Experimental removals reveal dietary niche partitioning facilitates coexistence between native and introduced species

Abstract Niche overlap between native species and ecologically similar invaders can lead to competitive exclusion of threatened native species, but if two such species also co‐occur naturally elsewhere, interactions between native and introduced populations may mirror coevolved niche partitioning that reduces competition and promotes coexistence. A single, insular population of Fremont's squirrel (Tamiasciurus fremonti) the Mount Graham red squirrel (MGRS; T. f. grahamensis) in the Pinaleño Mountains, Arizona, USA, is critically endangered and resource competition with introduced Abert's squirrels (Sciurus aberti) may threaten its long‐term persistence. The species are naturally synoptic in other mountain sites, and both consume diets comprised primarily of conifer seeds and fungi. We conducted experimental removals of introduced Abert's squirrels and used stable isotope analysis of diets before and after removals, and of diets in naturally syntopic populations to test the hypothesis that dietary niche partitioning can facilitate coexistence between native and introduced species. We also developed a novel approach to determine the influence of fluctuating food availability on carbon enrichment in consumers. Mount Graham red squirrels and introduced Abert's squirrels partitioned the dietary niche similarly to naturally syntopic populations. Removals had no apparent effect. Diet of MGRS was more closely linked to availability of resources than to presence of Abert's squirrels. Flexible dietary niche of introduced Abert's squirrels may have allowed them to exploit a resource opportunity in syntopy with MGRS. Variable food production of MGRS habitat may intensify competition in poor years, and territorial defense against non‐native Abert's squirrels likely imposes fitness costs on individual MGRS. Similarity in our model species’ diets may make MGRS more vulnerable to competition if climate change eliminates the advantages of larder‐hoarding. Where introduced populations of ecologically similar species are better adapted to changing conditions, they may ultimately replace southern peripheral populations of native species.


