Graph-theoretic modeling of functional habitat connectivity for lynx on the Okanogan Highlands, northern Washington
Jones, Aaron Paul
MetadataShow full item record
Maintenance of habitat connectivity over various geographic scales is a conservation imperative for the long-term persistence of many species. Functional linkages for a given focal species may not be readily apparent or spatially coincident with explicit structural patterns in a landscape. In species-level habitat connectivity modeling, therefore, techniques derived from empirical data are needed to sustain a species-centric perspective on landscape mosaics. This study demonstrates an empirically-based, combined Bayesian and graph-theoretic approach to modeling functional habitat connectivity. The case study focused on a subpopulation of Canadian lynx resident to a subalpine plateau in north-central Washington. Using data collected during two prior radiotelemetry studies, habitat selection by individual lynx was analyzed at two spatial scales using a combination of compositional analysis, chi-squared tests, and Neu's method. Significant habitat associations were used to create probability models for lynx presence using Bayes theorem. These provided a composite index of habitat suitability and enabled data-driven definition of habitat patches and an impedance surface for lynx movement in the study landscape. To address gradual variation in habitat quality, multiple study-wide and within-home-range patch surfaces were derived in accord with minimum percentages of areas of known lynx presence. Patches and cost-weighted interpatch distances were then converted to lattice data structures (graphs) and used as frameworks for hierarchical analyses of functional connectivity. Mapped probabilities of lynx presence indicated higher quality habitat in the northward interior of the study area and in several drainages along its northern periphery. Areas most significant to study-wide connectivity were identified in terms of core linkages and those along which the habitat network is most susceptible to disruption. A general divergence between these areas suggests that the former, "parsimoniously-connected" cores of lynx habitat may also be the areas most resilient to fragmentation. Places where these areas co-occur, however, are "hotspots" supported by dual justifications for conservation prioritization. In measuring global connectivity, three indices were relatively insensitive to changes in patch surface definition until the graph based on seven percent areas of known presence was reached, after which point fragmentation was marked. A critique of the Bayesian graph-theoretic modeling approach concludes the study, with emphasis given to its applicability for identifying subregional habitat linkages for far-ranging carnivores.