Extended Data Fig. 3: Edge effect measurement based on tree cover.

A spatial log-linear regression model, as described in the Methods section, was conducted per grid cell with tree canopy cover for year 2000 from Hansen et al.³⁰ as a dependent variable and log₁₀-transformed distance to the edge as an independent variable. We refer to estimate value of β₁ as \(\frac{\Delta Tree\,cover}{\Delta D}\), which is an abbreviated form of \(\frac{\Delta Tree\,canopy\,cover}{\Delta log_{10}(Distance\,to\,edge)}\). a, Map of \(\frac{\Delta Tree\,cover}{\Delta D}\) on tree cover. Each geographical grid cell with a scale of 100 km is shaded according to the\(\frac{\Delta Tree\,cover}{\Delta D}\). b, Map of the log₁₀-transformed coefficients of variation of \(\frac{\Delta Tree\,cover}{\Delta D}\). Each geographical grid cell with a scale of 100 km is shaded according to the log₁₀-transformed coefficient of variation.