Landscape Connectivity Modeling for Conservation Planning

Abstract

Landscape connectivity modelling -- the quantitative assessment of the degree to which landscape structure facilitates or
impedes animal movement between habitat patches -- is an essential tool in conservation planning. This study
benchmarked six connectivity models -- least-cost path (LCP), Circuitscape (CS), UNICOR, Linkage Mapper,
Omniscape, and graph-theoretic betweenness centrality (GBC) -- against empirical GPS telemetry movement data and
population genetic FST from 12 species across Iberian Peninsula landscapes (284 GPS-tracked individuals; 48
population pairs; 28,484 SNPs per species). Circuitscape with empirically calibrated resistance surfaces achieved the
highest corridor prediction overlap (CPO = 0.74 +- 0.08) and IBR correlation (r(S) = +0.84 +- 0.06). Omniscape
performed comparably (CPO = 0.72; r(S) = +0.82) while producing spatially continuous maps better suited to land-use
planning. LCP performed lowest (CPO = 0.48; r(S) = +0.54). Step selection function (SSF) resistance surface calibration
improved all model performance by 18.4-24.4% over expert-estimated generic surfaces. These results confirm
Circuitscape as the optimal tool for corridor identification, Omniscape for planning without patch data, and demonstrate
that empirical resistance calibration is the highest-return investment for connectivity model improvement.