Evidence of natal habitat preference induction within one habitat type

1. Introduction

Natal dispersal is an important ecological process to promote gene flow, reduce inbreeding, and contribute to the viability of populations and survival of individuals [1], and is composed of three distinct phases: emigration, transience, and immigration [2]. Much work is dedicated to understanding the proximate and ultimate factors driving emigration [2–4] and has provided a framework in which we can form and test hypotheses regarding the initiation of natal dispersal. By contrast, fewer studies focus on transience and immigration (settlement), which are difficult to document, yet of critical importance for understanding how individuals select features on the landscape and make settlement decisions. Ultimately choices during these phases influence individual fitness [5–7]. The transience stage of natal dispersal can be costly both in terms of energy expenditure and risk of predation during dispersal, and in deferred costs to survival and fitness post-dispersal [8,9], therefore, strategies to aid in identifying quality settlement locations and reduce time spent in transience may be favoured [10]. A disperser's decision to stop moving and select a settlement location may be among the most important choices made in its lifetime, akin to selecting mates [5,6,10], yet what environmental cues play a role in this decision are poorly understood. Several factors may serve as reliable cues that influence or trigger settlement during natal dispersal, including an individual's physical condition, social cues such as the presence of conspecifics, and environmental cues that signify quality habitat complete with food, shelter, and mates [10,11].

Recognition of and preference for habitat features that resemble those of the natal area by dispersing individuals is termed habitat imprinting or natal habitat preference induction (NHPI) [12–14]. This mechanism for habitat selection is thought to have adaptive significance because a natal area that supports offspring to dispersal age likely typifies quality habitat [15]. Selection may favour NHPI as the use of reliable cues facilitates rapid identification of quality habitat and reduces search time prior to settlement [12], and may reduce the dependence on specific phenological cues only present during narrow temporal windows. NHPI, in turn, may drive observed variation in post-dispersal habitat selection [12]. Rapid habitat alteration via anthropogenic and climate-driven disturbances may complicate or reduce the current adaptive value of NHPI, effectively decoupling the cues that once conferred fitness benefits. Situations where once reliable habitat cues become decoupled from fitness benefits are known as ecological traps [16–18], and if NHPI is a common mechanism used by dispersers to identify suitable settlement locations, habitat selection may appear maladaptive in landscapes that have experienced rapid change from anthropogenic disturbances such as fire, fragmentation, non-native species, parasites, and disease [18,19]. Further, should rapid change alter important habitat cues without affecting the overall fitness benefits of a site, NHPI may prevent settlers from identifying sites as suitable for occupancy [20]. Thus NHPI is an important consideration for the conservation of threatened species and the habitats required to sustain them.

Evidence of natal habitat preference during dispersal has been documented in insects, fish, amphibians, reptiles, birds, and mammals [11,17]. NHPI has rarely been satisfactorily documented in observational field-based studies due to three strict criteria that must be satisfied: (i) genetic variation in habitat preference must be controlled, (ii) experience with natal environments should occur at the appropriate age or ontogenetic stage that dispersal occurs, and (iii) tests for habitat preference should occur at the age or ontogenetic stage that settlement occurs [12]. Several recent studies appear to document NHPI in wild animal populations, and suggest NHPI may be important. Young North American red squirrels (hereafter red squirrels), Eurasian red squirrels (Sciurus vulgaris), brush mice (Peromyscus boylii), and red wolves (Canis rufus) selected settlement sites of a similar habitat type to their natal area or natal home range [14,21–24]. Siberian flying squirrels (Pteromys volans) settled in forest patches of similar size to the natal patch [23], and dispersing common loons (Gavia immer) selected lakes of similar size and pH to their natal lake for breeding [25]. Although most tests of NHPI examine post-dispersal habitat selection among distinct, contrasting habitat types, if NHPI is a successful strategy we might also expect that animals have the capacity to resolve more subtle structural or compositional differences within a single habitat type, particularly in the case of habitat specialists.

