Habitat parameters influencing the distribution of a geographically isolated flying squirrel

Abstract

The San Bernardino flying squirrel (SBFS) is an isolated subspecies of Humboldt's flying squirrel, occurring in montane sky islands in the San Bernardino and San Jacinto Mountains in Southern California, USA. Recent small mammal surveys in the San Jacinto Mountains suggest the squirrel is extirpated. Our objectives were (1) determine habitat features, including forest metrics and topographical factors, that influence SBFS presence, in the San Bernardino Mountains; (2) use information collected in the San Bernardino Mountains to confirm squirrel occurrence and habitat preference in the San Jacinto Mountains; and (3) assess habitat and climatic differences between the two mountain ranges that could explain species persistence in one mountain range but not the other. We surveyed for SBFS using camera traps at 54 sites in the San Bernardino Mountains and 34 sites in the San Jacinto Mountains using both camera traps and acoustics. In the San Bernardino Mountains, we detected squirrels in sites that were more mesic, had higher structural heterogeneity, and had greater amounts of downed woody material compared to non-detection sites. Habitat parameters were similar between the two ranges; however, squirrels were not detected in the San Jacinto Mountains. Conditions in the San Jacinto Mountains were hotter and drier. Increased temperatures due to climate change could potentially explain the absence of flying squirrels in the San Jacinto Mountains.

INTRODUCTION

Worldwide, species are threatened by anthropogenic activities, leading to range shifts, population declines, and extinctions (Ceballos & Ehrlich, 2002; Dirzo et al., 2014; Wiens, 2016). This is especially true for endemic or isolated populations as they are more susceptible to higher extinction rates due to inbreeding depression, smaller population sizes, and reduced dispersal capabilities (Flaherty et al., 2008; Frankham, 1998; Smith & Person, 2007). Mountainous species are extremely vulnerable to climate warming as upward elevational shifts in response to warming may be impossible or lead to further isolation and decreasing patch size (Ali et al., 2021; Parmesan & Yohe, 2003; Smith, 2020). Catastrophic events (e.g., wildfires, insect outbreaks) can exacerbate the vulnerability of isolated populations due to the loss or alteration of limited habitat or food resources (Koprowski et al., 2005; Lindenmayer et al., 2013). Because isolated populations are more susceptible to habitat degradation or loss (Weigl, 2007), they often require specific conservation actions to ensure species persistence (Bosso et al., 2016; Marrero-Gómez & Bañares-Baudet, 2022).

North American flying squirrels (Glaucomys spp.) are arboreal, nocturnal Sciurids considered to be keystone species due to their role as dispersers of hypogeous fungi and as prey (Maser & Maser, 1988; Smith, 2007, 2012). Therefore, managing forests for flying squirrel presence is important to long-term forest health (Carey, 2000). To understand the vulnerability of a species to potential forest disturbance and climate change, understanding differences in climatic and ecological conditions between populations may highlight factors influencing persistence.

In the western United States, Humboldt's flying squirrels (Glaucomys oregonensis) are associated with range of coniferous forests (Arbogast et al., 2017; Smith, 2007). Habitat preferences of Humboldt's flying squirrels are well studied in the Sierra Nevada and Cascade Mountains (e.g., Carey et al., 1999; Pyare & Longland, 2002; Smith et al., 2003). Flying squirrels in these mountain ranges select large, contiguous stands of old growth coniferous forests with large diameter trees and snags that provide denning habitat as well as structural complexity that provides cover and food resources (e.g., truffles; Carey et al., 1999; Loeb et al., 2000; Meyer et al., 2007). Riparian corridors are known to be important to flying squirrels in semi-arid and arid forests, as increased microclimate moisture can influence food availability (Meyer et al., 2005, 2007). Although information influencing the persistence of Humboldt's flying squirrel is generally well studied, less is known about the habitat preferences of its southernmost subspecies.

The San Bernardino flying squirrel (Glaucomys oregonensis californicus; hereafter referred to as SBFS) is the most southern population of Humboldt's flying squirrel. SBFS is disjunct from the contiguous populations of Humboldt's flying squirrel populations in the Sierra Nevada by 245 km of the Mojave Desert (Arbogast et al., 2017; Wells-Gosling & Heaney, 1984; Yuan et al., 2022). SBFS are restricted to the San Bernardino Mountains and San Jacinto Mountains of southern California, two ranges separated by 12 km of alluvial fan and chapparal scrub. The SBFS is not well studied, other than surveys conducted in the early 20th century (Grinnell & Swarth, 1913; Sumner, 1927) and fire management, genetic and occurrence studies during the 21st century (Clark et al., 2021; Mazzela & Koprowski, 2020; Yuan, 2020). Whether SBFS selects the same habitat features as flying squirrels farther north in the range is unknown, as the two southern California ranges are lower in elevation and latitude, formed by different geological processes (Harden, 2004) and have different vegetation classifications (Sawyer et al., 2009).

