Spatial Ecology of Re-introduced American Martens in the Northern Lower Peninsula of Michigan.
American martens (Martes Americana; hereafter, martens) were extirpated from Michigan's Lower Peninsula by 1911 due to habitat loss and over harvest (Earle et al, 2001). During 1985-1986 the Michigan Department of Natural Resources (MDNR) and U.S. Forest Service (USFS) re-introduced 49 martens (25 M and 24 F) in Pigeon River Country State Forest and 36 martens (20 M and 16 F) in Manistee National Forest of the northern Lower Peninsula of Michigan (NLP; Williams et al., 2007; Fig. 1). Re- introduction sites were located >120 km apart and have been genetically isolated since re-introduction (Bicker, 2007). Martens are not trapped in the NLP, although they are trapped in the Upper Peninsula of Michigan. Martens are listed as a Regional Forester's Sensitive species for Region 9 USFS lands in Michigan and Wisconsin and Management Indicator Species on Ottawa National Forest lands (USDA, 2006). The NLP population of martens is the southern-most distributed marten population in eastern North America, relative to their historic range (Williams et al., 2007). This population offers unique perspective of a re-introduced and unharvested population that lacks fishers (Pekania pennanti) as a potential competitor. It also provides useful insight into challenges of maintaining a fragmentation-sensitive carnivore in a landscape that contains private holdings, two federally-managed national forests and >800,000 ha of state-managed forest (MDNR, 2013).
Martens have historically been associated with late-seral coniferous forests (e.g., Buskirk and Powell, 1994). However, studies in the eastern U.S. and Canada and northern Great Lakes Region suggest martens select deciduous and mixed-wood forests in late-seral stages (Chapin et al., 1997; Potvin et al, 2000; Dumyahn et al, 2007). Within-stand structural characteristics such as complex vertical and horizontal components are important and may supersede age or dominant cover type species (Chapin et al., 1997; Potvin et al., 2000). Regardless of dominant cover type, martens are sensitive to habitat fragmentation (Hargis et al, 1999) and typically avoid open habitat gaps and open regenerating forest stands (Potvin et al, 2000). Martens require large patches of habitat and establish home ranges in landscapes comprised of [greater than or equal to] 70% habitat (Bissonette et al, 1997; Chapin et al., 1998; Hargis et al., 1999; Potvin et al, 2000; Dumyahn et al., 2007). Further, spatial configuration of selected cover types influences establishment patterns of martens (Fuller and Harrison, 2005). Small areal cover types embedded within a dominant and dissimilar cover type (i.e., inclusions) are also important components of marten habitat (Cheveau et al., 2013; McCann et al, 2014).
Spatially-explicit habitat models have become a valuable tool that wildlife managers can use to understand species' habitat distribution. Martens are a good candidate species for landscape-scale habitat models due to their sensitivity to landscape fragmentation (Bissonette et al, 1997). Martens also are rare and have a patchy distribution, making them difficult to detect and monitor. Kirk and Zielinski (2009) developed a landscape-based model for martens in California. Their model did not perform well with a seasonal validation data set; however, they did find martens selected habitat based on broad-scale landscape features (Kirk and Zielinski, 2009). We used a hierarchical approach to depict a multi-scale framework of cover type (hereafter, habitat) selection decisions (Bissonette and Broekhuizen, 1995) to best inform management (Lindenmayer, 2000). During April 2005 to July 2006, we conducted a radiotelemetry study on a re-introduced population of martens in the NLP. We used these data to develop a landscape-scale model depicting areas most suitable for martens by comparing habitat of marten core-area home ranges to potentially available habitat. Our objective was to develop a spatial model to identify marten habitat based on estimates of marten core area requirements and assessed marten habitat and landscape requirements. Subsequently, we used an independent data set to validate this spatial model to identify potential marten habitat.
We defined two study areas near the original marten re-introduction sites in Pigeon River Country State Forest (45[degrees]10'N, 84[degrees]30'W; PRSF) and Manistee National Forest (43[degrees]51'N, 85[degrees]57'W; MNF; Fig. 1). The PRSF study site (1077 [km.sup.2]) was located in Antrim, Charlevoix! Cheboygan, Emmet, Montmorency, and Otsego counties. The area was comprised of upland deciduous forest (30%), upland coniferous forest (19%), upland mixed forest (16%), lowland forest (18%), agriculture/openland (10%), wetland/water (6%), and urban developement (1%). Major upland tree species included red maple (Acer rubrum), sugar maple (A. saccharum), red oak (Quercus rubra), white oak (Q. alba), basswood (Tilia americana), beech (Fagus grandifolia), white spruce (Picea glauca), and white pine (Pinus strobus; McFadden, 2007; Buchanan, 2008). The MNF site was located in the western portion of the NLP, 129 km southwest of PRSF, and was separated from the latter by a landscape fragmented by private landholdings, agriculture, and transportation networks. The MNF study area was comprised of 537 [km.sup.2] of forest within Lake, Wexford, and Manistee counties. The area consisted of upland deciduous forest (32%), upland coniferous forest (25%), upland mixed forest (25%), lowland forest (8%), agriculture/openland (5%), wetland/ water (4%), and urban development (1%). Major tree species included red maple, sugar maple, quaking aspen (Populus tremloides), big tooth aspen (P. grandidentata), red oak, white oak, black cherry (Prunus serotina), white cedar (Thuja ocddentalis), and hemlock (Tsuga canadensis-, Harden, 1998). The MNF also contained red pine (Pinus resinosa) and jack pine (Pinus banksiana) plantations with inclusion components of mature (i.e., >80 y) red and white oak (Buchanan, 2008). We defined each study area as the area in which we live-trapped martens, plus a 2.5 km buffer surrounding outermost locations of traps as a maximal measure of the radius of marten home ranges based on our radiotelemetry data (McFadden, 2007; reviewed in Fagerstone, 1987; Fig. 1).
