A variety of landscape models conceptualize and interpret human impacts on ecosystems and their biodiversity. The simplest, a 'patch-matrix' model, is rooted in Island Biogeography Theory and assumes a dichotomy between human-defined habitat patches and a surrounding matrix that is completely inhospitable. This dichotomy between patch and matrix habitats has been recently relaxed, with the 'continuum' model taking this relaxation to its extreme and logical endpoint - a species-based model with no a priori definition of habitat or matrix. Yet, because few empirical comparisons exist of these bookending conceptual models, we lack understanding of their relative utility or the merits of hybrid approaches that combine attributes of patch-matrix and continuum models. To guide such considerations, we empirically compared how continuum, patch-matrix, and hybrid models explain beetle distributions across two contrasting fragmented landscapes. We predicted that patch-matrix and hybrid models would best explain species abundances within the Hope River Forest Fragmentation Project (New Zealand), owing to strong landscape contrast and distinct (“hard”) structural edges between forest fragments and grassland. Conversely, we predicted that hybrid and continuum models would best explain species abundances within the Wog Wog Habitat Fragmentation Experiment (Australia) owing to weak landscape contrast and soft structural edges between Eucalyptus and pine plantation forests.
Our results largely support our predictions and illustrate the differing merits of patch-matrix, continuum, and hybrid models across contrasting landscapes and species. Within the Hope River Forest Fragmentation Project, we found support for patch-matrix and hybrid models, with models varying in their merits among beetle species that differed in trophic level and habitat specificity. Conversely, within the Wog Wog Habitat Fragmentation Experiment, we found comparable support for continuum and hybrid models, and no support for patch-matrix models, among beetle species, regardless of trophic level and habitat specificity. Together, these results point to the broad utility of hybrid models (10 of 12 species) and the more limited utility of patch-matrix models (2 of 12 species), for explaining patterns of species abundance. Yet, considering the details of when particular models perform best may be contingent on attributes of the landscape and focal species. We conclude by considering new research needed to advance the application of models to understand species responses to land-use change.