Use of constructed coarse woody debris corridors in a clearcut by American martens (Martes americana) and their prey

Use of constructed coarse woody debris corridors in a clearcut by American martens (Martes... Abstract American martens (Martes americana) are typically associated with mature coniferous forests because they provide overhead cover and coarse woody debris (CWD) that martens require for protection and hunting. Therefore, clearcuts are considered poor marten habitat because they contain no overhead cover and relatively little CWD. We examined the efficacy of retaining CWD and constructing CWD corridors in a recently harvested clearcut to promote the use of the area by martens and small mammals, a major prey resource. We installed remote cameras in corridors, the surrounding clearcut and forest to monitor the distribution of martens and small mammals. Martens and red squirrels did not use CWD corridors more frequently than clearcut areas in summer; forest-floor small mammals, however, used corridors approximately three times as frequently as other habitats (x2 = 13.374, P = 0.001). Marten presence was positively associated with mature, dense forest and long pieces of CWD. In winter, red squirrels used corridors more frequently than other clearcut areas, and limited data suggested that martens preferred the corridors to other clearcut areas. Consequently, we recommend that forest managers retain CWD and construct CWD corridors within clearcuts to provide small mammal habitat, and to enhance marten habitat. Introduction Coarse woody debris (CWD) is an important component of forest ecosystems because of the many ecological functions it provides, including nutrient cycling, energy flow and habitat for wildlife (Harmon et al., 1986). Many mammal species use CWD as a habitat element because it can provide cover from predators, insulation, subnivean habitat and hunting opportunities (McComb, 2003). CWD is defined as any piece of wood >7.5 cm in diameter, and can be created by natural or anthropogenic disturbances (British Columbia Forest Analysis and Inventory Branch, 2015). Retention of CWD in clearcuts may mitigate habitat loss by providing cover for many wildlife species that require complex forest structure (Hargis and McCullough, 1984; Andruskiw et al., 2008). American martens (Martes americana) are small carnivores from the family Mustelidae that inhabit most of the coniferous forests of Canada, Alaska and the Rocky Mountain states of the USA (Powell et al., 2003). Habitat for American martens has been characterized as mature coniferous and mixed-wood forests that contain old trees and large amounts of CWD (Lofroth, 1993; Thompson and Colgan, 1994; Poole et al., 2004; Andruskiw et al., 2008; Thompson et al., 2012). Martens select these forests because they provide overhead cover for protection from aerial predators (Poole et al., 2004), and contain high densities of mice (Peromyscus spp. and Zapus spp.), voles (Microtus spp. and Myodes spp.), shrews (Sorex spp.) and red squirrels (Tamiasciurus hudsonicus), their major prey items (Sherburne, 1993; Buskirk and Ruggiero, 1994; Martin, 1994). CWD also provide martens with subnivean habitat during winter (Wiebe et al., 2014) such as resting sites, thermal insulation and refugia from predators (Buskirk et al., 1989) as well as access to prey (Sherburne and Bisonette, 1994). Martens generally avoid non-forested areas because of their lack of structural complexity and overhead cover (Buskirk et al., 1989; Potvin et al., 2000; Wiebe et al., 2013) but there is a paucity of knowledge about whether martens select for CWD within recently harvested areas. Forest-floor small mammals such as mice, voles and shrews are found in higher densities in areas with large amounts of CWD (Fauteux et al., 2012), including constructed CWD corridors in clearcuts (Sullivan et al., 2011, 2012), while red squirrels are reported to avoid harvested areas (Bakker and van Vuren, 2004). Because of the known associations of martens and their forest-floor small mammal prey with CWD, retention of CWD in clearcuts may help to mitigate the adverse effects of forest harvest on martens. Starting in the 1990s, an unprecedented outbreak of mountain pine beetle (Dendroctonus ponderosae) in British Columbia, Canada, resulted in the death of some 13 million hectares of mature lodgepole pine (Pinus contorta Douglas ex Louden; Lindenmayer et al., 2012). This prompted an increase in clearcut harvest to salvage dead lodgepole pine. This strategy was designed to maximize the value of dead standing trees and mitigate economic losses associated with a significant decline in timber supply (Lindenmayer et al., 2012). Following clearcut harvest, woody debris is often removed from harvested areas to reduce the risk of wildfire and further insect outbreaks (Agee, 1993). The debris is either burned onsite or transported to bioenergy plants to create electricity (Mabee and Saddler, 2009) but retaining CWD in clearcuts to create habitat for wildlife such as martens and their prey may be more beneficial. In an effort to understand the relationship between CWD and these wildlife species, we used presence–absence camera data to address two research objectives: (1) to determine if martens and/or small mammals use constructed CWD corridors more frequently than areas with dispersed CWD in a clearcut and (2) to determine the fine-scale habitat characteristics that best predict marten and forest floor small mammal presence within and surrounding a recent clearcut. Methods Study area Initial research was conducted from May to August 2015, and supplemental data were collected from February to April 2017 within the John Prince Research Forest (JPRF), located ~50 km north of Fort St. James in central British Columbia, Canada (54°36.417′N; 124°21.237′W). The research forest is located within the Sub-Boreal Spruce biogeoclimatic zone and is characterized by severe snowy winters, short warm summers and moderate annual precipitation (Meidinger and Pojar, 1991). The study site was located within and adjacent to Block 67, a 93-ha clearcut that was harvested in the winter of 2014 (Figure 1). The block had a primarily southeastern aspect with undulating topography. Figure 1 View largeDownload slide Map of study area within and surrounding a clearcut, with inset map indicating the location of the John Prince Research Forest in central British Columbia, Canada. Figure 1 View largeDownload slide Map of study area within and surrounding a clearcut, with inset map indicating the location of the John Prince Research Forest in central British Columbia, Canada. Prior to harvest, the stand was dominated by a mixture of paper birch (Betula papyrifera Marsh.), hybrid interior spruce (Picea glauca (Moench) Voss × Picea engelmannii Parry ex Elngelm.) and Douglas-fir (Pseudotsuga menziesii (Mirb.) Franco) with average diameters of 37–45 cm and average heights of 26–29 m. The silvicultural prescription for the clearcut was to harvest marketable hybrid interior spruce and Douglas-fir while retaining all deciduous trees and Douglas-fir with diameters >55 cm. Several wildlife tree patches were retained to protect an ungulate mineral lick, a potential fisher (Pekania pennanti) denning habitat, a previous bear denning site and to maintain connectivity throughout the cut area. The harvest prescription called for high CWD retention targets post-harvest in the clearcut (30–50 m3 ha−1). This level is significantly higher than the legal minimum requirement for interior British Columbia which has been set at four logs per hectare, each being a minimum 2 m in length and 7.5 cm at one end (British Columbia, 2004). Industrial standards, however, commonly exceed this target as suggested by the British Columbia Chief Forester (2010); average CWD retention by industrial operators was ~5 m3 ha−1 for habitats similar to the study area. Much of the residual CWD was used to construct nine corridors using an excavator backhoe within the boundaries of the clearcut. The treatment was installed over a period of 3 days during September 2014 and was completed separately from logging operations to ensure the prescription was appropriate for future monitoring. Cost for constructing the corridors was ~$200 CAD/100 m of corridor. The corridors were composed of a variety of types and sizes of wood, were ~1 m high × ~5 m wide, and ranged between 100 m and 200 m in length. The corridors were intended to connect areas of surrounding forest to wildlife tree patches within the clearcut (Figure 1) and to serve as resting and hunting habitat for mesocarnivores such as marten. Survey design In summer 2015, 42 passive infra-red remote video cameras (Bushnell 8MP Trophy Cam HD Black LED Trail Camera with Night Vision model #119676CN, Bushnell Outdoor Products) were installed systematically throughout the sampling area on a hexagonal grid with 200 m spacing between cameras (Figure 1). In addition, one camera was installed in each of the nine corridors, as close to the midpoint as possible while maintaining a minimum distance of 50 m from cameras located in adjacent plots. Systematically placed cameras were set to sensor level: medium; LED control: medium and 30 s video recording. Corridor cameras were set to sensor level: low because the contrast of vegetation movement against the woody debris on medium sensor level resulted in memory cards filled with videos of vegetation movement. Cameras were baited weekly with Sharpe’s marten lure (a short-range, strawberry and anise-scented lure) and 5 ml of peanut butter placed 2 m in front of the camera on a stick or log. Cameras were positioned 0.5 m from the ground on a tree or wooden stake, pointed north-west to minimize backlighting, and aimed towards the bait. A 0.5-m tall stick was placed 1 m behind the bait so that the approximate size of animals could be inferred from videos. In winter 2017, 18 cameras were installed in the harvested area: 9 in the corridors and 9 randomly in the clearcut, a minimum of 50 m away from corridors. The cameras were baited with a small (~75 g) piece of hanging meat (American beaver; Castor canadensis Kuhl) and a commercial lure (a combination of Sharpe’s lynx and mink lures) placed on a small log 2–3 m in front of the camera. The cameras were set to sensor level: medium; LED control: medium and 30 s video recording. Sampling schedule In summer 2015, cameras were installed between May 24 and 28 and remained active until August 20th. This sampling period allowed for observation of seasonal variation in the presence of martens and small mammals from before vegetation flushed until leaves were in full flush. In winter 2017, cameras were active from February 3 to April 6 and baited/lured at the beginning of three 3-week sessions. Animals captured in the videos were identified to species if possible, and the length of time that each animal stayed at the camera was recorded. Visits were considered separate if the time from the end of one video to the beginning of the next exceeded 1 min. Habitat assessment At each of the camera sites in summer 2015, slope (%), slope position, elevation and aspect were measured. We also assessed shrub cover within a 3.99-m radius plot (50 m2) around each camera. A 1.5-m cover pole divided into 50 cm sections and 10 cm bands was used to assess cover. If shrubs obscured >25% of a band it was considered covered (Toledo et al., 2010). We recorded cover measurements weekly, and as we found no seasonal trend in measurements we averaged them over the summer study period. Within an 11.28- m radius plot (400 m2), we counted the number of tree stems >12.5 cm diameter at breast height (DBH, British Columbia Forest Analysis and Inventory Branch, 2015), squirrel middens (piles of cone bracts discarded by red squirrels), tip-up mounds (exposed roots of fallen trees), stumps and saplings. For all trees >12.5 cm DBH, we measured DBH and identified trees to species level. We also categorized all trees into a wildlife tree type: Types 1 and 2: live; Types 3–5: dead with intact top or Types 6–9: dead with a broken top (Keisker, 2000). At each camera, three 50-m transects oriented at 0°, 120° and 240° from the camera were used to measure CWD. All CWD > 7.5 cm diameter (British Columbia Forest Analysis and Inventory Branch, 2015) that intersected the transect line was classified by decomposition class: Type 1: little to no decay; Type 2: significant decay but wood retained structure or Type 3: structure of wood was lost. For each piece of CWD, we recorded the diameter, length, orientation to the transect