TRAPPING AND RADIOTELEMETRY
We used model 108 Tomahawk cage traps (Tomahawk Live Trap, Tomahawk, Wis.) to trap martens during January through mid-March 2005 and 2006. Captured animals were restrained with trap combs (Thomasma and Peterson, 1998) and immobilized with an intramuscular injection of 25 mg/kg ketamine hydrochloride (HCl) plus 5 mg/kg xylazine HCl (Belant, 1992). Age (juv: <1 y; ad: >1 y), sex, and weight were determined for each marten. Age class was based on plaque build-up on teeth and weight (Poole et al., 1994). We gave each marten uniquely numbered ear tags and affixed a radiotransmitter with a mortality sensor (MOD-080, Telonics Inc., Mesa, Ariz.). After handling we allowed each marten to recover inside a trap lined with wool blankets and covered for insulation. We remained at the release site until each animal fully recovered and dispersed from the trap. Our research was conducted under a trapping and handling permit from the Michigan Department of Natural Resources (MDNR; no. SC 1172) and the Institutional Animal Care and Use Committee at Central Michigan University (no. 27-04).
During April -July each year, radiocollared martens were located using a truck mounted four element Yagi directional antenna, electronic compass (Lovallo et al, 1994) and triangulation with [greater than or equal to] 3 bearings (White and Garrott, 1990). We chose April -July to represent marten habitat selection during the nonbreeding period and assumed both sexes would be using resources similarly (Sandell, 1989). Telemetry bearing error (2.5[degrees]) was determined by taking bearings to 11 reference transmitters placed at known locations within the study area. Bearings for marten location were collected within 20 min to reduce error related to animal movement. Locations were collected daily and at randomly assigned times over a 24 h period to sample representative habitat and marten home-range use. We estimated locations and error polygons using the maximum likelihood estimator (Lenth, 1981) in Locate III (Nams, 2005).
We estimated 95% contour (hereafter, 95% home range) and 50% contour (hereafter, core area) home ranges for radiocollared adult martens using the fixed kernel home-range estimator (Worton, 1989) and Animal Movement Extension in ArcView (Hooge and Eichenlaub, 1997). The 95% home ranges were used to estimate the percentage of marten habitat within home ranges compared to the landscape. We modeled marten movements at the scale of core areas given they are estimated more reliably than home ranges (Seaman et al., 1999). Core areas were also used preferentially given martens showed little intersexual or intrasexual overlap and core-area size did not differ between sexes or between study sites (McFadden, 2007). We performed incremental analysis for each marten by plotting locations against home-range area to determine minimum number of locations required to reach a stable home range (Kenward, 2001). We excluded individuals with insufficient number of locations to reach home-range size stabilization.
ESTIMATING PATTERNS OF HABITAT USE
We used 2001 IFMAP/GAP Lower Peninsula Land Cover land-cover grid data with 30 m resolution to describe available habitat. These data had overall omission and commission accuracy of 81 % (Space Imaging, 2004). Land-cover data were reclassified using ArcMap 9.0 (ESRI, Redlands, Calif.) from 34 original land-cover types into seven unique cover-type classes: upland deciduous forest, upland coniferous forest, upland mixed forest, lowland forest, agriculture/openland, urban, and wetland/water (Table 1).
We estimated marten habitat use as proportion of used habitat at the home-range level compared to total available habitat in the study area landscapes (2nd order selection, sensu Johnson, 1980). We also classified marten radio locations into used classes of cover types for comparison to available habitat in the home range (3rd order selection, sensu Johnson, 1980). Instead of using the coordinate point of radio locations to define habitat used by martens, we placed a 95% telemetry error ellipse (as calculated in Locate III) around each radio location. We overlaid error ellipses onto land-cover data and determined each habitat type used as a percentage of total error ellipse area to obtain proportion of used marten habitat (i.e., proportion of one location) within each ellipse. Because a majority of error ellipses encircled multiple habitat types in which the location could actually be found, this method offered a less biased approach to assessing habitat associated with each location. Analyses of both levels of selection were completed for 95% home ranges and core areas. We used compositional analysis (Aebischer et al, 1993) in SAS (Ott and Hovey, 1997) to analyze patterns of habitat use, with upland deciduous forest as a reference category. Null values were assigned 0.0001 (Aebischer et al., 1993).