Of conservation concern, the SBFS is assumed to be extirpated from the San Jacinto Mountains (Federal Register, 2016). Surveys conducted by Clark et al. (2021) did not detect SBFS, suggesting this subspecies may be extirpated from the San Jacinto Mountains. Given that the San Jacinto Mountains support the southernmost population of Humboldt's flying squirrel in the western United States, the extirpation may be a harbinger. Therefore, our objectives were (1) to determine forest habitat and topographical factors that might influence SBFS presence in the San Bernardino Mountains, and (2) to confirm if SBFS still occurs in the San Jacinto Mountains and if any factors that might differ between these two mountain ranges could explain species persistence in one range but not the other. Our study may provide insight into the potential impacts of climate warming in this biodiversity hotspot (Myers et al., 2000).

MATERIALS AND METHODS Study area

Our study area consisted of all locations on forested public lands in the San Bernardino Mountains (34°15′ N, 117°1′ W), San Bernardino County, and San Jacinto Mountains (33°46′ N, 116°43′ W), Riverside County, CA, USA (Figure S1). The San Bernardino Mountains are part of the Transverse Range, oriented east–west, rising-up from the San Bernardino Valley and descending into the Mojave Desert. These mountains have a broad valley leveling the top of the range. The San Jacinto Mountains, part of the Peninsular Range orientated north–south, gradually rise from Hemet Valley to the south and precipitously fall into the Palm Springs Valley, creating one of the largest gains in elevation in the contiguous United States, raising 1219 m in 1600 m. The descending northern slope is steep and dry, supporting shallow soils, and thus generally void of forest habitat (USGS, 2007). Both mountain ranges from cooler, more mesic sky islands surrounded by lower elevation coastal plains or desert. Precipitation in the San Bernardino Mountains and San Jacinto Mountains can exceed 100 and 81 cm, respectively, with most precipitation occurring during the winter months. Summer temperatures in the San Bernardino Mountains and San Jacinto Mountains range between 22 to 7°C and 23 to 7°C, respectively, with winter temperatures range between 5 to −8°C and 6 to −7°C, respectively. We surveyed seven forest types: Eastside Pine, Jeffery Pine, Montane Hardwood Conifer, Montane Hardwood, Ponderosa Pine, Subalpine Pine, and Sierran Mixed Conifer (CDFW, 2014; USFWS, 2016). We included lands in our sample frame that were managed by the San Bernardino National Forest, the state of California and local jurisdictions. Areas surveyed in the San Bernardino Mountains ranged in elevation from 1250 to 2870 m and from 1570 to 2870 m in the San Jacinto Mountains.

Survey methods San Bernardino Mountains

Within the study area, we used ArcMap® (Environmental Systems Research Institute, Inc., Redlands, CA, USA) to generate 100,000 random points spaced at a distance of 10 m. To avoid confounding factors created by wildfires, we excluded areas affected by fire after 1994, because prior to 1994 fire maps were not accurately drawn and were not available as digitized data. To minimize effects of roads, we did not survey within 100 m of any road as reported average daily movements of flying squirrels were ≤100 m (Smith, 2007). To improve our ability to detect habitat relationships specific to forest types, we also did not survey within 100 m of habitat edges (e.g., where two different forest types were mapped adjacent to each other). We randomly selected a maximum of 12 survey points stratified by forest types, depending on availability within the study area (Table S1). The difference in sample size among forest types was a result of the amount of area represented by each forest type not confounded by roads, habitat edge, or fire. We placed survey points >240 m apart, creating an 18 ha buffer, and assumed they were independent as flying squirrel home ranges are typically 3–15 ha (Weigl, 2007).

After locating points, we established survey sites and surveyed for SBFS presence by placing a MCG-12635 motion sensor camera (Moultrie Feeders; LLC., Calera, AL, USA) approximately 1.5–2 m up the bole of a tree at each survey point. We mounted two wire bait cages (15 cm × 15 cm × 2 cm) on a tree bole ~2 m opposite of the camera. One cage had the front covered with aluminum flashing to reduce the availability of bait to birds. Each cage was filled with equal parts rolled oat, black oil sunflower seeds, and peanut butter. We deployed cameras at each site for 13 consecutive nights. Camera trap surveys were conducted from March through June in 2016–2019 and only one survey was completed per site. Cameras were set to trigger on motion, with a 30 s delay between bursts of three pictures. No animals were handled during our study.

We recorded forest metrics and environmental parameters within a 30 m radius plot from each surveyed point. Within the plot, we recorded the diameter at breast height (DBH) and species for all trees with a living crown with a DBH ≥ 50 cm using a steel diameter tape (Forestry Suppliers, Inc., Jackson, MS, USA). Tree height was measured using TruPulse® 200B Professional Laser Rangefinder (0.3 m accuracy; Laser Technology, Inc., Centennial, CO, USA). Similarly, we measured all freestanding snags ≥25 cm DBH. We recorded the amount of downed, dead woody material (DWM) within the plot. The length, mid-point diameter, and decomposition stage (Maser et al., 1979) of DWM was measured for each log with a diameter ≥15 cm at mid-point. We also calculated the maximum cross-sectional area contacting the ground.

