Abstract

Little is known about the behavior of mammals moving on unfamiliar ground, yet this information could be critical to assessing and enhancing landscape connectivity. I investigated the movements of adult red squirrels (Tamiasciurus hudsonicus) on unfamiliar ground to determine if squirrels selected specific microhabitat features and if manipulations of preferred features influenced movement choices. Rather than selecting for cover from predators or territorial conspecifics, 25 squirrels released outside their home ranges used microhabitat features that appeared to allow rapid, efficient, and inconspicuous travel (logs, open vegetation, low slopes, and high shrub cover per stem) while maintaining proximity to arboreal escape routes. Similarly, in 73 trials with 55 individuals, squirrels released in experimental plots moved preferentially through areas of greater log cover and cover per stem. Manipulating microhabitat features in plots within forested habitat enhanced red squirrel movements, underscoring the possibility of altering microhabitats in 2nd-growth forests or corridors to increase landscape connectivity for forest-associated mammals.

Natural habitats are becoming increasingly fragmented by human activities, and in many areas only small patches remain. Populations can persist in patches if local extinctions or declines are balanced by recolonization (Brown and Kodric-Brown 1977; Hanski and Gilpin 1991). Consequently, many conservation biologists advocate connecting patches with corridors or with a matrix suitable for animal movement (Franklin 1993; Meffe and Carroll 1994; Noss 1987; Ricketts 2001; Rosenberg et al. 1997). Maintaining or increasing connectivity requires an understanding of the movement behavior of species of concern. Unfortunately, little is known about the scales at which most animals perceive habitat and make decisions about movement or the features important to them during movement (Goodwin 2003; Lima and Zollner 1996).

Selection of movement paths may occur at several spatial scales ranging from the landscape (e.g., travel through patch A or patch B) to the microhabitat (e.g., under shrub or out in the open). In the short term, addition of habitat patches to a landscape to increase connectivity may be difficult, but altering microhabitats to soften matrix or restore degraded corridors may be possible (Franklin 1993; Ricketts 2001). For such techniques to succeed, mammals must select specific microhabitat features while moving and creation of preferred microhabitats must enhance movements. Several studies (Bennett et al. 1994; Lorenz and Barrett 1990; Merriam and Lanoue 1990; Ruefenacht and Knight 1995) have shown that microhabitat features are important determinants of corridor use, but few studies have investigated whether manipulations of microhabitats can alter movements (but see Planz and Kirkland 1992).

Identifying travel routes of mammals leaving their home ranges is logistically difficult because of the unpredictable nature of dispersal and excursions. Hence, most analyses of microhabitat selection by mammals have focused on movements within the home range (but see Bennett et al. 1994; Lorenz and Barrett 1990; Merriam and Lanoue 1990; Ruefenacht and Knight 1995). Mammals outside their home range must choose novel travel routes and reassess the relative importance of behaviors such as foraging, vigilance, and territorial defense. For example, animals on unfamiliar ground experience a heightened risk of predation because of increased encounter rates with predators and a lack of knowledge of escape routes (Ambrose 1972; Belichon et al. 1996; Metzgar 1967; Sakai and Noon 1997; Smith 1968). Territorial animals also face an increased likelihood of agonistic encounters with conspecifics, interactions that reduce predator vigilance (Bernays and Wcislo 1994; Diaz-Uriarte 1999; Price et al. 1990) and can result in attack and injury (Lair 1990). Thus, movement behavior may differ when an animal is off its home range.

I investigated microhabitat-scale movements of red squirrels (Tamiasciurus hudsonicus), which are territorial and forest dependent, by observing the paths of individuals translocated off their home ranges. Dispersal by juvenile red squirrels occurs as a series of forays and returns to the natal territory until a vacant territory is found (Larsen and Boutin 1994). Homing movements of translocated red squirrels appear analogous to movements after self-initiated excursions (Bovet 1984, 1992, 1995; Thibault and Bovet 1999). Thus, translocation may yield insights into microhabitat use patterns of dispersing juveniles or adults. My objectives were to determine if red squirrels select specific microhabitat features when traveling on unfamiliar ground, and if manipulation of microhabitat features preferred by red squirrels influences their movements.

Materials and Methods

Study area.—The study was conducted on Mitkof Island (56°N, 133°W) in southeastern Alaska from June 1998 to September 2001. The 518-km2 island is mountainous, has a wet maritime climate, and supports perhumid rainforest (Alaback 1994). Old-growth forests range from low-stature forests (10–20 m) containing shore pine (Pinus contorta var. contorta), Alaska-cedar (Chamaecyparis nootkatensis), western redcedar (Thuja plicata), mountain hemlock (Tsuga mertensiana), western hemlock (T. heterophylla), or Sitka spruce (Picea sitchensis) to forests dominated by tall (30- to 40-m) western hemlock and Sitka spruce. A well-developed ericaceous shrub layer is found in most late-seral forests.

Overview of releases.—I used a 2-step process to investigate whether microhabitat features influence squirrel movement paths. First, to determine if squirrels selected for specific microhabitat features, and if so, which features they favored, I released squirrels in forest on unfamiliar ground and compared their movement paths to random points. Next, to corroborate the results of this observational study and to test whether manipulation of microhabitat features influenced movement paths, I released squirrels in experimental plots offering a choice between favorable and unfavorable features.

