Assessing landscape suitability for reintroduction of American marten (Martes americana) using an agent-based model

Posted 02/4/13

In order to successfully reintroduce a species, one must both identify suitable habitat and integrate information on the ecology of that species. Post-doc Steve Wangen did this with his agent-based modeling for the American marten in Wisconsin. Recent reintroductions in Wisconsin have failed to establish. Steve's results suggest that competition and territorialism may restrict this species from full re-establishment but increasing connectivity between viable habitats may overcome this problem.


American marten (courtesy of US Fish and Wildlife Service)

The American marten (Martes americana) once thrived in Wisconsin, but is currently classified as a state-endangered species after expiration from the state due to loss of habitat and overexploitation from fur trapping. This species depends on large and contiguous stands of late successional pine and mixed forest habitat, a habitat which is currently rare in northern Wisconsin. To date, successful reintroduction of individuals has occurred in northern Wisconsin, but with limited population expansion or growth occurring. In order to establish viable populations through reintroduction efforts, an adequate assessment of priority marten habitat is needed. A successful reintroduction of the marten, an umbrella and indicator species, into Wisconsin would be a positive indication of habitat restoration to a pre-settlement landscape and would inevitably help to conserve other species. In order to evaluate the potential success of the reintroduction efforts of the marten into northern Wisconsin, Steve has developed an agent-based modeling approach. His approach goes beyond the traditional modeling methods, which typically only identify the amount and location of suitable habitat, by evaluating the functionality of the habitat through an assessment of individual interactions with the ecosystem. This individual-based approach to habitat suitability, with higher resolution interaction of individuals with habitat and other components of the ecosystem, produces a dynamic bottom-up approach to habitat modeling. The detail provided will better estimate the current success of, or identification of new, reintroduction sites. The model developed by Steve incorporates three components; the ecology of the species, resource availability, and the location of viable habitat across the landscape. This includes detailed information on the movement, foraging success, reproduction, and behaviorally driven interactions such as territorialism of martens in addition to other interactions including predation. Resource availability is represented as population fluctuations of voles, the species main prey source. These interactions are then modeled in varying simulations within the landscape using the National Land Cover Database (NLCD) across Northern Wisconsin in order to gain insight into the population level effects each factor has.

Individual time-step of simulations showing male marten territories defined by scent markings left behind as they navigate the landscape. Each color represents the territory of a different individual, while the black, white, and gray represent different cover type classifications from NLCD data.

Steve’s initial results suggest that, at present, marten habitat in northern Wisconsin is too fragmented and isolated to restore historic level population functionality, although remnant populations may persist. Although viable habitat is available, competition for resources and territorial exclusion may restrict full successful reintroduction across Northern Wisconsin. Steve’s modeling efforts suggest that increasing connectivity between viable habitats as the best solution to improving current reintroduction efforts, providing a clear management option for the conservation of this species. In addition, Steve will assess how landscape factors and different land-uses may affect marten populations. With change of land-use likely to occur, he will integrate future land-use change scenarios to determine those circumstances that would be the most detrimental to the long-term viability of this species within Wisconsin. He will also examine how segmentation analysis using NLCD and other data sources might better reflect how the spatial distribution of available habitat can be identified using remotely sensed data products. Ultimately, Steve would like to integrate this agent-based model with Game Play in the trails forward program ( .”

Story by Brooke Bateman