We established a 60 m transect centered on the survey point to record additional metrics. Each transect was positioned along an azimuth randomly selected within each plot. We used a GRS Densiometer® (Geographic Resource Solutions, Arcata, CA, USA) to measure canopy cover every 5 m along each transect and a convex spherical crown densiometer to determine canopy closure every 15 m along each transect (Korhonen et al., 2006; Lemmon, 1956). Every 5 m along each transect, we also recorded the distance to the nearest tree, its DBH, height and species on both sides of the transect. The nearest tree was determined by running a perpendicular plane to each transect and identifying the first intercept of leaf or branch material. If a tree canopy crossed each transect, that tree was recorded only once. We classified trees along the transect into four categories: ≤3 cm DBH, >3 to ≤15 cm, >15 to ≤30 cm, and >30 cm. We defined a plot as being physically heterogenous (also called structural, vertical complexity) when its habitat had trees in all four categories. Plots defined as being not heterogeneous had at least one missing category. Finally, at these 5 m increments along each transect, we recorded the depth of litter and depth of organic material (duff) below the litter (Brown, 1974).

To determine topographic parameters, we used ArcMap to calculate slope, aspect, and elevation using 10 m digital elevation maps. Because soil moisture is linked to truffle production, we determined soil type to estimate the soil's ability to hold moisture (H₂O Soil Storage; SSURGO, 2019). Riparian areas are known to be important to flying squirrels, therefore we also calculated the distance to the nearest Type 1 stream (Strahler, 1957) for the area encompassed by each 30 m radius circle.

San Jacinto Mountains

Following surveys in the San Bernardino Mountains, we conducted SBFS and habitat surveys in the San Jacinto Mountains. Methods and techniques followed were the same, except point selection was not randomized and length of camera operation was extended to 25 days. We used habitat parameters reported in the literature that northern flying squirrels (Glaucomys sabrinus) and Humboldt's flying squirrel select (Carey et al., 1999; Lehmkuhl et al., 2006; Pyare & Longland, 2002; Smith, 2007) and combined it with our preliminary findings in the San Bernardino Mountains to place 24 cameras and bait stations in locations believed to yield the highest probability of detection. Three cameras were placed at locations trapped by Grinnell and Swarth (1913). We completed camera trap surveys from April through October in 2019 and 2021. Ten of the 24 sites were randomly selected for forest metric measurements. Additionally, 10 passive acoustic Pettersson D500x® detectors (Pettersson Eleckrontik AB, Uppsala, Sweden) were deployed following a protocol by Diggins et al. (2016), yielding a total of 34 sites during August 2018. We did not record forest metrics at acoustic survey sites.

Climatic variations between mountain ranges

We downloaded monthly climate summary data (available from 1960 to 2009) from the NOAA (2020) database for the two mountain ranges. Climate data for both mountain ranges were not available prior to 1960. Glaucomys spp. live on average ≤4 years in the wild (Jackson, 1961). We set our end date to 2009, because a time span of three generations (2009–2018) should be sufficient to ascertain extirpation if the effects of climate change were to contribute to local extirpation in the San Jacinto Mountains (Staerk et al., 2019; Wells-Gosling & Heaney, 1984).

Data analysis

We analyzed camera trap and acoustic data by visual inspection. We sorted through camera trap data and identified all photos with positive SBFS detections using reference photos and field guides. For acoustic data, we used a SBFS captive call library and reference calls from Humboldt's and northern flying squirrels (Farwell et al., 2024; Gilley et al., 2019) to identify SBFS call files manually in SonoBat 2.9.8 (DND Design, Arcata, CA, USA). We estimated nightly detection using only camera data, wherein one positive detection resulted in species presence at that site for the corresponding night. To estimate detection probability (p), we used the methodology of MacKenzie et al. (2002), which incorporates all 13 nights of survey effort at each point. Because probability of detection was estimated to be 0.9997 after 10 nights (see Section 3), we used our raw detection (presence)/non-detection data for further analyses.

We used ordinations based on non-metric multidimensional scaling (NMDS in the ‘vegan’ package in R (Oksanen et al., 2012)) to visualize habitat differences between (i) sites with and without SBFS detections in the San Bernardino Mountains, and (ii) sites with detections in the San Bernardino Mountains and all sites in the San Jacinto Mountains. NMDS ordination is a commonly employed dimension reduction technique to represent differences between sets of multivariate data (Anderson et al., 2011). Ordinations used here were based on Jaccard distances, which are suitable for binary (detection/non-detection) data (Anderson et al., 2011). We used permutational analysis of variance (PERMANOVA in the ‘vegan’ package in R (Oksanen et al., 2012)) to determine differences in the comparisons (i and ii) listed above. Lastly, we used an indicator species analysis to identify the specific variables that contribute most to the differences between sites with detections in the San Bernardino Mountains and sites in the San Jacinto Mountains.