Releases testing microhabitat selection.—I trapped red squirrels in a range of old-growth forest types at 4 study sites adjacent to clear-cuts ≤10 years old using Tomahawk live traps (Tomahawk Live Trap Co., Tomahawk, Wisconsin) and placed radiocollars (models MD-2C and PD-2C, Holohil Systems Ltd., Carp, Ontario, Canada) on 65 adults (≥175 g). I calculated home-range centers (bivariate arithmetic mean) of collared squirrels based on a mean of 17 (range 4–33) telemetry locations obtained by homing in on individuals; centers calculated from the first 4 locations were within 23.3 m ± 2.4 SE of centers calculated from ≥20 locations (n = 17). Locations were always =3 h apart (time to independence = 63 min—Swihart et al. 1988).

To obtain movement paths for evaluating microhabitat selection, from July through September I translocated (132–887 m) squirrels off their home ranges to the opposite sides of clear-cuts and released them 20–30 m into the forest in conjunction with a gap-crossing study (Bakker and Van Vuren 2004). Translocation induces homing behavior in red squirrels (Bakker and Van Vuren 2004; Bovet 1984). Squirrels had an unobstructed view of the landscape during transport (Bovet 1991). At the release site, I attached tracking spools (2.4 g, model 40-2 7B, Danfield Thread, Inc., Winsted, Connecticut), which contained approximately 250 m of thread that dispensed from the spool interior, to the rumps of 25 squirrels (11 males and 14 females) using cyano-acrylate glue. I then released squirrels in a random direction from the transport trap, which was covered with a dark cloth so that the open door would not be immediately apparent to the squirrel. After releasing squirrels, I moved ≥20 m away opposite the release direction. Tracking spools recorded initial movement paths (e.g., Brock and Kelt 2004; Cox et al. 2000; Cunha and Vieira 2002; Key and Woods 1996; Zollner 2000). Simultaneous telemetry by 2 or 3 observers at 3-min intervals was used to obtain approximate overall paths and homing times. To estimate speed during homing, I censored elapsed time between successive telemetry locations in which little movement (<20 m) toward home was evident from telemetry (i.e., stops).

I sampled spooled paths at 10-m intervals for 110 m beginning 10 m from the release point. I recorded the release point with a global positioning system receiver (Geoexplorer II, Trimble Navigation Ltd., Sunnyvale, California) and measured the azimuth and slope of each 10-m segment using a compass and clinometer. To test whether paths were oriented toward home, I compared the azimuth from release to home-range center and from release to 50 m along the path (i.e., closest sampled point = 45 m) with a V-test (Zar 1999).

To assess selection, I also sampled paired random points located a random distance between 2.5 and 5.0 m from path points either 90° or 270° (chosen randomly) from the line of travel. At both path and random points, I visually estimated percentage cover (0–100%) in the shrub layer (vegetation 0.3–2.5 m tall), low shrub layer (0.3–1.0 m), and herb layer (<0.3 m) and counted the number of woody stems and large nonwoody stems (i.e., skunk cabbage [Lysichiton americanum] and ferns =0.3 m) in a 1-m quadrat frame. I measured slope across the quadrat and distance to and size of the nearest log (≥10-cm midpoint diameter and ≥l-m long) and noted whether the sampling point was on a log. To pose a conservative test of selection for log surfaces, random points at logs were assumed to be on top of logs; thus, no random points were located under logs although some path points were. I measured distance to and size of the nearest tree (conifer = 2.5 m tall) and the nearest codominant tree providing direct access to the canopy. Residuals of regressing shrub cover on stem density estimated shrub cover not explained by stem density (hereafter “cover per stem”). Finally, I recorded net distance traveled in the canopy. When sampling points occurred where the thread was in a tree (8 of 267 points), I sampled paths a random distance 0–5 m after the thread returned to the ground.

Variables predicting movement paths were identified with conditional logistic regression (PROC PHREG, SAS 8.1–SAS Institute Inc. 2000), which is logistic regression for paired data using differences between cases and controls as predictors (Hosmer and Lemeshow 1989; Stokes et al. 1995). Each squirrel path was a stratum, and each path had 11 paired subsamples of microhabitat. I built conditional logistic regression models with stepwise selection (P = 0.05 to enter and P = 0.10 to remove) after univariate screening for candidate variables (P ≤ 0.25). To identify collinearity problems, I examined Spearman's rank correlations (PROC CORR, SAS 8.1, SAS Institute Inc. 2000) and tolerance values (PROC REG, SAS 8.1, SAS Institute Inc. 2000; Menard 1995). I rescaled 2 variables to achieve linearity of the logit (Hosmer and Lemeshow 1989). Distance to nearest tree had positive log odds for 1st quartile distances and negative log odds with overlapping confidence intervals for the remaining quartiles. It was replaced with the binary variable “tree proximity,” which had a value of 1 for 1st quartile distances (≤82 cm) and 0 otherwise. Similarly, cover in the low shrub layer was replaced with “dense low shrub cover” (0: ≤50%, 1: =50% cover). I evaluated models that included main effects and their interactions. To determine if absolute or relative distances to arboreal escape routes was important, I tested interactions between tree proximity and tree density (trees/m2). Further assessment of model goodness-of-fit was achieved by examining residual chi-square and residual diagnostics (Hosmer and Lemeshow 1989; Stokes et al. 1995).

Releases in experimental plots.—From preliminary analyses of data from spooled paths, I identified 3 microhabitat features that squirrels used preferentially and that I could manipulate: low woody stem density, high cover per stem, and high log cover. For each of these microhabitat features, I established three 8-m-radius plots in which I manipulated vegetation to create more favorable and less favorable areas, or subplots (Fig. 1a).