| INTRODUC TI ON
Invasive species are permanent components of most ecosystems (Vitousek, Mooney, Lubchenco, & Melillo, 1997) and understanding the effects of competition between invaders and ecologically similar natives is important to conserve biodiversity (Davis, 2003). Unlike invasive predators, which have directly caused extinctions of native species (Sax & Gaines, 2008), competitive exclusion by invaders is challenging to quantify because it may not be directly observable and population declines may have multiple causes (Sax et al., 2007).
Native wildlife population declines due to invasive species have been documented (Herbold & Moyle, 1986;Petren & Case, 1996), but competition could also lead to niche partitioning and coexistence (Davis, 2003;Shea & Chesson, 2002). The ability to predict the outcome of biological invasions would be valuable to conservation of threatened populations.
In systems with variable food production, competing native species may adapt by altering their diets and feeding strategies in order to avoid direct competition (Pulliam, 1986). In two closely related species of charr (Salvelinus sp.) that occur in zones of sympathy, niche partitioning varied over time, and according to species and size of individuals (Nakano, Fausch, & Kitano, 1999). Niche partitioning may also be flexible in novel interactions where an introduced species occupies habitat similar to its native range, which could promote coexistence with native competitors. We studied a model system involving interactions between Abert's squirrels (Sciurus aberti) and Fremont's squirrels (Tamiasciurus fremonti) in Arizona, USA, where the species co-occur both naturally and in anthropogenic syntopy. One Fremont's squirrel subspecies, the Mount Graham red squirrel (MGRS; T. f. grahamensis) occurs in a single, critically endangered population in the Pinaleño Mountains, an isolated mountain range surrounded by desert grassland. Abert's squirrels introduced in the 1940s (Davis & Brown, 1988) have established throughout all forested areas of the Pinaleño Mountains (Hutton et al., 2003). Elsewhere in Arizona, T. fremonti (hereafter, red squirrel) and Abert's squirrels are naturally syntopic in zones of overlap between preferred forest types (Ferner, 1974;Hall, 1981); red squirrels inhabit primarily mixed-conifer and spruce-fir forest (Smith, 1981), and Abert's squirrels are typically associated with ponderosa pine (Pinus ponderosa) forest (Keith, 1965). Red squirrels are small (200-250 g) larder-hoarders that display vigorous movements and vocalizations in defending territories (Koprowski, 2005;Smith, 1968a) whereas the larger (500-750 g) Abert's squirrels are secretive, occasional scatter-hoarders (Allred, 2010;Keith, 1965).
A potentially important effect of interaction is through competition for food. Diets of Abert's and red squirrels consist primarily of conifer seeds and fungi (Nash & Seaman, 1977;Steele, 1998). Both species also consume pollen from staminate cones of conifers during late spring and early summer (Keith, 1965;Smith, 1968a). During periods of scarcity, Abert's squirrels also depend on phloem stripped from conifer twigs (Allred, 2010;Patton, 1974), and red squirrels rely on food cached in their larder-hoards or middens (Steele, 1998).
Such divergent strategies, in addition to differential use of available food in periods of relative abundance may facilitate coexistence in natural syntopy (Chesson, 1994). As Abert's squirrel are 2-3 times larger than red squirrels, their ability to depress availability of diet items could have negative effects on MGRS especially during years of low food production. It is unknown, to what extent dietary resource partitioning occurs in the invaded community of the Pinaleño Mountains. However, the source population for Abert's squirrel translocations was naturally syntopic with red squirrels (Davis & Brown, 1988), and niche separation may be maintained by coevolved differences in habitat selection (Rosenzweig, 1981).
To examine the potential for coexistence between ecologically similar native and introduced species, we combined an experimental removal of introduced Abert's squirrels with a stable-isotopic study of dietary niche partitioning among naturally syntopic red squirrels and Abert's squirrels. Stable isotope analysis is a well-established method for understanding dietary resource partitioning among carnivores (Merkle, Polfus, Derbridge, & Heinemeyer, 2017) and herbivores (Bovendorp, Libardi, Saramento, Camargo, & Percequillo, 2017). We tested the hypothesis that dietary niche partitioning facilitated coexistence. We predicted: (a) low overlap in proportions of common items in MGRS and Abert's squirrel diets; (b) Abert's squirrel removals would not cause MGRS diet shifts; (c) evidence of dietary to changing conditions, they may ultimately replace southern peripheral populations of native species.

K E Y W O R D S
coexistence, diet, introduced species, niche, wildlife niche partitioning, specifically in total proportion of fungi consumed; (d) overlaps in diet proportions would be related to variability in food production; and (e) dietary niche interactions between MGRS and Abert's squirrels would be similar to those in sites of natural syntopy.

| Study site
We collected data from study sites in coniferous forest of 3 mountain ranges in Arizona, USA, where red squirrels and Abert's squirrels were common. We conducted experimental removals on Abert's squirrels at our primary site, Mount Graham (MG), in the Pinaleño Mountains, Graham County. This site was ≈200 ha of mixed-conifer forest composed of Engelmann spruce (Picea engelmannii), Douglasfir (Pseudostuga menziesii), corkbark fir (Abies lasiocarpa var. arizonica), southwestern white pine (Pinus strobiformis), and ponderosa pine at 2,800-3,000-m elevation (Hutton et al., 2003). Two sites of natural syntopy between our model species were in the San Francisco Peaks (SF), Coconino County, and the White Mountains (WM), Greenlee County. The SF site was ≈100 ha in mixed-conifer forest dominated by Douglas-fir and ponderosa pine at 2,500-m elevation. The WM site was divided into 2 ≈50-ha areas 2,800-and 2,400-m elevation, dominated by Douglas-fir and corkbark fir, and ponderosa pine and Gambel oak (Quercus gambelii), respectively. Red squirrel population densities at the upper and lower elevation WM sites were ≈0.3/ ha and ≈0.1/ha, respectively (Derbridge & Koprowski, unpublished data). We also collected samples from introduced non-syntopic (INS) Abert's squirrels at a ponderosa pine forest area, ≈3 km southeast from the MG site at 2,750-m elevation, where MGRS were not present. inhalation of 100% isoflurane. We restrained captured animals in a cloth handling cone (Koprowski, 2002), and held a plastic bag containing cotton balls partially soaked with isoflurane over protruding nostrils and mouth for 30 s or until heartbeat was no longer detectable. We coordinated with local hunters to remove other Abert's squirrels and document removal effort.