North American red squirrels are a small forest obligate mammal with a broad distribution and are unique in their territoriality and food hoarding behaviour; territories are centred around a central larder hoard or midden [26]. Because of their close affinity to and habitat specificity within forests [26], red squirrels make excellent model systems for the study of habitat cues that influence settlement decisions. Red squirrel territoriality and larder hoarding behaviour help buffer against fluctuating resource availability, and also require that individuals make informed settlement choices within a narrow window of time, as the chosen settlement location must supply adequate conifer cones and suitable habitat structure to maintain cone stores through the following year. Previous work has shown that red squirrels inhabiting mature and highly thinned mixed-conifer forest exhibit a preference for settlement sites that resemble their natal areas [22]. We studied natal dispersal and post-dispersal habitat selection, and test for evidence of NHPI in the Mount Graham red squirrel (Tamiasciurus hudsonicus grahamensis) inhabiting mixed-conifer forest by measuring microhabitat variables at two spatial scales. Mt Graham red squirrels are an isolated, endangered subspecies restricted to the Pinaleño Mountains, in southeastern Arizona, USA, and represent the southernmost population of red squirrels in North America [27] (figure 1). Recent genetic analyses suggest that southwestern Tamiasciurus belong to a separate species T. fremonti [28], further distinguishing Mt Graham red squirrels genetically from other red squirrel populations in which NHPI has been investigated. The Mt Graham red squirrel is uniquely suited to study NHPI and its influence on habitat use and selection for many reasons: (i) Mt Graham red squirrels (hereafter red squirrels) are intensively monitored via mark–recapture, radiotelemetry, and population surveys, hence many individuals of known age are followed during the period when both dispersal and settlement are known to occur, (ii) recent analyses of nuclear and mitochondrial DNA indicate that little genetic variation exists, with mean inter-individual relatedness = 0.75 [27,29], which may allow us to more adequately disentangle the influence of genetics or NHPI on inter-individual differences in habitat selection observed in the field [12], (iii) recent disturbance events including insect infestation beginning in 1996 [30] and subsequent wildfires in 1996 and 2004 [31] have altered forest composition and continuity, in some areas leaving structure intact, but substantially reducing living biomass, and (iv) despite high mortality and reduced reproductive success in insect damaged spruce–fir forest, red squirrel settlement cues appear to remain unchanged, suggesting a potential ecological trap for individuals that settle there [16,32,33].

• Download figure
• Open in new tab
• Download powerpoint

Figure 1.

Overview of Mt Graham red squirrel habitat above 2 348 m in elevation, burn severity and associated study areas in the Pinaleño Mountains, Arizona, USA. Natal and settlement sites obtained from radio-collared juveniles shown in grey and white circles, respectively, and straight-line dispersal vectors are indicated by black lines.

If NHPI occurs in juvenile red squirrels born and settling within mixed-conifer forest, then intra-individual differences in forest structure and composition between natal and settlement locations should be less than that expected from random pairwise comparisons of natal and settlement sites. We determined natal and post-dispersal site characteristics from ground-based plots and remotely sensed data at two spatial scales and assessed habitat selection in juveniles by comparing forest structure, composition, food availability, and microclimate at natal and settlement sites to random locations. We then tested for evidence of NHPI within mixed-conifer forest across all individuals and in short- versus long-distance dispersers by quantification of intra-individual differences in forest structure and composition between natal and settlement locations compared with expected differences obtained from randomization.

3. Results

(c) Evidence of natal habitat preference induction in mixed-conifer forest

Across plot types, intra-individual differences in forest structure and composition, estimated simultaneously via Euclidean distance between natal and settlement habitat vectors, were smaller than expected based upon randomized pairwise differences (mean Euclidean distance score (ED score) plot (n = 47) = 3.8, mean expected ED score of randomizations = 4.2, p = 0.002; mean ED score 30 m buffer (n = 63) = 2.5, mean expected ED score of randomizations = 3.1, p < 0.001; mean ED score 100 m buffer (n = 63) = 2.2, mean expected ED score of randomizations = 3.1, p < 0.001; figure 2). Intra-individual Euclidean distances were also smaller than expected in short-distance dispersers (less than or equal to 150 m from their natal nest) across all plot types (mean ED score plot (n = 27) = 3.3, mean expected ED score of randomizations = 4.3, p < 0.001; mean ED score 30 m buffer (n = 37) = 1.9, mean expected ED score of randomizations = 3.1, p < 0.001; mean ED score 100 m buffer (n = 37) = 1.4, mean expected ED score of randomizations = 3.1, p < 0.001; figure 2).

• Download figure
• Open in new tab
• Download powerpoint

Figure 2.

Distribution of observed and expected intra-individual Euclidean distance scores between forest structure and composition at natal and settlement sites for juvenile Mt Graham red squirrels in the Pinaleño Mountains, Arizona, USA. Intra-individual Euclidean distance scores within 30 m ground-based plots, and within 30 and 100 m buffers summarized via lidar data are shown in relation to 10 000 iterations of randomized pairwise comparisons between natal and settlement sites for all animals (a), short-distance dispersers (b), and long-distance dispersers (c). Mean empirical intra-individual Euclidean distances between natal and settlement locations are indicated with a vertical line. (Online version in colour.)