Fig. 1

Paths of red squirrels released in 8-m-radius forest plots in southeastern Alaska in which single microhabitat variables were experimentally manipulated, a) Squirrels were released along the axis between subplots (arrow), with direction of release chosen randomly. Subplots with favorable microhabitat (shaded) contain: b) higher log cover, c) higher shrub cover per stem, or d) lower stem densities. The figure combines paths from 3 replicate plots per microhabitat treatment. Results of binomial tests on choices are shown in Table 2.

Fig. 1

Paths of red squirrels released in 8-m-radius forest plots in southeastern Alaska in which single microhabitat variables were experimentally manipulated, a) Squirrels were released along the axis between subplots (arrow), with direction of release chosen randomly. Subplots with favorable microhabitat (shaded) contain: b) higher log cover, c) higher shrub cover per stem, or d) lower stem densities. The figure combines paths from 3 replicate plots per microhabitat treatment. Results of binomial tests on choices are shown in Table 2.

Table 2

Binomial tests of choices made by red squirrels released in the center of plots in which one-half of the plot was manipulated to enhance a single microhabitat feature. The favorable microhabitat condition is the condition selected by red squirrels along spooled paths when released in forest. The experiment included 3 replicate plots per microhabitat treatment. P values are corrected using sequential Bonferroni adjustments. Actual paths are shown in Fig. 1.

Manipulated microhabitat n Proportion choosing homeward direction P Favorable microhabitat condition Proportion choosing favorable microhabitat Corrected P 
Log presence 25 0.28 0.986 High 0.76 0.014 
Cover per stem 25 0.32 0.964 High 0.72 0.028 
Stem density 23 0.39 0.851 Low 0.52 0.583 
Manipulated microhabitat n Proportion choosing homeward direction P Favorable microhabitat condition Proportion choosing favorable microhabitat Corrected P 
Log presence 25 0.28 0.986 High 0.76 0.014 
Cover per stem 25 0.32 0.964 High 0.72 0.028 
Stem density 23 0.39 0.851 Low 0.52 0.583 

For each stem density plot, I 1st removed all leaf cover from the entire plot by cutting off stem parts that bore leaves, but retaining stem bases ≥0.3 m tall. Then, in randomly assigned subplots, I created favorable microhabitat by cutting off all stems at ground level and unfavorable microhabitat by leaving defoliated stem bases. For each cover per stem plot, I removed approximately 80% of leaf cover from a randomly assigned subplot to create an unfavorable microhabitat and left the other subplot unaltered. Because of the difficulty of moving large logs, I constructed log plots at locations with existing log cover. I created favorable microhabitat in one-half of the plot by augmenting and rearranging logs until 8 logs radiated from the center like spokes of a wheel. Often several logs were placed end to end to reach 8 m, but each log was ≥10 cm at midpoint diameter and ≥1 m long. I removed as many logs as possible from the unfavorable subplot.

To characterize subplots, I measured percentage cover in the shrub, low shrub, and herb layers; densities of logs and stems; and maximum slope in 8 systematically spaced 1-m quadrats. I then compared favorable and unfavorable subplots of each plot separately using analysis of variance (PROC GLM, SAS 8.1, SAS Institute Inc. 2000), applying sequential Bonferroni adjustments to P values (Rice 1989). I also counted numbers of trees and mapped all logs to calculate log cover.

A total of 55 squirrels (24 males and 31 females) was released, each in ≤2 plots located outside its home range and always in different microhabitat treatments. To limit visual information about homeward orientation, I carried squirrels to manipulated plots in traps enclosed in a cloth bag (Bovet 1991). I then transferred squirrels to a trap retrofitted with a remote-release mechanism, in which the back of the trap was replaced with Plexiglas attached to a 15-m string fed through a pulley suspended above the trap. This trap was placed in the plot center oriented along the axis between subplots (Fig. 1a). Before release, each squirrel acclimated for =20 min while I sat in a concealed location opposite the release direction, which was selected randomly. Upon release, I noted the time the squirrel left the trap and the approximate path it took. I assigned all squirrel paths to unfavorable subplots unless paths included <l-m net distance in unfavorable subplots.

I conducted this research humanely in accordance with guidelines of the American Society of Mammalogists (Animal Care and Use Committee 1998) and with the approval of the University of California, Davis, Animal Care and Use Committee. Values in results are given as mean ± SE.

Results

Releases testing microhabitat selection.—Most travel ( = 97.9% ± 0.7%) was on the ground or on logs, and 10 of 25 paths did not include trees within 110 m of the release site. Spooled paths were not oriented toward home (u = −0.241, P = 0.594; Fig. 2), but all translocated squirrels homed successfully within 24 h. Travel speeds were well below the squirrel's capability ( = 25.6 ± 6.4 m/min versus maximum of 250 m/min—Layne and Benton 1954), and stops were common ( = 7.4 ± 1.7 min/100 m).

Fig. 2

Path orientations from release to 50 m (net distance) for red squirrels on unfamiliar ground in forest in southeastern Alaska relative to the homeward orientation indicated by solid circles (n = 21, 4 paths did not reach 50 m from the release along the 110 m of spool).

Fig. 2

Path orientations from release to 50 m (net distance) for red squirrels on unfamiliar ground in forest in southeastern Alaska relative to the homeward orientation indicated by solid circles (n = 21, 4 paths did not reach 50 m from the release along the 110 m of spool).