Based on our estimates of marten core areas and use of habitats, we developed a spatial model of potential marten habitat across the NLP. We created a grid theme covering the NLP consisting of 29,807 nonoverlapping hexagons the size of a marten's average core area (1.6 [km.sup.2]). The hexagon grid was overlaid onto the reclassified landcover theme and Spatial Statistics by Regions in Patch Analyst Grid 3.0 (Rempel, 2003) was used to quantify habitat covariates within each hexagon. A total of seven habitat covariates were identified to use in modeling marten habitat throughout the NLP based on biological importance to martens. Based on our analysis of habitat use by martens in our study area, we identified the following five habitat covariates: proportion of upland deciduous forest; proportion of upland coniferous forest; proportion of upland mixed forest; proportion of wetlands; and proportion of urban. These covariates included habitats selected and avoided by martens. We quantified two other habitat covariates, number of habitat (upland deciduous, upland conifer, and upland mixed forest) patches and area-weighted mean habitat patch fractal dimension given martens are sensitive to habitat fragmentation (Chapin et al., 1998; Potvin et al., 2000). We found no multi-collinearity among habitat covariates based on a Spearman rank correlation analysis and r [greater than or equal to] 0.50. To approximate normality, the proportional habitat variables were transformed using an arcsine of square root transformation and [log.sub.10]-transformation for number of patches and patch fractal dimension.
The Penrose distance statistic (Nielsen and Woolf, 2002; Manly, 2005; Preuss and Gehring, 2007) was used to measure similarity between marten core areas and the remainder of hexagons in the NLP. We calculated a mean marten vector as mean values of seven habitat covariates within marten core areas. We compared each hexagon in the NLP to the mean marten vector by calculating Penrose distance. A Penrose distance of zero indicated a hexagon with habitat most similar to marten core areas, whereas greater values were more dissimilar to marten core areas. We classified each hexagon into three Penrose distance classes (PDC). PDC 1 was defined as the lowest Penrose distance to an upper value equal to the 95% confidence interval value for marten-occupied hexagons (i.e., this class included Penrose distances for >95% of all marten core areas). PDC 2 was set with an upper limit as the maximal value of Penrose distance for marten-occupied hexagons. Therefore, PDC 1 corresponded to the range of habitat conditions used by most martens, whereas PDC 2 identified less suitable marten habitat. PDC 3 was defined as hexagons with Penrose distance values greater than values for marten-occupied hexagons (i.e., not suitable marten habitat).
All calculations were performed using a spreadsheet and the final output values were appended to the hexagon grid for display. PDC 1 estimated primary marten habitat throughout the NLP. Spatial Statistics in ArcMap was used to measure number of PDC 1 patches, area, and connectivity. A frequency distribution of PDC 1 patch area determined if marten habitat was highly fragmented. The estimated average distance to the nearest neighboring PDC 1 habitat patch measured the isolation of marten habitat. Patch connectivity and potential barriers for martens dispersing were estimated by measuring number of gaps between habitat patches along a Euclidean-distance path between PRSF and MNF.
Independent (nonbreeding season) track surveys were performed on traversable roads in MNF and PRSF during winter 2001-2007 using 4x4 truck, ATV, or snowmobile. Track surveys were conducted in coordination with existing state, federal, and tribal track surveys in our area and traversed all roads and forest trails within the entire extent of the study areas. These routes also remained the same during the sampling period. Track surveys were performed at least 24-48 h after a snow or wind event and marten track locations were recorded in real time with a GPS receiver. For each survey we plotted track locations on the Penrose habitat model in ArcGIS and calculated frequency distributions of percentage of track locations that occurred in each PDC. Over the time-series of track surveys, we counted a recorded track only once in a PDC hexagon, without replacement on an annual basis. A chi-square test was performed to determine if track occurrences were recorded more or less than expected within each PDC (Zar, 2010). The expected number of marten tracks was based on proportional length of forest trail or road surveyed within each PDC. All statistical analyses were performed using SAS software (SAS Institute, Inc., Caiy, N.C.) and [alpha] = 0.05.
We captured a total of 40 martens (25 M and 15 F), including 22 individuals in PRSF (13 M and 9 F) and 18 individuals in MNF (12 M and 6 F). We obtained 949 total locations from April 2005 to July 2006. We documented five marten mortalities and seven individuals had transmitters that expired before adequate number of locations was obtained (McFadden, 2007). Based on incremental analysis, we were able to use 28 adult martens (18 M and 10 F) for estimating home-range size and habitat use patterns. Number of locations for each marten averaged 26 (SE = 2). The 95% error ellipses had an average error of 0.07 [km.sup.2] (SE = 0.04). Average male 95% home range and core area was 11.2 [km.sup.2] (se= 1.4 [km.sup.2]) and 1.7 [km.sup.2] (SE = 0.3 [km.sup.2]), respectively. Average female 95% home range and core area was 6.9 [km.sup.2] (SE = 1.1 [km.sup.2]) and 1.4 [km.sup.2] (SE = 0.3 [km.sup.2]), respectively.