line, tree type (deciduous or conifer) and whether it was on or off the ground. We used a geographic information system to measure the proximity of each camera to CWD corridors, forest edge, wildlife tree patches, roads, streams and open water. Statistical analysis We used data from camera records to construct logistic regression models that represented the relationship between the presence-absence of martens and forest-floor small mammals during summer and habitat variables measured at each camera site (Table 1). The cameras functioned continuously, thus, we identified a sampling interval as a 24-h period. Using exploratory analysis of the habitat variables measured at each camera site, we constructed candidate models for martens (Table 2) and forest-floor small mammals (Table 3), which we then ranked using the differences in Akaike’s information criteria for small sample sizes (ΔAICc) (Burnham and Anderson, 2002) and associated weights (AICcw) to select the most parsimonious logistic regression models. We used the cluster option in Stata (ver.12.1 StataCorp LP, 2011; Rogers, 1993) to correct standard errors for multiple observations at the same site. We used the receiver operating characteristic to evaluate the predictive accuracy of the most parsimonious models. Models with an area under the curve (AUC) score of 0.7–0.9 were considered useful (Boyce et al., 2002). The relative strengths of the coefficients within the most parsimonious model were assessed using 95% confidence intervals. Coefficients were considered significant factors predicting presence of martens or forest floor small mammals if the interval did not overlap zero. These analyses were completed using Stata. Frequencies of habitat use were compared using a chi-square test. A Kruskal–Wallis test in R version 3.2.4 (R Core Team, 2016) was used to determine significant differences in the presence of martens and small mammals in different habitats, and during the day vs night. Table 1 Variables tested to construct logistic regression models describing the presence of martens and small mammals within and surrounding a recently harvested clearcut in central British Columbia, Canada during summer 2015. Variable Type Description Slope Continuous % Slope Aspect Continuous Aspect (degrees) Elevation Continuous Elevation (m) Slope position Categorical Slope position (1–5) Location Categorical Forest (1) or clearcut (0) Corridor Categorical Camera in corridor (1) or not (0) #Middens Continuous # of middens present in 11.28 m plot #Tipups Continuous # of root tip-up present in 11.28 m plot #Stumps Continuous # of stumps present in 11.28 m plot #Saplings Continuous # of conifer saplings (<1.3 m height) present in 11.28 m plot #Stems Continuous # trees >12.5 cm DBH present in 11.28 m plot Mean DBH Continuous Mean DBH of trees >12.5 cm DBH in 11.28 m plot (cm) % Conifer trees Continuous Per cent of conifer trees of trees >12.5 cm DBH in 11.28 m plot #WLT3–5 Continuous Number of wildlife trees Types 3–5 in 11.28 m plot #WLT6–9 Continuous Number of wildlife trees Types 6–9 in 11.28 m plot #PiecesCWD Continuous Total # of pieces of CWD on transects MeanDBHCWD Continuous Mean diameter of CWD >7.5 cm on transects (cm) MeanlengthCWD Continuous Mean length of CWD on transects (m) % CWD C1,2 Continuous Per cent of CWD in decay class 1 or 2 on transects % CWD conifer Continuous Per cent of CWD conifer on transects % CWD ON Continuous Per cent of CWD on the ground on transects Volume CWD Continuous Total volume of CWD on transects (m3) AV % cover Continuous Average per cent shrub cover <1.5 m CorDist Continuous Distance to nearest corridor (m) EdgeDist Continuous Distance to forest edge (m) WTPDist Continuous Distance to wildlife tree patch (m) RoadDist Continuous Distance to road (m) LakeDist Continuous Distance to lake (m) CreekDist Continuous Distance to creek (m) JulDay Continuous Julian day Variable Type Description Slope Continuous % Slope Aspect Continuous Aspect (degrees) Elevation Continuous Elevation (m) Slope position Categorical Slope position (1–5) Location Categorical Forest (1) or clearcut (0) Corridor Categorical Camera in corridor (1) or not (0) #Middens Continuous # of middens present in 11.28 m plot #Tipups Continuous # of root tip-up present in 11.28 m plot #Stumps Continuous # of stumps present in 11.28 m plot #Saplings Continuous # of conifer saplings (<1.3 m height) present in 11.28 m plot #Stems Continuous # trees >12.5 cm DBH present in 11.28 m plot Mean DBH Continuous Mean DBH of trees >12.5 cm DBH in 11.28 m plot (cm) % Conifer trees Continuous Per cent of conifer trees of trees >12.5 cm DBH in 11.28 m plot #WLT3–5 Continuous Number of wildlife trees Types 3–5 in 11.28 m plot #WLT6–9 Continuous Number of wildlife trees Types 6–9 in 11.28 m plot #PiecesCWD Continuous Total # of pieces of CWD on transects MeanDBHCWD Continuous Mean diameter of CWD >7.5 cm on transects (cm) MeanlengthCWD Continuous Mean length of CWD on transects (m) % CWD C1,2 Continuous Per cent of CWD in decay class 1 or 2 on transects % CWD conifer Continuous Per cent of CWD conifer on transects % CWD ON Continuous Per cent of CWD on the ground on transects Volume CWD Continuous Total volume of CWD on transects (m3) AV % cover Continuous Average per cent shrub cover <1.5 m CorDist Continuous Distance to nearest corridor (m) EdgeDist Continuous Distance to forest edge (m) WTPDist Continuous Distance to wildlife tree patch (m) RoadDist Continuous Distance to road (m) LakeDist Continuous Distance to lake (m) CreekDist Continuous Distance to creek (m) JulDay Continuous Julian day Table 1 Variables tested to construct logistic regression models describing the presence of martens and small mammals within and surrounding a recently harvested clearcut in central British Columbia, Canada during summer 2015. Variable Type Description Slope Continuous % Slope Aspect Continuous Aspect (degrees) Elevation Continuous Elevation (m) Slope position Categorical Slope position (1–5) Location Categorical Forest (1) or clearcut (0) Corridor Categorical Camera in corridor (1) or not (0) #Middens Continuous # of middens present in 11.28 m plot #Tipups Continuous # of root tip-up present in 11.28 m plot #Stumps Continuous # of stumps present in 11.28 m plot #Saplings Continuous # of conifer saplings (<1.3 m height) present in 11.28 m plot #Stems Continuous # trees >12.5 cm DBH present in 11.28 m plot Mean DBH Continuous Mean DBH of trees >12.5 cm DBH in 11.28 m plot (cm) % Conifer trees Continuous Per cent of conifer trees of trees >12.5 cm DBH in 11.28 m plot #WLT3–5 Continuous Number of wildlife trees Types 3–5 in 11.28 m plot #WLT6–9 Continuous Number of wildlife trees Types 6–9 in 11.28 m plot #PiecesCWD Continuous Total # of pieces of CWD on transects MeanDBHCWD Continuous Mean diameter of CWD >7.5 cm on transects (cm) MeanlengthCWD Continuous Mean length of CWD on transects (m) % CWD C1,2 Continuous Per cent of CWD in decay class 1 or 2 on transects % CWD conifer Continuous Per cent of CWD conifer on transects % CWD ON Continuous Per cent of CWD on the ground on transects Volume CWD Continuous Total volume of CWD on transects (m3) AV % cover Continuous Average per cent shrub cover <1.5 m CorDist Continuous Distance to nearest corridor (m) EdgeDist Continuous Distance to forest edge (m) WTPDist Continuous Distance to wildlife tree patch (m) RoadDist Continuous Distance to road (m) LakeDist Continuous Distance to lake (m) CreekDist Continuous Distance to creek (m) JulDay Continuous Julian day Variable Type Description Slope Continuous % Slope Aspect Continuous Aspect (degrees) Elevation Continuous Elevation (m) Slope position Categorical Slope position (1–5) Location Categorical Forest (1) or clearcut (0) Corridor Categorical Camera in corridor (1) or not (0) #Middens Continuous # of middens present in 11.28 m plot #Tipups Continuous # of root tip-up present in 11.28 m plot #Stumps Continuous # of stumps present in 11.28 m plot #Saplings Continuous # of conifer saplings (<1.3 m height) present in 11.28 m plot #Stems Continuous # trees >12.5 cm DBH present in 11.28 m plot Mean DBH Continuous Mean DBH of trees >12.5 cm DBH in 11.28 m plot (cm) % Conifer trees Continuous Per cent of conifer trees of trees >12.5 cm DBH in 11.28 m plot #WLT3–5 Continuous Number of wildlife trees Types 3–5 in 11.28 m plot #WLT6–9 Continuous Number of wildlife trees Types 6–9 in 11.28 m plot #PiecesCWD Continuous Total # of pieces of CWD on transects MeanDBHCWD Continuous Mean diameter of CWD >7.5 cm on transects (cm) MeanlengthCWD Continuous Mean length of CWD on transects (m) % CWD C1,2 Continuous Per cent of CWD in decay class 1 or 2 on transects % CWD conifer Continuous Per cent of CWD conifer on transects % CWD ON Continuous Per cent of CWD on the ground on transects Volume CWD Continuous Total volume of CWD on transects (m3) AV % cover Continuous Average per cent shrub cover <1.5 m CorDist Continuous Distance to nearest corridor (m) EdgeDist Continuous Distance to forest edge (m) WTPDist Continuous Distance to wildlife tree patch (m) RoadDist Continuous Distance to road (m) LakeDist Continuous Distance to lake (m) CreekDist Continuous Distance to creek (m) JulDay Continuous Julian day Table 2 Logistic regression models for presence–absence data of American martens at camera traps within and near a recently harvested clearcut in central British Columbia, Canada during summer 2015. Model Rank k1 AICc2 ∆AICc3 AICcw4 Stems+julday+lengthcwd+wlt69 1 4 1007.30 0.00 0.65 Stems+julday+lengthcwd 2 3 1008.90 1.60 0.29 Stems+julday+wlt69 3 3 1012.96 5.66 0.04 Julday+lengthcwd+wlt69 4 3 1015.15 7.85 0.01 Stems+julday 5 2 1016.59 9.29 0.01 Julday+lengthcwd 6 2 1022.84 15.54 <0.01 Julday+wlt69 7 2 1034.18 26.88 <0.01 Julday 8 1 1057.47 50.17 <0.01 Stems+lengthcwd+wlt69 9 3 1154.75 147.45 <0.01 Stems+lengthcwd 10 2 1155.99 148.69 <0.01 Stems+wlt69 11 2 1160.18 152.88 <0.01 Lengthcwd+wlt69 12 2 1162.69 155.39 <0.01 Stems 13 1 1163.41 156.11 <0.01 Lengthcwd 14 1 1169.56 162.26 <0.01 Wlt69 15 1 1182.28 174.98 <0.01 Model Rank k1 AICc2 ∆AICc3 AICcw4 Stems+julday+lengthcwd+wlt69 1 4 1007.30 0.00 0.65 Stems+julday+lengthcwd 2 3 1008.90 1.60 0.29 Stems+julday+wlt69 3 3 1012.96 5.66 0.04 Julday+lengthcwd+wlt69 4 3 1015.15 7.85 0.01 Stems+julday 5 2 1016.59 9.29 0.01 Julday+lengthcwd 6 2 1022.84 15.54 <0.01 Julday+wlt69 7 2 1034.18 26.88 <0.01 Julday 8 1 1057.47 50.17 <0.01 Stems+lengthcwd+wlt69 9 3 1154.75 147.45 <0.01 Stems+lengthcwd 10 2 1155.99 148.69 <0.01 Stems+wlt69 11 2 1160.18 152.88 <0.01 Lengthcwd+wlt69 12 2 1162.69 155.39 <0.01 Stems 13 1 1163.41 156.11 <0.01 Lengthcwd 14 1 1169.56 162.26 <0.01 Wlt69 15 1 1182.28 174.98 <0.01 1 Number of parameters used in the model. 2 Akaike’s information criterion for small sample sizes value. 3 Difference in Akaike’s information criterion for small sample sizes value from the most parsimonious model. 4 Weight of Akaike's information criterion value for small sample sizes. Table 2 Logistic regression models for presence–absence data of American martens at camera traps within and near a recently harvested clearcut in central British Columbia, Canada during summer 2015. Model Rank k1 AICc2 ∆AICc3 AICcw4 Stems+julday+lengthcwd+wlt69 1 4 1007.30 0.00 0.65 Stems+julday+lengthcwd 2 3 1008.90 1.60 0.29 Stems+julday+wlt69 3 3 1012.96 5.66 0.04 Julday+lengthcwd+wlt69 4 3 1015.15 7.85 0.01 Stems+julday 5 2 1016.59 9.29 0.01 Julday+lengthcwd 6 2 1022.84 15.54 <0.01 Julday+wlt69 7 2 1034.18 26.88 <0.01 Julday 8 1 1057.47 50.17 <0.01 Stems+lengthcwd+wlt69 9 3 1154.75 147.45 <0.01 Stems+lengthcwd 10 2 1155.99 148.69 <0.01 Stems+wlt69 11 2 1160.18 152.88 <0.01 Lengthcwd+wlt69 12 2 1162.69 155.39 <0.01 Stems 13 1 1163.41 156.11 <0.01 Lengthcwd 14 1 1169.56 162.26 <0.01 Wlt69 15 1 1182.28 174.98 <0.01 Model Rank k1 AICc2 ∆AICc3 AICcw4 Stems+julday+lengthcwd+wlt69 1 4 1007.30 0.00 0.65 Stems+julday+lengthcwd 2 3 1008.90 1.60 0.29 Stems+julday+wlt69 3 3 1012.96 5.66 0.04 Julday+lengthcwd+wlt69 4 3 1015.15 7.85 0.01 Stems+julday 5 2 1016.59 9.29 0.01 Julday+lengthcwd 6 2 1022.84 15.54 <0.01 Julday+wlt69 7 2 1034.18 26.88 <0.01 Julday 8 1 1057.47 50.17 <0.01 Stems+lengthcwd+wlt69 9 3 1154.75 147.45 <0.01 Stems+lengthcwd 10 2 1155.99 148.69 <0.01 Stems+wlt69 11 2 1160.18 152.88 <0.01 Lengthcwd+wlt69 12 2 1162.69 155.39 <0.01 Stems 13 1 1163.41 156.11 <0.01 Lengthcwd 14 1 1169.56 162.26 <0.01 Wlt69 15 1 1182.28 174.98 <0.01 1 Number of parameters used in the model. 2 Akaike’s information criterion for small sample sizes value. 3 Difference in Akaike’s information criterion for small sample sizes value from the most parsimonious model. 