Distance to and size of the nearest codominant tree, size of the nearest tree, and percentage cover of shrubs were not significant path predictors in univariate conditional logistic regression and were not considered in multivariate model building (Fig. 3). Distance to nearest log was eliminated because it was strongly correlated with log travel (r = 0.738, P < 0.001) and because it was not significant when only ground travel was considered (P = 0.639). All other variables used together in multivariate model building had lower correlations (all r ≤ 0.531).

Fig. 3

Microhabitat variables measured along red squirrel paths and at paired random points in forest in southeastern Alaska. Mean difference (±SE) between microhabitat features on paths and paired random points shown. Solid circles (◅) denote variables significant (P ≤ 0.03) in multivariate conditional logistic regression. Open circles (○) indicate variables significant in univariate screening (P ≤ 0.25). Variables that were not significant (P = 0.25) predictors and not used in model building are denoted with a horizontal line (–). Actual values are shown for tree proximity and low shrub cover; these variables were converted to binary variables to meet test assumptions. SE is too small to be visible for some variables.

Fig. 3

Microhabitat variables measured along red squirrel paths and at paired random points in forest in southeastern Alaska. Mean difference (±SE) between microhabitat features on paths and paired random points shown. Solid circles (◅) denote variables significant (P ≤ 0.03) in multivariate conditional logistic regression. Open circles (○) indicate variables significant in univariate screening (P ≤ 0.25). Variables that were not significant (P = 0.25) predictors and not used in model building are denoted with a horizontal line (–). Actual values are shown for tree proximity and low shrub cover; these variables were converted to binary variables to meet test assumptions. SE is too small to be visible for some variables.

Six microhabitat characteristics best predicted red squirrel paths. Paths were positively associated with log travel, cover per stem, and tree proximity, and negatively associated with dense low shrub cover, herb cover, and slope (model χ2 =120.645, d.f. = 8, P < 0.001; Table 1; Fig. 3). Odds ratios indicate that squirrels were nearly 8 times more likely to travel on logs than on ground. Similarly, squirrels were almost 4 times more likely to use locations close to trees (within 82 cm) rather than more distant locations and were only about half as likely to use dense cover (=50%) in the low shrub layer relative to open areas. A 10-unit increase in cover per stem increased odds of use by 20% whereas a 10% increase in herb cover decreased it by 16%. Finally, an increase in slope of 10° decreased odds of use by nearly one-third. Densities of woody and nonwoody stems and log size were candidate predictors after univariate tests but were not significant in the multivariate model (Table 1; Fig. 3). Proximity to the nearest tree had a strong negative interaction with tree density (Table 1); the importance of staying close to trees was much greater when tree densities were low. Herb cover interacted negatively with log travel, indicating that low herb cover was a more important predictor of squirrel paths when travel was on logs. Nonetheless, herb cover remained a significant predictor when only ground travel was considered (odds ratio 0.982, P = 0.016).

Table 1

Selection of microhabitat features along red squirrel movement paths in southeastern Alaska based on stepwise selection for conditional logistic regression to predict path locations from paired random locations. Predictors not included in the model made univariate screening for significance (P ≤ 0.25), but were not selected in multivariate model building (selection criteria P < 0.05 to enter, P = 0.10 to remove). Nonsignificant interactions are not shown.

Variable Parameter estimate Standard error P Odds ratio 
Predictors included in the model 
Log travel 2.052 0.400 ≪0.001 7.785 
Cover per stem 0.018 0.005 ≪0.001 1.018 
Tree proximity (<82 cm) 1.350 0.299 ≪0.001 3.858 
Slope (°) −0.035 0.008 ≪0.001 0.966 
Tree proximity × density −2.559 0.738 <0.001 0.077 
Dense low shrub cover (50% at <1 m) −0.588 0.263 0.025 0.556 
Herb cover (%) −0.018 0.008 0.021 0.982 
Log travel × herb cover −0.078 0.033 0.017 0.925 
Predictors not included in the model 
Nonwoody stems (m−2  0.154  
Woody stems (m−2  0.346  
Log size (diameter in cm)   0.616  
Variable Parameter estimate Standard error P Odds ratio 
Predictors included in the model 
Log travel 2.052 0.400 ≪0.001 7.785 
Cover per stem 0.018 0.005 ≪0.001 1.018 
Tree proximity (<82 cm) 1.350 0.299 ≪0.001 3.858 
Slope (°) −0.035 0.008 ≪0.001 0.966 
Tree proximity × density −2.559 0.738 <0.001 0.077 
Dense low shrub cover (50% at <1 m) −0.588 0.263 0.025 0.556 
Herb cover (%) −0.018 0.008 0.021 0.982 
Log travel × herb cover −0.078 0.033 0.017 0.925 
Predictors not included in the model 
Nonwoody stems (m−2  0.154  
Woody stems (m−2  0.346  
Log size (diameter in cm)   0.616  

Releases in manipulated plots.—Squirrels were more likely to choose subplots in which log cover had been enhanced and in which cover per stem was greater (Table 2; Fig. 1). Low stem density, however, did not influence the direction squirrels moved (Table 2; Fig. 1). Choices in manipulated plots did not reflect homeward orientation (Table 2). After the trap door was triggered, squirrels generally did not immediately run out (mean time in open trap was 4.7 ± 0.9 min). Although I sat opposite the release direction, 30% of squirrels traveled in the 2 quadrants closest to me, and 4 squirrels (6% of releases) passed within 1 m.

