Analysis of spatial heterogeneity and driving factors of land surface temperature using spatial regression models

Subject Areas :

Zahra Parvar ¹ , Marjan Mohammadzadeh ² , Sepideh Saeidi ³

1 - PhD. Student, Department of Environmental Sciences, Faculty of Fisheries and Environmental Sciences, Gorgan University of Agricultural Sciences and Natural Resources, Iran.
2 - Associate Professor, Department of Environmental Sciences, Faculty of Fisheries and Environmental Sciences, Gorgan University of Agricultural Sciences and Natural Resources, Iran.
3 - Assistant Professor, Department of Environmental Sciences, Faculty of Fisheries and Environmental Sciences, Gorgan University of Agricultural Sciences and Natural Resources, Iran.

Received: 2022-12-20 Accepted : 2023-03-04 Published : 2023-02-20

Keywords: Geographic weighted regression, Ordinary least square, Spatial heterogeneity, landscape,

Abstract :

Land surface temperature is a significant factor affecting thermal variation and balance in global studies. In the last two decades, the great necessity for LST data in environmental studies and land resource management activities has made the measurement of LST as a major scientific debate. Discovering the spatial heterogeneity of land surface temperature and analyzing the key factors and specific effective spatial relationships that are affected by time series have great importance in land management. The aim of this study is to analysis of land surface temperature driving factors and spatial heterogeneity using spatial regression models. To review this issue, daily LST maps were prepared by the radiative transfer equation method using Landsat 7 and 8 data for 2002, 2013, and 2021 years in Bojnord city. The analysis of land surface temperature in areas where barren lands prevail requires nighttime temperature data. Therefore, MODIS night LSTs were also prepared as auxiliary maps. Pearson correlation, spatial autocorrelation, ordinary least square, and geographically weighted regression models were used for data analysis. Then, the performance of the models was compared using the coefficient of determination and the Akaike information criterion. The results showed that the GWR approach had a better prediction accuracy and a better ability to describe spatial non-stationarity than the OLS approach. The spatial response of LST and different influencing variables from 2002 to 2021 showed that the development of green space plays an important role in modulating land surface temperatures. Since LST is influenced by various variables, including topography, climatic and atmospheric variables, and vegetation, therefore, understanding spatial relationships and analyzing the areas with high LST can be useful as a way forward in the planning strategies.