| Experimental removals
We estimated efficacy of removal by surveys of Abert's squirrel feeding sign (i.e., remains of conifer cones, stripped twigs, clipped branches marked and scattered in patterns not seen in MG red squirrel sign) on 16 50-m line transects each in the treatment and control areas. We marked transects along randomly selected azimuths taken at 16 points selected at random from a list of historically mapped points used for long-term monitoring of the endangered MGRS. To ensure somewhat even spacing of transects, we selected 4 mapped points in each quadrant. We surveyed transects 9 times at 7-day intervals, May-July, 2012, and 8 times at ≈14-day intervals, May-August 2014. We were unable to survey transects as often in 2014 due to time constraints related to concurrent studies. We also used remote cameras (Bushnell Trophy Two cameras faced inwards at opposite ends of the transect, and a third camera faced perpendicular to the line at 25 m. We removed vegetation that had the potential to move in wind to limit the possibility of cameras being triggered by non-target activity. We set cameras to "Normal" sensitivity and highest image resolution (8 megapixels). As Abert's squirrels were unmarked and exclusive territories are not maintained (Farentinos, 1972), we assumed all single Abert's squirrel photographs for each hour of transect operation represented unique individuals. Where >1 Abert's squirrel F I G U R E 1 The Mount Graham study area in the Pinaleño Mountains, Arizona, USA. The black line indicates a road that bisected the study area; the Abert's squirrel treatment area was west of the road. Black crosses (2012) and white circles (2014) indicate Abert's squirrel removal locations. Animals removed from the control area were known to be primarily resident in the treatment area occurred in a single image, we counted two animals. We used Mann-Whitney U tests (α = 0.05) to compare counts of Abert's squirrel feeding sign and photographs between the treatment and control areas.

| Diet sample collection and preparation
We collected known diet items of our model species from the MG site during 2011-2014, and from the SF and WM sites in 2014 and 2015, including seeds from ovulate cones of conifers (Dodd, States, & Rosenstock, 2003;Edelman & Koprowski, 2005;Froehlich, 1990), epigeous fungi (i.e., mushrooms) and hypogeous fungi (i.e., truffles and false truffles [hereafter truffles]; Keith, 1965;Smith, 1968a). We used freshly dug shallow (<50 mm) pits created by squirrels when unearthing truffles to guide our own searches, and probed the peripheries of these pits to find remaining sporocarps. We also collected conifer twigs from branches clipped by Abert's squirrels, and pollen cones from conifers at which we found squirrel feeding sign.
Both squirrel species molt in spring and fall; tails are molted only in fall (Keith, 1965;Layne, 1954) We used an ultrasonic cleaner (Model FS20H, Fisher Scientific, Pittsburgh, PA) to remove coarse debris from hair samples in glass vials of deionized water. After air-drying for 24 hr, we rinsed samples under a ventilation hood in a 2:1 chloroform/methanol solution to remove fine debris and oils (Derbridge, Krausman, & Darimont, 2012). We removed and air-dried kernels from seed husks of conifer cones. We air-dried fungi samples and used a scalpel to cut out clean sporocarp interiors. We used a scalpel to strip outer bark from twigs, and separately removed the exposed phloem layer. We ground diet items and hairs to powder in a Wig-L-Bug ® DS-80 amalgamator (Crescent Dental Co., Chicago, IL). We agitated staminate cones inside a clear plastic sample bag to separate pollen from cones. We We sent samples to the Environmental Isotope Laboratory (Department of Geosciences, University of Arizona) for analysis of C and N stable isotopes on a continuous-flow gas-ratio mass spectrometer (Finnigan Delta PlusXL) coupled to a Costech elemental analyzer. We express isotope values in delta notation (δ) as: where Z is 13 C or 15 N, and R is 13 C/ 12 C or 15 N/ 14 N. Standardization was based on acetanilide for elemental concentration, NBS-22 and USGS-24 for δ 13 C, and IAEA-N-1 and IAEA-N-2 for δ 15 N. Based on repeated internal standards, precision was better than ±0.10 for δ 13 C and ±0.2 for δ 15 N.