Stem density and cover per stem plots differed in stem density and cover per stem, respectively (all corrected P ≤ 0.027). In log plots, log cover in favorable subplots ( = 18.3 ± 1.8 m2) exceeded that in unfavorable subplots ( = 0.6 ± 0.7 m2). For all other plots, stem density and cover per stem did not differ (all corrected P ≥ 0.852), and log cover differences averaged only 0.2 ± 2.4 m2 ( = 5.6 ± 1.3 m2 logs/subplot). No subplots differed in herb cover or slope (all corrected P ≥ 0.082). The mean difference between subplots in number of trees was 0.6 ± 1.5 ( = 12.6 ± 2.4 trees per subplot). Two cover per stem plots had more cover in the low shrub layer, an unfavorable feature, in favorable subplots (corrected P ≤ 0.008).

Discussion

Downed logs were the single most important microhabitat feature for moving red squirrels. As travel surfaces, logs typically have few obstacles and a low slope and thus can be traversed quickly, efficiently, and inconspicuously. Auditory concealment is frequently cited as a reason that small mammals might use logs rather than ground, where dry leaf litter rustles when traversed (Barnum et al. 1992; Barry and Francq 1980; Graves et al. 1988; Hayes and Cross 1987; Planz and Kirkland 1992; Roche et al. 1999). Peromyscus leucopus increased use of logs after leaf fall (Barnum et al. 1992), and in laboratory conditions selected wet over dry litter (Fitzgerald and Wolff 1988), presumably to avoid noise produced when litter is abundant or dry. Similarly, Peromyscus maniculatus traveled on coniferous needles more than deciduous leaves, but only when litter was dry, and favored travel on logs more when on a leaf substrate (Roche et al. 1999). The coniferous forests of southeastern Alaska permit relatively quiet travel because they are almost continuously damp and have little leaf litter. Consequently, other benefits of travel on logs, such as speed and efficiency, may be more important in this region. Logs also may serve as familiar objects to translocated squirrels; I frequently observed red squirrels scanning and traveling on logs within the home range and Tamiasciurus are known to use logs for caching and foraging (McComb 2003; Smith and Mannan 1994).

Like logs, other microhabitats favored by red squirrels appear to facilitate rapid, efficient, and inconspicuous travel. Rather than selecting shrub cover, which could provide visual concealment from predators, red squirrels preferentially used areas with open vegetation and low slopes. Various hypotheses have been proposed for the avoidance of vegetation by some small mammals inside their home range. Dense vegetation is likely energetically expensive to move through because mammals must either physically push stems aside or use tortuous paths to avoid them (Simonetti 1989). Incidental movement of stems and leaves during travel creates visual and auditory cues for predators (Barnum et al. 1992; Barry and Francq 1980; Planz and Kirkland 1992), and dense vegetation can simultaneously impede predator detection (Schooley et al. 1996). Potential predators in the study area are northern goshawks (Accipiter gentilis) and red-tailed hawks (Buteo jamaicensis), which are primarily visual hunters, and American marten (Martes americana) and ermine (Mustela erminea), which are terrestrial predators with limited fields of view that probably rely on both visual and auditory cues (Zielinski 2000). For red squirrels, cues from moving vegetation also may alert territorial conspecifics and lead to chases and attacks. Finally, dense vegetation can reduce escape speeds (Schooley et al. 1996) and, for displaced or exploring red squirrels, increase return times to undefended territories. Only herb cover showed significant interactions with log travel, and squirrels still strongly avoided herbs when on the ground. Selection for low slopes and avoidance of dense low shrub cover was comparable whether squirrels were on logs or on the ground.

Some mice and voles preferentially use areas with greater cover when on their home range or in a laboratory (e.g., Eadie 1953; Harestad and Shackleton 1990; Kaufman et al. 1985), where individuals are likely traveling short distances and have knowledge of escape routes. In contrast, translocated red squirrels were moving long distances on unfamiliar ground, crossing territories of conspecifics, and returning to their own undefended territories. Seeking paths with fewer obstacles rather than dense vegetation reduces energetic costs, limits cues to predators and conspecifics caused by vegetation, maintains detection distances, and maximizes travel and escape speeds, and thus may be a more effective strategy for long-distance movements.

Despite avoidance of dense vegetation, the amount of overall cover along squirrel paths was disproportionately high relative to the number of stems encountered, suggesting a trade-off between unobstructed movement and visual concealment. This behavior of trading off obstacles and cover was exhibited over a range of stem densities and varying levels of low shrub cover (i.e., neither interacted with cover per stem; P ≥ 0.759). High shrub cover relative to stem densities is characteristic of the low-light environments of old-growth forests. In contrast, clear-cuts have high stem densities (forest = 15.1 ± 0.8 m−2; clear-cut = 27.8 ± 3.3 m−2) and thus low cover per stem (residuals: forest = −1.7 ± 1.6 m−2; clear-cut = −22.9 ± 3.9 m−2Bakker 2003), whereas dense 2nd-growth forest has few shrubs.

Red squirrels selected for close proximity to trees. Tree squirrels use arboreal escape routes when encountering aerial or ground predators (Lima et al. 1985; Smith 1995; Temple 1987) and numerous studies have shown that foraging squirrels (Sciurus carolinensis and S. niger) perceive predation risk to be lower near trees (Bowers and Breland 1996; Lima et al. 1985; Newman and Caraco 1987; Tuen and Brown 1996). Interestingly, red squirrels varied the intensity of this behavior in response to changes in tree density. Red squirrels were more likely to run from tree to tree when tree density was low, as indicated by the negative interaction between tree proximity and tree density (Table 1). When tree density was high, arboreal escape routes were never far away, and squirrels did not need to alter their movements to maintain proximity. P. maniculatus, which uses shrubs as cover, responded similarly to variation in shrub density in short-grass prairie (Stapp and Van Home 1997). Despite maintaining ready access to trees, however, only 40% of red squirrels actually climbed trees along spooled paths, and almost 98% of travel was on the ground.