ESTIMATING PATTERNS OF HABITAT USE
Martens used cover types in the landscape (2nd order selection) nonrandomly at both the home range ([LAMBDA] = 0.167, P < 0.001) and core area ([LAMBDA] = 0.158, P < 0.001). Marten home ranges and core areas were both comprised of proportionately more upland deciduous forest and upland mixed forest than upland coniferous forest, agriculture/open land, lowland forest, water/wetland, and urban land cover types. Marten 95% home ranges and core areas had an average of 76% (SE = 3%) and 79% (SE = 4%) marten habitat (upland deciduous and mixed forest), respectively compared to 59% (SE = 1%) available in the landscape. Within home ranges (3rd order selection), marten habitat use was nonrandom at both the 95% home range ([LAMBDA] = 0.333, P = 0.001) and core area ([LAMBDA] = 0.522, P = 0.028). Within 95% home ranges, martens selected upland deciduous and upland mixed forests, and avoided urban, lowland forest, agriculture/openland, and water/wetland. Within core areas martens selected upland conifer forest and avoided urban, lowland forest, agriculture/ openland, and water/wetland.
Mean Penrose distance for hexagons containing marten core areas was 1.81 (SE = 0.29), range 0.15-8.19. We did not find any marten core areas or radiolocations within hexagons with Penrose distance values >8.19. Average Penrose distance for NLP hexagons was 47.99 (SE = 1.15). The lowest Penrose distance value hexagons (i.e., most similar to marten core areas) clustered in north-central and west-central portions of the NLP (Fig. 2). Penrose Distance Class 1 comprised 10,864 [km.sup.2] of the NLP (22% of total area) dispersed among 574 patches. Among PDC 1 patches, 70% were <1000 ha (Fig. 3). Average nearest neighbor distance of PDC 1 patches was 5.3 km (SE = 2.0 km). Along the Euclidean distance path between PRSF and MNF, we found average distance between habitat patches was 7.0 km (se = 0.8 km). This path included 10 gaps between habitat patches and six of these gaps included state or interstate highways (Fig. 1). Our model indicated that 46% and 32% of the NLP was comprised of PDC 2 and PDC 3 hexagons, respectively.
Marten tracks were not found in proportion to available Penrose distance classes ([chi square] = 8.41, df=2, P = 0.015). There were more tracks in PDC 1 and fewer tracks in PDC 3 relative to expectation given availability. We recorded 28 of 32 marten tracks in PDC 1, which provides independent support for our definition of PDC 1 as primary marten habitat. No tracks were recorded in PDC 3, which may provide further evidence that hexagons in this PDC were likely not usable habitat for martens (Table 2).
Similar to Chapin et al. (1997), Potvin et al. (2000), and Dumyahn et al. (2007), we found martens selected for mature upland deciduous and upland mixed forest at the home-range scale. We also determined that marten home ranges were comprised of [greater than or equal to] 70% habitat compared to the landscape, which concurs with past studies (Bissonette et al, 1997; Chapin et al., 1998; Hargis et al., 1999; Potvin et al, 2000; Dumyahn et al., 2007). Dumyahn et al. (2007) reported high variance in proportion of cover-type habitat in core areas and suggested core areas best reflected finer-scale habitat features. Within core areas we found martens used upland conifer stands most frequently. Buchanan (2008) found greater numbers of conifer patches and greater availability of coarse woody debris within marten core areas in our study areas. Cheveau et al. (2013) and McCann (2014) noted the importance of inclusions within marten home ranges. In our study upland conifer stands existed as inclusions within predominantly upland deciduous/upland mixed forest stands. Pine plantations within MNF were used extensively by martens; however, these plantations were unique because they had prevalent mature oak inclusions within them (Buchanan, 2008). In the NLP logging during the 19th and early 20th centuries converted these forests from a pine-dominated to a deciduous-mixed forest landscape (Whitney, 1987). As these forests continue to age, they provide important within-stand structural habitat (Payer and Harris, 2004) components required by martens (Buchanan, 2008).
Our habitat model identified <25% of the NLP as primary marten habitat and independent track survey data supported our definition of PDC 1 as the habitat class where most martens were distributed over a 6 y period. Marten habitat in the NLP was fragmented and consisted of mostly smaller (i.e., <1000 ha) habitat patches isolated by 3.3 times and 5.9 times the average radius of a male and female home range, respectively (Powell, 1994). Connectivity between PRSF and MNF was poor based on overall distance, number of habitat gaps, and potential highway filter barriers. Fragmentation of habitat in the NLP likely constrains marten habitat selection (Bissonette et al, 1997) and hinders future viability of martens in the NLP without direct conservation and management actions. Martens are sensitive to homogenization of landscapes which can compress niches and increase competitive relationships with intraguild predators, such as fishers (Manlick et al., 2017). However, fishers were extirpated from the NLP and never re-introduced, therefore competition with fishers did not constrain our marten population. Bobcats (Lynx rufus) were not considered competitors of martens, because bobcat habitat did not spatially overlap in the NLP (Preuss and Gehring, 2007). Therefore, fragmentation of habitat, along with genetic constraints, in the NLP appears to be the primary driver of marten population dynamics.