Individuals that dispersed long distances (greater than 150 m from their natal nest) did not exhibit strong evidence of NHPI when all variables were considered simultaneously, as intra-individual Euclidean distances between natal and settlement sites were not different from the expected distribution of random pairwise differences (mean long-distance ED score plot (n = 19) = 4.6, mean expected long-distance ED score of randomizations = 4.3, p = 0.94; mean long-distance ED score 30 m buffer (n = 23) = 3.4, mean expected long-distance ED score of randomizations = 3.1, p = 0.91; mean long-distance ED score 100 m buffer (n = 23) = 3.5, mean expected long-distance ED score of randomizations = 3.1, p = 0.94; figure 2). Each year, 40% of females and 65% of males dispersed greater than 150 m from the natal area. Across scales and methods to summarize forest structure and composition (ground-based plot and 30 and 100 m buffers with lidar data), on average 38% (37% based on ground plot variables, 43% based on lidar variables within 30 m buffers, and 35% based on lidar variables within 100 m buffers) of these individuals did show evidence of NHPI and selected settlement locations that were more similar than expected (i.e. proportion of long-distance dispersers with ED score < random ED score), whereas 62% of long-distance dispersers settled in areas structurally and compositionally dissimilar to their natal area.

There is some evidence that females tended to select settlement sites that were more similar to their natal areas compared with males, particularly at the 30 m scale (ground-based plot mean female ED score: 3.39, mean male ED score 4.17, Welch t_44.9 = −1.77, p = 0.08; lidar 30 m buffer mean female ED score: 2.00, mean male ED score 3.01, Welch t_53.8 = −2.54, p = 0.01), including ground-based plot variables for females that dispersed long distances (long-distance ground-based plot mean female ED score: 3.66, mean male ED score: 4.93, Welch t_2.8 = −2.51, p = 0.09). There were no differences in ED scores between males and females at 100 m buffer scales.

(i) Evidence of natal habitat preference induction in plot-based and lidar-derived variables

Across all individuals, basal area, live basal area, dead basal area, number of living trees, and species diversity measured at ground-based plots were more similar between natal and settlement locations than expected (table 1). However, when only long-distance dispersers were considered, natal and settlement sites were not different from expected random pairwise differences based upon ground-based plot variables (table 1). Among lidar-derived variables summarized at the plot and home range scale (30 and 100 m buffers), canopy cover, tree height, live basal area, total basal area, variability in tree height, and the ratio of living to total basal area were more similar between natal and settlement locations than expected at the home range scale (table 1). Long-distance dispersers settled in locations that had canopy cover and live basal area more similar to their place of birth than expected at the plot scale; at the home range scale canopy cover tended to be more similar than expected approximately 80% of the time (table 1).

View this table:

• View inline
• View popup

Table 1.

Observed versus expected absolute differences in forest structure and composition between natal and settlement locations for juvenile Mt Graham red squirrels measured from ground-based plots and lidar data within 30 and 100 m buffers. Shown are the absolute differences in forest structure and composition at settlement locations compared with natal locations (natal–settlement, with sign of the actual difference indicated by +/−) for all animals and for long-distance dispersers that settled >150 m from the natal area survival to first spring in the Pinaleño Mountains, Arizona, USA. Instances where observed differences were less than expected with probability p < 0.2 are in italics to indicate variables that may represent important cues for NHPI.

Although natal and settlement locations were more similar than expected for many variables, matching was not exact, and mismatches tended to occur such that natal forest variables > settlement forest variables and individuals settled in locations with less forest structure than their natal area. At ground-based plots, mean intra-individual differences (natal − settlement) suggest that all individuals and long-distance dispersers settled at locations with less basal area, large trees and species richness, more live and dead trees, and higher species diversity compared with their natal area (indicated by sign of actual differences, table 1). For lidar-derived variables summarized at the plot scale, all individuals and long-distance dispersers tended to settle at locations with less canopy cover, mean tree height, variability in tree height, and basal area compared with their natal area. At the home range scale, all individuals and long-distance dispersers settled at locations with more canopy cover and lower mean tree height, variability in tree height, and live to total basal area ratio compared with their natal area; long-distance dispersers settled in areas with more total basal area (table 1).

(ii) Euclidean distance scores versus individual and environmental variables

At ground-based plots, and within 30 and 100 m buffers characterized via lidar, intra-individual ED scores between natal and settlement locations across all animals and for short- and long-distance dispersers were not correlated with individual or environmental attributes such as sex, individual body mass index, number of litter mates, individual behavioural tendency for locomotion, cone availability at the natal site, cone availability at the settlement site, or year (all Pearson's linear correlations, p > 0.12; range 0.12–0.72). Intra-individual ED scores between natal and settlement locations were correlated with dispersal distance, and the strength and direction of the correlation depended upon the scale (30 m plot or 100 m home range) and the group of individuals in consideration (all, short-, and long-distance dispersers; figure 3).

