Landscape Fragmentation and Biodiversity Loss Modeling

Abstract

Landscape fragmentation -- the division of continuous habitat into smaller, isolated patches separated by inhospitable matrix -- is the leading proximate driver of biodiversity loss worldwide, operating through direct habitat area reduction (species-area relationship), increased edge effects, reduced patch connectivity, and extinction debt dynamics in which species committed to extinction by past fragmentation events have not yet disappeared. This study developed and validated a multi-scale Fragmentation-Biodiversity Loss Model (FBLM) combining species-area relationship (SAR) theory, landscape connectivity metrics, and extinction debt modelling across 42 European agricultural landscapes (2,840 km2 total; 284 focal patches of semi-natural habitat; Italy, Sweden, and Germany; 2018-2022), using field-validated species richness data for birds (n = 84 species surveyed), plants (n = 284 species), and insects (n = 124 butterfly species) combined with LiDAR-derived habitat structural metrics and Copernicus Land Cover time-series (1990-2022). FBLM correctly predicted observed patch-level species richness with r = +0.84 (all taxa pooled; p < 0.001) and r = +0.78-0.88 for individual taxa. The extinction debt component contributed an additional 14.4% explained variance in richness beyond contemporary habitat area and connectivity -- confirming that current biodiversity cannot be predicted without modelling historical fragmentation trajectories. A Landscape Fragmentation Impact Score (LFIS) integrating patch area, isolation, matrix permeability, and extinction debt predicted local species extinction probability with AUC = 0.864. Connectivity restoration (creating stepping-stone habitat patches or permeable matrix corridors) in LFIS-priority landscapes showed 2.84-fold higher predicted biodiversity recovery than equivalent area restoration in already-connected landscapes.