Martens in the NLP remain clustered around each of the re-introduction sites. Failure to expand from those localized areas may be related to habitat conditions, but also could be related to conspecific attraction. Conspecific attraction can lead to a clustered pattern of territories embedded within a large area of unoccupied territories, particularly at lower population densities (Greene and Stamps, 2001). We did not include conspecific attraction or other social covariates in our model (sensu Campomizzi et al, 2008), though our model performed well without social covariates. Further, conspecific attraction did not remain in the best marten habitat model reported by Kirk and Zielinski (2009). Kirk and Zielinski's (2009) best habitat model was broader-scaled (80 [km.sup.2]) than our habitat model; however, similar variables were important (i.e., proportion of marten habitat and number of patches). Their model predicted poorly against validation data collected in a different season (Kirk and Zielinski, 2009), whereas our model performed well when applied to a different season (i.e., winter-collected validation data).
Our model used broader-scale (30 x 30 m resolution) cover-type data to characterize and model marten habitat which may have introduced some bias (McCann et al., 2014). Broad-scale habitat spatial data may accurately reflect structural habitat but may not identify functional habitat which may require finer-scale data (McCann et al, 2014). We attempted to correct for this bias by using 50% core areas which may better reflect finer-scale habitat selection (Dumyahn et al., 2007) and weighting each radiolocation based on proportion of cover type in location error polygons. Our spatial data also does not incorporate age-structure of cover types. We advise some caution when applying the model for field use, as local, fine-scale surveys may be necessary to fully assess suitable marten habitat. For example resting site characteristics within stands may be important (Sanders et al., 2017). Our model likely overestimates primary marten habitat in the NLP based on these potential biases (McCann et al., 2014), however the model does identify' suitable habitat patches to further survey for within-stand habitat features needed by martens. Habitat selection patterns are best examined using a multi-scale approach (Bissonette et al., 1997; Lindenmayer, 2000), because fragmentation is perceived at multiple scales among mammals (Gehring and Swihart, 2003). Our modeling approach relied on marten habitat use at multiple home-range scales with the extrapolation to a broader landscape. Our model could be used as a tool for wildlife managers to identify potential marten habitat patches and assess connectivity at a landscape scale before and after re-introductions. Forest management must limit fragmentation of habitat patches which may include establishing habitat linkages, reducing barrier effects of highways, and special consideration of silvicultural practices (Godbout and Ouellet, 2008; Hearn et al., 2010). Our modet also may be combined with genetic models to assess gene flow and landscape barriers (Howell et al., 2016; Hillman et al., 2017). However, forest management also must maintain fine-scale habitat features required by martens.
Our habitat model identifies areas of suitable habitat for martens in the NLP of Michigan and potential distribution of marten populations. It identifies focal areas for surveying and monitoring marten population, including fine-scale habitat survey efforts. This model also could be useful for identifying habitat for further supplementation of the current population or re-introduction in more areas of the NLP. However, it could also be adapted and evaluated in larger portions of the geographic range of martens. Managers can use this model to inform landscape-scale forest management with special consideration for increasing amount of marten habitat and reducing isolation of fragmented patches of marten habitat.
Acknowledgments.--We thank Central Michigan University, Grand Traverse Band of Ottawa and Chippewa Indians, U.S. Fish & Wildlife Service, Citgo Petroleum, Inc., and Safari Club International for funding. We thank A. Gregory for field assistance. We thank Grand Traverse Band of Ottawa and Chippewa Indians, Little River Band of Ottawa Indians, Little Traverse Bay Band of Odawa, Michigan Department of Natural Resources, and USDA Forest Service for logistical support.
AEBISCHER, N.J., P. A. ROBERTSON, AND R. E. KENWARD. 1993. Compositional analysis of habitat use from animal radio-tracking data. Ecology, 74:1313-1325.
BELANT, J. L. 1992. Field immobilization of American martens (Martes americana) and short-tailed weasels (Mustela erminea). J. Wildlife Dis., 28:662-665.
BICKER, S. A. 2007. Genetic bottleneck analysis of Michigan's Lower Peninsula reintroduced populations of American marten (Martes americana). Thesis, Central Michigan University, Mt. Pleasant.
BISSONETTE.J. A. AND S. BROEKHUIZEN. 1995. Martes populations as indicators of habitat spatial patterns: the need for a multiscale approach, p. 95-121. In: W. Z. Lidicker, Jr. (ed.). Landscape approaches in mammalian ecology' and conservation. Minnesota University Press, Minneapolis.