4 Weight of Akaike's information criterion value for small sample sizes. Table 3 Logistic regression models for presence–absence data of forest-floor small mammals at camera traps within and near a recently harvested clearcut in central British Columbia, Canada during summer 2015. Model Rank k1 AICc2 ∆AICc3 AICcw4 Corridor+cover+stems 1 3 5095.44 0.00 0.76 Corridor+cover 2 2 5099.12 3.68 0.12 Corridor+stems 3 2 5099.45 4.01 0.10 Corridor 4 1 5103.95 8.51 0.01 Cover+stems 5 2 5271.27 175.83 <0.01 Stems 6 1 5274.09 178.65 <0.01 Cover 7 1 5351.03 255.59 <0.01 Model Rank k1 AICc2 ∆AICc3 AICcw4 Corridor+cover+stems 1 3 5095.44 0.00 0.76 Corridor+cover 2 2 5099.12 3.68 0.12 Corridor+stems 3 2 5099.45 4.01 0.10 Corridor 4 1 5103.95 8.51 0.01 Cover+stems 5 2 5271.27 175.83 <0.01 Stems 6 1 5274.09 178.65 <0.01 Cover 7 1 5351.03 255.59 <0.01 1 Number of parameters used in the model. 2 Akaike’s information criterion for small sample sizes value. 3 Difference in Akaike’s information criterion for small sample sizes value from the most parsimonious model. 4 Weight of Akaike’s information criterion value for small sample sizes. Table 3 Logistic regression models for presence–absence data of forest-floor small mammals at camera traps within and near a recently harvested clearcut in central British Columbia, Canada during summer 2015. Model Rank k1 AICc2 ∆AICc3 AICcw4 Corridor+cover+stems 1 3 5095.44 0.00 0.76 Corridor+cover 2 2 5099.12 3.68 0.12 Corridor+stems 3 2 5099.45 4.01 0.10 Corridor 4 1 5103.95 8.51 0.01 Cover+stems 5 2 5271.27 175.83 <0.01 Stems 6 1 5274.09 178.65 <0.01 Cover 7 1 5351.03 255.59 <0.01 Model Rank k1 AICc2 ∆AICc3 AICcw4 Corridor+cover+stems 1 3 5095.44 0.00 0.76 Corridor+cover 2 2 5099.12 3.68 0.12 Corridor+stems 3 2 5099.45 4.01 0.10 Corridor 4 1 5103.95 8.51 0.01 Cover+stems 5 2 5271.27 175.83 <0.01 Stems 6 1 5274.09 178.65 <0.01 Cover 7 1 5351.03 255.59 <0.01 1 Number of parameters used in the model. 2 Akaike’s information criterion for small sample sizes value. 3 Difference in Akaike’s information criterion for small sample sizes value from the most parsimonious model. 4 Weight of Akaike’s information criterion value for small sample sizes. Results Throughout the summer 2015 sampling period, there were a total of 151 marten visits, 6834 forest-floor small mammal visits and 245 red squirrel visits to camera traps. The frequency of marten visits to camera traps decreased considerably from 49 visits in the week of June 5–11 to only 1–6 visits/week from July 3 to August 20. In winter 2017, there were 16 marten visits, 6 forest-floor small mammal visits and 51 red squirrel visits to camera traps. Marten habitat use During the summer, frequency of marten visits was highest in forested locations and lowest in clearcut and corridor locations (x2 = 12.188, P = 0.002, Figure 2). In winter 2017, although sample sizes were too small to test statistically, there were nearly twice as many detections of martens in corridors compared with random sites (Figure 2). Additionally, all marten visits to corridor sites occurred before March 22nd, while all visits by marten to random sites were recorded on March 18th or later. In summer, martens used corridors significantly more frequently at night than during the day – marten visits to corridors during the day were significantly less than to forest and clearcut sites (x2 = 12.154, P = 0.002, Figure 3), whereas at night martens used corridors as frequently as clearcut sites (x2 = 4.904, P = 0.086, Figure 3). Figure 2 View largeDownload slide Average number of marten visits (±1 SE) to camera traps in forest, clearcut, and corridor locations during summer 2015 and winter 2017 within and surrounding a clearcut in central British Columbia, Canada. Figure 2 View largeDownload slide Average number of marten visits (±1 SE) to camera traps in forest, clearcut, and corridor locations during summer 2015 and winter 2017 within and surrounding a clearcut in central British Columbia, Canada. Figure 3 View largeDownload slide Average number of marten visits (±1 SE) to camera traps in forest, clearcut and corridor locations during day and night in summer 2015, within and surrounding a clearcut in central British Columbia, Canada. Figure 3 View largeDownload slide Average number of marten visits (±1 SE) to camera traps in forest, clearcut and corridor locations during day and night in summer 2015, within and surrounding a clearcut in central British Columbia, Canada. The highest ranked logistic regression model explaining the presence of martens included parameters for number of stems, Julian day, mean length of CWD and number of Type 6–9 wildlife trees (dead with a broken top; Table 2). This model had useful predictive accuracy (AUC = 0.83). The coefficients generated from that model suggested the presence of martens was positively associated with number of stems, number of Type 6–9 wildlife trees and greater than average mean lengths of CWD (Figure 4). A negative coefficient for Julian day suggested that martens were more likely to be present at camera traps earlier in the summer study period. The probability of the presence of martens increased exponentially (y = 4E − 5e0.1457x, R2 = 1) with increasing mean length of CWD at a site. Figure 4 View largeDownload slide Coefficients and 95% confidence intervals of the top-ranked logistic regression model representing the influence of environmental and temporal variables on martens' presence within and surrounding a clearcut in central British Columbia, Canada during summer 2015. Figure 4 View largeDownload slide Coefficients and 95% confidence intervals of the top-ranked logistic regression model representing the influence of environmental and temporal variables on martens' presence within and surrounding a clearcut in central British Columbia, Canada during summer 2015. Small mammal habitat use Mice were the most common type of forest-floor small mammal detected, followed by voles and shrews, respectively (Figure 5). All forest-floor small mammals were significantly more likely to be detected in corridors compared with the forest or clearcut (x2 = 13.374, P = 0.001, Figure 5). There were approximately three times as many mice, twice as many voles and seven times as many shrews present in the corridors when compared with the forest or clearcut sites. Only summer data are presented for forest-floor small mammals, as detection in winter was low because they spend most of their time in subnivean habitat. Figure 5 View largeDownload slide Average number of forest-floor small mammal visits (±1 SE) to camera traps in forest, clearcut and corridor locations within and surrounding a clearcut in central British Columbia, Canada during summer 2015. Figure 5 View largeDownload slide Average number of forest-floor small mammal visits (±1 SE) to camera traps in forest, clearcut and corridor locations within and surrounding a clearcut in central British Columbia, Canada during summer 2015. The most parsimonious model for forest-floor small mammals included parameters for corridor, shrub cover and number of stems (Table 3). This model had a poor predictive accuracy (AUC = 0.61). During the summer, frequency of red squirrel visits was highest in forested locations and there was no difference in use between clearcut and corridor locations (Figure 6). In winter, however, red squirrels used corridors ~10 times more frequently than other clearcut locations (Figure 6). Figure 6 View largeDownload slide Average number of red squirrel visits (±1 SE) to camera traps in forest, clearcut, and corridor locations during summer 2015 and winter 2017 within and surrounding a clearcut in central British Columbia, Canada. Figure 6 View largeDownload slide Average number of red squirrel visits (±1 SE) to camera traps in forest, clearcut, and corridor locations during summer 2015 and winter 2017 within and surrounding a clearcut in central British Columbia, Canada. Discussion Marten habitat use Frequency of visits by martens to camera traps decreased considerably from 49 visits in the first week of June to only 1–6 visits/week during the months of July and August. This change in visitation rate could be related to the small spatial scale of the study. For example, the average home range size in the study area calculated from radio-collared male and female marten was 1012 ha and 626 ha, respectively (John Prince Research Forest, 2017), thus the death or emigration of an individual marten could greatly influence the frequency of visits across the 93 ha study area. As breeding season occurs in July and August (Powell et al. 2003), it is possible that martens dispersed away from the area if there were no females of reproductive age present in the study area. We could not individually identify all individual martens, and therefore cannot test the effect of immigration and emigration on visitation rates. Another possible explanation for the change in visitation rate is that martens were less drawn to the lure when food sources such as berries and bird or small mammal nests became more abundant throughout the summer. During the summer, martens used forested habitats most frequently and used the constructed CWD corridors with the same frequency as other areas of the clearcut. Limited data suggest that martens preferred the corridors to other areas in the clearcut during winter; although not statistically significant, there were nearly twice as many visits by martens to corridor vs random clearcut locations in winter 2017. Marten detections at corridor sites during the winter occurred on March 22nd or earlier but on March 18th or later at random sites. The temporal differences in detections between corridor and random sites may be explained by early spring conditions, as from the end of March to early April, the study area transitioned from being dominated by snow cover to being largely barren of snow with remnant patches. The behaviour and habitat use of martens was also likely transitioning during this time period. Martens may use CWD corridors more frequently in winter than in summer because they provide a high density of small mammal prey when other food sources such as berries and nests are not available. The high density of CWD in the corridors also provides marten with subnivean habitat such as resting sites, thermal insulation and refugia from predators (Buskirk et al., 1989). During the day in summer, martens used forested habitat most frequently, while at night martens were equally likely to use clearcut and forested habitats. This could be because there is an increased risk of predation from raptors in clearcuts during the day (Poole et al., 2004). Also, martens are more likely to use the corridors and clearcut locations at night to hunt nocturnal forest-floor mammals (Buskirk, 1983) that are most abundant in the CWD corridors within the clearcut. It is important to note that the study area had much greater amounts of dispersed CWD and higher levels of green tree retention than typically is present in clearcuts found across central British Columbia. Increased structural complexity and overhead cover along with the constructed CWD corridors may have provided sufficient habitat for martens such that they used the clearcut and corridor locations as frequently as the forest during the night. Given this pattern, it is possible that in clearcuts with little to no green tree retention and legal minimum amounts of CWD, constructed CWD corridors may be even more beneficial for martens and their prey. The presence of martens was best explained by the number of stems, number of Type 6–9 wildlife trees and mean length of CWD. These findings are consistent with the general understanding of the habitat ecology of American martens. It is well documented that martens select areas with dense forest (Spencer et al., 1983; Poole et al., 2004; Baldwin and Bender, 2008). For our study area, martens likely used sites with a high number of stems because these areas provide canopy cover that reduces the risk from aerial predators (Poole et al., 2004). Martens are known to select for mature forests because of the associated overhead cover and understory complexity (Thompson and Colgan, 1994). Type 6–9 wildlife trees are more likely to be present in older forest stands suggesting a similar pattern of use for martens in our study area (Keisker, 2000). Martens also use Type 6–9 wildlife trees as resting sites (Keisker, 2000). Although it is well-known that martens select for areas with high volumes of CWD (Buskirk et al., 1989; Sherburne and Bisonette, 1994; Poole et al., 2004; Andruskiw