OUP user menu

Habitat Associations of Small Mammals at Two Spatial Scales in the Northern Sierra Nevada

Stephanie A. Coppeto, Douglas A. Kelt, Dirk H. Van Vuren, James A. Wilson, Seth Bigelow
DOI: http://dx.doi.org/10.1644/05-MAMM-A-086R1.1 402-413 First published online: 21 April 2006


Effective management strategies require an understanding of the spatial scale at which fauna use their habitat. Toward this end, we sampled small mammals in the northern Sierra Nevada, California, over 2 years at 18 livetrapping grids among 5 forest types. Forest types were defined by overstory tree composition, and 19 microhabitat variables were measured at all trap stations. Forest type and year explained 93% of variation in abundance of North American deermice (Peromyscus maniculatus), whereas only 40% was explained by microhabitat and year. Similarly, variation in abundance of long-eared and shadow chipmunks (Neotamias quadrimaculatus and N. senex) was better explained by forest type and year (67%) than by microhabitat and year (30%). Red fir (Abies magnifica) forests supported more P. maniculatus and Neotamias species than mixed-conifer and pine–cedar forests, and more Neotamias species than mixed-fir forests. Five of 6 uncommon species were significantly associated with forest type; golden-mantled ground squirrels (Spermophilus lateralis), northern flying squirrels (Glaucomys sabrinus), and long-tailed and montane voles (Microtus longicaudus and M. montanus) were captured almost exclusively in red fir forests, whereas dusky-footed woodrats (Neotoma fuscipes) and California ground squirrels (Spermophilus beecheyi) were found in pine–cedar, mixed-fir, and mixed-conifer forests. The first 2 axes of a canonical correspondence analysis on microhabitat variables explained 71% of variation in combined small mammal abundance. Microhabitat associations varied among species but were driven primarily by canopy openness, shrub cover, and shrub richness. Although much of the small mammal fauna appeared to select habitat at both spatial scales studied, canonical correspondence analysis using forest type as a covariate revealed that microhabitat explained much less variation in small mammal abundance than did forest type.

Key words
  • California
  • canonical correspondence analysis
  • coniferous forests
  • macrohabitat
  • microhabitat
  • Sierra Nevada
  • small mammals
  • spatial scale

Small mammals play vital roles in forest ecosystems, serving as dispersers of fungal spores (Maser and Maser 1988; Pyare and Longland 2001) and seeds (Vander Wall 1993; Vander Wall et al. 2001); consumers of plants, seeds, and fruits (Carey et al. 1999a; Gunther et al. 1983); and as prey for mammalian and avian predators (Carey et al. 1992; Forsman et al. 1984; Zielinski et al. 1983). Of 474 mammals living in North America north of Mexico (Baker et al. 2003), more than 25% (n = 122) are small mammals (< 1 kg) that occur in western coniferous forests and associated habitats (Lawlor 2003). Nearly 75% (n = 90) of these species occur in montane woodlands (Lawlor 2003) where area of younger, managed forests now exceed the areas once covered by old-growth forests (Carey and Johnson 1995). Given the diversity of small mammals in these heavily impacted western forests (Barbour et al. 2002; Beesley 1996; Lawlor 2003) and their essential interactions with flora and fauna across multiple trophic levels (e.g., Carey et al. 1992; Forsman et al. 1984), management of these lands should be based in part on an understanding of the ecology of small mammals.

Although western coniferous forests and wildlife have evolved in the presence of natural disturbance, fire suppression and silvicultural practices have radically altered the landscape. Most western forests are now composed of younger trees at higher densities and are structurally less diverse than historic forest conditions (Carey and Harrington 2001; Carey and Johnson 1995; McKelvey et al. 1996). This forest structure is particularly evident in low- to mid-elevation coniferous forests of the Sierra Nevada (Miller and Urban 2000), where vegetative density likely increased because of rapid recruitment of shade-tolerant species (e.g., white fir [Abies concolor]) at the expense of less shade-tolerant species (e.g., ponderosa pine [Pinus ponderosa] —Ansley and Battles 1998; Barbour et al. 2002; Kilgore and Taylor 1979). As a result, live and dead fuels have accumulated in the understory of Sierra forests, homogenizing the landscape (McKelvey et al. 1996), and an increase in shade-tolerant trees has closed the once patchy canopy, decreasing understory plant diversity (Halpern and Spies 1995). Although current logging practices open up the canopy and allow for the development of understory vegetation, they frequently remove downed wood and snags naturally found in old-growth forests. Understory vegetation and downed wood may be key habitat elements for small mammals (Carey and Johnson 1995; Manning and Edge 2004; Morrison and Anthony 1989; Sullivan et al. 1999). Some mammal species occur in lower numbers in these younger, managed forests (Carey and Johnson 1995); for example, although northern flying squirrels (Glaucomys sabrinus) may tolerate logging practices, they often occur in greater abundance in old-growth forests than in younger, managed forests lacking old-growth habitat components (Carey 1995; Cote and Ferron 2001; Waters and Zabel 1995; but see Smith et al. 2003). In contrast, generalist species such as North American deermice (Peromyscus maniculatus), golden-mantled ground squirrels (Spermophilus lateralis), and many chipmunks (Neotamias species) may capitalize on disturbed conditions and increase in managed stands (Morrison and Anthony 1989; Sullivan and Klenner 2000; Sullivan et al. 1999; Waters and Zabel 1998).