--, D.J. HARRISON, C. D. HARGIS, AND T. G. CHAPIN. 1997. The influence of spatial scale and scale sensitive properties on habitat selection by American martens, p. 368-385. In:J. A. Bissonette (ed.). Wildlife and landscape ecology. Springer, New York.
BUCHANAN, C. 2008. Multi-scale habitat selection by American marten Martes americana in Michigan's Lower Peninsula. Thesis, Central Michigan University, Mt. Pleasant.
BUSKIRK, S. W. AND R. A. POWELL. 1994. Habitat ecology' of fishers and American martens, p. 283-296. In: S.W. Buskirk, A. S. Harestad, M. G. Raphael, and R. A. Powell (eds.). Martens, sables, and fishers: biology and conservation. Cornell University Press, Ithaca, N.Y.
CAMPOMIZZI, A. J., J. A. BUTCHER, S. I. FARRELL, A. G. SNELGROVE, B. A. COLLIER, K. J. GUTZWILLER, M. L. MORRISON, AND R. N. WILKINS. 2008. Conspecific attraction is a missing component in wildlife habitat modeling. J Wildlife Manage., 72:331-336.
CHAPIN, T. G., D.J. HARRISON, AND D. M. PHILLIPS. 1997. Seasonal habitat selection bv marten in an untrapped forest preserve. J Wildlife Manage., 53:191-196.
--, --, AND D. D. KATNIK. 1998. Influence of landscape pattern on habitat use bv American marten in an industrial forest. Conservation Biology. 12:1327-1337.
CHEAVEAC, M., L. IMBEAU, P. DRAPEAU, AND L. BELANGER. 2013. Marlen space use and habitat selection in managed coniferous boreal forests of eastern Canada. J. Wildlife Manage.. 77:749-760.
DUMYAHN, J. B., P. A. ZOLLNER, AND J. H. GILBERT. 2007. Winter home-range characteristics of American marten (Martes americana) in northern Wisconsin. Am. Midi, Nat., 158:382-394.
EARLE, R. D., L. H. MASTENBROOK, AND T. F. REIS. 2001. Distribution and abundance of marten in northern Michigan. Michigan Department of Natural Resources Wildlife Report No. 3321, Lansing.
FAGERSTONE, K. A. 1987. Black-footed ferret, long-tailed weasel, short-tailed weasel, and least weasel, p. 549-573 In: M. Novak, J. A. Baker, M.E. Obbard, and B. Malloch (eds). Wild furbearer management and conservation in North America Ontario Ministry of Natural Resources.
FULLER, A. K. AND D.J. HARRISON. 2005. Influence of partial timber harvesting on American martens in north-central Maine. J Wildlife Manage.. 69:710-722.
GEHRING, T. M. AND R. K. SWIHART. 2003. Body size, niche breadth, and ecologically scaled responses to habitat fragmentation: mammalian predators in an agricultural landscape. Biol. Consent., 109:283-295.
GODBOUT, G. AND J-P. OUELLET. 2010. Fine-scale habitat selection of American marten at the southern fringe of the boreal forest. Ecoscience, 17:175-185.
GREENE, C. M. AND J. A. STAMPS. 2001. Habitat selection at low population densities. Ecology, 82:2091-2100.
HARDEN, J. M. 1998. Status of the American marten Martes americana in the Manistee national forest on lower Michigan. Ms. Thesis, Central Michigan University, Mt. Pleasant.
HARGIS, C. D., J. A. BISSONETTE, AND D. L. TURNER. 1999. The influence of forest fragmentation and landscape pattern on American martens. J. Appl. Ecol., 36:157-172.
HEARN, B.J., D.J. HARRISON, A. K. FILLER. C. G. LUNDRIGAN, AND W.J. CURRAN. 2010. Paradigm shifts in habitat ecology of threatened Newfoundland martens J Wildlife Manage., 74:719-728.
HILLMAN, T. L., P. KEENLANCE, J. A. MOORE, B.J. SWANSON, J.JACQUOT, J. C. WITT, AND A. CORNMAN. 2017. Genetic diversity of reintroduced American martens in Michigan's Lower Peninsula. J Mamm., 98:1489-1496.
HOOGE, P. N. AND B. EICHENLAUB. 1997. Animal movement extension to arcview. ver. 1.1. Alaska Science Center--Biological Science Office, LT.S. Geological Survey, Anchorage, Alaska.
HOWELL, P. E., E. L. ROEN, B. W. WILLIAMS, G. J. ROLOFF, AND K.T. SCRIBNER. 2016. Contiguity of landscape features pose barriers to gene flow among American marten (Martes americana) genetic clusters in the Upper Peninsula of Michigan. Landscape Ecol, 31:1051-1062.
JOHNSON, D. H. 1980. The comparison of usage and availability measurements for evaluating resource preference. Ecology, 61:65-71.
KENWARD, R. E. 2001. A manual of wildlife radio tagging. Academic Press, San Diego, Calif.