We tested variables separated by NMDS using exploratory statistics to see if these variables may account for detection/non-detection results, with an interest to help understand if factors promoting fungal growth could explain sites with and without SBFS detections in the San Bernardino Mountains with the intention to understand relationships that could lead to solutions important to management. These variables were DWM and ground cover (i.e., seedlings, ≤3 cm DBH). We also tested factors indicating arid conditions, such as presence of xeric vegetation and aspect. Finally, we tested habitat heterogeneity. We analyzed continuous variables using a two-sample t-test, whereas we evaluated categorical variables using chi-square. We set p ≤ .01, .05, and .10 as the observed probability level for significance, letting the reader assess their own standard of significance.

We used R (R Core Team, 2016) to conduct ordination and logistic regression for variables that tested highly significant (p ≤ .01) for presence. McFadden's pseudo R² was used to estimate model fit for logistic regressions (Menard, 2000).

To determine differences in habitat between the San Bernardino and the San Jacinto Mountains, we used PERMANOVA and NMDS. To evaluate potential change in regional climate we calculated average temperature for each month for the decade 1960–1969 and compared it to monthly averages for the decade 2000–2009. California statewide temperatures by decade relative to long-term average did not show an increase until 1981–1990 (CalEPA, 2018). Comparing the decades 1960–1969 to 2000–2009 provided a comparison before and after the measurable changes in climate statewide.

RESULTS

We sampled a total of 54 survey sites in the San Bernardino Mountains. SBFS were detected at 19 sites. Detection probability was 0.56 per night. During a survey period with ≥10 survey nights, the probability of not detecting SBFS on a plot where SBFS was present was <0.03%. In the San Jacinto Mountains, we conducted 600 nights of camera trap surveys at 24 sites and 146 nights of acoustic surveys at 10 sites but did not detect SBFS at any of the sites during our surveys.

Eastside Pine, Montane Hardwood Conifer, and Sierran Mixed Conifer were the dominate forest types; with 256, 106, and 111 points available for survey selection respectively. Of those points, 10 Montane Hardwood Conifer were randomly selected, of which three had SBFS detections and 12 Sierran Mixed Conifer were randomly selected, of which seven had SBFS detections. All forest types, with selections and detections, are listed in Table S1.

In the San Bernardino Mountains, all habitat types except Eastside Pine, had SBFS present (Table S1). Every Eastside Pine plot recorded juniper (Juniper spp.), curl leaf mahogany (Cercocarpus ledifolius), or pinyon pine (Pinus spp.); all species adapted to dry microclimates (CNPS, 1995). Moreover, our data also showed little structural heterogeneity across this forest type, with >60% of our Eastside Pine plots lacking heterogeneity. Due to the combination of these two factors, we excluded Eastside Pine from further analysis.

To differentiate between detection and non-detection sites for the SBFS, we compared the habitat characteristics of the remaining 44 plots (i.e., excluding eastside pine) in the San Bernardino Mountains. The results of the NMDS ordination did not reveal significant differences between sites where SBFSs were detected or not detected. Comparisons of sites with and without SBFSs in the San Bernardino Mountains revealed considerable overlap in terms of grouped habitat characteristics (Figure 1; PERMANOVA F_1,42 = 1.537, p = .193). The PERMANOVA was not significant, and the ordination yielded overlapping polygons (Figure 1).

FIGURE 1. PERMANOVA ordination depicting habitat characteristics between plots with (1–18) and without (19–44) San Bernardino flying squirrel (SBFS; Glaucomys oregonensis californicus) detections in the San Bernardino Mountains. Comparisons of sites with and without SBFSs in the San Bernardino Mountains revealed considerable overlap in terms of grouped habitat characteristics. The larger trapezoid encompasses plots where no squirrels were detected; the smaller polygon encompasses plots where flying squirrels were detected. Stress = 0.103.

The t-tests evaluating single parameters showed area of DWM tested significant at p ≤ .01 (t₄₂ = 4.031, $\alpha$= 2.704, N = 44, d.f. = 42). Figure 2 illustrates the relationship between area of DWM and presence of SBFS on a plot. As the area of DWM exceeds 60 m² (~200 m²/ha), the probability of a plot being utilized by SBFS increases. To support this statement, we developed a logistic regression model which showed an increasing area of DWM significantly predicts flying squirrel non-detection/detection (logistic regression: z = 2.990, p = .0028, N = 44). To estimate model fit we used McFaddens's pseudo R² (R² = 0.28), indicating the model predicted the data well (Menard, 2000). Additionally, the percentage of a plot covered by seedlings (≤3.0 cm DBH) tested significant at p ≤ .05 (t₄₂ = 2.12, $\alpha$= 2.021, N = 44, d.f. = 42).

FIGURE 2. Relationship between the area of down woody material (DWM) touching the ground and the detection of San Bernardino flying squirrels (Glaucomys oregonensis californicus) on a plot (total of 63 plots). The flying squirrels selected plots with higher areas of DWM touching the ground (p ≤ .01). Squirrel detections favored habitat with ≥60 m2 DWM on a 30 m2 radius plot.

