GEOAI-BASED SPATIOTEMPORAL ASSESSMENT OF SURFACE URBAN HEAT ISLAND DYNAMICS AND LAND COVER CHANGE USING GOOGLE EARTH ENGINE: EVIDENCE FROM MANADO CITY, INDONESIA

Abstract

Rapid urbanization has significantly transformed land cover patterns and intensified the Surface Urban Heat Island (SUHI) phenomenon in many tropical cities, posing increasing environmental and public health challenges. However, comprehensive long-term assessments integrating land cover dynamics, land surface temperature (LST), and SUHI using cloud-based geospatial platforms remain limited, particularly in rapidly developing cities of eastern Indonesia. This study investigates the spatiotemporal relationship between land cover change and SUHI dynamics in Manado City, Indonesia, between 2000 and 2023 by integrating Landsat imagery, Google Earth Engine (GEE), and a GeoAI-based Random Forest classification approach. Landsat 7 ETM+ and Landsat 9 OLI/TIRS datasets were processed to generate land cover maps and retrieve LST, while SUHI intensity was derived from spatial thermal anomalies. Classification accuracy was evaluated using confusion matrices, achieving Overall Accuracy of 95.28% (Kappa = 0.915) for 2000 and 92.00% (Kappa = 0.870) for 2023. Results reveal a substantial decline in vegetation from 10.844,523 ha (75,69%) to 9.724,045 ha (67,87%), accompanied by an expansion of built-up land from 3.374,629 ha (23,55%)  to 3.939,957 ha (27,49%). These land cover transformations corresponded with an increase in average LST from 33.79°C to 37.08°C and a spatial expansion of high-intensity SUHI zones from the urban core toward developing suburban districts. Subdistrict-level analysis further shows considerable spatial heterogeneity in thermal conditions, with the greatest warming occurring in the most intensively developed areas. The results confirm that urban expansion and vegetation loss are the principal drivers of increasing SUHI intensity in Manado, supporting climate-responsive urban planning and sustainable land-use management