Forests at the urban–mountain interface moderate local climate, yet the spatial coupling between thermal regimes and vegetation dynamics is not well quantified. Focusing on Namyangju (Republic of Korea), we analyzed eight cloud-free Landsat-8/9 scenes from 2023 over forested pixels (30 m). Land surface temperature (LST) was retrieved from Lev-el-1 thermal Band 10, and NDVI from Bands 4–5. For each pixel we computed annual co-efficients of variation (CV) for LST and NDVI, then applied Getis–Ord Gi* to derive Z-scores (GiZ) that identify statistically significant clusters of intra-annual variability. We compared two formulations of the LST–NDVI linkage across ~308,000 grid points: regres-sion of variability magnitudes (CV–CV) and regression of local clustering intensities (GiZ–GiZ). Both relationships were positive, but GiZ–GiZ exhibited a tighter fit (R² = 0.623) than CV–CV (R² = 0.464), indicating stronger co-location of thermal and greenness clusters than raw variability alone suggests. Hotspots for both fields coincided with upland ridges and dissected slopes, whereas coldspots concentrated in valleys, highlighting terrain’s first-order control on microclimate and phenology. Methodologically, coupling time-series CVs with Gi* and simple OLS on GiZScores provides a transparent, spatially informed indicator of thermal–greenness coupling. Practically, GiZ-based belts offer actionable tar-gets for monitoring and management in urban-adjacent forests.