Effective strategies for conservation of small mammals in heavily managed systems require an understanding of how organisms use their habitat and resources. Interpretation of such ecological patterns is complicated by spatial scale; studies asking the same question at different scales may yield very different patterns (see Wiens 1989). Despite intense research efforts, no single spatial scale is most suitable for explaining patterns of abundance and distribution of small mammals. However, it has become clear that single-scale studies fail to address the hierarchical way species use their habitat, and therefore risk misinterpreting potentially important associations (Bissonette et al. 1997; Johnson 1980; Martin and McComb 2002; Wiens et al. 1986). Nevertheless, segregation at one scale, the level of microhabitat, has been suggested as a means of coexistence in sympatric small mammals (Price and Kramer 1984), and responses to microhabitat variability have been confirmed in numerous studies (Bellows et al. 2001; Bowman et al. 2001; Butts and McComb 2000; Carey and Harrington 2001; Castleberry et al. 2002; Jorgensen 2004; Martin and McComb 2002). Hence, some authors contend that microhabitat characteristics influence the distributions of small mammals more than macrohabitat characteristics (Bellows et al. 2001; Castleberry et al. 2002). However, other studies have found that variation in abundance of small mammals has been better explained by simple macrohabitat characteristics (e.g., habitat type) than microhabitat variables (Jorgensen and Demarais 1999; Morris 1984, 1987), leading some authors to suggest that microhabitat associations are constrained within, or are less important than, characteristics of the larger spatial scale (Jorgensen 2004; Jorgensen and Demarais 1999; Kelt et al. 1994, 1999; Morris 1984, 1987).

The objective of this study was to describe and compare macro- and microhabitat associations of small mammal species in diverse forest types of the northern Sierra Nevada of California. Specifically, we asked 2 questions. First, what vegetative characteristics best explain abundance of small mammals at the macrohabitat (e.g., forest type) and microhabitat (e.g., trap station) scales? Second, does one spatial scale excel at explaining variation in small mammal abundance? This study was designed to enhance our understanding of small mammal habitat associations at distinct spatial scales and assist resource managers in building realistic strategies to preserve small mammals and overall biodiversity in western coniferous forests.

Materials and Methods

Study area.—This study was conducted in the Plumas National Forest, near Quincy, California (Fig. 1). Eighteen livetrapping grids were established in 3 general locations in the Mt. Hough Ranger District, with topography ranging from relatively flat to steeply sloping or mixed terrain, and elevation ranging from 1,180 to 2,250 m. Mean annual precipitation from 1997 through 2003 was 975.5 mm and mean temperatures ranged from 2.8°C (range of means: −9.3°C to 13.2°C) in January to 20.6°C (range of means: 3.3–38.5°C) in July (National Climatic Data Center-NOAA, Quincy weather station; http://www.ncdc.noaa.gov/oa/ncdc.html).

Fig. 1

Map of study area in Plumas National Forest (2003–2004) with a) locations of 18 small mammal livetrapping grids in 5 forest types and b) trap configuration within a livetrapping grid. Inset shows the location of the forest in California.

We established trapping grids in 5 principal forest types of the Plumas National Forest. Forest types were defined by the dominant live tree species representing ≥70% of total tree composition, and included white fir (Abies concolor, n = 4 grids), red fir (A. magnifica, n = 3), mixed-fir (codominant mix of white fir and Douglas-fir [Pseudotsuga menziesii], n = 5), mixed-conifer (n = 3), and pine-cedar (codominant mix of ponderosa pine [Pinus ponderosa] and Jeffrey pine [P. jeffreyi] and incense cedar [Calocedrus decurrens], n = 3; Fig. 1a). Overall, the Plumas National Forest is dominated by A. concolor and P. menziesii, so these forest types had proportionally more trapping grids. Common shrubs in the region include mountain rose (Rosa woodsii), Sierra gooseberry (Ribes roezlii), serviceberry (Amelanchier utahensis), bush chinquapin (Chrysolepis sempervirens), green-leaf and white-leaf manzanita (Arctostaphylos patula and A. viscida), mountain whitethorn and deerbrush (Ceanothus cordulatus and C. intigerrimus), bitter cherry (Prunus emarginata), and huckleberry oak (Quercus vaccinifolium). Pinemat manzanita (Arctostaphylos nevadensis) occurred almost exclusively in red fir forests, and buckbrush (Ceanothus cuneatus) predominantly in pine-cedar forests.

Small mammal sampling.—Seventeen livetrapping grids consisted of a 10 × 10 array of Sherman live traps (Model XLK, 7.6 × 9.5 × 30.5 cm, H. B. Sherman Traps, Inc., Tallahassee, Florida) with 10-m spacing, nested within a larger 6×6 grid of 72 Tomahawk traps (Model 201, 40.6 × 12.7 × 12.7 cm, Tomahawk Live Trap Co., Tomahawk, Wisconsin; 1 ground, 1 arboreal) with 30-m spacing (Fig. 1b). A final grid was constrained by road configuration, such that the array of Sherman traps was nested within a 4 × 9 grid of 72 Tomahawk traps (30-m trap spacing, 1 ground, 1 arboreal). Thus, all grids had 120 trap stations and covered 2.25 ha (3.24 ha with a one-half [15-m] intertrap distance buffer) of contiguous forest. Arboreal Tomahawk traps were removed from all grids on 1 August 2004 because of consistently poor capture rates. Minimum distance among trapping grids was 1 km with the exception of 2 red fir grids that were approximately 700 m apart. No small mammals were documented to move between trapping grids.

