Multiple linear and non-linear regression analyses of various soil and terrain indices with regard to their efficiency in the determination of temporal changes in LST values within Trichy district of Tamil Nadu, India.

Abstract

Land surface temperature (LST) of the terrain is the result of thermal energy received from the sun as the source. The study site was Trichy district of Tamil Nadu state in India wherein the LST values pertaining to all the belonging pixels have been calculated using generalized split window (GSW) algorithm from the Landsat-8 satellite images on two different dates with a gap of 5years between them. Along with the LST calculation on these 2 dates, corresponding NDWI, NDVI, NDBI, NDMI, BSI, DBI, DBSI, LSE, and albedo values have also been calculated. Then, changes in the dependent variable (LST) and in the each of the respective explanatory or independent variables have been undergone through the process of multiple linear and non-linear regression analysis to find out the best-fitting model set of explanatory variables that best describes the variation in LST values at a location. In this research, it has been found that a non-linear model set comprising five independent variables like change in BSI, LSE, NDWI, NDBI, and albedo seem to be the best-fitting model for predicting the variations in the dependent variable to the maximum degree and also with acceptable redundancy. Also, it has been observed that changes in LSE or albedo values at any place within the study-site could play the most significant role in having an influence in the LST value change that could be recorded in the same over any given time period, thus proving the vastness of the effect that could be made by an increase in urbanization or structural development on the extent of positive change in LST.