Land Surface Temperature (LST) is a vital satellite remote sensing-driven indicator of earth heat studies. LST can provide information about urban heat emission, urban climate, and human activities in urban areas. In recent years, the calculated LST for a satellite image pixel has been studied as a parameter affected by urban environment factors such as available land cover types in the same pixel. However, in this study, a scenario in which the calculated LST for a pixel is not only affected by the factors in the same pixel but also by the factors in the neighbor pixels is studied. Firstly, required maps for the calculated LST and influential factors (indicators of vegetation, building, and water surfaces) are produced from satellite remote sensing images. Secondly, the relationship between the LST and influential factors is modeled using the Ordinary Least Squares (OLS) model. Thirdly, Moran’s I and Lagrange Multiplier tests are used to analyze the existence of spatial dependency and correlation in residuals of the OLS model. Fourthly, three spatial regression models (Spatially Lagged X (SLX), Spatial Lag (SL), and Spatial Error (SE) models) are used to model the spatial dependency and correlation between the LST and influential factors. Finally, the outcomes of the models are compared and evaluated. Results present related maps for the variables besides maps for spatial residuals in the spatial regression models. The outcomes of the models are investigated by p-values, log-likelihood, and RMSE. To conclude, the spatial regression models fitted the relation between the dependent and independent variables better than the OLS model.
