Land Surface Temperature Patterns and Their Statistical Association with Land Use/Land Cover in Sundernagar Tehsil, Himachal Pradesh: A Multi-Date Remote Sensing Analysis

Source abstract

Land surface temperature (LST) is an important parameter for monitoring the thermal condition of the land surface, especially in rapidly urbanizing mountain areas. This study examines the spatial pattern of LST and its statistical association with land use/land cover (LULC) in Sundernagar Tehsil, Himachal Pradesh, India, using Landsat 8 OLI/TIRS Collection 2 Level 2 scenes of 23 March 2015, 4 March 2020 and 18 March 2025. The three scenes were acquired on different dates within March; therefore, each scene is treated as a separate cross-sectional observation, and differences in absolute LST among dates are not interpreted as long-term change. LST was taken from the Surface Temperature band (ST_B10) and summarised for five LULC classes: Protected Area, Agricultural Area, Build-up Area, Barren Land and Water Bodies. Zonal statistics gave the same thermal order in all three scenes, with Barren Land showing the highest mean LST (29.74 °C, 22.20 °C, 25.21 °C) and Water Bodies the lowest (24.05 °C, 18.18 °C, 20.72 °C). Neighboring pixels are spatially autocorrelated and cannot be treated as independent observations. Therefore, the pixel-level Pearson correlations between LST and the spectral indices (NDVI, NDWI, and NDBI) are interpreted as descriptive measures of association, while nominal p-values and confidence intervals are not treated as inferential statistics. On the full raster, NDBI had positive correlations with LST (r = 0.3671, 0.2406, and 0.3154), while NDWI had negative correlations (r = −0.2790, −0.2988, and −0.2801). Across the three spectral indices, the unadjusted relationships explained approximately 2–13% of LST variance. NDVI showed weak positive unadjusted correlations with LST (r = 0.1315–0.1879), but its regression coefficient became negative in all three scenes after adjustment for elevation, slope, and aspect. Adding terrain variables to the LULC models increased R2 from 0.148–0.315 to 0.598–0.632, indicating that interpretation of land cover thermal associations was sensitive to terrain adjustment. The results show a consistent association between land cover and surface temperature within each date and can support land use planning in Himalayan towns, but they do not show a causal effect or a long-term trend.