Experimental alteration of preferred microhabitat features influenced red squirrel movements, corroborating selection along spooled paths and highlighting the potential for manipulations of microhabitat to enhance connectivity. Presence of logs, the most important microhabitat variable along spooled paths, was also the most influential manipulated feature. Most squirrels (76%) released in log plots chose subplots with high log cover and traveled atop these logs, whereas only 2 of 25 squirrels traveled entirely in subplots with scant log cover. Similarly, red squirrels selected for greater cover per stem along spooled paths and also chose subplots with greater cover per stem. Although univariate analysis suggested red squirrels avoided high woody stem densities along spooled paths, stem density was not chosen in multivariate model building, indicating that avoidance of areas with dense stems is better explained by other variables, such as cover per stem and cover in the low shrub layer. Likewise, red squirrels showed no preference for subplots with experimentally reduced stem densities.

In all experiments, red squirrels released away from their home range did not initially orient toward home, suggesting that other factors, such as habitat selection, governed movement decisions. Similarly, displaced squirrels did not appear to exhibit flight behavior, instead moving at relatively slow speeds with frequent stops, consistent with active decision-making. Thus, these posttranslocation movements may yield insights into movement decisions made during excursions and dispersal. Although microhabitats played a clear role in path selection within forested habitat, the decision to enter and cross matrix may be strongly influenced by key habitat elements and landscape features. For example, red squirrels are reluctant to enter clear-cuts on unfamiliar ground (Bakker 2003), perhaps because of the absence of arboreal escapes, and landscape configuration influences their decisions to cross clear-cuts (Bakker and Van Vuren 2004). Nonetheless, this study suggests that enhancing favored microhabitat features in 2nd-growth forests or along corridors could increase connectivity for forest-associated mammals.

Acknowledgments

I thank D. Van Vuren, M. Johnson, D. Kelt, P. Stapp, and 4 anonymous reviewers for helpful critiques and G. DeGayner, T. Hartshorn, and K. Hastings for logistical and intellectual support. A. Bernstein, J. Compton, K. Freeman, S. Mussulman, and K. Thorne provided expert field assistance in demanding conditions. This project received financial support from the Environmental Protection Agency Science to Achieve Results program, the Graduate Group in Ecology at the University of California, Davis, and the Switzer Foundation. The United States Department of Agriculture Forest Service provided generous in-kind support.

Literature Cited

Alaback
P. B.
1994
.
Biodiversity patterns in relation to climate: the coastal temperate rainforests of North America
. Pp.
105
133
in
High-latitude rainforests and associated ecosystems of the west coast of the Americas
  (
Lawford
R. G.
Alaback
P. B.
Fuentes
E.
, eds.).
Springer-Verlag
,
New York
.
Ambrose
H. W.
1972
.
Effect of habitat familiarity and toe-clipping on rate of owl predation in Microtus pennsylvanicus
.
Journal of Mammalogy
 
53
:
909
912
.
Animal Care, Use Committee
.
1998
.
Guidelines for the capture, handling, and care of mammals as approved by the American Society of Mammalogists
.
Journal of Mammalogy
 
79
:
1416
1431
.
Bakker
V. J.
2003
.
Behavior and habitat relations of forest-associated sciurids in a fragmented landscape
.
Ph.D. dissertation
 ,
University of California
,
Davis
.
Bakker
V. J.
Van Vuren
D. H.
.
2004
.
Gap-crossing decisions by the red squirrel (Tamiasciurus hudsonicus), a forest-dependent small mammal
.
Conservation Biology
 
18
:
689
697
.
Barnum
S. A.
Manville
C. J.
Tester
J. R.
Carmen
W. J.
.
1992
.
Path selection by Peromyscus leucopus in the presence and absence of vegetative cover
.
Journal of Mammalogy
 
73
:
797
801
.
Barry
R. E.
Jr.
Francq
E. N.
.
1980
.
Orientation to landmarks within the preferred habitat by Peromyscus leucopus
.
Journal of Mammalogy
 
61
:
292
303
.
Belichon
S.
Clobert
J.
Massot
M.
.
1996
.
Are there differences in fitness components between philopatric and dispersing individuals?
Acta Oecologica
 
17
:
503
517
.
Bennett
A. F.
Henein
K.
Merriam
G.
.
1994
.
Corridor use and the elements of corridor quality: chipmunks and fencerows in a farmland mosaic
.
Biological Conservation
 
68
:
155
165
.
Bernays
E. A.
Wcislo
W. T.
.
1994
.
Sensory capabilities, information processing, and resource specialization
.
Quarterly Review of Biology
 
69
:
187
204
.
Bovet
J.
1984
.
Strategies of homing behavior in the red squirrel, Tamiasciurus hudsonicus
.
Behavioral Ecology and Sociobiology
 
16
:
81
88
.
Bovet
J.
1991
.
Route based visual information has limited effect on homing performance of red squirrels, Tamiasciurus hudsonicus
.
Ethology
 