et al., 2008), there has been no empirical evidence relating the length of CWD pieces with the presence of martens. We found that the probability of a marten being present at a camera site increased exponentially with increasing mean length of CWD. This association could be because longer pieces of CWD provide greater habitat complexity, as they are more likely to stack on each other and become elevated instead of sinking into the ground. Small mammal habitat use All types of forest-floor small mammals (mice, voles and shrews) were observed more frequently in the corridors than in the forest or clearcut. There were three times as many mice, twice as many voles and seven times as many shrews present in the constructed corridors when compared with the forest or clearcut sites. This finding is consistent with work by Sullivan et al. (2011, 2012) and Sullivan and Sullivan (2014) who found that constructed woody debris corridors in clearcuts contained a higher abundance of forest-floor small mammals. The high density of small mammals in the corridors provides an abundant prey source for martens, as small mammals make up a significant portion of martens’ diets (Buskirk and Ruggiero, 1994; Martin, 1994). However, the relationship between the environmental factors that we measured and the presence of small mammals was weak. This is despite a large sample of observations. For our study area, forest-floor small mammals may be super abundant resulting in a near ubiquitous distribution of those species. Red squirrels are commonly known as a forest-dependent species that select for mature, closed-canopy forest and avoid clearcuts (Bakker and van Vuren, 2004; Haughland and Larsen, 2004). This was reflected in our results from summer 2015, as red squirrels were ~10 times more likely to be found in forested sites compared with clearcut or corridor sites. During winter, red squirrels used the CWD corridors ~10 times more frequently than other clearcut locations and as frequently as they used forested locations in summer. This indicates that red squirrels may be using the corridors to travel across or forage in clearcuts that they otherwise would have avoided. The corridors likely provide protection from predators and subnivean access for forage (Bakker and van Vuren, 2004). This high use of the corridors by red squirrels in winter increases the prey base for martens and results in the improvement of clearcut areas as habitat for both species. Conclusion Our results suggest that constructed CWD corridors are effective in enhancing clearcuts as habitat for martens and their prey. Corridors provide martens with resting sites, thermal insulation, refugia from predators and subnivean access to high densities of prey. Although martens did not use these corridors more than other clearcut locations during the second summer post-harvest, limited data suggest that martens preferred the corridors to other clearcut areas during winter. Furthermore, it is likely that the corridors will provide high-quality habitat for martens 30–40 years post-harvest when there is sufficient overhead cover to reduce risk from aerial predators. We recommend that forest managers retain areas of mature, dense forest as these areas provide overhead cover that martens require to mitigate predation risk and that long pieces of CWD be retained post-harvest as this may provide increased habitat complexity that martens require for prey capture and predator avoidance. Acknowledgements We gratefully acknowledge the people of the Tl’azt’en Nation for allowing us to conduct this research on their traditional territory. We would also like to thank D. Powe and the many student research assistants from the Tl’azt’en Nation who provided fieldwork assistance. We are also grateful for the contributions of S. Grainger of the John Prince Research Forest. We thank M. Gillingham and two anonymous reviewers for their helpful comments. Funding Natural Sciences and Engineering Research Council of Canada, the University of Northern British Columbia and the John Prince Research Forest. Conflict of interest statement None declared. References Agee , J.K. 1993 Fire ecology of Pacific Northwest forests . Island Press , p. 505 . Andruskiw , M. , Fryxell , J.M. , Thompson , I.D. and Baker , J.A. 2008 Habitat-mediated variation in predation risk by the American marten . Ecology 89 , 2273 – 2280 . Google Scholar CrossRef Search ADS PubMed Bakker , V.J. and van Vuren , D.H. 2004 Gap-crossing decisions by the red squirrel, a forest-dependent small mammal . Conserv. Biol. 18 , 689 – 697 . Google Scholar CrossRef Search ADS Baldwin , R.A. and Bender , L.C. 2008 Distribution, occupancy, and habitat correlates of American martens (Martes americana) in Rocky Mountain National Park, Colorado . J. Mammal. 89 , 419 – 427 . Google Scholar CrossRef Search ADS Boyce , M.S. , Vernier , P.R. , Nielsen , S.E. and Schmiegelow , F.K.A. 2002 Evaluating resource selection functions . Ecol. Model. 157 , 281 – 300 . Google Scholar CrossRef Search ADS British Columbia 2004 Forest and Ranges Practices Act Forest Planning and Practices Regulation . Queen’s Printer . British Columbia Chief Forester . 2010 Chief Forester’s Guidance on Coarse Woody Debris Management. www.for.gov.bc.ca/ftp/hfp/external/!publish/frep/extension/. British Columbia Forest Analysis and Inventory Branch 2015 Vegetation Resources Inventory Sampling Procedures. Version 5.2 . BC Ministry of Forests, Lands and Natural Resource Operations . Burnham , K.P. and Anderson , D.R. 2002 Model selection and multimodel inference: a practical information-theoretic approach . 2nd edn . Springer-Verlag , p. 488 . Buskirk , S.W. 1983 The Ecology of Marten in Southcentral Alaska . University of Alaska , p. 131 . Buskirk , S.W. , Forrest , S.C. , Raphael , M.G. and Harlow , H.J. 1989 Winter resting site ecology of marten in the central Rocky Mountains . J. Wildl. Manage. 53 , 191 – 196 . Google Scholar CrossRef Search ADS Buskirk , S.W. and Ruggiero , L. 1994 American marten. In The scientific basis for conserving large carnivores: American marten, fisher, lynx, and wolverine in the western United States . Ruggiero L. , Aubry K. , Buskirk S. , Lyon J. and Zielinski W. (eds). United States Department of Agriculture , pp. 17 – 37 . Fauteux , D. , Imbeau , L. , Drapeau , P. and Mazerolle , M.J. 2012 Small mammal responses to coarse woody debris distribution at different spatial scales in managed and unmanaged boreal forests . For. Ecol. Manage. 266 , 194 – 205 . Google Scholar CrossRef Search ADS Hargis , C.D. and McCullough , D.R. 1984 Winter diet and habitat selection of marten in Yosemite National Park . J. Wildl. Manage. 48 , 140 – 146 . Google Scholar CrossRef Search ADS Harmon , M.E. , Franklin , J.F. , Swanson , F.J. , Sollins , P. , Gregory , S.V. , Lattin , J.D. , et al. 1986 Ecology of coarse woody debris in temperate ecosystems . Adv. Ecol. Res. 15 , 133 – 267 . Google Scholar CrossRef Search ADS Haughland , D.L. , Larsen , K.W. 2004 Ecology of North American red squirrels across contrasting habitats: relating natal dispersal to habitat . J. Mammal. 85 , 225 – 236 . Google Scholar CrossRef Search ADS John Prince Research Forest 2017 . Unpublished data. Keisker , D.G. 2000 Types of wildlife trees and coarse woody debris required by wildlife of north-central British Columbia . Resource Branch Ministry of Forests . Lindenmayer , D.B. , Burton , P.J. and Franklin , J.F. 2012 Salvage logging and its ecological consequences . Island Press , p. 246 . Lofroth , E.C. 1993 Scale dependent analyses of habitat selection by marten in the Sub-Boreal Spruce biogeoclimatic zone, British Columbia. Master’s Thesis, Simon Fraser University, British Columbia. Mabee , W.E. and Saddler , J.N. 2009 Bioethanol from lignocellulosics: status and perspectives in Canada . Bioresour. Technol. 101 , 4806 – 4813 . Google Scholar CrossRef Search ADS PubMed Martin , S.K. 1994 Feeding ecology of American martens and fishers. In Martens, sables and fishers: biology and conservation . Buskirk S. , Harestad A. , Raphael M. and Powell R. (eds). Cornell University , pp. 297 – 315 . McComb , W.C. 2003 Ecology of coarse woody debris and its role as a habitat for mammals. In Mammal Community Dynamics. Management and Conservation in the Coniferous Forests of Western North America . Zabel C.J. and Anthony R.G. (eds). Cambridge University Press , pp. 374 – 404 . Google Scholar CrossRef Search ADS Meidinger , D.V. and Pojar , J. 1991 The Ecosystems of British Columbia . BC Ministry of Forests . Special report series no. 6. Poole , K.G. , Porter , A.D. , de Vries , A. , Maundrell , C. , Grindal , S.D. and St. Clair , C.C. 2004 Suitability of a young deciduous-dominated forest for American marten and the effects of forest removal . Can. J. Zool. 82 , 423 – 435 . Google Scholar CrossRef Search ADS Potvin , F. , Belanger , L. and Lowell , K. 2000 Marten habitat selection in a clearcut boreal landscape . Conserv. Biol. 14 , 844 – 857 . Google Scholar CrossRef Search ADS Powell , R.A. , Buskirk , S.W. and Zielinski , W.J. 2003 Fisher and Marten. In Wild Mammals of North America: Biology, Management, and Conservation . Feldhamer G.A. , Thompson B.C. and Chapman J.A. (eds). The John Hopkins University Press , pp. 635 – 646 . R Core Team . 2016 R: A language and environment for statistical computing. R Foundation for Statistical Computing, Vienna, Austria. http://www.R-project.org/. Rogers , W.H. 1993 . sg17: Regression standard errors in clustered samples. Stata Technical Bulletin 13: 19–23. Reprinted in Stata Technical Bulletin Reprints, vol. 3, pp. 88–94. College Station, TX: Stata Press. Sherburne , S.S. 1993 Squirrel middens influence marten (Martes americana) use of subnivean access points . Am. Midl. Nat. 129 , 204 – 207 . Google Scholar CrossRef Search ADS Sherburne , S.S. and Bisonette , J.A. 1994 Marten subnivean access point use: response to subnivean prey levels . J. Wildl. Manage. 58 , 400 – 405 . Google Scholar CrossRef Search ADS Spencer , W.D. , Barrett , R.H. and Zielinski , W.J. 1983 Marten habitat preferences in the northern Sierra Nevada . J. Wildl. Manage. 47 , 1181 – 1186 . Google Scholar CrossRef Search ADS StataCorp . 2011 Stata Statistical Software: Release 12. College Station, TX: StataCorp LP. Sullivan , T.P. and Sullivan , D.S. 2014 Diversifying clearcuts with green-tree retention and woody debris structures: conservation of mammals across forest ecological zones . Silva Fenn 48.5 , id 1219 . Sullivan , T.P. , Sullivan , D.S. , Lindgren , P.M.F. and Ransome , D.B. 2012 If we build habitat, will they come? Woody debris structures and conservation of forest mammals . J. Mammal. 93 , 1456 – 1468 . Google Scholar CrossRef Search ADS Sullivan , T.P. , Sullivan , D.S. , Lindgren , P.M.F. , Ransome , D.B. , Bull , J.G. and Ristea , C. 2011 Bioenergy or biodiversity? Woody debris structures and maintenance of red-backed voles on clearcuts . Biomass Bioenergy 35 , 4390 – 4398 . Google Scholar CrossRef Search ADS Thompson , I.D. and Colgan , P.W. 1994 Marten activity in uncut and logged boreal forests in Ontario . J. Wildl. Manage. 58 , 280 – 288 . Google Scholar CrossRef Search ADS Thompson , I.D. , Fryxell , J.M. and Harrison , D.J. 2012 Improved insights into use of habitat by American martens. In Biology and Conservation of Martens, Sables, and Fishers . Aubry K.B. , Zielinski W.J. , Raphael M.J. , Proulx G. and Buskirk S.W. (eds). Cornell University Press , pp. 209 – 230 . Toledo , D.P. , Herrick , J.E. and Abbott , L.B. 2010 A comparison of cover pole with standard vegetation monitoring methods . J Wildl. Manage. 74 , 600 – 604 . Google Scholar CrossRef Search ADS Wiebe , P.A. , Fryxell , J.M. , Thompson , I.D. , Börger , L. and Baker , J.A. 2013 Do trappers understand marten habitat? J. Wildl. Manage. 77 , 379 – 391 . Google Scholar CrossRef Search ADS Wiebe , P.A. , Thompson , I.D. , McKague , C.I. , Fryxell , J.M. and Baker , J.A. 2014 Fine-scale winter resource selection by American martens in boreal forests and the effect of snow depth on access to coarse woody debris . Écoscience 21 , 123 – 132 . Google Scholar CrossRef Search ADS © Institute of Chartered Foresters, 2018. All rights reserved. For Permissions, please e-mail: journals.permissions@oup.com. This article is published and distributed under the terms of the Oxford University Press, Standard Journals Publication Model (https://academic.oup.com/journals/pages/about_us/legal/notices) http://www.deepdyve.com/assets/images/DeepDyve-Logo-lg.png Forestry: An International Journal Of Forest Research Oxford University Press