We used a chi-square to test two categorical variables that affect how arid a site was. Pinyon/Juniper Present (X² = 10.154, p ≤ .01, $\alpha$= 6.35, N = 42, d.f. = 1) and Northern Aspect (X² = 4.116, p ≤ .05, $\alpha$ = 3.841, N = 42, d.f. = 1) tested significant. We used the same test to evaluate Habitat Heterogeneity, which also tested significant (X² = 2.946, p ≤ .10, $\alpha$= 2.71, N = 42, d.f. = 1).

Differences between San Bernardino and San Jacinto Mountains

Forest habitats on the two mountain ranges differed significantly from one another (Figure 3; PERMANOVA: F_1,25 = 8.99, p < .001). We used an indicator species analysis to identify the specific variables that contribute most to the differences between the two sites. This analysis identified six significant parameters: Nearest Stream and Percentage of Small Trees (30.1–50.0 cm DBH) were greater in the San Bernardino Mountains, whereas Number Conifers >50 cm along the transect, Total Trees >50 cm (density within the 30 m radius plot), Total Number Snags >20 cm, and % Total Closure were greater in the San Jacinto Mountains. The six habitat parameters that differentiated the two mountain ranges were not the parameters identified by NMDS when analyzing SBFS detections in the San Bernardino Mountains. We then selected two parameters that tested significant between detection and non-detection plots in the San Bernardino Mountains (Area of DWM and Habitat Heterogeneity), to test if these parameters were different between the two ranges and might explain why no detections were made in the San Jacinto Mountains. The Area of DWM and Habitat Heterogeneity were not significantly different (p > .10) between the two ranges. These parameters do not appear to differentiate habitats between plots in San Bernardino and San Jacinto mountains. Individual pinyon or juniper trees, which did test significant in the San Bernardino Mountains (p ≤ .01) when found scattered within a forest type, were not observed in the San Jacinto Mountains while searching for plots. Thus, the absence of these xeric adapted trees is like habitat where SBFS were detected in the San Bernardino Mountains. Aspect, which was significant (p ≤ .05) between detection/non-detection in the San Bernardino Mountains, was different between the two ranges. Of the 18 plots where SBFS were observed in the San Bernardino Mountains, 15 (83%) of the plots faced north. In the San Jacinto Mountains, of 29 total plots, only 8 (28%) plots faced north.

FIGURE 3. PERMANOVA ordination depicting habitat characteristics between plots in the San Jacinto Mountains (left polygon), where no San Bernardino flying squirrels (SBFS; Glaucomys oregonensis californicus) were detected, and plots in the San Bernardino Mountains (right polygon), where SBFS were detected. Stress = 0.093. Forest habitats on the two mountain ranges differed significantly from one another in terms of grouped habitat characteristics.

Average monthly temperatures for both mountain ranges show an increase between the 1960 and 2000 decades (Table S2). Overall, the increase in the San Jacinto Mountains is slightly less than the increase recorded in the San Bernardino Mountains. The yearly average temperature increase in the San Bernardino Mountains is 1.64°C (±SD = 0.67) and for summer months (June, July, August, and September) it is 2.08°C (±SD = 0.51). In the San Jacinto Mountains, the increase is 1.47°C (±SD = 0.79) and 1.85°C (±SD = 0.39), respectively.

DISCUSSION

During our surveys, we had high detection probabilities for SBFS in the San Bernardino Mountains. SBFS were detected in habitats representing all forest types above 1500 m in this mountain range, except for Eastside Pine habitats. SBFS were not limited to stands of old-growth or second growth conifers. Sierran Mixed Conifer habitat was occupied by SBFS at higher rates than either Montane Hardwood or Mixed Hardwood Conifers, highlighting this species preferences for more conifer dominant stands. However, the variety of forest types where SBFS were detected demonstrates habitat plasticity within this isolated mountain range. These results suggest that other habitat factors, besides forest type, determine the distribution of SBFS.

We found SBFS were absent in xeric forest conditions present in Eastside Pine habitat and at sites with juniper or pinyon. Eastside Pine occurs on well-drained basaltic soils in drier conditions compared to other conifer habitats and grows in even aged stands, representing little habitat heterogeneity (CDFW, 2014). Juniper and pinyon are drought tolerant and persist in xeric habitats characterized by poor soils, with arid to semi-arid conditions and low humidity (Evans, 1988). Xeric conditions combined with paucity of structural diversity probably created a lack of suitable habitat for SBFS, explaining their absence.

The most important forest parameter predicting presence of SBFS is the amount of DWM (Figure 2). There is a positive association between DWM, flying squirrel presence, and various truffle species (Amaranthus et al., 1994; Carey et al., 1999). Where DWM contacts the ground it creates moist, organic soils supporting hypogenous and epigeous fungi, which are important food sources for flying squirrels (Amaranthus et al., 1994; Maser et al., 1978; Weigl, 2007). DWM also provides denning habitat for flying squirrels (Carey et al., 1997; Diggins et al., 2015).

