Assessing urban heat island patterns in Bharatpur metropolitan city of Nepal using multi-temporal remote sensing and machine learning

Abstract

Rapid urbanization in Nepal’s Terai is driving Land Use and Land Cover (LULC) change and intensifying the Urban Heat Island (UHI) effect. This study quantifies UHI dynamics in Bharatpur Metropolitan City (BMC), Nepal, across six reference years from 2000 to 2024 using multi-temporal Landsat data in Google Earth Engine (GEE). LULC was classified using Random Forest (RF), achieving Overall Accuracy of 89.00–95.50% and Kappa coefficients of 0.853–0.940. Land Surface Temperature (LST) was derived from thermal infrared bands and validated against Department of Hydrology and Meteorology (DHM) ground stations; UHI intensity was computed by pixel-wise normalization; and the Normalized Difference Vegetation Index (NDVI), Normalized Difference Built-up Index (NDBI), and Normalized Difference Moisture Index (NDMI) were correlated with LST and UHI by LULC class using a stratified sample of 800 pixels per class. A Cellular Automata-Markov model with an RF-derived suitability surface projected 2030 LULC (Figure of Merit = 0.302); RF regression predicted the standardized 2030 thermal field (R2 = 0.578, RMSE = 2.07 °C); and 2030 UHI was derived via Urban Thermal Field Variance Index (UTFVI) zoning and Getis-Ord Gi* hotspot analysis. Built-up area expanded from 21.63% in 2000 to 45.11% in 2024 (+ 23.48 percentage points), while mean LST ranged from 30.46 °C to 37.72 °C and peaked at 57.75 °C in 2010. NDMI was the dominant thermal correlate (r = − 0.84, p < 0.001), significantly exceeding NDVI in built-up (r = − 0.64) and vegetated (r = − 0.46) areas, while NDBI correlated most strongly with warming (r = +0.81). By 2030, built-up area is projected to reach 56.21% (+ 11.10 percentage points from 2024), and the statistically significant heat-hotspot area is projected to expand by 12.63 percentage points (36.69% to 49.32%), even as peak-to-baseline UHI intensity narrows from +3.41 °C to +1.62 °C, indicating spatial expansion and thermal homogenization of the heat island rather than simple intensification. The UTFVI-classified heat-affected area is projected to grow from 38.84% to 48.95% of the study area over the same period. These quantitative findings provide an evidence base for urban planning in Bharatpur and comparable fast-urbanizing South Asian Terai cities.