• Download figure
• Open in new tab
• Download powerpoint

Figure 3.

Relationship between intra-individual pairwise Euclidean distance scores between natal and settlement sites for juvenile Mt Graham red squirrels in the Pinaleño Mountains, Arizona, USA, and dispersal distance (m) for short-distance dispersers ≤150 m (a), long-distance dispersers moving > 150 m (b), and all animals (c). Lines represent linear models of the relationship between Euclidean distance scores and dispersal distance at three scales of forest structure and composition characterization: ground-based plot data (solid line), lidar data summarized within 30 m buffers (dotted line), and lidar data summarized within 100 m buffers (dashed line). (Online version in colour.)

For short-distance dispersers, intra-individual ED scores tended to be positively correlated with dispersal distance (ground-based plot: r = 0.14, p = 0.50; lidar 30 m buffer: r = 0.35, p = 0.04; lidar 100-m buffer: r = 0.27, p = 0.11; figure 3). For long-distance dispersers, ED scores at the plot scale tended to be negatively correlated with dispersal distance, evidence that individuals that dispersed the farthest settled at locations more similar to their natal area in terms of variables measured at the plot scale (ground-based plot: r = −0.34, p = 0.10; lidar 30 m buffer: r = 0.07, p = 0.74). At the home range scale, ED scores tended to be positively correlated with dispersal distance (lidar 100 m buffer: r = 0.53, p = 0.01; figure 3). When we considered all animals together, intra-individual ED scores tended to be positively correlated with dispersal distance overall (ground-based plot: r = 0.22, p = 0.13; lidar 30 m buffer: r = 0.47, p < 0.001; lidar 100 m buffer: r = 0.71, p < 0.001; figure 3). In summary, we observed different kinds of evidence in support of NHPI for short- and long-distance dispersers. Intra-individual natal and settlement habitat characteristics were most similar among short-distance dispersers (figure 2), but habitat similarity may be partially explained by spatial autocorrelation (figure 3). Among long-distance dispersers at the plot scale, lack of spatial autocorrelation in post-settlement habitat selection and sustained evidence of NHPI for key red squirrel habitat components provide further evidence for NHPI despite the influence of spatial autocorrelation.

4. Discussion

Red squirrels appear to select forest structure and composition that differs from that randomly available on the landscape, and individuals tend to select components of forest structure that resemble their own natal area. Evidence for NHPI has been documented in free-living mammals born in strongly contrasting vegetation communities or structural types such as woodland and prairie [42], grassland and oak savannah [14], an urban, forest, shrub, and agricultural matrix [24], or in thinned versus mature forest [22], and here we test the hypothesis that NHPI may also function in single vegetation communities occupied by habitat specialists. Our results provide additional support for NHPI as a mechanism for post-dispersal habitat selection [14,22,25,42], in general, and specifically in an organism that is dependent upon particular vegetation community types such as mixed-conifer forest.

Ethics

All fieldwork was conducted under the University of Arizona Institutional Animal Care and Use Committee protocol no. 08-024, Arizona Game and Fish Department scientific collecting permit no. SP654189, US Fish and Wildlife Service permit no. TE041875-0, and adhered to the American Society of Mammalogists guidelines for the use of wild mammals in research (Sikes, Gannon, and The Animal Care and Use Committee of the American Society of Mammalogists 2011).

Data accessibility

Merrick M, Koprowski J. Data from: Evidence of natal habitat preference induction within one habitat type. Dryad Digital Repository: http://dx.doi.org/10.5061/dryad.v087b.

Authors' contributions

M.J.M. and J.L.K. jointly conceived the study. M.J.M. carried out the data collection, analysis, and drafting of the manuscript. J.L.K. provided additional funding, logistical support, served as project advisor, and contributed to manuscript revisions.

Competing interests

We declare we have no competing interests.

Funding

This research was supported by grants from the University of Arizona, US Fish and Wildlife Service, USDA Forest Service, Arizona Game and Fish Department, the Arizona Agricultural Experiment Station to J.L.K. and funds to M.J.M. from The Joint Fire Sciences Program Graduate Research Innovations award no. 3005940, The University of Arizona NASA Space Grant Consortium Fellowship, The University of Arizona Institute of the Environment Carson Scholars Fellowship, The American Society of Mammalogists Grant in Aid of Research and ASM Fellowship, The American Museum of Natural History Theodore Roosevelt Graduate Student Research Award, The Southwestern Association of Naturalists Howard McCarley Student Research Award, and T & E Inc. Grants for Conservation Biology.