Naufal Aryansyah Erlangga, Halmar Halide
The Surface Urban Heat Island (SUHI) phenomenon in Gowa Regency was analyzed based on the influence of urban vegetation density (NDVIu), rural vegetation density (NDVIr), and global climate variability represented by ENSO during the 2016–2025 period using 120 monthly observations and the Leave-One-Out Cross Validation (LOOCV) approach. The results indicate that land surface temperature (LST) in the urban area (Somba Opu) ranged from approximately 31–40°C, while the rural area (Parangloe) ranged from 27–36°C, producing an average temperature difference of about 2–5°C and monthly SUHI values fluctuating from near 0°C to above 10°C during extreme periods. Rural NDVI values were generally higher (0.60–0.78) than urban NDVI (0.40–0.70), confirming the role of vegetation in reducing surface temperature. Model evaluation shows that the Naive model was not suitable (R = -1; RMSE = 2.2644), Multiple Linear Regression produced a weak correlation (R = 0.2722; RMSE = 2.1710), while Random Forest achieved the best performance (R = 0.3235; RMSE = 2.1388), although the relationship strength remains weak to moderate. Predictor importance analysis indicates that in linear regression, rural NDVI had the highest contribution (normalized importance = 1), followed by urban NDVI (0.9461), whereas in Random Forest, ENSO became the most dominant variable (1), followed by rural NDVI (0.33875), suggesting that SUHI dynamics are influenced by a combination of local and global factors with predominantly non-linear relationships.
Keywords: Surface Urban Heat Island (SUHI); NDVI; Land Surface Temperature (LST); Multiple Linear Regression; Random Forest