Use of constructed coarse woody debris corridors in a clearcut by American martens (Martes americana) and their prey

Loading next page...
 
/lp/ou_press/use-of-constructed-coarse-woody-debris-corridors-in-a-clearcut-by-lEtPHDcRLo
Publisher
Institute of Chartered Foresters
Copyright
© Institute of Chartered Foresters, 2018. All rights reserved. For Permissions, please e-mail: journals.permissions@oup.com.
ISSN
0015-752X
eISSN
1464-3626
D.O.I.
10.1093/forestry/cpy010
Publisher site
See Article on Publisher Site

Abstract

Abstract American martens (Martes americana) are typically associated with mature coniferous forests because they provide overhead cover and coarse woody debris (CWD) that martens require for protection and hunting. Therefore, clearcuts are considered poor marten habitat because they contain no overhead cover and relatively little CWD. We examined the efficacy of retaining CWD and constructing CWD corridors in a recently harvested clearcut to promote the use of the area by martens and small mammals, a major prey resource. We installed remote cameras in corridors, the surrounding clearcut and forest to monitor the distribution of martens and small mammals. Martens and red squirrels did not use CWD corridors more frequently than clearcut areas in summer; forest-floor small mammals, however, used corridors approximately three times as frequently as other habitats (x2 = 13.374, P = 0.001). Marten presence was positively associated with mature, dense forest and long pieces of CWD. In winter, red squirrels used corridors more frequently than other clearcut areas, and limited data suggested that martens preferred the corridors to other clearcut areas. Consequently, we recommend that forest managers retain CWD and construct CWD corridors within clearcuts to provide small mammal habitat, and to enhance marten habitat. Introduction Coarse woody debris (CWD) is an important component of forest ecosystems because of the many ecological functions it provides, including nutrient cycling, energy flow and habitat for wildlife (Harmon et al., 1986). Many mammal species use CWD as a habitat element because it can provide cover from predators, insulation, subnivean habitat and hunting opportunities (McComb, 2003). CWD is defined as any piece of wood >7.5 cm in diameter, and can be created by natural or anthropogenic disturbances (British Columbia Forest Analysis and Inventory Branch, 2015). Retention of CWD in clearcuts may mitigate habitat loss by providing cover for many wildlife species that require complex forest structure (Hargis and McCullough, 1984; Andruskiw et al., 2008). American martens (Martes americana) are small carnivores from the family Mustelidae that inhabit most of the coniferous forests of Canada, Alaska and the Rocky Mountain states of the USA (Powell et al., 2003). Habitat for American martens has been characterized as mature coniferous and mixed-wood forests that contain old trees and large amounts of CWD (Lofroth, 1993; Thompson and Colgan, 1994; Poole et al., 2004; Andruskiw et al., 2008; Thompson et al., 2012). Martens select these forests because they provide overhead cover for protection from aerial predators (Poole et al., 2004), and contain high densities of mice (Peromyscus spp. and Zapus spp.), voles (Microtus spp. and Myodes spp.), shrews (Sorex spp.) and red squirrels (Tamiasciurus hudsonicus), their major prey items (Sherburne, 1993; Buskirk and Ruggiero, 1994; Martin, 1994). CWD also provide martens with subnivean habitat during winter (Wiebe et al., 2014) such as resting sites, thermal insulation and refugia from predators (Buskirk et al., 1989) as well as access to prey (Sherburne and Bisonette, 1994). Martens generally avoid non-forested areas because of their lack of structural complexity and overhead cover (Buskirk et al., 1989; Potvin et al., 2000; Wiebe et al., 2013) but there is a paucity of knowledge about whether martens select for CWD within recently harvested areas. Forest-floor small mammals such as mice, voles and shrews are found in higher densities in areas with large amounts of CWD (Fauteux et al., 2012), including constructed CWD corridors in clearcuts (Sullivan et al., 2011, 2012), while red squirrels are reported to avoid harvested areas (Bakker and van Vuren, 2004). Because of the known associations of martens and their forest-floor small mammal prey with CWD, retention of CWD in clearcuts may help to mitigate the adverse effects of forest harvest on martens. Starting in the 1990s, an unprecedented outbreak of mountain pine beetle (Dendroctonus ponderosae) in British Columbia, Canada, resulted in the death of some 13 million hectares of mature lodgepole pine (Pinus contorta Douglas ex Louden; Lindenmayer et al., 2012). This prompted an increase in clearcut harvest to salvage dead lodgepole pine. This strategy was designed to maximize the value of dead standing trees and mitigate economic losses associated with a significant decline in timber supply (Lindenmayer et al., 2012). Following clearcut harvest, woody debris is often removed from harvested areas to reduce the risk of wildfire and further insect outbreaks (Agee, 1993). The debris is either burned onsite or transported to bioenergy plants to create electricity (Mabee and Saddler, 2009) but retaining CWD in clearcuts to create habitat for wildlife such as martens and their prey may be more beneficial. In an effort to understand the relationship between CWD and these wildlife species, we used presence–absence camera data to address two research objectives: (1) to determine if martens and/or small mammals use constructed CWD corridors more frequently than areas with dispersed CWD in a clearcut and (2) to determine the fine-scale habitat characteristics that best predict marten and forest floor small mammal presence within and surrounding a recent clearcut. Methods Study area Initial research was conducted from May to August 2015, and supplemental data were collected from February to April 2017 within the John Prince Research Forest (JPRF), located ~50 km north of Fort St. James in central British Columbia, Canada (54°36.417′N; 124°21.237′W). The research forest is located within the Sub-Boreal Spruce biogeoclimatic zone and is characterized by severe snowy winters, short warm summers and moderate annual precipitation (Meidinger and Pojar, 1991). The study site was located within and adjacent to Block 67, a 93-ha clearcut that was harvested in the winter of 2014 (Figure 1). The block had a primarily southeastern aspect with undulating topography. Figure 1 View largeDownload slide Map of study area within and surrounding a clearcut, with inset map indicating the location of the John Prince Research Forest in central British Columbia, Canada. Figure 1 View largeDownload slide Map of study area within and surrounding a clearcut, with inset map indicating the location of the John Prince Research Forest in central British Columbia, Canada. Prior to harvest, the stand was dominated by a mixture of paper birch (Betula papyrifera Marsh.), hybrid interior spruce (Picea glauca (Moench) Voss × Picea engelmannii Parry ex Elngelm.) and Douglas-fir (Pseudotsuga menziesii (Mirb.) Franco) with average diameters of 37–45 cm and average heights of 26–29 m. The silvicultural prescription for the clearcut was to harvest marketable hybrid interior spruce and Douglas-fir while retaining all deciduous trees and Douglas-fir with diameters >55 cm. Several wildlife tree patches were retained to protect an ungulate mineral lick, a potential fisher (Pekania pennanti) denning habitat, a previous bear denning site and to maintain connectivity throughout the cut area. The harvest prescription called for high CWD retention targets post-harvest in the clearcut (30–50 m3 ha−1). This level is significantly higher than the legal minimum requirement for interior British Columbia which has been set at four logs per hectare, each being a minimum 2 m in length and 7.5 cm at one end (British Columbia, 2004). Industrial standards, however, commonly exceed this target as suggested by the British Columbia Chief Forester (2010); average CWD retention by industrial operators was ~5 m3 ha−1 for habitats similar to the study area. Much of the residual CWD was used to construct nine corridors using an excavator backhoe within the boundaries of the clearcut. The treatment was installed over a period of 3 days during September 2014 and was completed separately from logging operations to ensure the prescription was appropriate for future monitoring. Cost for constructing the corridors was ~$200 CAD/100 m of corridor. The corridors were composed of a variety of types and sizes of wood, were ~1 m high × ~5 m wide, and ranged between 100 m and 200 m in length. The corridors were intended to connect areas of surrounding forest to wildlife tree patches within the clearcut (Figure 1) and to serve as resting and hunting habitat for mesocarnivores such as marten. Survey design In summer 2015, 42 passive infra-red remote video cameras (Bushnell 8MP Trophy Cam HD Black LED Trail Camera with Night Vision model #119676CN, Bushnell Outdoor Products) were installed systematically throughout the sampling area on a hexagonal grid with 200 m spacing between cameras (Figure 1). In addition, one camera was installed in each of the nine corridors, as close to the midpoint as possible while maintaining a minimum distance of 50 m from cameras located in adjacent plots. Systematically placed cameras were set to sensor level: medium; LED control: medium and 30 s video recording. Corridor cameras were set to sensor level: low because the contrast of vegetation movement against the woody debris on medium sensor level resulted in memory cards filled with videos of vegetation movement. Cameras were baited weekly with Sharpe’s marten lure (a short-range, strawberry and anise-scented lure) and 5 ml of peanut butter placed 2 m in front of the camera on a stick or log. Cameras were positioned 0.5 m from the ground on a tree or wooden stake, pointed north-west to minimize backlighting, and aimed towards the bait. A 0.5-m tall stick was placed 1 m behind the bait so that the approximate size of animals could be inferred from videos. In winter 2017, 18 cameras were installed in the harvested area: 9 in the corridors and 9 randomly in the clearcut, a minimum of 50 m away from corridors. The cameras were baited with a small (~75 g) piece of hanging meat (American beaver; Castor canadensis Kuhl) and a commercial lure (a combination of Sharpe’s lynx and mink lures) placed on a small log 2–3 m in front of the camera. The cameras were set to sensor level: medium; LED control: medium and 30 s video recording. Sampling schedule In summer 2015, cameras were installed between May 24 and 28 and remained active until August 20th. This sampling period allowed for observation of seasonal variation in the presence of martens and small mammals from before vegetation flushed until leaves were in full flush. In winter 2017, cameras were active from February 3 to April 6 and baited/lured at the beginning of three 3-week sessions. Animals captured in the videos were identified to species if possible, and the length of time that each animal stayed at the camera was recorded. Visits were considered separate if the time from the end of one video to the beginning of the next exceeded 1 min. Habitat assessment At each of the camera sites in summer 2015, slope (%), slope position, elevation and aspect were measured. We also assessed shrub cover within a 3.99-m radius plot (50 m2) around each camera. A 1.5-m cover pole divided into 50 cm sections and 10 cm bands was used to assess cover. If shrubs obscured >25% of a band it was considered covered (Toledo et al., 2010). We recorded cover measurements weekly, and as we found no seasonal trend in measurements we averaged them over the summer study period. Within an 11.28- m radius plot (400 m2), we counted the number of tree stems >12.5 cm diameter at breast height (DBH, British Columbia Forest Analysis and Inventory Branch, 2015), squirrel middens (piles of cone bracts discarded by red squirrels), tip-up mounds (exposed roots of fallen trees), stumps and saplings. For all trees >12.5 cm DBH, we measured DBH and identified trees to species level. We also categorized all trees into a wildlife tree type: Types 1 and 2: live; Types 3–5: dead with intact top or Types 6–9: dead with a broken top (Keisker, 2000). At each camera, three 50-m transects oriented at 0°, 120° and 240° from the camera were used to measure CWD. All CWD > 7.5 cm diameter (British Columbia Forest Analysis and Inventory Branch, 2015) that intersected the transect line was classified by decomposition class: Type 1: little to no decay; Type 2: significant decay but wood retained structure or Type 3: structure of wood was lost. For each piece of CWD, we recorded the diameter, length, orientation to the transect line, tree type (deciduous or conifer) and whether it was on or off the ground. We used