Configuration coefficients can mislead conservation inference from beta diversity in fragmented landscapes

Juan Andrés Martínez Lanfranco 

Beta diversity informs reserve design and connectivity planning because compositional turnover links local assemblages to regional diversity. However, the conservation meaning of a beta-diversity pattern depends on the mechanism that generates it, not on the sign of a landscape-configuration coefficient alone. I use a recent Atlantic Forest tree beta-diversity analysis (Kramer et al. 2026, Global Change Biology, 32, e70855) as a worked case to show how a negative patch-number coefficient on total beta diversity can document a real landscape association without identifying fragmentation per se. In the dataset, habitat cover and patch number pass conventional collinearity screening but retain predictor geometry that limits same-amount, different-configuration contrasts. Patch number has a near-zero marginal correlation with total beta diversity yet becomes significant in the joint model, consistent with marginal-to-conditional amplification within a shared amount–configuration gradient. Under the alpha-corrected Raup-Crick response, patch number loses unique conditional support whereas spatial distance remains supported. Independent extraction across five operational extents further shows that explicit isolation, edge-exposure, interior-condition, and arrangement metrics do not provide convergent conditional support. The patch-number coefficient therefore documents a response- and metric-specific subdivision association, not an identified fragmentation-per-se mechanism. Conservation prescriptions such as several-small reserve design or corridor investment require mechanism-matched inference, response-estimand alignment, predictor separability, and biologically justified scale, not habitat-amount control alone.