SBFS showed a preference for plots on north facing slopes in the San Bernardino Mountains. Factors influencing moisture gradients across the landscape are responsible for variation in truffle production (Claridge et al., 1999; O'Dell et al., 1999). Slopes with northern aspects in southern California, influenced by a Mediterranean climate, support denser vegetation thus are shaded, cooler and maintain higher humidity levels compared to southern aspect slopes (Schoenherr, 1992). This may provide more favorable conditions for a variety of truffle species. Therefore, managing sites with a northern aspect for high levels of DWM should increase the probability of SBFS presence (Claridge et al., 1999; O'Dell et al., 1999).

Another factor our study found that contributed to increased SBFS detection at a site was structural heterogeneity. In the Olympic Peninsula, flying squirrel abundance was correlated with midstory and understory development, which made them less susceptible to predation (Carey, 1995). Flying squirrels are at risk of predation when foraging on the ground (Smith, 2012; Wilson & Carey, 1996), thus cover close to the ground, formed by seedlings, is important to survival. Although large conifer trees and snags are important to flying squirrels (Carey, 2003; Patterson, 2012), we did not find that those factors determined SBFS detection. Presence of large trees and snags were common among all our plots across the San Bernardino Mountains, which may have been why these factors were not associated to detecting squirrels. However, managing for structural heterogeneity at sites where large trees exist with little understory may increase the area available for habitat. Examples would be areas where low intensity fires have removed understory or areas where fuel management treatments removed understory and limbed up trees. The later may be difficult, but not impossible, if biologist work with Wildland Urban Interface firefighting personnel to create a Community Wildfire Protection Plan (FEMA, 2020).

We measured both percent canopy closure and cover (Korhonen et al., 2006). Neither metric was associated with SBFS detections. However, open canopies, may negatively influence truffle production in certain species. Canopy closure is defined as the proportion of sky hemisphere obscured by vegetation when viewed from a single point (Jennings et al., 1999). There appears to be a relatively narrow range between closed and open canopy (canopy closure) where SBFS were present (Figure 4). SBFS appear absent in open canopies (<55% cover), possibly due to drier microclimate (Waters et al., 1994) and increased exposure to avian predators. Conversely, too much canopy closure (≥80%) may impede gliding, limit escape behaviors and permit predators to chase prey from tree canopy to tree canopy (Hackett & Pagels, 2003; Keefe & Giuliano, 2004). A management suggestion would be to fell some trees when canopy closure exceeds 80% and use the downed trees to increase DWM.

FIGURE 4. Relationship between the detection of San Bernardino flying squirrels (SBFS; Glaucomys oregonensis californicus) and percent of canopy closure at each plot (total of 63 plots). Sixty meter transects were randomly positioned along an azimuth in each plot, and closure was measured every 15 m. SBFS selected habitats with a narrow range of percent canopy closure, 58%–77%, as shown in the smaller box/whisker plot on the left.

During our study, we did not detect any SBFS in the San Jacinto Mountains even though these surveys were >10 days, which reduced the probability of a false absence, and we focused on habitat types similar to sites with detections in the San Bernardino Mountains. Clark et al. (2021) used citizen scientists to deploy cameras at backyard feeder stations to detect SBFS in each mountain range. SBFS were recorded at all residences in the San Bernardino Mountains but were never recorded in the San Jacinto Mountains. Considering the high detection probability in the San Bernardino Mountains, the fact that the last known sighting of SBFS was several decades ago (Federal Register, 2016), and several efforts have now unsuccessfully detected SBFS, we believe the SBFS is likely extirpated from the San Jacinto Mountains.

Factors leading to the presumed extirpation of SBFS from the San Jacinto Mountains are most likely not related to structural features of the habitat. Two significant parameters for detection, DWM and habitat heterogeneity, are similar between the San Bernardino and the San Jacinto Mountains. The San Jacinto Mountains appear to be an older growth forest, with a higher density of large trees and greater canopy closure compared to the San Bernardino Mountains. These factors favor Humboldt's flying squirrel occupancy in the northern portion of its range (Carey et al., 1999; Meyer et al., 2007; Waters & Zabel, 1995). Although certain factors varied between the mountain ranges, such as large trees and snags, tree density, canopy cover, and distance to stream, none of these factors influenced SBFS detections despite being important variables in other parts of their range (Carey et al., 1999; Maser & Maser, 1988; Meyer et al., 2007; Pyare & Longland, 2002).