Small mammals were sampled monthly from July through October 2003 and May through August 2004. Heavy snow precluded sampling on 3 red fir grids in May 2004, and unanticipated problems precluded sampling on 12 grids in July 2004. Therefore, to equalize sampling efforts among trapping grids in 2004, July captures were excluded from analyses except for samples from 2 red fir forests. Thus, with the exception of 1 red fir trapping grid not sampled in May or July, all grids had 3 trapping sessions in 2004. To equalize sampling efforts between years, we excluded data for October 2003 in analyses. Thus, data were used from July, August, and September 2003, and from May, June, July (2 red fir sites only), and August 2004.

Monthly trapping sessions consisted of 4 consecutive trap-nights. Sherman and Tomahawk traps were set and baited every evening just before dusk and checked just after dawn. Sherman traps were closed until dusk, whereas Tomahawk traps were rebaited and checked again at midday, at which point they were closed until dusk. Traps were baited with rolled oats and black oil sunflower seeds coated in peanut butter. Coverboards and synthetic bedding material were provided as needed for protection from heat and cold.

All animals were weighed, sexed, examined for reproductive status, and identified to species before ear-tagging (National Band and Tag Co., Newport, Kentucky) and release at the point of capture. Although 2 chipmunk species (Neotamias senex and N. quadrimaculatus) occur in the study area, they overlap greatly in external characteristics (e.g., Ingles 1965; Jameson and Peeters 2004; Sutton 1995) and are treated here as Neotamias species. Similarly, 2 species of voles (Microtus longicaudus and M. montanus) occur at the sites but were not reliably distinguished; hence they are grouped as Microtus species. All handling procedures were approved by the Animal Use and Care Administrative Advisory Committee of the University of California, Davis, and met guidelines recommended by the American Society of Mammalogists (Animal Care and Use Committee 1998).

Macrohabitat associations.—Macrohabitat was defined by forest type. Overstory vegetation was quantified in July and August 2003 using point-centered quarter sampling (Mueller-Dombois and Ellenberg 1974) at 18 stratified Tomahawk trap stations per grid. Trees sampled had a diameter at breast height ≥ 10 cm.

Microhabitat associations.—Microhabitat characteristics were sampled during July-August 2003. All measurements were recorded within a 1-m-radius (3.14-m2) circular plot centered at every trap station. We visually estimated percentage cover of 12 ground-cover variables (Table 1). Tree canopy was quantified by taking a single, color digital photograph with a hemispherical lens mounted at 1.4 m aboveground at every trap station, and calculating percent canopy openness using Gap Light Analyzer, version 2 (Frazer et al. 2000). Percentage canopy openness represents the proportion of horizon-to-horizon view that is open sky. Aspect was measured with a compass by estimating the direction water would flow from the center of a trap station and was converted to north-south (e.g., −90° to +90°) and west-east (−90° to +90°) components. Slope was measured with a clinometer as the general decline of the substrate within each circular plot. Substrate (ground) hardness was measured using a soil penetrometer (Pocket penetrometer, Geotest Instrument Corp., Evanston, Illinois) as the mean of 4 measurements (1 per quadrant) within each circular plot. Thick duff layers at ≥50% of trapping grids (up to 15 cm deep) made digging for true soil measurements impractical and somewhat meaningless; therefore, substrate hardness was measured after removing only the surface duff layer composed of litter, downed wood, and rocks. Microhabitat vegetation (excluding canopy) was resampled in July 2004 at 25% (n = 30) of trap stations in 6 randomly chosen grids representing all forest types; because there was no change in these metrics (paired t-tests; all P > 0.05), measurements recorded in 2003 were used in comparisons with small mammal data from both 2003 and 2004.

View this table:
Table 1

Description of microhabitat variables measured in 1-m-radius (3.14-m2) circular plots at all trap stations in the Plumas National Forest, California (2003–2004).

Microhabitat variableDescription
Ground cover (%)
RocksExposed large rocks and stones
Bare groundExposed soil
Forbs and grassesHerbaceous and flowering vegetation and grasses
LitterDead leaves, pine needles, wood chips, sawdustlike debris
BranchesTwigs with diameter < 10 cm
Small logsLogs and stumps with diameter (within plot) = 10−50 cm
Large logsLogs and stumps with diameter (within plot) > 50 cm
Live shrubsWoody vegetation not considered sapling; height < 2 ma
Dead shrubsAs for live shrub but with no living foliage or no foliage
Vegetation matsNear ground surface shrub cover
(Ceanothus prostratus)
SaplingsSmall trees with height ≤ 2 m
Nonwoody perennialsbShrub- and forblike vegetation lacking woody stems
Canopy openness (%)Percentage open sky above breast height (1.4 m)
Shrub species richnessNumber of distinct, live shrub species
Sapling species richnessNumber of distinct, live sapling species
Substrate hardnessGround hardness averaged across 4 randomly sampled points
SlopeDegree of ground surface decline or incline
AspectProbable direction of water flow from center of trap station
  • a Mountain rose (Rosa woodsii) counted as shrub if height ≥ 50 cm, otherwise considered nonwoody perennial. Bitter cherry (Prunus emerginata) counted as shrub if height ≤ 2 m.

  • b Predominant species was snowberry (Symphoricarpos) but also included bracken fern (Pteridium aquilinum), thimbleberry (Rubus parviflorus), and Prince's pine (Chimaphila umbellatd).

Macrohabitat analyses.—We quantified small mammal assemblages on each trapping grid using 3 metrics. Total abundance (N) was calculated as the number of individuals captured in each session; repeat captures of individuals at given trap stations were not included in macro- and microhabitat analyses. Species richness (S) was simply the number of species documented to occur there. Diversity (H') was calculated using the Shannon-Wiener index (H' = −Σpi log pi), where pi is proportional representation of species i. Additionally, we obtained sufficient captures of P. maniculatus and Neotamias species for parametric analyses of macrohabitat associations. We examined the effects of forest type and year on the means of these metrics of community structure and the abundances of P. maniculatus and Neotamias species using repeated-measures multivariate analyses of variance (MANOVAs) and subsequent univariate analyses (ANOVAs). Abundance of P. maniculatus, Neotamias species, and all species (AO were square-root transformed to meet assumptions of normality. Variances were univariate homogeneous but were not examined at the multivariate level. The combination of univariate homogeneity and relatively balanced sample sizes (the smallest cell size was >50% the size of the largest cell—Scheiner 2001) suggested that statistical assumptions were met within reason; nevertheless, Pillai's trace was used to test the null hypothesis as it is robust to violations of assumptions (Scheiner 2001). Post hoc comparisons were conducted using Scheffé's test (Day and Quinn 1989) and considered significant at α = 0.05.

Six taxa (G. sabrinus, Microtus species, Neotoma fuscipes [dusky-footed woodrat], Spermophilus beecheyi [California ground squirrel], S. lateralis, and Tamiasciurus douglasii [Douglas's squirrel]) were captured at <50% of sampling grids, precluding the use of parametric tests. Because a Wilcoxon signed-rank test documented no significant differences in abundance of these species between sample years, a Kruskal-Wallis test was applied to species counts pooled from 2003 and 2004 to evaluate differences in abundance of each species among forest types. All macrohabitat analyses were conducted using SAS version 8 (SAS Institute Inc. 2000).

Microhabitat analyses.—We used canonical correspondence analysis (CCA; CANOCO version 4.5—ter Braak and Smilauer 2002) to describe associations between abundance of small mammals (n = 8 species) pooled across years and all microhabitat variables. CCA is a constrained ordination that directly and simultaneously relates species composition to environmental variables, unlike unconstrained ordinations (e.g., detrended correspondence analysis) that perform sequential analyses. CCA is an extension of multivariate multiple regression that is robust to moderate violations of normality assumptions (Lepš and Šmilauer 2003; Palmer 1993), performing well even with skewed species distributions (Palmer 1993). Small mammal counts were square-root transformed before ordination. Default options (e.g., biplot scaling focusing on interspecies distances) were used because they were appropriate for these analyses. Monte Carlo permutations (n = 500) were performed to test the significance of the contribution by each canonical axis to explanations of variation in small mammal abundance. We used forward selection with unrestricted Monte Carlo permutations (n = 500) to determine the relative importance of each measured microhabitat variable to species abundance. We qualitatively confirmed the overall ordination results by inspecting species-specific contour plots that illustrate the fitted values of species abundance (using Loess regression) and microhabitat variables in CCA space. Finally, we compared the power of the micro-and macrohabitat scales in explaining local variation of small mammal abundance by performing 2 additional CCA tests. In each test, the habitat characteristics of 1 scale was used as a covariate and the analysis then explained the remaining variation in abundance of small mammals due to characteristics of the other scale (Lepš and Šmilauer 2003).


From July through September 2003, a total of 37,152 trap-nights of effort (21,600 Sherman trap-nights and 15,552 Tomahawk trap-nights) yielded 1,201 captures of 464 small mammals. With only slightly greater effort (22,400 Sherman trap-nights and 15,840 Tomahawk trap-nights; 38,240 total trap-nights) between May and August 2004 we recorded 4,204 captures of 1,647 small mammals. This represented a 355% and 350% increase in total abundance and captures, respectively. Total abundance among trapping grids showed a linear increase between years (F = 96.34, d.f. = 1, P < 0.0001, adjusted r2 = 0.849). Neotamias species and P. maniculatus were the most frequently captured mammals in 2003 (49%) and 2004 (69%; Table 2); the latter species increased more than 8-fold between years. Incidental captures (<5 captures across both years) included Trowbridge's shrew (Sorex trowbridgii), snowshoe hare (Lepus americanus), long-tailed weasel (Mustela frenata), western spotted skunk (Spilogale gracilis), and striped skunk (Mephitis mephitis).

View this table:
Table 2

Change in abundance of small mammals in the Plumas National Forest, California, from 2003 to 2004, showing abundance and proportion of total captures (as percentage).

Species2003%2004%Change (%)
Glaucomys sabrinus512<1−60
Microtus species92231156
Neotamias species217473772374
Neotoma fuscipes163131−19
Peromyscus maniculatus119261,13069850
Spermophilus beecheyi306171−43
Spermophilus lateralis661475514
Tamiasciurus douglasii21101400

Forest structure.—Previous logging and fire history are not well documented for these trapping grids but the general vegetative structure was indicative of active fire suppression and silvicultural practices. For example, mixed-fir, white fir, and mixed-conifer forests were characterized by high tree density (440, 512, and 645 stems/ha, respectively), low proportion of open sky above trap points (mean canopy openness 12%, 11%, and 11%, respectively), deep duff layers (up to 15 cm), and heavy fuel and litter loads. However, 3 of the 5 mixed-fir sites, supported more heterogeneous understories characterized by high shrub cover and richness and highest cover of all forest types by nonwoody perennials. Pine-cedar and red fir forests had an open stand structure (178 and 166 stems/ha, respectively) with high proportion of open sky above trap points (mean canopy openness 40% and 47%, respectively) and high cover by rocks, exposed soils, and live shrubs. However, shrubs in pine-cedar forests were spatially clumped, whereas those in red fir formed a continuous ground cover. One trapping grid in pine-cedar forest likely experienced a fire in 1970 yet was structurally indistinguishable from other pine-cedar trapping grids.

Macrohabitat associations.—Small mammal abundance (N), species richness (S), and diversity (H') varied among forest types (MANOVA, Pillai's trace = 1.06, F = 3.58, d.f. = 12, 78, P < 0.001) and between sample years (MANOVA, Pillai's trace = 0.79, F = 29.68, d.f. = 3, 24, P < 0.001). Overall, the ANOVA models with forest type and sample year explained 84% of the variation in mean abundance (F = 14.86, d.f. = 9, P < 0.001), 49% of variation in mean species richness (F = 2.72, d.f. = 9, P = 0.022), and 31% of variation in mean species diversity (F = 1.32, d.f. = 9, P = 0.277; Table 3). Mean abundance and richness also differed among forest types, and abundance increased between years (Table 3). A posteriori multiple comparisons showed that red fir forests had higher abundances of small mammals than any other forest type (Scheffé's test, P < 0.05; Fig. 2a) and greater mean species richness than all types except mixed-fir (Scheffé's test, P < 0.05; Fig. 2b).

View this table:
Table 3

Analyses of variance of effects of forest type and sample year on mean small mammal abundance (N), richness (S), diversity (H'), abundance of Peromyscus maniculatus, and abundance of Neotamias species in the Plumas National Forest, California (2003–2004).

Total abundance
Forest type439.4615.08<0.001
Forest type × year43.091.180.343
Species richness
Forest type46.046.000.002
Forest type × year40.11,0.110.979
Species diversity
Forest type40.0040.300.875
Forest type × year40.021.380.267
Peromyscus maniculatus
Forest type46.337.22<0.001
Forest type × year43.854.390.008
Neotamias species
Forest type439.3911.97<0.001
Forest type × year41.880.570.686
  • a MS = mean square.

Fig. 2

Mean a) total abundance (N), b) species richness (S), c) abundance of Peromyscus maniculatus, and d) abundance of Neotamias species among 5 forest types in the Plumas National Forest, California, in 2003 and 2004. Means are calculated from 3–5 trapping grids per forest type with vertical lines indicating the standard error. Columns with the same letter are not different (Scheffé's test, P > 0.05).

Abundance of P. maniculatus and Neotamias species differed among forest types (MANOVA, Pillai's trace = 0.73, F = 3.78, d.f. = 8, 52, P = 0.0015) and between sample years (MANOVA, Pillai's trace = 0.91, F = 133.57, d.f. = 2, 25, P < 0.0001). Overall, the ANOVA models with forest type and sample year explained 93% of variation in abundance of P. maniculatus (F = 36.52, d.f. = 9, P < 0.0001) and 67% of variation in Neotamias species (F = 5.78, d.f. = 9, P = 0.0002) and indicated that abundance of P. maniculatus was influenced by forest type, year, and forest type × year interaction, and abundance of Neotamias species was influenced significantly by forest type (Table 3). P. maniculatus was more abundant in red fir, white fir, and mixed-fir sites than in mixed-conifer and pine-cedar forests (Scheffé's test, P < 0.05; Fig. 2c). Macrohabitat affinities of Neotamias species were similar to those of P. maniculatus but more narrowly focused with preferences for red fir and white fir sites over all other types (Scheffé's test, P < 0.05; Fig. 2d).

Abundances of G. sabrinus, Microtus species, N. fuscipes, S. beecheyi, and S. lateralis were influenced by forest type (Kruskal-Wallis tests; Fig. 3). For example, G. sabrinus2 = 13.62, d.f. = 4, P = 0.009), Microtus species (χ2 = 21.43, d.f. = 4,P = 0.0003), and S. lateralis2 = 28.04, d.f. = 4, P < 0.0001) were found almost exclusively in red fir forests. In contrast, N. fuscipes2 = 11.61, d.f. = 4, P = 0.021) and S. beecheyi2 = 16.62, d.f. = 4, P = 0.002) were associated with pine-cedar, mixed-fir, and mixed-conifer forests. T. douglasii was not common and its abundance did not differ among forest types (χ2 = 5.07, d.f. = 4, P = 0.28).