KIRK, T. A. AND W.J. ZIELINSKI. 2009. Developing and testing a landscape habitat suitability model for the American marten (Martes Americana) in the Cascades mountains of California. Landscape Ecol, 24:759-773.
LENTH, R. V. 1981. On finding the source of a signal. Technometrics, 23:149-124.
LINDENMAYER, D. B. 2000. Factors at multiple scales affecting distribution patterns and their implications for animal conservation: leadbeater's possum as a case study. Biodivers. Conserv., 9:15-35.
LOVALLO, M. J., K. C. VERCAUTEREN, N. C. HEDGE, E. M. ANDERSON, AND S. E. HYGNSTROM. 1994. An evaluation of electronic versus hand-held compasses for telemetry studies. Wildlife Soc. B., 22:662-667.
MANLICK, P. J., J. E. WOODFORD, 15. ZUCKERBERG, AND J. N. PAULL. 2017. Niche compression intensifies competition between reintroduced American martens (Martes americana) and fishers (Pekania pennant). J. Mamm.. 98:690-702.
MANLY, B. F. J. 2005. Multivariate statistical methods: a primer. 3rd ed. Chapman and Hall/CRC Press, Boca Raton, Fla.
MCCANN, N. P., P. A. ZOLLNER, ANDJ. H. GILBERT. 2014. Bias in the use of broadscale vegetation data in the analysis of habitat selection. J. Mammal.. 95:369-381.
MCFADDEN, L. M. 2007. Home-range dynamics and habitat selection of American martens Martes americana in Michigan's northern lower peninsula. Ms. Thesis, Central Michigan University, Ml. Pleasant.
MICHIGAN DEPARTMENT OF NATURAL RESOURCES. 2013. Northern Lower Peninsula regional state forest management plan. Lansing.
NAMS, V. O. 2005. Locate III user's guide. Pacer, Truro, Nova Scotia, Canada.
NIELSEN, C. K. AND A. WOOLF. 2002. Habitat-relative abundance relationship for bobcats in southern Illinois. Wildlife Soc. B., 30:222-230.
OTT, P. AND F. HOVEY. 1997. BYCOMP.SAS & MACOMP.SAS. Version 1.0. British Columbia Forest Service, Victoria, Canada.
PAYER, D. AND D. HARRISON. 2004. Relationships between forest structure and habitat use by American martens in Maine, USA, p. 173-186. In: D. J. Harrison, A. K. Fuller, and G. Proulx (eds.). Martens and fishers (Martes) in human-altered environments: an international perspective. Springer, New York.
POOLE, K. G., G. M. MATSON, M. A. STRICKLAND, A.J. MAGOUN, R. P. GRAF, AND L. M. DIX. 1994. Age and sex determination for American martens and fishers, p. 204-223. In: S. W. Buskirk, A. S. Harestad, M. G. Raphael, and R. A. Powell (eds.). Martens, sables, and fishers: biology and conservation. Cornell University Press, Ithaca, N.Y.
POTVIN, F. L. BELANGER, AND K. LOWELL. 2000. Marten habitat selection in a clearcut boreal landscape. Conserv. Biol.. 14:844-857.
POWELL, R. A. 1994. Structure and spacing of Martes populations, p. 101-121. In: S. W. Buskirk, A. S. Harestad, M. G. Raphael, and R. A. Powell (eds.). Martens, sables, and fishers: biology and conservation. Cornell University Press, Ithaca, N.Y.
PREUSS, T. S. AND T. M. GEHRING. 2007. Landscape analysis of potential bobcat habitat in the northern Lower Peninsula of Michigan./ Wildlife Manage., 71:2699-2706.
REMPEL, R. 2003. Patch Analyst 3.1. Centre for Northern Forest Ecosystem Research, Lakehead University Campus, Thunder Bay, Ontario, Canada.
SANDELL, M. 1989. The mating tactics and spacing patterns of solitary carnivores, p. 164-182. In:). L. Cattleman (ed.). Carnivore behavior, ecology and evolution. Cornell University Press, Ithaca N.Y.
SANDERS, D. E., A. CORXMAN, P. KEENLANCE, J. J. JACQUOT, D. E. UNGER, AND M. SPRIGGS. Resting site characteristics of American marten in the northern lower peninsula of Michigan. Am. Midi. Nat., 177:211-225
SEAMAN, D. E., J.J. MILLSPAUGH, B.J. KERNOHAN, G. C. BRUNDIGE, K. J. RAEDEKF., AND R. A. GITZEN. 1999. Effects of sample size on kernel home range estimates. J Wildlife Manage.. 63:739-747.
SPACE IMAGING. 2004. IFMAP: review of remote sensing technologies used in the IFMAP project. Final report. Space Imaging, Ann Arbor, Mich.
THOMASMA, L. E. AND R. O. PETERSON. 1998. Tool and technique for restraining live-captured martens and fishers. USDA Forest Service, Research Note NE-365.
UNITED STATES DEPARTMENT OF AGRICULTURE, FOREST SERVICE. 2006. Land and resource management plan, Ottawa National Forest, Ironwood, Mich.
WHITE, G. C. AND R. A. GARROTT. 1990. Analysis of wildlife radio-tracking data. Academic Press, New York, N.Y.
WHITNEY, G. G. 1987. An ecological history of the Great Lakes forest of Michigan. J. Ecol, 75:667-684.
WILLIAMS, B. W., J. H. GILBERT, AND P. A. ZOLLNER. 2007. Historical perspective on the reintroduction of the fish and American marten in Wisconsin and Michigan. General Technical Report NRS-5. USDA Forest Service, Northern Research Stations, Newtown Square, Penn.
WORTON, B. J. 1989. Using Monte Carlo simulation to evaluate kernel-based home range estimators. J. Wildlife Manage.. 59:794-800.
ZAR, J. H. 2010. Biostatistical analysis. 5th ed. Prentice-Hall, Inc., Upper Saddle River, N.J.
SUBMITTED 30 NOVEMBER 2018
ACCEPTED 21 JUNE 2019
THOMAS M. GEHRING, (1) LYNNEA M. MCFADDEN, (2) SARA A. PRUSSING, SARA A. BICKERSMITH, (3) CLAY BUCHANAN, (4) ERIC NELSON (5) and BRADLEY J. SWANSON
Department of Biology, Central Michigan University, Ml. Pleasant 48859
(1) Corresponding author: e-mail: email@example.com
(2) Present address: St. Clair County Community College, Port Huron, Michigan 48061
(3) Present address: Wadsworth Center, New York State Department of Health, Albany 12201
(4) Present address: Michigan Department of Natural Resources, 530 W Allegan St #4, Lansing 48933
(5) Present address: Minnesota Department of Natural Resources, 1601 Minnesota Dr, Brainerd 56401
Caption: FIG. 1.--Marten re-introduction sites (1985-1986) in Pigeon River Country State Forest (PRSF) and Manistee National Forest (MNF) and radiotelemetry study areas (shaded polygons) in the northern Lower Peninsula of Michigan, U.S.A., 2005-2006. State and interstate highways are shown as black lines to highlight potential barriers
Caption: FIG. 2.--Penrose distance model depicting areas of similar habitat quality to core areas of radiocollared martens and the northern Lower Peninsula of Michigan, U.S.A., 2003-2006. Lower values indicate higher similarity to marten core areas, with Penrose distance class 1 (0.15-2.39) considered primary marten habitat
Caption: FIG. 3.--Number of primary marten habitat patches (i.e., Penrose distance class 1) of various patch size categories in the northern Lower Peninsula of Michigan, U.S.A., 2005-2006
TABLE 1.--Reclassified landcover types based on IFMAP land-cover data and used for modeling marten habitat throughout the NLP. Classification rules adapted from Space Imaging (2004) Reclassification IFMAP classification Classification rules Urban Airports >10% human-made structures Road/parking lot Includes roads High intensity urban Low intensity urban Agriculture/openland Nonvegetated farmland Intensively managed crops Row crops Excludes forestry or Forage crops/no-till <25% vegetation crop Orchards/vineyards/ nursery Parks/golf courses Upland shrub/low density trees Sand/soil Exposed rock Mudflats Other bare/sparsely vegetated Upland deciduous Northern hardwood [greater than or association equal to] 60% Oak association canopy is deciduous Aspen association trees Other upland deciduous Upland coniferous Pines [greater than or equal to] 60% canopy is coniferous Other upland conifers trees Upland mixed forest Mixed upland Conifer to deciduous deciduous tree Mixed upland conifers Ratio is >40%:<60% Upland mixed forest to <60%:>40% Lowland forest Lowland deciduous Flooded in past 5 yr forest or >60% canopy is Lowland coniferous lowland tree forest Lowland mixed forest Water/wetland Water >75% area is water Floating aquatic or >60% wetland Lowland shrub vegetation Emergent wetland Mixed nonforest wetland TABLE 2.--Marten tracks recorded along forest trails and roads within each Penrose distance class in the northern Lower Peninsula of Michigan, U.S.A (2001-2007) Penrose Length Number Number distance of road of marten of marten class surveyed, km tracks tracks per km 1 396 28 0.07 2 118 4 0.03 3 98 0 0
|Printer friendly Cite/link Email Feedback|
|Author:||Gehring, Thomas M.; McFadden, Lynnea M.; Prussing, Sara A.; Bickersmith, Sara A.; Buchanan, Clay; Ne|
|Publication:||The American Midland Naturalist|
|Date:||Oct 1, 2019|
|Previous Article:||Life History Variation Within and Among Populations of the Ninespine Stickleback (Pungitius pungitius) in Alaska: Lake-Stream Contrasts.|
|Next Article:||A Wandering Black Bear (Ursus americanus, Pallas 1780) in the Sierra Gorda Biosphere Reserve, Queretaro.|