a geographic information system to measure the proximity of each camera to CWD corridors, forest edge, wildlife tree patches, roads, streams and open water. Statistical analysis We used data from camera records to construct logistic regression models that represented the relationship between the presence-absence of martens and forest-floor small mammals during summer and habitat variables measured at each camera site (Table 1). The cameras functioned continuously, thus, we identified a sampling interval as a 24-h period. Using exploratory analysis of the habitat variables measured at each camera site, we constructed candidate models for martens (Table 2) and forest-floor small mammals (Table 3), which we then ranked using the differences in Akaike’s information criteria for small sample sizes (ΔAICc) (Burnham and Anderson, 2002) and associated weights (AICcw) to select the most parsimonious logistic regression models. We used the cluster option in Stata (ver.12.1 StataCorp LP, 2011; Rogers, 1993) to correct standard errors for multiple observations at the same site. We used the receiver operating characteristic to evaluate the predictive accuracy of the most parsimonious models. Models with an area under the curve (AUC) score of 0.7–0.9 were considered useful (Boyce et al., 2002). The relative strengths of the coefficients within the most parsimonious model were assessed using 95% confidence intervals. Coefficients were considered significant factors predicting presence of martens or forest floor small mammals if the interval did not overlap zero. These analyses were completed using Stata. Frequencies of habitat use were compared using a chi-square test. A Kruskal–Wallis test in R version 3.2.4 (R Core Team, 2016) was used to determine significant differences in the presence of martens and small mammals in different habitats, and during the day vs night. Table 1 Variables tested to construct logistic regression models describing the presence of martens and small mammals within and surrounding a recently harvested clearcut in central British Columbia, Canada during summer 2015. Variable Type Description Slope Continuous % Slope Aspect Continuous Aspect (degrees) Elevation Continuous Elevation (m) Slope position Categorical Slope position (1–5) Location Categorical Forest (1) or clearcut (0) Corridor Categorical Camera in corridor (1) or not (0) #Middens Continuous # of middens present in 11.28 m plot #Tipups Continuous # of root tip-up present in 11.28 m plot #Stumps Continuous # of stumps present in 11.28 m plot #Saplings Continuous # of conifer saplings (<1.3 m height) present in 11.28 m plot #Stems Continuous # trees >12.5 cm DBH present in 11.28 m plot Mean DBH Continuous Mean DBH of trees >12.5 cm DBH in 11.28 m plot (cm) % Conifer trees Continuous Per cent of conifer trees of trees >12.5 cm DBH in 11.28 m plot #WLT3–5 Continuous Number of wildlife trees Types 3–5 in 11.28 m plot #WLT6–9 Continuous Number of wildlife trees Types 6–9 in 11.28 m plot #PiecesCWD Continuous Total # of pieces of CWD on transects MeanDBHCWD Continuous Mean diameter of CWD >7.5 cm on transects (cm) MeanlengthCWD Continuous Mean length of CWD on transects (m) % CWD C1,2 Continuous Per cent of CWD in decay class 1 or 2 on transects % CWD conifer Continuous Per cent of CWD conifer on transects % CWD ON Continuous Per cent of CWD on the ground on transects Volume CWD Continuous Total volume of CWD on transects (m3) AV % cover Continuous Average per cent shrub cover <1.5 m CorDist Continuous Distance to nearest corridor (m) EdgeDist Continuous Distance to forest edge (m) WTPDist Continuous Distance to wildlife tree patch (m) RoadDist Continuous Distance to road (m) LakeDist Continuous Distance to lake (m) CreekDist Continuous Distance to creek (m) JulDay Continuous Julian day Variable Type Description Slope Continuous % Slope Aspect Continuous Aspect (degrees) Elevation Continuous Elevation (m) Slope position Categorical Slope position (1–5) Location Categorical Forest (1) or clearcut (0) Corridor Categorical Camera in corridor (1) or not (0) #Middens Continuous # of middens present in 11.28 m plot #Tipups Continuous # of root tip-up present in 11.28 m plot #Stumps Continuous # of stumps present in 11.28 m plot #Saplings Continuous # of conifer saplings (<1.3 m height) present in 11.28 m plot #Stems Continuous # trees >12.5 cm DBH present in 11.28 m plot Mean DBH Continuous Mean DBH of trees >12.5 cm DBH in 11.28 m plot (cm) % Conifer trees Continuous Per cent of conifer trees of trees >12.5 cm DBH in 11.28 m plot #WLT3–5 Continuous Number of wildlife trees Types 3–5 in 11.28 m plot #WLT6–9 Continuous Number of wildlife trees Types 6–9 in 11.28 m plot #PiecesCWD Continuous Total # of pieces of CWD on transects MeanDBHCWD Continuous Mean diameter of CWD >7.5 cm on transects (cm) MeanlengthCWD Continuous Mean length of CWD on transects (m) % CWD C1,2 Continuous Per cent of CWD in decay class 1 or 2 on transects % CWD conifer Continuous Per cent of CWD conifer on transects % CWD ON Continuous Per cent of CWD on the ground on transects Volume CWD Continuous Total volume of CWD on transects (m3) AV % cover Continuous Average per cent shrub cover <1.5 m CorDist Continuous Distance to nearest corridor (m) EdgeDist Continuous Distance to forest edge (m) WTPDist Continuous Distance to wildlife tree patch (m) RoadDist Continuous Distance to road (m) LakeDist Continuous Distance to lake (m) CreekDist Continuous Distance to creek (m) JulDay Continuous Julian day Table 1 Variables tested to construct logistic regression models describing the presence of martens and small mammals within and surrounding a recently harvested clearcut in central British Columbia, Canada during summer 2015. Variable Type Description Slope Continuous % Slope Aspect Continuous Aspect (degrees) Elevation Continuous Elevation (m) Slope position Categorical Slope position (1–5) Location Categorical Forest (1) or clearcut (0) Corridor Categorical Camera in corridor (1) or not (0) #Middens Continuous # of middens present in 11.28 m plot #Tipups Continuous # of root tip-up present in 11.28 m plot #Stumps Continuous # of stumps present in 11.28 m plot #Saplings Continuous # of conifer saplings (<1.3 m height) present in 11.28 m plot #Stems Continuous # trees >12.5 cm DBH present in 11.28 m plot Mean DBH Continuous Mean DBH of trees >12.5 cm DBH in 11.28 m plot (cm) % Conifer trees Continuous Per cent of conifer trees of trees >12.5 cm DBH in 11.28 m plot #WLT3–5 Continuous Number of wildlife trees Types 3–5 in 11.28 m plot #WLT6–9 Continuous Number of wildlife trees Types 6–9 in 11.28 m plot #PiecesCWD Continuous Total # of pieces of CWD on transects MeanDBHCWD Continuous Mean diameter of CWD >7.5 cm on transects (cm) MeanlengthCWD Continuous Mean length of CWD on transects (m) % CWD C1,2 Continuous Per cent of CWD in decay class 1 or 2 on transects % CWD conifer Continuous Per cent of CWD conifer on transects % CWD ON Continuous Per cent of CWD on the ground on transects Volume CWD Continuous Total volume of CWD on transects (m3) AV % cover Continuous Average per cent shrub cover <1.5 m CorDist Continuous Distance to nearest corridor (m) EdgeDist Continuous Distance to forest edge (m) WTPDist Continuous Distance to wildlife tree patch (m) RoadDist Continuous Distance to road (m) LakeDist Continuous Distance to lake (m) CreekDist Continuous Distance to creek (m) JulDay Continuous Julian day Variable Type Description Slope Continuous % Slope Aspect Continuous Aspect (degrees) Elevation Continuous Elevation (m) Slope position Categorical Slope position (1–5) Location Categorical Forest (1) or clearcut (0) Corridor Categorical Camera in corridor (1) or not (0) #Middens Continuous # of middens present in 11.28 m plot #Tipups Continuous # of root tip-up present in 11.28 m plot #Stumps Continuous # of stumps present in 11.28 m plot #Saplings Continuous # of conifer saplings (<1.3 m height) present in 11.28 m plot #Stems Continuous # trees >12.5 cm DBH present in 11.28 m plot Mean DBH Continuous Mean DBH of trees >12.5 cm DBH in 11.28 m plot (cm) % Conifer trees Continuous Per cent of conifer trees of trees >12.5 cm DBH in 11.28 m plot #WLT3–5 Continuous Number of wildlife trees Types 3–5 in 11.28 m plot #WLT6–9 Continuous Number of wildlife trees Types 6–9 in 11.28 m plot #PiecesCWD Continuous Total # of pieces of CWD on transects MeanDBHCWD Continuous Mean diameter of CWD >7.5 cm on transects (cm) MeanlengthCWD Continuous Mean length of CWD on transects (m) % CWD C1,2 Continuous Per cent of CWD in decay class 1 or 2 on transects % CWD conifer Continuous Per cent of CWD conifer on transects % CWD ON Continuous Per cent of CWD on the ground on transects Volume CWD Continuous Total volume of CWD on transects (m3) AV % cover Continuous Average per cent shrub cover <1.5 m CorDist Continuous Distance to nearest corridor (m) EdgeDist Continuous Distance to forest edge (m) WTPDist Continuous Distance to wildlife tree patch (m) RoadDist Continuous Distance to road (m) LakeDist Continuous Distance to lake (m) CreekDist Continuous Distance to creek (m) JulDay Continuous Julian day Table 2 Logistic regression models for presence–absence data of American martens at camera traps within and near a recently harvested clearcut in central British Columbia, Canada during summer 2015. Model Rank k1 AICc2 ∆AICc3 AICcw4 Stems+julday+lengthcwd+wlt69 1 4 1007.30 0.00 0.65 Stems+julday+lengthcwd 2 3 1008.90 1.60 0.29 Stems+julday+wlt69 3 3 1012.96 5.66 0.04 Julday+lengthcwd+wlt69 4 3 1015.15 7.85 0.01 Stems+julday 5 2 1016.59 9.29 0.01 Julday+lengthcwd 6 2 1022.84 15.54 <0.01 Julday+wlt69 7 2 1034.18 26.88 <0.01 Julday 8 1 1057.47 50.17 <0.01 Stems+lengthcwd+wlt69 9 3 1154.75 147.45 <0.01 Stems+lengthcwd 10 2 1155.99 148.69 <0.01 Stems+wlt69 11 2 1160.18 152.88 <0.01 Lengthcwd+wlt69 12 2 1162.69 155.39 <0.01 Stems 13 1 1163.41 156.11 <0.01 Lengthcwd 14 1 1169.56 162.26 <0.01 Wlt69 15 1 1182.28 174.98 <0.01 Model Rank k1 AICc2 ∆AICc3 AICcw4 Stems+julday+lengthcwd+wlt69 1 4 1007.30 0.00 0.65 Stems+julday+lengthcwd 2 3 1008.90 1.60 0.29 Stems+julday+wlt69 3 3 1012.96 5.66 0.04 Julday+lengthcwd+wlt69 4 3 1015.15 7.85 0.01 Stems+julday 5 2 1016.59 9.29 0.01 Julday+lengthcwd 6 2 1022.84 15.54 <0.01 Julday+wlt69 7 2 1034.18 26.88 <0.01 Julday 8 1 1057.47 50.17 <0.01 Stems+lengthcwd+wlt69 9 3 1154.75 147.45 <0.01 Stems+lengthcwd 10 2 1155.99 148.69 <0.01 Stems+wlt69 11 2 1160.18 152.88 <0.01 Lengthcwd+wlt69 12 2 1162.69 155.39 <0.01 Stems 13 1 1163.41 156.11 <0.01 Lengthcwd 14 1 1169.56 162.26 <0.01 Wlt69 15 1 1182.28 174.98 <0.01 1 Number of parameters used in the model. 2 Akaike’s information criterion for small sample sizes value. 3 Difference in Akaike’s information criterion for small sample sizes value from the most parsimonious model. 4 Weight of Akaike's information criterion value for small sample sizes. Table 2 Logistic regression models for presence–absence data of American martens at camera traps within and near a recently harvested clearcut in central British Columbia, Canada during summer 2015. Model Rank k1 AICc2 ∆AICc3 AICcw4 Stems+julday+lengthcwd+wlt69 1 4 1007.30 0.00 0.65 Stems+julday+lengthcwd 2 3 1008.90 1.60 0.29 Stems+julday+wlt69 3 3 1012.96 5.66 0.04 Julday+lengthcwd+wlt69 4 3 1015.15 7.85 0.01 Stems+julday 5 2 1016.59 9.29 0.01 Julday+lengthcwd 6 2 1022.84 15.54 <0.01 Julday+wlt69 7 2 1034.18 26.88 <0.01 Julday 8 1 1057.47 50.17 <0.01 Stems+lengthcwd+wlt69 9 3 1154.75 147.45 <0.01 Stems+lengthcwd 10 2 1155.99 148.69 <0.01 Stems+wlt69 11 2 1160.18 152.88 <0.01 Lengthcwd+wlt69 12 2 1162.69 155.39 <0.01 Stems 13 1 1163.41 156.11 <0.01 Lengthcwd 14 1 1169.56 162.26 <0.01 Wlt69 15 1 1182.28 174.98 <0.01 Model Rank k1 AICc2 ∆AICc3 AICcw4 Stems+julday+lengthcwd+wlt69 1 4 1007.30 0.00 0.65 Stems+julday+lengthcwd 2 3 1008.90 1.60 0.29 Stems+julday+wlt69 3 3 1012.96 5.66 0.04 Julday+lengthcwd+wlt69 4 3 1015.15 7.85 0.01 Stems+julday 5 2 1016.59 9.29 0.01 Julday+lengthcwd 6 2 1022.84 15.54 <0.01 Julday+wlt69 7 2 1034.18 26.88 <0.01 Julday 8 1 1057.47 50.17 <0.01 Stems+lengthcwd+wlt69 9 3 1154.75 147.45 <0.01 Stems+lengthcwd 10 2 1155.99 148.69 <0.01 Stems+wlt69 11 2 1160.18 152.88 <0.01 Lengthcwd+wlt69 12 2 1162.69 155.39 <0.01 Stems 13 1 1163.41 156.11 <0.01 Lengthcwd 14 1 1169.56 162.26 <0.01 Wlt69 15 1 1182.28 174.98 <0.01 1 Number of parameters used in the model. 2 Akaike’s information criterion for small sample sizes value. 3 Difference in Akaike’s information criterion for small sample sizes value from the most parsimonious model. 4 Weight of Akaike's information criterion value for small sample sizes. Table 3 Logistic regression models for presence–absence data of forest-floor small mammals at camera traps within and near a recently harvested clearcut in central British Columbia, Canada during summer 2015. Model Rank k1 AICc2 ∆AICc3 AICcw4 Corridor+cover+stems 1 3 5095.44 0.00 0.76 Corridor+cover 2 2 5099.12 3.68 0.12 Corridor+stems 3 2 5099.45 4.01 0.10 Corridor 4 1 5103.95 8.51 0.01 Cover+stems 5 2 5271.27 175.83 <0.01 Stems 6 1 5274.09 178.65 <0.01 Cover 7 1 5351.03 255.59 <0.01 Model Rank k1 AICc2 ∆AICc3 AICcw4 Corridor+cover+stems 1 3 5095.44 0.00 0.76 Corridor+cover 2 2 5099.12 3.68 0.12 Corridor+stems 3 2 5099.45 4.01 0.10 Corridor 4 1 5103.95 8.51 0.01 Cover+stems 5 2 5271.27 175.83 <0.01 Stems 6 1 5274.09 178.65 <0.01 Cover 7 1 5351.03 255.59 <0.01 1 Number of parameters used in the model. 