Fig. 3

Mean abundance of small mammal species in 5 forest types in the Plumas National Forest, California (for 2003 and 2004). Means are calculated from 3–5 trapping grids per forest type with vertical lines indicating the standard error. Peromyscus maniculatus and Neotamias species are shown separately because their numbers were substantially greater than those of the other species.

Microhabitat associations.—Canonical correspondence analysis on microhabitat associations was based on 4,503 individuals captured at 1,424 trap stations (Fig. 4). The first 2 canonical axes cumulatively explained a large proportion (71%) of variation in local abundance of small mammals. The 1st canonical axis alone explained more variation (53%) than axes 2, 3, and 4 combined (37%) and was positively correlated with canopy openness, cover by live shrubs, and shrub species richness (Table 4). Although most variation in local abundance of small mammals was explained by the 1st canonical axis (Monte Carlo permutation test, F = 66.09, P = 0.002), the 2nd and 3rd axes also contributed to explanations of community variation (Monte Carlo permutation test, F = 22.22, P = 0.002, and F = 18.37, P = 0.002, respectively) but were less highly correlated with microhabitat characteristics (Table 4). Forward selection included the following variables (at P ≤ 0.05) in the final model explaining overall small mammal abundance: canopy openness, cover by rocks, bare ground, branches, large logs, live shrubs, shrub richness, substrate hardness, slope, and south-facing aspects.

View this table:
Table 4

Weighted correlations between microhabitat variables measured in the Plumas National Forest, California (2003–2004), and first 3 species axes of canonical correspondence analysis.

Weighted correlations
Microhabitat variableSpecies axis 1Species axis 2Species axis 3
Ground cover (%)
Bare ground0.1486−0.00340.0601
Forbs and grasses−0.08180.0201−0.0472
Small logs−0.03290.00410.0264
Large logs0.02180.07170.0530
Live shrubs0.29600.1001−0.0103
Dead shrubs−0.04500.0215−0.0296
Vegetation mats−0.0754−0.02070.0392
Nonwoody vegetation−0.10590.0610−0.0002
Canopy openness0.4637−0.04180.0543
Shrub species richness0.20170.2057−0.0171
Sapling species richness−0.0772−0.00060.0641
Substrate hardness−0.0810−0.09380.1225
South aspect0.08980.1233−0.0085
East aspect−0.18780.07130.0704
Fig. 4

Biplot of canonical correspondence analysis of small mammal trap-scale abundances and microhabitat variables in the Plumas National Forest, California (2003–2004). Vector length indicates the strength of correlation between given variables and the canonical axes. The symbol a is dead shrubs, b is forbs and grasses, c is saplings, d is sapling richness, and e is small logs. Species acronyms are: Glsa = Glaucomys sabrinus, Misp = Microtus species, Nesp = Neotamias species, Nefu = Neotoma fuscipes, Pema = Peromyscus maniculatus, Spbe = Spermophilus beecheyi, Spla = Spermophilus lateralis, Tado = Tamiasciurus douglasii.

Canonical correspondence analysis described diverse micro-habitat affinities for many species (Fig. 4). For example, G. sabrinus, S. lateralis, and Microtus species exhibited strong microhabitat preferences for open canopy, high cover by shrubs, bare ground, and rocks; these characteristics dominate the understories of red fir forests where these species reached their highest abundance. Neotamias species were captured across many microhabitats but affinities were best described by high shrub cover and richness, open canopies, bare ground, rocks, large logs, and south-facing aspects, characteristics associated predominantly with red fir forests but also representative of mixed-fir forest understories. Local captures of N. fuscipes, S. beecheyi, and T. douglasii were not restricted to narrow microhabitats; rather, these species exhibited broader affinities for similar microhabitat features. The location of P. maniculatus in the center of CCA space likely is artifactual because this species was found at 99% of trap stations used in analyses and more than half of all analyzed trap stations were located in closed canopy forests (i.e., left side of Fig. 4). Inspection of a species-specific contour plot reveals that P. maniculatus reached its highest abundance in traps characterized by open canopy and cover by live shrubs, rocks, and bare ground (Fig. 5).

Fig. 5

Contour plot of canonical correspondence analysis of trap-scale abundance of Peromyscus maniculatus and microhabitat variables in the Plumas National Forest, California (2003–2004). Contour lines are fitted values of P. maniculatus abundance from a Loess regression model (i.e., number of animals captured per trap, across both years) and vectors are as in Fig. 4. The symbol a is dead shrubs, b is forbs and grasses, c is saplings, d is sapling richness, and e is small logs.

We conducted parallel CCAs using either forest type (macro-habitat) or station-level habitat associations (microhabitat) as covariates to further distinguish the relative importance of the 2 spatial scales studied here. When forest type was used as a covariate in CCA, microhabitat characteristics explained only 34.5% of the variation in local abundance of small mammals on the 1st canonical axis. In the reciprocal test (using micro-habitat variables as covariates), forest type explained 72.1% of the variation in small mammal abundance on the 1st axis.


Small mammal abundances on our sampling grids varied markedly between years. It is not entirely clear whether these abundances represented unusually low numbers for 2003 or high counts for 2004, or simply reflected a system characterized by temporal variation in population sizes. In spite of more than a century of detailed study of mammals in California, there are few published studies of small mammal demography in coniferous forests of the Sierra Nevada; however, examination of the limited data available suggests that abundances at other sites were similar to those found here in 2004. With comparable trapping effort, Waters and Zabel (1998) captured similar numbers of small mammals (N = 1700) in the Lassen National Forest, although their captures were not dominated by deermice as in this study. The winter preceding our 2003 season was characterized by 2 late and severe storms, and we speculate that these may have contributed to low abundance in that year.

Six of 8 small mammal species reached their highest mean abundance in red fir forests. Compared to mixed-conifer forests, Sierran red fir forests have not been as greatly affected by human activities such as logging, and consequently they have received less research attention until recently (Barbour et al. 1998; Fernau et al. 1998; Laacke and Tappeiner 1996; Taylor and Halpern 1991). In contrast to previous studies documenting moderately low cover and richness of herbs and shrubs in red fir forests (Barbour and Woodward 1985), the red fir forests sampled in this study generally had a dense shrub layer dominated by pinemat manzanita. With the exception of the high-elevation mixed-fir grid that also supported high small mammal abundance, red fir forests were the only type to have pinemat manzanita, which provides cover and food for small mammals (Laacke and Tappeiner 1996). Many mammals, particularly S. lateralis, were frequently observed eating the fruits of this species. In addition, ripening of manzanita fruits in August coincided with the emergence of young S. lateralis from burrows. The presence of a dense vegetation layer in the understory, especially of pinemat manzanita, evidently contributed to high species richness and abundance of small mammals at red fir forests.

White fir and mixed-fir forests also supported high small mammal abundances, intermediate between that of red fir forests and mixed-conifer and pine-cedar forests. White fir and mixed-fir forest types dominate the Plumas National Forest where our study was conducted; the disproportionate presence of these types has been attributed to the rapid recruitment of white fir and low mortality of Douglas-fir trees (Ansley and Battles 1998). For small mammal species particularly dependent upon cone production (e.g., Tamiasciurus), these forests likely offer plentiful food resources. Nonetheless, high abundance at white fir grids was puzzling given the lack of understory complexity—many grids were covered in litter and branches. Notably, 1 white fir grid had more than twice the cover by understory vegetation (e.g., shrubs, forbs, and grasses) as the other grids in this forest type and also had more small mammals. P. maniculatus and Neotamias species composed the vast majority of captures at this forest type; both species have been found in equal or greater abundance in disturbed systems as compared to old growth, appearing tolerant of heavily impacted systems (Sullivan and Klenner 2000; Sullivan et al. 1999, 2000; Von Trebra et al. 1998; Waters and Zabel 1998). In contrast to white fir stands, 3 of the 5 mixed-fir forests had more complex understories characterized by non-woody vegetation, diverse shrub communities, and downed wood of all size classes. The more structurally complex understories of these sites may have contributed to the increased species richness seen at sites dominated by white fir alone.

Compared to all other types, mixed-conifer and pine-cedar forests supported low abundances and few species. Only N. fuscipes reached its greatest abundance in pine-cedar and mixed-conifer forests. N. fuscipes in northern California and Oregon occurs predominantly in stands with dense shrub cover (Carey et al. 1999b; Linsdale and Tevis 1951; Sakai and Noon 1993) and is rare in either very open or closed (old-growth) stands (Linsdale and Tevis 1951; Sakai and Noon 1993). This species also forages on incense cedar (M. B. McEachern, pers. comm.) and a variety of oaks (QuercusAtsatt and Ingram 1983; Hemmes et al. 2002), both of which are common in pine-cedar and mixed-conifer stands in our study area. The lower abundances of N. fuscipes in these forest types likely reflected the marginal nature of the habitat; for example, mixed-conifer trapping grids lacked a dense shrub layer and most pine-cedar forests had a dense but clumped shrub layer within an otherwise very open area.

Microhabitat affinities, as with macrohabitat, varied among species but were driven primarily by canopy openness, shrub cover, and shrub richness. Previous studies of small mammal habitat associations in the Pacific Northwest have converged on models emphasizing cover by understory vegetation (predominantly shrubs) and secondarily by cover and volume of downed wood (Carey and Harrington 2001; Carey and Johnson 1995; Manning and Edge 2004; Morrison and Anthony 1989). These features generally also are important for small mammals in the northern Sierra Nevada.

It may be important to note that the vast majority of captures in this study were of generalist taxa such as P. maniculatus, N. fuscipes, and Neotamias species. We had relatively few or no captures of more specialized taxa that are known to occur in this region; for example, Zapus princeps, Neotoma cinerea, possibly Phenacomys intermedius, and both soricid and talpid insectivores (e.g., Sorex palustris, Neurotrichus gibbsii, and Scapanus latimanus). Clearly, our inferences for these species are limited by small sample sizes. Further efforts focused on species with more restricted habitat requirements would be highly useful to understanding and managing this system. An important but likely unanswerable question is whether or not these species are naturally rare, or if historical management of these forests, emphasizing fire suppression and timber removal, has altered the distribution of habitats needed by specialist species. If the latter is correct, then it may be that a century of human activities has selected for ecological generalists in this forest, a form of biotic homogenization (McKinney and Lockwood 1999; Olden and Poff 2003).