| Stable isotope diet analysis
We calculated mean and standard deviation δ 13 C and δ 15 N for all diet items and squirrel species, and used t-tests to determine if diet items and squirrel species were isotopically distinct (α = 0.05). Because Abert's squirrels rarely consume spruce seeds (Allred, 2010;Edelman & Koprowski, 2005), we calculated separate δ 13 C and δ 15 N that omitted spruce seeds from the Abert's squirrel diet analysis. We used a Bayesian stable isotope mixing model with rodent-specific diet-hair trophic discrimination factors of 3.4‰ for C and 2.4‰ for N (Kurle, Koch, Tershy, & Croll, 2014) to estimate the proportional contribution of each diet item to the fall and summer diets (Moore and Semmens 2008) of red squirrels and Abert's squirrels. We used

| Effect of food production on diet
We examined the relationship between annual variation in availability of common diet items and composition of MGRS and introduced Abert's squirrel diets by taking advantage of the fact that δ 13 C of seeds and mushrooms were isotopically distinct -seeds being relatively enriched and mushrooms relatively depleted in δ 13 C ( Figure 3). We used 21 years (1994-2014) of conifer seed and mushroom production data collected by the University of Arizona's MGRS Monitoring Program (Koprowski, Alanen, & Lynch, 2005) to establish long-term food availability means and standard deviations.
Conifer seeds were collected from 9 10-m x 10-m seed plots each containing three randomly placed 0.25-m 2 seed traps. Viable seeds of Engelmann spruce, Douglas-fir, and corkbark fir were quantified as seeds/m 2 . All mushrooms of genera known to be consumed by red squirrels (Smith, 1968a(Smith, , 1968b were collected from 9 1-m × 100-m plots, at 2-week intervals over 12 weeks from late July-October each year and quantified as dry weight kg/ha. We used analysis of variance (ANOVA) to test for among-year differences in conifer seed and mushroom production during our study, 2011-2014. We used δ 13 C as an index of the effect of variability in conifer seed and mushroom production on δ 13 C of MGRS and introduced Abert's squirrels. We used δ 13 C of individual squirrels as the response variable in linear regression models with predictors of seed and mushroom availability in each year 2011-2014. We scaled seed and mushroom data by subtracting our 21-year mean from each year of study and dividing by the 21-year standard deviation. We assumed that because mushrooms and truffles formed a distinct δ 13 C group ( Figure 3) and truffle production may be somewhat proportional to mushroom production (Fogel & Hunt, 1979), our model would provide insight on fungi availability in general. We included the covariates season (summer/fall), group (before/after removal), area (treatment/control), and sex. To take into account the potential effect of timing of removals on variation in δ 13 C, we included a group*area interaction in models. We used a backwards elimination process to remove non-significant variables one at time to choose a final model by Akaike's Information Criterion (AIC) when only significant variables or the main variables of biological interest (i.e. conifer seeds and mushrooms) remained. We conducted this analysis using the lm function in R 3.4.2 (R Development Core Team 2017).

