Extended Data Fig. 4: SHAP-based interpretation of environmental drivers of edge effects across biomes.

SHAP summary plots and dependence plots showing the contribution of each environmental variable to the estimation of edge effect (\(\frac{\Delta AGB}{\Delta D}\)) in tropical (a–d), temperate (e–h), and boreal forests (i–l). For each biome, panels show the SHAP summary plot (a, e, i) and dependence plots for the top three predictors (b–d for tropical, f–h for temperate, and j–l for boreal). Some variables were excluded due to collinearity: in tropical forests, soil moisture and elevation were excluded due to correlation with MAP and MAT, respectively; in temperate forests, elevation was excluded due to collinearity with slope; and in boreal forests, agriculture and elevation were excluded due to collinearity with MAT and slope, respectively.