The "Den Mother" marten habitat quality models were created to provide insight into American marten habitat selection behavior and to promote the recovery of the Newfoundland marten (Martes americana atrata) population. Although these objectives are typical of most wildlife habitat modeling projects, the marten's idiosyncratic habitat ecology and apparently intractable conflicts associated with timber harvesting motivated a unique, process-oriented approach to appraising landscapes. The Den Mother models used optimal decision-making principles to synthesize critical resources (den sites and foraging opportunities) and constraints (adverse thermal situations and exposure to predations) into a single hierarchical framework. The resulting spatially explicit, combinatorial optimization models depend on a complex array of interacting assumptions. However, in mechanistic models, explicit assumptions provide the means by which insights are gained. For example, manipulating prey population parameters provided a clear demonstration of how resource conditions confound the relationship between landscape configuration and marten fitness, thereby challenging conventional definitions of habitat based on vegetation alone. Likewise, the models' sensitivity to spatial circumstances argued against the concept of an "optimal landscape," a traditional objective for wildlife habitat analyses. Although the model analyses did not refute the conventional wisdom that marten are strongly associated with (and may depend on) large contiguous blocks of senescing and defoliated forests, they did suggest that the marten is an opening-sensitive, rather than coresensitive, species. The models also suggested new avenues for research addressing marten den site selection, predator avoidance behavior, foraging efficiency, and space use strategies, as well as new techniques for assessing the trade-offs that govern marten habitat selection behavior. Finally, the models also suggested new guidelines for promoting marten recovery in an adaptive management context, including recommendations for placing artificial resting structures; creating favorable landscape mosaics; managing ephemeral resources such as transition old-growth forests, defoliation, and coarse woody debris; and developing alternative, competing management scenarios that address both forest and prey conditions simultaneously.

Habitat fragmentation occurs when large tracts of an orginal habitat are replaced by smaller patches of two or more habitat types, largely through human activities. I studied the behavior of six measures of landscape pattern that seemed appropriate for quantifying fragmentation, and used these measures to investigate the effects of forest fragmentation on American martens (Martes americana) and their prey. The measures I selected were edge density, contagion, mean nearest neighbor distance between patches, mean proximity index, perimeter-area fractal dimension, and mass fractal dimension. To test the behavior of these measures with a variety of landscape patterns, I used a computer program to create nine series of increasingly fragmented landscapes that differed in the size and shape of patches, and in the way fragmentation was allowed to increase. Patch size changed the range of attainable values for all measures examined, and patch shape affected all measures except nearest neighbor distance and mean proximity index. The method in which fragmentation increased within each landscape series also affected all measures. None of the measures was able to differentiate between different spatial distributions of patches. To investigate the effects of forest fragmentation on martens and their prey, I selected 18 areas of mature forest habitat in Utah that differed in the amount of landscape heterogeneity due to natural openings and timber clearcuts. I conducted a live-trap survey of martens within each site over three summers from 1991-1993, and a 7-week snap-trap survey of small mammals within 12 of the sites in 1992. Martens were negatively correlated with increasing fragmentation, and mean proximity index was the strongest correlate with reductions in marten captures across sites (x2= 9.48, df= 1, P = 0.04). Capture rates of red-backed voles (Clethrionomys gapperi) also declined with increasing fragmentation (x2 = 4.66, df = 1, P = 0.03), while deer mice (Peromyscus maniculatus) capture rates increased (x2 = 6.12, df= 1, P = 0.01). Martens and voles both appeared sensitive to landscape pattern, with low numbers in areas having large, closely spaced patches of unforested habitat.