87
:
59
65
.
Bovet
J.
1992
.
Mammals
. Pp.
321
361
in
Animal homing
  (
Papi
F.
, ed.).
Chapman & Hall
,
London, United Kingdom
.
Bovet
J.
1995
.
Homing in red squirrels (Tamiasciurus hudsonicus): the importance of going straight
.
Ethology
 
101
:
1
9
.
Bowers
M. A.
Breland
B.
.
1996
.
Foraging of gray squirrels on an urban-rural gradient: use of the GUD to assess anthropogenic impact
.
Ecological Applications
 
6
:
1135
1142
.
Brock
R. E.
Kelt
D. A.
.
2004
.
Influence of roads on the endangered Stephens' kangaroo rat (Dipodomys stephensi): are dirt and gravel roads different?
Biological Conservation
 
118
:
633
640
.
Brown
J. H.
Kodric-Brown
A.
.
1977
.
Turnover rates in insular biogeography: effect of immigration on extinction
.
Ecology
 
58
:
445
449
.
Cox
M. P. G.
Dickman
C. R.
Cox
W. G.
.
2000
.
Use of habitat by the black rat (Rattus rattus) at North Head, New South Wales: an observational and experimental study
.
Austral Ecology
 
25
:
375
385
.
Cunha
A. A.
Vieira
M. V.
.
2002
.
Support diameter, incline, and vertical movements of four didelphid marsupials in the Atlantic forest of Brazil
.
Journal of Zoology (London)
 
258
:
419
26
.
Diaz-Uriarte
R.
1999
.
Anti-predator behaviour changes following an aggressive encounter in the lizard Tropidurus hispidus
.
Proceedings of the Royal Society of London, B. Biological Sciences
 
266
:
2457
2464
.
Eadie
W. R.
1953
.
Response of Microtus to vegetative cover
.
Journal of Mammalogy
 
34
:
263
264
.
Fitzgerald
V. J.
Wolff
J. O.
.
1988
.
Behavioral responses of escaping Peromyscus leucopus to wet and dry substrata
.
Journal of Mammalogy
 
69
:
825
828
.
Franklin
J. F.
1993
.
Preserving biodiversity: species, ecosystems, or landscapes
.
Ecological Applications
 
3
:
202
205
.
Goodwin
B. J.
2003
.
Is landscape connectivity a dependent or independent variable?
Landscape Ecology
 
18
:
687
699
.
Graves
S.
Maldonado
J.
Wolff
J. O.
.
1988
.
Use of ground and arboreal microhabitats by Peromyscus leucopus and Peromyscus maniculatus
.
Canadian Journal of Zoology
 
66
:
277
278
.
Hanski
L
Gilpin
M.
.
1991
.
Metapopulation dynamics: brief history and conceptual domain
.
Biological Journal of the Linnean Society
 
42
:
3
16
.
Harestad
A. S.
Shackleton
D. M.
.
1990
.
Cover and use of travel routes by female Townsend−s voles in a laboratory arena
.
Biology of Behaviour
 
15
:
196
204
.
Hayes
J. P.
Cross
S. P.
.
1987
.
Characteristics of logs used by western red-backed voles, Clethrionomys californicus, and deer mice, Peromyscus maniculatus
.
Canadian Field-Naturalist
 
101
:
543
546
.
Hosmer
D. W.
Lemeshow
S.
.
1989
.
Applied logistic regression
 .
John Wiley & Sons, Inc.
,
New York
.
Kaufman
D. W.
Peak
M. E.
Kaufman
G. A.
.
1985
.
Peromyscus leucopus in riparian woodlands: use of trees and shrubs
.
Journal of Mammalogy
 
66
:
139
143
.
Key
G. E.
Woods
R. D.
.
1996
.
Spool-and-line studies on the behavioural ecology of rats (Rattus spp.) in the Galapagos Islands
.
Canadian Journal of Zoology
 
74
:
733
737
.
Lair
H.
1990
.
The calls of the red squirrel: a contextual analysis of function
.
Behaviour
 
115
:
254
282
.
Larsen
K. W.
Boutin
S.
.
1994
.
Movements, survival, and settlement of red squirrel (Tamiasciurus hudsonicus) offspring
.
Ecology
 
75
:
214
223
.
Layne
J. N.
Benton
A. H.
.
1954
.
Some speeds of small mammals
.
Journal of Mammalogy
 
35
:
103
104
.
Lima
S. L.
Valone
T. J.
Caraco
T.
.
1985
.
Foraging-efficiency-predation-risk trade-off in the gray squirrel (Sciurus carolinensis)
.
Animal Behaviour
 
33
:
155
165
.
Lima
S. L.
Zollner
P. A.
.
1996
.
Towards a behavioral ecology of ecological landscapes
.
Trends in Ecology and Evolution
 
11
:
131
135
.
Lorenz
G. C
Barrett
G. W.
.
1990
.
Influence of simulated landscape corridors on house mouse (Mus musculus) dispersal
.
American Midland Naturalist
 
123
:
348
356
.
McComb
W. C.
2003
.
Ecology of coarse woody debris and its role as habitat for mammals
. Pp.
374
404
in
Mammal community dynamics; management and conservation in the coniferous forests of western North America
  (
Zabel
C. J.
Anthony
R. G.
, eds.).
Cambridge University Press
,
New York
.
Meffe
G. K.
Carroll
C. R.
.
1994
.
The design of conservation reserves
. Pp.
265
304
in
Principles of conservation biology
  (
Meffe
G. K.
Carroll
C. R.
, eds.).
Sinauer Associates, Inc., Publishers
,
Sunderland, Massachusetts
.
Menard
S. W.
1995
.
Applied logistic regression analysis. Quantitative applications in the social sciences no. 106
 .
Sage Publications
,
Thousand Oaks, California
.
Merriam
G.
Lanoue
A.
.
1990
.
Corridor use by small mammals: field measurement for three experimental types of Peromyscus leucopus
.
Landscape Ecology
 
