Impact of land cover classes on surface temperature in the vicinity of urban lakes and vegetation patches: A non-parametric regression analysis over decadal data

Rapid urban expansion drives significant land cover changes, which impacts land surface temperature (LST). Understanding the trends in LST is never more relevant in the warming world. In the present study, we obtained the land use and land cover (LULC) change information over the last two decades (2000-2020) and related it with the LST, especially in the vicinity of lakes and vegetation patches. Support Vector Regression (SVR) and Locally Estimated Scatter-plot Smoothing (LOESS) non-parametric regression analyses were fit to explain the LST with the predictor variables consisting of LULC ratios in the vicinity. Results infer that SVR and LOESS models fit the observed LST with the least errors at a correlation of 0.818 and 0.816 for lakes and 0.706 and 0.712 for vegetation patches, respectively. Further, Sobol sensitivity analysis reveals that vegetation cover in the lake buffer zones plays a significant role in LST in the vicinity. Whereas vegetation, barren, and agricultural cover affects LST in vegetation buffer zones as they modify albedo, evapotranspiration, and provide shade that affects LST. The findings provide a new dimension on the role of vegetation in lake buffer areas, which significantly contributes to lowering the LST. This emphasizes re-designing nature-based infrastructure and buffer areas to reap maximum cooling benefits in the warming world.