| Experimental removals
We removed 18 Abert's squirrels from the treatment area dur-

| Stable isotope data
We collected 110 and 71 hair samples from MGRS and Abert's squirrels, respectively, at the MG site, and 7 from INS Abert's F I G U R E 3 Individual δ 13 C and δ 15 N of Mount Graham red squirrels and Abert's squirrels in fall and summer, and mean and SD of δ 13 C and δ 15 N of their fall and summer diet items. Stable isotope values were calculated from squirrel hairs and samples of diet items collected 2011-2014 in the Pinaleño Mountains, Arizona, USA. Mushrooms, phloem, and pollen were not isotopically distinct, and diet models used combinations of mushroom and phloem data (epphlo Figure 6) for fall diet, and mushroom and pollen data (epipol Figure 7) for summer diet squirrels, and 47 from red and Abert's squirrels at the SF and WM sites (Table 1). From 71 samples, δ 13 C and δ 15 N of diet items were different for ≥1 isotope in most comparisons, but epigeous fungi were not isotopically distinct from phloem or pollen ( Figure 3); therefore, we created two grouped diet sources from individual isotope values of epigeous fungi and phloem (i.e., epphlo), and epigeous fungi and pollen (i.e., epipol) to be used in fall and summer diet estimates, respectively. Although this grouping process limits inference on single diet items, it is recommended over discarding known diet items from analysis (Phillips et al., 2014). We commonly observed Abert's squirrels consuming pollen, and stomachs of 8 Abert's squirrels removed in May-July during the study were filled with a homogenous thick pollen-yellow paste (Derbridge, personal observation).

| Stable isotope diet analysis
Mount Graham red squirrel diets consisted of ≥60% conifer seeds and ≤40% fungi, pollen, and phloem in fall and summer, while Abert's squirrel diet was ≤52% conifer seeds and ≥48% fungi, pollen, and phloem at the MG site. Credible intervals for consumption of conifer seeds by introduced Abert's squirrels and MGRS did not overlap in fall or summer, and credible intervals for truffle consumption did not overlap in summer. The INS Abert's squirrels consumed a lower proportion of conifer seeds and higher proportion of truffles than MGRS and Abert's squirrels at the MG site ( Table 2).
The proportional contribution of all diet items to MGRS fall diet was the same in treatment and control areas before and after the first removal; all 95% credible intervals overlapped (Table 3a; Figure 6).
Mean conifer seed consumption decreased in both areas after the second removal, and mean fungi consumption increased (Table 3b; Figure 6). Abert's squirrels consumed more fungi and phloem than MGRS throughout the study ( Figure 6). Mount Graham red squirrel summer diet did not differ before and after the second removal; all 95% credible intervals overlapped (Table 3c; Figure 7).
Red squirrels at the SF site consumed primarily conifer seeds in fall (x = 58%) and SF Abert's squirrels consumed primarily truffles (x = 68%). At the WM site, fungi comprised ≥70% of red squirrel fall diet, and conifer seeds comprised ≥55% of WM Abert's squirrel diet.
Summer diet estimates of red squirrels at the SF and WM sites followed the same pattern as fall diets (Table 2).

