Abstract:

Land surface temperature (LST) is a direct impact of urbanization and a crucial factor in global climate and land cover changes. In this research, we aim to identify the impact of land use/land cover (LULC) on LST as well as analyze the relationship between LST and three spectral indices using linear, polynomial and multiple regression models. The LST was first retrieved from Landsat imagery using single-channel algorithm. Afterwards, LULC maps were developed using maximum likelihood (ML) classifier and three spectral indices, namely Normalized Difference Vegetation Index (NDVI), Normalized Difference Built-up Index (NDBI) and Normalized Difference Water Index (NDWI). Finally, regression analysis was conducted to model the relationship between LST and the three spectral indices. Landsat 8 OLI/TIRS imagery of year 2019 of Dakahlia Governorate in Egypt was processed for LST retrieval as well as LULC classification. The ML classifier achieved an overall accuracy and kappa coefficient of 95.14% and 0.857, respectively, while of those based on spectral indices were 94.86% and 0.777, respectively. The results demonstrated an average temperature of 35.8°C, 31.2°C and 27.6°C for urban, vegetation and water, respectively. The LST statistics difference between classification methods of the three land covers was less 2°C. Based on the regression analysis, the NDVI and NDWI indicated a negative correlation with LST, while the NDBI indicated a positive correlation with LST. The polynomial regression analysis of LST against NDVI and NDWI demonstrated a better coefficient of determination (R2) than linear regression analysis of 0.341 and 0.305, respectively. For NDBI, linear and polynomial regression analysis demonstrated very close R2 of 0.624 and 0.628, respectively. The multiple regression analysis of LST against NDVI, NDBI and NDWI revealed R2 of 0.699. Consequently, the three spectral indices can be used as effective indicators for separating terrain into different classes, and hence relate their LST.  

Keywords: