Fig. 2: Relationship between the temperature gradient from edge to interior and macroclimatic Tsurf;.
a Full dataset including all sampled locations (n = 39,455). b Subset of locations inside the forest and where cooling increased towards the forest interior (n = 28,141). Each point represents the strength of the temperature edge effect in one satellite scene, calculated as the change in Tsurf (∆T) with log-transformed distance to forest edge (Methods). Points are colour-coded by density (blue = high density, brown = low density), with larger points indicating greater weight in the inverse variance-weighted quadratic regression. The black curve represents the fitted regression line, with shaded areas denoting confidence intervals. In both panels, the positive quadratic trend indicates a stronger cooling effect from edge to interior under higher macroclimatic temperatures. R² values indicate that macroclimatic temperature explains 19.7% of the variation in (a) and 16.3% in (b). All regression coefficients (intercept, Tsurf, and quadratic term) are statistically significant (p < 0.001).For seasonal analyses, see Supplementary Fig. 3; for analyses using MODIS-derived Tsurf and annual mean temperature, see Supplementary Figs. 4 and 5.