Considering our study indicates loss of suitable habitat is not responsible for extirpation of SBFS from the San Jacinto Mountains, four circumstances that could explain this phenomenon are predation, disease (Schloegel et al., 2006), fire and climate change (Román-Palacios & Wiens, 2020). Predation is unlikely, as SBFS primary predator, the California spotted owl (Strix occidentalis occidentalis), has abandoned nesting sites and territories in the San Jacinto Mountains (USFS, 2020). The number of extinctions attributed to infectious disease is relatively small (McCallum, 2012) with an estimated ≤4% of extinctions worldwide attributed to disease (Smith et al., 2006). There is no historic evidence of disease in the San Jacinto Mountains negatively affecting Rodentia, although plague was investigated as a potential cause of declining chipmunk populations in the 1980s. However, no data or direct evidence could be assembled to substantiate this claim (AMNH, 1991; Lang & Wills, 1991) and flying squirrels are not known to carry plague. Therefore, we believe disease is an unlikely factor.

Both mountain ranges have been affected similarly by fire, with approximately 21% of the forest types we studied burned since 1994. The San Jacinto Mountains is the smaller of the two ranges where the cumulative effects of fire charred 4662 ha of forest land, leaving 18,390 ha untouched. Not all burnt areas were affected equally, both in size and intensity. Mazzela and Koprowski (2020) reported SBFS occupied lower burn severities, although they occurred in higher burn severity areas if sufficient canopy cover remained. Therefore, we believe the remaining intact habitat is of sufficient enough to sustain a population of SBFS.

Rising temperatures from climate change are a more probable explanation. Ambient air temperature has increased in both mountain ranges, following parallel paths (Figure 5), with the San Jacinto Mountains always maintaining a higher ambient temperature (Table S2). Increases in temperature are expected to alter the distribution of ectomycorrhizae (ECM) and potentially influence truffle production by increasing soil dehydration, fine root mortality and microbial respiration rates (Hyvönen et al., 2007). The temperature increases in the San Bernardino Mountains and the San Jacinto Mountains between the 1960 and 2000 decades averaged 1.64 and 1.47°C, respectively. Yet, the average monthly temperature for the San Bernardino Mountains in the 2000s never approached the average monthly temperature of the San Jacinto Mountains in the 1960s, with the exception of May where the difference was 0.28°C (Figure 5). In the early 1990s SBFS were observed in the San Jacinto Mountains (USFWS, 2016), thus we assume presence in the 1960s along with conditions being acceptable to support ECM.

FIGURE 5. Average monthly temperature in the San Bernardino Mountains for 2000–2009 compared to the average monthly temperature in the San Jacinto Mountains for 1960–1969. The San Jacinto Mountains maintain higher temperatures. Data were downloaded from the NOAA database.

Comparing the temperature increases between the two mountain ranges for the 2000 decade yield a different response. The San Jacinto Mountains average difference in monthly temperature was 3.31°C (±SD = 0.66) higher than the San Bernardino Mountains. Thus, the San Jacinto Mountains are warmer than the San Bernardino Mountains and averaged an increase of 1.85°C for summer months, or 1.47°C for the year. The higher yearly average temperature (3.31°C, Table 1), combined with an increase in average monthly temperature, may have a greater effect on the ecosystem. Our study surveyed for SBFS 20 years after the last sighting (USFWS, 2016) and 40 years after the effect of climate change was first recorded in California CalEPA (2018). This may be enough time for ECM communities to decrease to levels not sustainable for SBFS and enough generational cycles (Wells-Gosling & Heaney, 1984) for SBFS to become locally extinct.

TABLE 1 Monthly average temperatures in the San Bernardino Mountains and San Jacinto Mountains, comparing the decades 1960–1969 and 2000–2009.

Month	San Bernardino Mountains average temperature (°C)	San Jacinto Mountains average temperature (°C)	San Jacinto Mountains versus San Bernardino Mountains average temperature difference (°C)	San Bernardino Mountains average temperature (°C)	San Jacinto Mountains average temperature (°C)	San Jacinto Mountains versus San Bernardino Mountains average temperature difference (°C)
	1960–1969	1960–1969	1960–1969	2000–2009	2000–2009	2000–2009
January	0.11	4.22	4.11	1.61	5.72	4.11
February	1.22	4.06	2.83	1.67	5.39	3.72
March	1.94	5.61	3.67	3.89	7.78	3.89
April	4.83	9.00	4.17	6.00	8.72	2.72
May	9.11	11.22	2.11	11.67	13.89	2.22
June	13.00	15.67	2.67	15.61	17.89	2.28
July	16.50	20.06	3.56	18.89	21.39	2.50
August	16.61	19.67	3.06	18.11	21.50	3.39
September	12.94	16.28	3.33	14.78	18.28	3.50
October	9.00	12.67	3.67	9.94	13.72	3.78
November	3.44	7.61	4.17	4.94	8.72	3.78
December	0.61	5.11	4.50	1.89	5.78	3.89
Average difference for decade			3.49			3.31
Standard deviation			0.71			0.66

Note: Temperature data was downloaded from the NOAA database (2020). Table illustrates San Jacinto Mountains averaged 3.49°C higher monthly temperatures for years 1960–1969 and averaged 3.31°C higher monthly temperatures for years 2000–2009. The temperature difference between the two decades is similar.