| D ISCUSS I ON
We found broad support for the hypothesis that dietary niche partitioning can facilitate coexistence between ecologically similar native and introduced species. Experimental removals of Abert's squirrels had no effect on MGRS diet, and stable isotope analysis revealed consistent differences in proportions of primary food items in diets.
As predicted, consumption of fungi differed, diets were related to availability, and dietary niche interactions resembled those of the naturally co-occurring populations at the SF and WM sites.
Our results from monitoring of removal success provided evidence that we achieved measureable population reduction. It was not possible to remove all Abert's squirrels from the treatment area and we expected recolonization to occur due to habitat connectivity and the introduced species' ubiquity throughout forested areas of the Pinaleño Mountains. Camera data indicated Abert's squirrels were common on treatment area transects after the first removal in 2012 likely because we initiated these transects 10 weeks after concluding removals when Abert's squirrels were already recolonizing the area. During the second removal in 2014, we initiated camera transects during the peak of removal activity, and camera data re-  (Nakano et al., 1999). Introduction to the Pinaleño Mountains in the 1940s likely represented a niche opportunity (Shea & Chesson, 2002) for Abert's squirrels because high resource availability, especially at lower elevations where MGRS were rare or absent, allowed a positive rate of increase. Abert's squirrels likely achieved dietary niche separation from MGRS in several ways. First, the introduced species consumed more truffles in general but more than double the amount consumed by MGRS in summer. Second, Abert's squirrels consumed more mushrooms throughout the year (Dodd et al., 2003;States & Wettstein, 1998), but also more phloem and pollen in fall and summer, respectively. Because mushrooms were isotopically identical to phloem and pollen, we were unable to quantify their individual contributions to niche partitioning. However, phloem is a rare food item for MGRS (Froehlich, 1990) and common for Abert's squirrels (Keith, 1965), thus any proportion of phloem that provided nutrition and contributed to δ 13 C and δ 15 N of Abert's squirrels likely represented partitioning.
Abert's squirrels spend much of the earliest period of summer consuming pollen (Keith, 1965) and it may have been the most abundant food source for Abert's squirrels at the MG site before fungi production followed summer rains in early July. Climate change has caused milder winters and earlier springs in many alpine environments, and reduced snow cover along with altered phenology of diet Note. Separate values for seeds reflect a difference in cones targeted by red and Abert's squirrels (i.e., Abert's squirrels were not observed processing spruce cones).
TA B L E 2 Proportional contributions to diet (mean and 95% credible intervals) of 3 diet items/groups based on posterior estimates from Bayesian stable isotope mixing models of δ 13 C and δ 15 N in hairs of red and Abert's squirrels ( Abert's squirrels to be superior competitors for these respective resources (Connell, 1980). Niche partitioning according to seed type or size plays an important role in community structure for many taxa including other mammals (Brown & Lieberman, 1973) and species of ants (Davidson, 1977), but if this trade-off has been important for resource partitioning between MGRS and Abert's squirrels, the decline of spruce in the Pinaleño Mountains due to insect damage and fire  may not favor MGRS. Die-offs of spruce (Picea sp.) and expansion of invasive species populations are already widespread in North American forests (Logan, Régnière, & Powell, 2003).
Although the absence of a removal effect on MGRS diet alone could not clearly demonstrate niche partitioning, our comparisons across sites were consistent with the experimental results. Red squirrels and Abert's squirrels in syntopy appear to consume different proportions of conifer seeds and fungi. This is further supported by the relative contributions of seeds to WM and SF Abert's F I G U R E 6 Mean and SD (black bars) proportional contribution of seeds, truffles, and mushrooms and phloem (i.e. epphlo) to the fall diets of Mount Graham red squirrels in treatment (squares) and control ( aggressive territorial species coexist with submissive species, but the latter species is restricted in abundance and to less-preferred food (Savolainen & Vepsäläinen, 1988).