2 Akaike’s information criterion for small sample sizes value. 3 Difference in Akaike’s information criterion for small sample sizes value from the most parsimonious model. 4 Weight of Akaike’s information criterion value for small sample sizes. Table 3 Logistic regression models for presence–absence data of forest-floor small mammals at camera traps within and near a recently harvested clearcut in central British Columbia, Canada during summer 2015. Model Rank k1 AICc2 ∆AICc3 AICcw4 Corridor+cover+stems 1 3 5095.44 0.00 0.76 Corridor+cover 2 2 5099.12 3.68 0.12 Corridor+stems 3 2 5099.45 4.01 0.10 Corridor 4 1 5103.95 8.51 0.01 Cover+stems 5 2 5271.27 175.83 <0.01 Stems 6 1 5274.09 178.65 <0.01 Cover 7 1 5351.03 255.59 <0.01 Model Rank k1 AICc2 ∆AICc3 AICcw4 Corridor+cover+stems 1 3 5095.44 0.00 0.76 Corridor+cover 2 2 5099.12 3.68 0.12 Corridor+stems 3 2 5099.45 4.01 0.10 Corridor 4 1 5103.95 8.51 0.01 Cover+stems 5 2 5271.27 175.83 <0.01 Stems 6 1 5274.09 178.65 <0.01 Cover 7 1 5351.03 255.59 <0.01 1 Number of parameters used in the model. 2 Akaike’s information criterion for small sample sizes value. 3 Difference in Akaike’s information criterion for small sample sizes value from the most parsimonious model. 4 Weight of Akaike’s information criterion value for small sample sizes. Results Throughout the summer 2015 sampling period, there were a total of 151 marten visits, 6834 forest-floor small mammal visits and 245 red squirrel visits to camera traps. The frequency of marten visits to camera traps decreased considerably from 49 visits in the week of June 5–11 to only 1–6 visits/week from July 3 to August 20. In winter 2017, there were 16 marten visits, 6 forest-floor small mammal visits and 51 red squirrel visits to camera traps. Marten habitat use During the summer, frequency of marten visits was highest in forested locations and lowest in clearcut and corridor locations (x2 = 12.188, P = 0.002, Figure 2). In winter 2017, although sample sizes were too small to test statistically, there were nearly twice as many detections of martens in corridors compared with random sites (Figure 2). Additionally, all marten visits to corridor sites occurred before March 22nd, while all visits by marten to random sites were recorded on March 18th or later. In summer, martens used corridors significantly more frequently at night than during the day – marten visits to corridors during the day were significantly less than to forest and clearcut sites (x2 = 12.154, P = 0.002, Figure 3), whereas at night martens used corridors as frequently as clearcut sites (x2 = 4.904, P = 0.086, Figure 3). Figure 2 View largeDownload slide Average number of marten visits (±1 SE) to camera traps in forest, clearcut, and corridor locations during summer 2015 and winter 2017 within and surrounding a clearcut in central British Columbia, Canada. Figure 2 View largeDownload slide Average number of marten visits (±1 SE) to camera traps in forest, clearcut, and corridor locations during summer 2015 and winter 2017 within and surrounding a clearcut in central British Columbia, Canada. Figure 3 View largeDownload slide Average number of marten visits (±1 SE) to camera traps in forest, clearcut and corridor locations during day and night in summer 2015, within and surrounding a clearcut in central British Columbia, Canada. Figure 3 View largeDownload slide Average number of marten visits (±1 SE) to camera traps in forest, clearcut and corridor locations during day and night in summer 2015, within and surrounding a clearcut in central British Columbia, Canada. The highest ranked logistic regression model explaining the presence of martens included parameters for number of stems, Julian day, mean length of CWD and number of Type 6–9 wildlife trees (dead with a broken top; Table 2). This model had useful predictive accuracy (AUC = 0.83). The coefficients generated from that model suggested the presence of martens was positively associated with number of stems, number of Type 6–9 wildlife trees and greater than average mean lengths of CWD (Figure 4). A negative coefficient for Julian day suggested that martens were more likely to be present at camera traps earlier in the summer study period. The probability of the presence of martens increased exponentially (y = 4E − 5e0.1457x, R2 = 1) with increasing mean length of CWD at a site. Figure 4 View largeDownload slide Coefficients and 95% confidence intervals of the top-ranked logistic regression model representing the influence of environmental and temporal variables on martens' presence within and surrounding a clearcut in central British Columbia, Canada during summer 2015. Figure 4 View largeDownload slide Coefficients and 95% confidence intervals of the top-ranked logistic regression model representing the influence of environmental and temporal variables on martens' presence within and surrounding a clearcut in central British Columbia, Canada during summer 2015. Small mammal habitat use Mice were the most common type of forest-floor small mammal detected, followed by voles and shrews, respectively (Figure 5). All forest-floor small mammals were significantly more likely to be detected in corridors compared with the forest or clearcut (x2 = 13.374, P = 0.001, Figure 5). There were approximately three times as many mice, twice as many voles and seven times as many shrews present in the corridors when compared with the forest or clearcut sites. Only summer data are presented for forest-floor small mammals, as detection in winter was low because they spend most of their time in subnivean habitat. Figure 5 View largeDownload slide Average number of forest-floor small mammal visits (±1 SE) to camera traps in forest, clearcut and corridor locations within and surrounding a clearcut in central British Columbia, Canada during summer 2015. Figure 5 View largeDownload slide Average number of forest-floor small mammal visits (±1 SE) to camera traps in forest, clearcut and corridor locations within and surrounding a clearcut in central British Columbia, Canada during summer 2015. The most parsimonious model for forest-floor small mammals included parameters for corridor, shrub cover and number of stems (Table 3). This model had a poor predictive accuracy (AUC = 0.61). During the summer, frequency of red squirrel visits was highest in forested locations and there was no difference in use between clearcut and corridor locations (Figure 6). In winter, however, red squirrels used corridors ~10 times more frequently than other clearcut locations (Figure 6). Figure 6 View largeDownload slide Average number of red squirrel visits (±1 SE) to camera traps in forest, clearcut, and corridor locations during summer 2015 and winter 2017 within and surrounding a clearcut in central British Columbia, Canada. Figure 6 View largeDownload slide Average number of red squirrel visits (±1 SE) to camera traps in forest, clearcut, and corridor locations during summer 2015 and winter 2017 within and surrounding a clearcut in central British Columbia, Canada. Discussion Marten habitat use Frequency of visits by martens to camera traps decreased considerably from 49 visits in the first week of June to only 1–6 visits/week during the months of July and August. This change in visitation rate could be related to the small spatial scale of the study. For example, the average home range size in the study area calculated from radio-collared male and female marten was 1012 ha and 626 ha, respectively (John Prince Research Forest, 2017), thus the death or emigration of an individual marten could greatly influence the frequency of visits across the 93 ha study area. As breeding season occurs in July and August (Powell et al. 2003), it is possible that martens dispersed away from the area if there were no females of reproductive age present in the study area. We could not individually identify all individual martens, and therefore cannot test the effect of immigration and emigration on visitation rates. Another possible explanation for the change in visitation rate is that martens were less drawn to the lure when food sources such as berries and bird or small mammal nests became more abundant throughout the summer. During the summer, martens used forested habitats most frequently and used the constructed CWD corridors with the same frequency as other areas of the clearcut. Limited data suggest that martens preferred the corridors to other areas in the clearcut during winter; although not statistically significant, there were nearly twice as many visits by martens to corridor vs random clearcut locations in winter 2017. Marten detections at corridor sites during the winter occurred on March 22nd or earlier but on March 18th or later at random sites. The temporal differences in detections between corridor and random sites may be explained by early spring conditions, as from the end of March to early April, the study area transitioned from being dominated by snow cover to being largely barren of snow with remnant patches. The behaviour and habitat use of martens was also likely transitioning during this time period. Martens may use CWD corridors more frequently in winter than in summer because they provide a high density of small mammal prey when other food sources such as berries and nests are not available. The high density of CWD in the corridors also provides marten with subnivean habitat such as resting sites, thermal insulation and refugia from predators (Buskirk et al., 1989). During the day in summer, martens used forested habitat most frequently, while at night martens were equally likely to use clearcut and forested habitats. This could be because there is an increased risk of predation from raptors in clearcuts during the day (Poole et al., 2004). Also, martens are more likely to use the corridors and clearcut locations at night to hunt nocturnal forest-floor mammals (Buskirk, 1983) that are most abundant in the CWD corridors within the clearcut. It is important to note that the study area had much greater amounts of dispersed CWD and higher levels of green tree retention than typically is present in clearcuts found across central British Columbia. Increased structural complexity and overhead cover along with the constructed CWD corridors may have provided sufficient habitat for martens such that they used the clearcut and corridor locations as frequently as the forest during the night. Given this pattern, it is possible that in clearcuts with little to no green tree retention and legal minimum amounts of CWD, constructed CWD corridors may be even more beneficial for martens and their prey. The presence of martens was best explained by the number of stems, number of Type 6–9 wildlife trees and mean length of CWD. These findings are consistent with the general understanding of the habitat ecology of American martens. It is well documented that martens select areas with dense forest (Spencer et al., 1983; Poole et al., 2004; Baldwin and Bender, 2008). For our study area, martens likely used sites with a high number of stems because these areas provide canopy cover that reduces the risk from aerial predators (Poole et al., 2004). Martens are known to select for mature forests because of the associated overhead cover and understory complexity (Thompson and Colgan, 1994). Type 6–9 wildlife trees are more likely to be present in older forest stands suggesting a similar pattern of use for martens in our study area (Keisker, 2000). Martens also use Type 6–9 wildlife trees as resting sites (Keisker, 2000). Although it is well-known that martens select for areas with high volumes of CWD (Buskirk et al., 1989; Sherburne and Bisonette, 1994; Poole et al., 2004; Andruskiw et al., 2008), there has been no empirical evidence relating the length of CWD pieces with the presence of martens. We found that the probability of a marten being present at a camera site increased exponentially with increasing mean length of CWD. This association could be because longer pieces of CWD provide greater habitat complexity, as they are more likely to stack on each other and become elevated instead of sinking into the ground. Small mammal habitat use All types of forest-floor small mammals (mice, voles and shrews) were observed more frequently in the corridors than in the forest or clearcut. There were three times as many mice, twice as many voles and seven times as many shrews present in the constructed corridors when compared with the forest or clearcut sites. This finding is consistent with work by Sullivan et al. (2011, 2012) and Sullivan and Sullivan (2014) who found that constructed woody debris corridors in clearcuts contained a higher abundance of forest-floor small mammals. The high density of small mammals in the corridors provides an abundant prey source for martens, as small mammals make up a significant portion of martens’ diets (Buskirk and Ruggiero, 1994; Martin, 1994). However, the relationship between the environmental factors that we measured and the presence of small mammals was weak. This is despite a large sample of observations. For our study area, forest-floor small mammals may be super abundant resulting in a near ubiquitous distribution of those species. Red squirrels are commonly known as a forest-dependent species that select for mature, closed-canopy forest and avoid clearcuts (Bakker and van Vuren, 2004; Haughland and Larsen, 2004). This was reflected in our results from summer 2015, as red squirrels were ~10 times more likely to be found in forested sites compared with clearcut or corridor sites. During winter, red squirrels used the CWD corridors ~10 times more frequently than other clearcut locations and as frequently as they used forested locations in summer. This indicates that red squirrels may be using the corridors to travel across or forage in clearcuts that they otherwise would have avoided. The corridors likely provide protection from predators and subnivean access for forage (Bakker and van Vuren, 2004). This high use of the corridors by red squirrels in winter increases the prey base for martens and results in the improvement of clearcut areas as habitat for both species. Conclusion Our results suggest that constructed CWD corridors are effective in enhancing clearcuts as habitat for martens and their prey. Corridors provide martens with resting sites, thermal insulation, refugia from predators and subnivean access to high densities of prey. Although martens did not use these corridors more than other clearcut locations during the second summer post-harvest, limited data suggest that martens preferred the corridors to other clearcut areas during winter. Furthermore, it is likely that the corridors will provide high-quality habitat for martens 30–40 years post-harvest when there is sufficient overhead cover to reduce risk from aerial predators. We recommend that forest managers retain areas of mature, dense forest as these areas provide overhead cover that martens require to mitigate predation risk and that long pieces of CWD be retained post-harvest as this may provide increased habitat complexity that martens require for prey capture and predator avoidance. Acknowledgements We gratefully acknowledge the people of the Tl’azt’en Nation for allowing us to conduct this research on their traditional territory. We would also like to thank D. Powe and the many student research assistants from the Tl’azt’en Nation who provided fieldwork assistance. We are also grateful for the contributions of S. Grainger of the John Prince Research Forest. We thank M. Gillingham and two anonymous reviewers for their helpful comments. Funding Natural Sciences and Engineering Research Council of Canada, the University of Northern British Columbia and the John Prince Research Forest. Conflict of interest statement None declared. References Agee , J.K. 1993 Fire ecology of Pacific Northwest forests . Island Press , p. 505 . Andruskiw , M. , Fryxell , J.M. , Thompson , I.D. and Baker , J.A. 2008 Habitat-mediated variation in predation risk by the American marten . Ecology 89 , 2273 – 2280 . Google Scholar CrossRef Search ADS PubMed Bakker , V.J. and van Vuren , D.H. 2004 Gap-crossing decisions by the red squirrel, a forest-dependent small mammal . Conserv. Biol. 18 , 689 – 697 . Google Scholar CrossRef Search ADS Baldwin , R.A. and Bender , L.C. 2008 Distribution, occupancy, and habitat correlates of American martens (Martes americana) in Rocky Mountain National Park, Colorado . J. Mammal. 89 , 419 – 427 . Google Scholar CrossRef Search ADS Boyce , M.S. , Vernier , P.R. , Nielsen , S.E. and Schmiegelow , F.K.A. 2002 Evaluating resource selection functions . Ecol. Model. 157 , 281 – 300 . Google Scholar CrossRef Search ADS British Columbia 2004 Forest and Ranges Practices Act Forest Planning and Practices Regulation . Queen’s Printer . British Columbia Chief Forester . 2010 Chief Forester’s Guidance on Coarse Woody Debris Management. www.for.gov.bc.ca/ftp/hfp/external/!publish/frep/extension/. British Columbia Forest Analysis and Inventory Branch 2015 Vegetation Resources Inventory Sampling Procedures. Version 5.2 . BC Ministry of Forests, Lands and Natural Resource Operations . Burnham , K.P. and Anderson , D.R. 2002 Model selection and multimodel inference: a practical information-theoretic approach . 2nd edn . Springer-Verlag , p. 488 . Buskirk , S.W. 1983 The Ecology of Marten in Southcentral Alaska . University of Alaska , p. 131 . Buskirk , S.W. , Forrest , S.C. , Raphael , M.G. and Harlow , H.J. 1989 Winter resting site ecology of marten in the central Rocky Mountains . J. Wildl. Manage. 53 , 191 – 196 . Google Scholar CrossRef Search ADS Buskirk , S.W. and Ruggiero , L. 1994 American marten. In The scientific basis for conserving large carnivores: American marten, fisher, lynx, and wolverine in the western United States . Ruggiero L. , Aubry K. , Buskirk S. , Lyon J. and Zielinski W. (eds). United States Department of Agriculture , pp. 17 – 37 . Fauteux , D. , Imbeau , L. , Drapeau , P. and Mazerolle , M.J. 2012 Small mammal responses to coarse woody debris distribution at different spatial scales in managed and unmanaged boreal forests . For. Ecol. Manage. 266 , 194 – 205 . Google Scholar CrossRef Search ADS Hargis , C.D. and McCullough , D.R. 1984 Winter diet and habitat selection of marten in Yosemite National Park . J. Wildl. Manage. 48 , 140 – 146 . Google Scholar CrossRef Search ADS Harmon , M.E. , Franklin , J.F. , Swanson , F.J. , Sollins , P. , Gregory , S.V. , Lattin , J.D. , et al. 1986 Ecology of coarse woody debris in temperate ecosystems . Adv. Ecol. Res. 15 , 133 – 267 . Google Scholar CrossRef Search ADS Haughland , D.L. , Larsen , K.W. 2004 Ecology of North American red squirrels across contrasting habitats: relating natal dispersal to habitat . J. Mammal. 85 , 225 – 236 . Google Scholar CrossRef Search ADS John Prince Research Forest 2017 . Unpublished data. Keisker , D.G. 2000 Types of wildlife trees and coarse woody debris required by wildlife of north-central British Columbia . Resource Branch Ministry of Forests . Lindenmayer , D.B. , Burton , P.J. and Franklin , J.F. 2012 Salvage logging and its ecological consequences . Island Press , p. 246 . Lofroth , E.C. 1993 Scale dependent analyses of habitat selection by marten in the Sub-Boreal Spruce biogeoclimatic zone, British Columbia. Master’s Thesis, Simon Fraser University, British Columbia. Mabee , W.E. and Saddler , J.N. 2009 Bioethanol from lignocellulosics: status and perspectives in Canada . Bioresour. Technol. 101 , 4806 – 4813 . Google Scholar CrossRef Search ADS PubMed Martin , S.K. 1994 Feeding ecology of American martens and fishers. In Martens, sables and fishers: biology and conservation . Buskirk S. , Harestad A. , Raphael M. and Powell R. (eds). Cornell University , pp. 297 – 315 . McComb , W.C. 2003 Ecology of coarse woody debris and its role as a habitat for mammals. In Mammal Community Dynamics. Management and Conservation in the Coniferous Forests of Western North America . Zabel C.J. and Anthony R.G. (eds). Cambridge University Press , pp. 374 – 404 . Google Scholar CrossRef Search ADS Meidinger , D.V. and Pojar , J. 1991 The Ecosystems of British Columbia . BC Ministry of Forests . Special report series no. 6. Poole , K.G. , Porter , A.D. , de Vries , A. , Maundrell , C. , Grindal , S.D. and St. Clair , C.C. 2004 Suitability of a young deciduous-dominated forest for American marten and the effects of forest removal . Can. J. Zool. 82 , 423 – 435 . Google Scholar CrossRef Search ADS Potvin , F. , Belanger , L. and Lowell , K. 2000 Marten habitat selection in a clearcut boreal landscape . Conserv. Biol. 14 , 844 – 857 . Google Scholar CrossRef Search ADS Powell , R.A. , Buskirk , S.W. and Zielinski , W.J. 2003 Fisher and Marten. In Wild Mammals of North America: Biology, Management, and Conservation . Feldhamer G.A. , Thompson B.C. and Chapman J.A. (eds). The John Hopkins University Press , pp. 635 – 646 . R Core Team . 2016 R: A language and environment for statistical computing. R Foundation for Statistical Computing, Vienna, Austria. http://www.R-project.org/. Rogers , W.H. 1993 . sg17: Regression standard errors in clustered samples. Stata Technical Bulletin 13: 19–23. Reprinted in Stata Technical Bulletin Reprints, vol. 3, pp. 88–94. College Station, TX: Stata Press. Sherburne , S.S. 1993 Squirrel middens influence marten (Martes americana) use of subnivean access points . Am. Midl. Nat. 129 , 204 – 207 . Google Scholar CrossRef Search ADS Sherburne , S.S. and Bisonette , J.A. 1994 Marten subnivean access point use: response to subnivean prey levels . J. Wildl. Manage. 58 , 400 – 405 . Google Scholar CrossRef Search ADS Spencer , W.D. , Barrett , R.H. and Zielinski , W.J. 1983 Marten habitat preferences in the northern Sierra Nevada . J. Wildl. Manage. 47 , 1181 – 1186 . Google Scholar CrossRef Search ADS StataCorp . 2011 Stata Statistical Software: Release 12. College Station, TX: StataCorp LP. Sullivan , T.P. and Sullivan , D.S. 2014 Diversifying clearcuts with green-tree retention and woody debris structures: conservation of mammals across forest ecological zones . Silva Fenn 48.5 , id 1219 . Sullivan , T.P. , Sullivan , D.S. , Lindgren , P.M.F. and Ransome , D.B. 2012 If we build habitat, will they come? Woody debris structures and conservation of forest mammals . J. Mammal. 93 , 1456 – 1468 . Google Scholar CrossRef Search ADS Sullivan , T.P. , Sullivan , D.S. , Lindgren , P.M.F. , Ransome , D.B. , Bull , J.G. and Ristea , C. 2011 Bioenergy or biodiversity? Woody debris structures and maintenance of red-backed voles on clearcuts . Biomass Bioenergy 35 , 4390 – 4398 . Google Scholar CrossRef Search ADS Thompson , I.D. and Colgan , P.W. 1994 Marten activity in uncut and logged boreal forests in Ontario . J. Wildl. Manage. 58 , 280 – 288 . Google Scholar CrossRef Search ADS Thompson , I.D. , Fryxell , J.M. and Harrison , D.J. 2012 Improved insights into use of habitat by American martens. In Biology and Conservation of Martens, Sables, and Fishers . Aubry K.B. , Zielinski W.J. , Raphael M.J. , Proulx G. and Buskirk S.W. (eds). Cornell University Press , pp. 209 – 230 . Toledo , D.P. , Herrick , J.E. and Abbott , L.B. 2010 A comparison of cover pole with standard vegetation monitoring methods . J Wildl. Manage. 74 , 600 – 604 . Google Scholar CrossRef Search ADS Wiebe , P.A. , Fryxell , J.M. , Thompson , I.D. , Börger , L. and Baker , J.A. 2013 Do trappers understand marten habitat? J. Wildl. Manage. 77 , 379 – 391 . Google Scholar CrossRef Search ADS Wiebe , P.A. , Thompson , I.D. , McKague , C.I. , Fryxell , J.M. and Baker , J.A. 2014 Fine-scale winter resource selection by American martens in boreal forests and the effect of snow depth on access to coarse woody debris . Écoscience 21 , 123 – 132 . Google Scholar CrossRef Search ADS © Institute of Chartered Foresters, 2018. All rights reserved. For Permissions, please e-mail: journals.permissions@oup.com. This article is published and distributed under the terms of the Oxford University Press, Standard Journals Publication Model (https://academic.oup.com/journals/pages/about_us/legal/notices)

Journal

Forestry: An International Journal Of Forest ResearchOxford University Press

Published: Oct 1, 2018

There are no references for this article.

You’re reading a free preview. Subscribe to read the entire article.


DeepDyve is your
personal research library

It’s your single place to instantly
discover and read the research
that matters to you.

Enjoy affordable access to
over 18 million articles from more than
15,000 peer-reviewed journals.

All for just $49/month

Explore the DeepDyve Library

Search

Query the DeepDyve database, plus search all of PubMed and Google Scholar seamlessly

Organize

Save any article or search result from DeepDyve, PubMed, and Google Scholar... all in one place.

Access

Get unlimited, online access to over 18 million full-text articles from more than 15,000 scientific journals.

Your journals are on DeepDyve

Read from thousands of the leading scholarly journals from SpringerNature, Elsevier, Wiley-Blackwell, Oxford University Press and more.

All the latest content is available, no embargo periods.

See the journals in your area

DeepDyve

Freelancer

DeepDyve

Pro

Price

FREE

$49/month
$360/year

Save searches from
Google Scholar,
PubMed

Create lists to
organize your research

Export lists, citations

Read DeepDyve articles

Abstract access only

Unlimited access to over
18 million full-text articles

Print

20 pages / month

PDF Discount

20% off