pattern is treated as one of the primary indications of environmental impacts of land cover change. Researchers continue to explore the potential contribution of land surface to temperature rising. The LST-land surface relationship is dynamic and varies spatially. Based upon the previous studies, this research assumes that such dynamics is manifested at two levels: (1) the phenomenon level, and (2) its formation mechanism level. The research presents a workflow of exploring such dynamics at both levels. The variogram of the phenomenon and multi-scale analysis of the LST-land surface relationship are mutually interpreted. In the case study of Wuhan, China, the variogram of the LST indicates that the operational scale of the phenomenon is 500–650m. It suggests the optimal scale to inspect the LST and its cause in the study area. This finding is verified and further inspected through multi-scale analysis of the LST–Impervious Surface Fraction (ISF) relationship at the formation mechanism level. The research also employs the Spatial Autocorrelation model to show how the ISF impacts the LST through scales. A flexible autocorrelation weight matrix is proposed and implemented in the model. The parameters of the model exhibit the thermal sensitivity of land surface and again represent the scale features. The Ordinary Least Square regression is used as the benchmark. Several implications are discussed.