While our study provided new information about SBFS, several questions are raised that deserve attention in the future. Our study did not survey the diversity, abundance, and importance of hypogeous fungi to SBFS. Butler et al. (1991) conducted microhistological analysis of fecal pellets collected from SBFS captured in the San Bernardino Mountains and reported spores from three genera of hypogeous fungi (Melanogaster, Hymenogaster, and Gymnomyces) were common, if not dominant, in their diets. The first step would be to research hypogenous fungi communities in the San Bernardino Mountains and compare those results to the San Jacinto Mountains.

Another presumed mammalian extirpation in the San Jacinto Mountains is the lodgepole chipmunk (Tamias speciosus speciosus). The lodgepole chipmunk and Humboldt's flying squirrel have similar ranges in the eastern Sierras south to the mountains we studied (Arbogast et al., 2017; Best et al., 1994). Both species consume truffles as a primary food source (Hall, 1991; Izzo et al., 2005; Meyer et al., 2005). Surveys conducted in 1991 were the first not to detect lodgepole chipmunks (AMNH, 1991). From 2008 to 2013, the San Diego Natural History Museum repeated Grinnell's 1908 surveys of the San Jacinto Mountains. Their surveys did not record lodgepole chipmunks at sites where they were previously detected during the early part of the 20th century, raising additional concerns that this species also is extirpated (NGS, 2012).

We found that SBFS in southern California are more plastic to habitat requirements than Humboldt's flying squirrels in the northern part of their range. We also found that SBFS are most likely extirpated from the San Jacinto Mountains. Habitat conditions between the two mountain ranges are similar. One plausible explanation for SBFS' local extinction from the San Jacinto Mountains is these mountains are warmer and drier, and climate warming has increased the average monthly summer temperature by 1.85°C which had a greater effect on the abundance and distribution of ECM fungi.

If our assertions are true, implementing management actions, including reintroduction, in the San Jacinto Mountains would be unsuccessful, shifting management concerns to the San Bernardino Mountains. Mammals occupying high elevation habitats in the western United States typically contract their ranges, as opposed to expanding their ranges upward (Moritz et al., 2008). Managers could identify preferred areas of habitat and set those areas aside as preserves. These areas would have topographic characteristics that favor cool, moist microclimates. Within those areas, management actions could be directed towards providing sufficient DWM, along with habitat heterogeneity and canopy cover. These forest conditions may not exist in some areas identified for preserves but could be managed for years in advance by timbering trees and planting understory. Such areas may provide refugia for populations to contract as conditions warm. Additionally, fuel management strategies could be developed to limit the impacts of wildfires, with particular emphasis to prevent the mega wildfires we are currently experiencing in the west, with a goal of returning forests to historic fire regimes. Managers could monitor local demographics and temperatures, to gain an estimate for the time needed for advance planning. Additionally, if climate warming isolates areas identified as preserves into pockets of habitat, making dispersal across the mountain range improbable, then potentially individuals could be translocated within the San Bernardino Mountains to maintain genetic diversity.

AUTHOR CONTRIBUTIONS

CSW conceived the study, obtained funding, developed the methods, collected data, conducted the research, analyzed data, and wrote the manuscript. DAH performed NMDS and PERMANOVA analyses along with writing related sections. CAD contributed to data collection, data analysis (for acoustics), and writing and editing the manuscript. PFD contributed to data analysis (detectability) and critically revised the manuscript. SCY collected data, conducted preliminary data analysis, and contributed to manuscript revisions. DBB collected data and conducted preliminary data analysis. SBT helped develop original questions and aided with data collection. SBT also conducted the original rodent surveys indicating concerns of extirpation.

ACKNOWLEDGMENTS

Robin Eliason, USFS, gave our team unwavering support and allowed us to move seamlessly through the forest of paperwork. Taylor Winchell, Denver Water, provided much needed insight to climate warming and directed us to the proper data sets. Reese Brand Philips, USFWS, provided advice blending study design with field logistics. Ed Turner, USFWS, provided excellent GIS support. Drew Farr, USFS and Jennifer Gee, UC Riverside, provided lodging facilities. Kim Boss, USFS, provided data on the California spotted owl. Jane Teranes, UC San Diego, provided high quality interns. Cherie Barns and Temor Iman aided in the collection of field data. Finally, our work would not have been possible without the support and financial foundation provided by Scott Sobiech, USFWS. The findings and conclusions in this article are those of the author(s) and do not necessarily represent the views of the U.S. Fish and Wildlife Service (USFWS). The authors complied with USFWS 212 FW 1 Standards of Conduct while conducting all aspects of this work.

FUNDING INFORMATION

Funding was provided by the Carlsbad Fish and Wildlife Office using Recovery Funds.

CONFLICT OF INTEREST STATEMENT

None of the authors have any potential sources of conflict of interest. There are no relationships or financial interests that might be perceived as influencing an authors' objectivity when conducting this work and writing the manuscript. We adhered to the highest level of ethics with our research. All data are available upon request.

DATA AVAILABILITY STATEMENT

Data are available upon request to the lead author.