Spatio-Temporal Assessment Of Urban Heat Islands And Future Prediction: A Case Study Of Kathmandu Valley Using Optical Remote Sensing

Abstract

Kathmandu Valley has experienced among the most rapid and concentrated urban expansion in South Asia over the past two decades, with direct but incompletely characterised consequences for surface thermal conditions. This study assesses the spatio-temporal evolution of Urban Heat Islands (UHI) across the valley from 2008 to 2023 using multi-temporal Landsat imagery processed in Google Earth Engine (GEE), and projects land surface conditions to 2028. Land Use Land Cover (LULC) was mapped using a Random Forest (RF) classifier at five-year intervals; Land Surface Temperature (LST) was retrieved from thermal bands via an emissivity-corrected inversion. The Normalised Difference Vegetation Index (NDVI) and Normalised Difference Built-up Index (NDBI) were used to quantify vegetation cover and impervious surface intensity and their relationship with LST. A CA-Markov model was used for LULC projection to 2028, followed by RF Regression to forecast LST under projected land-cover conditions. Pearson's correlation analysis yielded r = +0.751 for LST–NDBI and r = −0.759 for LST–NDVI, both statistically significant. Built-up area expanded from 8,720 ha (9.32%) in 2008 to 18,525 ha (19.82%) by 2023, with mean LST peaking at 27.34°C in 2018. UHI intensity was highest in the central built-up core (class mean 1.00–1.19) and negative over forested hillsides (−0.91 to −0.83). By 2028 built-up cover is projected to reach 25.47% of the valley, with mean LST rising to approximately 26.88°C and the central thermal hotspot expanding further. These findings demonstrate that vegetation management and targeted greening in high-density wards can partially counteract UHI intensification, providing quantitative evidence to support climate-resilient land-use planning in the Kathmandu Valley.