Given that small mammals in our study exhibited affinities at both spatial scales studied, is one scale better suited than the other? Variation in abundance of Neotamias species and P. maniculatus was explained moderately to very well with ANOVA on macrohabitat-scale information (70% and 93%, respectively). We conducted a parallel set of ANOVAs on microhabitat data that was less informative than the CCA and therefore is not presented here; it is informative, nonetheless, that only 30% and 40% of the variation in Neotamias species and P. maniculatus were explained by this analysis, suggesting that macrohabitat-scale characteristics may yield greater insights to patterns of small mammal distribution and abundance than microhabitat variables (Jorgensen and Demarais 1999; Morris 1984, 1987). The lower explanatory power of analyses on P. maniculatus using microhabitat data agrees with numerous earlier studies (Adler 1987; Bellows et al. 2001; Dueser and Shugart 1978; Hanley and Barnard 1999; Morrison and Anthony 1989; Yahner 1986) and supports interpretations that this species has very generalized habitat requirements, although recent studies in the Pacific Northwest suggest that P. maniculatus there may be more specialized than previously thought (Carey and Harrington 2001; Manning and Edge 2004).

Finally, CCA on microhabitat variables explained a large proportion of the variation in abundances of all small mammals, although the explanatory power of microhabitat was much less than that by macrohabitat characteristics despite macrohabitat being defined by far fewer variables (n = 1) than microhabitat (n = 19). However, similarly structured macro-habitats such as red fir and pine-cedar forests (e.g., open canopies, dense shrubs, bare ground, and rocks) yielded distinct small mammal communities. Thus, although different macrohabitats shared some microhabitat features, they also support understory characteristics that were unique, thereby allowing for habitat segregation. These results support the conjecture that habitat associations are hierarchical but largely determined by characteristics specific to various macrohabitat types (e.g., Jorgensen 2004; Kelt et al. 1999; Morris 1984, 1987), but these affinities are partly constrained by the understory communities unique to each forest type. Nevertheless, the influence of spatial scale on selection may be taxon-specific, as suggested by the dichotomy between P. maniculatus and Neotamias species. Alternatively, this difference may represent habitat associations of generalist versus specialist species with the former showing more coarse-grained associations than the latter.

Future studies of spatial scale and small mammal habitat associations should emphasize comparability across studies and spatial scales. A critical issue is the definition of scale; Jorgensen (2004) highlighted the challenges associated with comparing results among microhabitat studies because of differences in microhabitat definitions. Forty-nine percent of studies that he reviewed measured vegetation in an area ≥ 45 m2, whereas 24% did so in an area ≤ 11 m2; no study examined vegetation in areas between 11 and 45 m2. Of course, the appropriate scale at which to measure “microhabitat” will depend upon the life-history characteristics of the focal organism, but convergence on commonly accepted definitions (e.g., Morris 1987) is desperately needed. Choosing common scales at which to conduct studies of habitat associations can provide for more generalities and allow for successful meta-analyses among studies (e.g., Kelt et al. 1999). Moreover, comparing the same characteristics across multiple spatial scales would greatly enhance our understanding of habitat features potentially important to small mammals at different spatial scales.

Management and conservation implications.—Because small mammals are basal to many food webs, managers need realistic strategies for conserving and monitoring these faunas. Results from this study suggest that the most cost-effective approach is to gather data at the macrohabitat scale (e.g., forest type) and emphasize coarse-scale assessment of understory structure. Orrock et al. (2000) suggested that for southern red-backed voles (Myodes gapperi), coarse-grained information of habitat type (e.g., forest type) may allow for identification of suitable areas for more refined (e.g., meso-habitat) assessments that require minimal effort. Furthermore, the influential microhabitat variables identified in our study (e.g., shrub cover and canopy openness) also may be considered macrohabitat descriptors and can be readily quantified at a coarser scale. Hence, information at both scales is essential but can be gathered and quantified with minimal effort.

White fir and mixed-fir forests supported high small mammal abundance but white fir forests were species-poor; further, neither forest type yielded population levels as high as those in red fir forests. We postulate that white fir forests could support a more diverse fauna of small mammals if they were supplemented with more structurally diverse understories; those white fir and mixed-fir forests that yielded the highest captures and species richness had diverse understory characteristics including a dense shrub layer, larger downed woody debris (e.g., larger than branches), and abundant nonwoody vegetation, all of which are characteristics of old-growth forests. Silvicultural practices, then, may foster small mammal diversity by opening patches in the canopy, allowing sunlight to reach the forest floor, and promoting the development of understory vegetation (Carey 1995; Laacke and Tappeiner 1996; Morriss 1954).


We thank M. Johnson and P. Stine for administrative and financial work on this project; M. Rejmanek, E. Laca, M. Watnik, and J. Baldwin for statistical support; B. Holycross for technical assistance; and C. Salk, J. Katz, B. Rock, R. Innes, J. Brooks, K. Marsee, A. Wasiutynski, M. Gilbart, C. Morcos, D. Smith, and A. Goldman for their invaluable efforts in the field. Two anonymous reviewers provided critical feedback on an earlier draft. This work was supported by the Joint Fire Sciences Program and the United States Department of Agriculture Forest Service (Region 5).


  • Associate Editor was Eric C. Hellgren.

Literature Cited

View Abstract