The hypothesis that dietary niche partitioning facilitated coexistence was further supported by our carbon index models. The greatest difference in conifer seed consumption between MGRS and Abert's squirrels occurred in 2011, when seed production was lowest during the study. Mount Graham red squirrels likely relied on seeds stored in middens during the previous year to avoid starvation following this crop failure (Smith, 1968b), but the strong influence of mushrooms represented by a negative estimate (i.e. depleted δ 13 C) in our availability model confirms this diet item's importance in a highly variable environment. When the largest cone crop during the study occurred the following year, conifer seeds Notes. We used linear models with conifer seed production (c), mushroom production (m), season (s), group (g), area (a), sex (sx), and an interaction between group and area as predictors of variation in δ 13 C of Mount Graham red squirrels (M), and Abert's squirrels (A). We estimated a full model with an interaction between group and area, and removed non-significant predictors until only the diet items and significant predictors remained. We then used AIC to select the top model. contributed the largest proportions to diets of both species. The relative enrichment in δ 13 C of Abert's squirrels demonstrated the introduced species' ability to exploit the resource opportunity of abundant seed production. The contrasting years and diet estimates suggest resource partitioning reduced competition for conifer seeds when in limited supply, and that seed production maintained the introduced Abert's squirrel population in good years. Similar patterns of coexistence based on dietary niche separation were observed among gerbils (Gerbillus sp.) and jerboas (Jaculus jaculus), and the ability of jerboas to roam widely to take advantage of patchily available resources (Brown et al., 1994) may mirror the strategy employed by Abert's squirrels that move among red squirrel territories in search of food. Dietary niche partitioning between Mount Graham red squirrels and Abert's squirrels resembled partitioning in naturally syntopic populations of these species, which suggests competition for food does not preclude coexistence. This consistent separation may be a consequence of natural selection for a system of habitat selection that tends toward minimizing competition (Rosenzweig, 1981) between red squirrels and Abert's squirrels, as has been demonstrated through removal experiments on other rodents (Abramsky & Sellah, 1982). The Abert's squirrels introduced in the 1940s were translocated from the San Francisco Peaks, one of our sites of natural syntopy, and they may have been adapted for coexistence with red squirrels. Coevolution is expected to limit niche overlap and dampen effects of competition (Smith, Mooney, & Agrawal, 2008).
Other components of ecological niche overlap may differ among populations and provide evidence MGRS are negatively affected by introduced Abert's squirrels. Niche theory predicts per-capita growth rates to be limited by gains from resource consumption and losses from maintenance requirements, and native species could be competitively excluded if they have higher maintenance requirements than ecologically similar invaders (Shea & Chesson, 2002). In the Pinaleño Mountains, resources support invasive Abert's squirrels whose presence likely increases maintenance requirements for MGRS because they are forced to defend territories against the non-native intruders. In addition to this potential competition deficit for MGRS, daily interactions with Abert's squirrels may cause chronic stress leading to reduced individual fitness in the endangered native species. This possibility was recently demonstrated by a removal experiment where elevated glucocorticoid concentrations in Eurasian red squirrels dropped following removal of invasive eastern gray squirrels in Lombardy, Italy (Santicchia et al., 2018).
Future studies on other components of niche overlap could help determine the potential for stable coexistence between our model species and provide insight for other invaded communities.
Many of the challenges faced by MGRS are common to small populations at southern range peripheries. Such populations may be less resistant to invasion effects as they tend to possess lower genetic diversity (Vucetich & Waite, 2003) and occur in lower densities than populations at range centers (Lomolino & Channell, 1995).
Mount Graham red squirrels have extremely low genetic variation (Fitak, Koprowski, & Culver, 2013) which could inhibit their potential to adapt to a changing environment. They also have much larger home ranges than other red squirrel populations (Koprowski, King, & Merrick, 2008), and low population density may increase resource availability for Abert's squirrels whose introduced population in the Pinaleño Mountains is central to the species' native range. Rapid transformation of forests  and predicted declines in winter precipitation in the southwestern US (Williams et al., 2010) due to global climate change will not favor populations at southern extents of species' ranges. Resource partitioning between ecologically similar native and introduced species may not keep pace with rapid forest ecosystem change, and the risk of competitive exclusion may be greater for vulnerable native populations.

ACK N OWLED G M ENTS
We thank multiple field technicians and students associated with

CO N FLI C T O F I NTE R E S T
None declared.

AUTH O R CO NTR I B UTI O N
Both authors contributed to study design. J.D. collected and analyzed the data, and wrote the manuscript; J.K. procured funds and permits, provided edits, assisted with revisions, and gave approval for publication.

DATA ACCE SS I B I LIT Y
The raw data relevant to all analyses in this manuscript were archived with Dryad under: https://doi.org/10.5061/dryad.966ff04.