4
:
123
132
.
Metzgar
L. H.
1967
.
An experimental comparison of screech owl predation on resident and transient white-footed mice (Peromyscus leucopus)
.
Journal of Mammalogy
 
48
:
387
391
.
Newman
J. A.
Caraco
T.
.
1987
.
Foraging, predation hazard and patch use in gray squirrels
.
Animal Behaviour
 
35
:
1804
1813
.
Noss
R. F.
1987
.
Corridors in real landscapes: a reply to Simberloff and Cox
.
Conservation Biology
 
1
:
159
164
.
Planz
J. V.
Kirkland
G. L.
Jr.
1992
.
Use of woody ground litter as a substrate for travel by the white-footed mouse Peromyscus leucopus
.
Canadian Field-Naturalist
 
106
:
118
121
.
Price
K.
Boutin
S.
Ydenberg
R.
.
1990
.
Intensity of territorial defense in red squirrels: an experimental test of the asymmetric war of attrition
.
Behavioral Ecology and Sociobiology
 
27
:
217
222
.
Rice
W. R.
1989
.
Analyzing tables of statistical tests
.
Evolution
 
43
:
223
225
.
Ricketts
T. H.
2001
.
The matrix matters: effective isolation in fragmented landscapes
.
American Naturalist
 
158
:
87
99
.
Roche
B. E.
Schulte-Hostedde
A. I.
Brooks
R. J.
.
1999
.
Route choice by deer mice: reducing the risk of auditory detection by predators
.
American Midland Naturalist
 
142
:
194
198
.
Rosenberg
D. K.
Noon
B. R.
Meslow
E. C.
.
1997
.
Biological corridors: form, function, and efficacy
.
BioScience
 
47
:
677
687
.
Ruefenacht
B.
Knight
R. L.
.
1995
.
Influences of corridor continuity and width on survival and movement of deermice Peromyscus maniculatus
.
Biological Conservation
 
71
:
269
274
.
Sakai
H. F.
Noon
B. R.
.
1997
.
Between-habitat movement of dusky-footed woodrats and vulnerability to predation
.
Journal of Wildlife Management
 
61
:
343
350
.
SAS Institute Inc
.
2000
.
SAS/STAT user's guide. Release 8.1 ed
 .
SAS Institute Inc.
,
Cary, North Carolina
.
Schooley
R. L.
Sharpe
P. B.
Van Horne
B.
.
1996
.
Can shrub cover increase predation risk for a desert rodent?
Canadian Journal of Zoology
 
74
:
157
163
.
Simonetti
J. A.
1989
.
Microhabitat use by small mammals in central Chile
.
Oikos
 
56
:
309
318
.
Smith
A. A.
Mannan
R. W.
.
1994
.
Distinguishing characteristics of Mount Graham red squirrel midden sites
.
Journal of Wildlife Management
 
58
:
437
445
.
Smith
C.
1995
.
The niche of diurnal tree squirrels
. Pp.
209
225
in
Storm over a mountain island; conservation biology and the Mt. Graham affair
  (
Istock
C. A.
Hoffmann
R. S.
, eds.).
University of Arizona Press
,
Tucson
.
Smith
C. C.
1968
.
The adaptive nature of social organization in the genus of tree squirrels Tamiasciurus
.
Ecological Monographs
 
38
:
31
63
.
Stapp
P.
Van Horne
B.
.
1997
.
Response of deer mice (Peromyscus maniculatus) to shrubs in shortgrass prairie: linking small-scale movements and the spatial distribution of individuals
.
Functional Ecology
 
11
:
644
651
.
Stokes
M. E.
Davis
C. S.
Koch
G. G.
.
1995
.
Categorical data analysis using the SAS system
 .
SAS Institute Inc.
,
Cary, North Carolina
.
Swihart
R. K.
Slade
N. A.
Bergstrom
B. J.
.
1988
.
Relating body size to the rate of home range use in mammals
.
Ecology
 
69
:
393
399
.
Temple
S. A.
1987
.
Do predators always capture substandard individuals disproportionately from prey populations?
Ecology
 
68
:
669
674
.
Thibault
A.
Bovet
J.
.
1999
.
Homing strategy of the eastern chipmunk, Tamias striatus (Mammalia: Rodentia): validation of the critical distance model
.
Ethology
 
105
:
73
82
.
Tuen
A. A.
Brown
J. S.
.
1996
.
Evaluating habitat suitability for tree squirrels in a suburban environment
.
Malaysian Applied Biology
 
25
:
1
8
.
Zar
J. H.
1999
.
Biostatistical analysis
 .
Prentice-Hall, Inc.
,
Upper Saddle River, New Jersey
.
Zielinski
W. J.
2000
.
Weasels and martens: carnivores in northern latitudes
. Pp.
95
113
in
Activity patterns in small mammals; an ecological approach
  (
Halle
S.
Stenseth
N. C.
, eds.).
Springer-Verlag
,
New York
.
Zollner
P. A.
2000
.
Comparing the landscape level perceptual abilities of forest sciurids in fragmented agricultural landscapes
.
Landscape Ecology
 
15
:
523
533
.

Author notes

Associate Editor was Craig L. Frank.