Does Ungulate Foraging Behavior in Forest Canopy Gaps Produce a Spatial Subsidy with Cascading Effects on Vegetation?

Concentrated foraging in forest canopy gaps by large ungulates may produce a pulsed spatial resource subsidy with cascading effects on the composition and developmental trajectory of gap vegetation. To test this hypothesis, we investigated the influence of white-tailed deer (Odocoileus virginianus) use of 12 artificial canopy gaps in a hemlock-northern hardwood forest. Ground-layer vegetation was monitored and available reactive nitrogen was assayed using resin beads deployed under the snowpack (March–April) and soon after snowmelt (May). Deer use of openings was consistent with the forage maturation hypothesis, with the greatest levels of use occurring in small gaps. Allometric relationships suggest that mean localized winter pulses of deer-excreted N may be on par and/or in excess of annual atmospheric N deposition in the region. Correspondingly, deer access plots contained significantly more reactive N than exclosure plots soon after snowmelt (P = 0.036) in April. While the pulse was indistinguishable by May, our nonmetric multidimensional scaling ordination results suggest that plant community composition in exclosure and control plots reflects this pulsed gradient in N availability. Given the importance of canopy disturbances and gaps to the perpetuation of forest ecosystems, localized and/or heterogeneous impacts may be magnified as forests turn over.

deer yarding, forage maturation hypothesis, nitrogen, Odocoileus virginianus, Tsuga canadensis

Canopy gap dynamics have been linked to the maintenance and perpetuation of biological diversity in forested ecosystems through their influence on species composition and structural complexity (Runkle 1981, Clebsch and Busing 1989, Goldblum 1997, Kneeshaw and Prévost 2007). Forest ungulates interact with canopy disturbances and in some cases alter their foraging behavior and movement patterns to capitalize on temporal and spatial pulses in resource availability (Stewart et al. 2000, Kuijper et al. 2009). Preferential use of forest canopy gaps could have important implications for postdisturbance plant community dynamics (Kuijper et al. 2009, Holmes and Webster 2011, Kern et al. 2012). A broader understanding of ungulate-disturbance interactions and associated direct and indirect effects in forested ecosystems is needed to better balance biodiversity conservation with sustainable forest and wildlife management (Reimoser and Gossow 1996, Kraft et al. 2004, Murray et al. 2013).

While originally developed to predict patterns of aggregation by large grazers (Hobbs and Swift 1988, Fryxell 1991, Wilmshurst et al. 1995), the “forage-maturation hypothesis” has shown promise for predicting foraging patch selection and partial migration by browsers and generalist herbivores (Stewart et al. 2000, Mysterud et al. 2011). Herbivores are predicted to select foraging patches of intermediate biomass, which are at early stages of maturity and contain higher quality forage. They should spend less time in patches that contain less biomass as well as patches high in biomass but dominated by poorly digestible, mature vegetation (Hobbs and Swift 1988, Stewart et al. 2000). Stewart et al. (2000) successfully applied this theory to white-tailed deer (Odocoileus virginianus) use of clearings in subtropical thorn woodlands. Their results showed strong seasonality of use associated with forage maturation (forb versus shrub spout; Stewart et al. 2000). A logical extension of this theory would predict that canopy gaps in closed canopy forests, which typically contain greater herb cover and colonizing sapling biomass than the surrounding forest matrix (Webster and Lorimer 2002, Shields and Webster 2007, but see Moser et al. 2008), should experience greater use. Use should be highest in small to intermediate size gaps, which provide abundant forage of high nutritional quality. Nutritional quality declines with increasing light in larger gaps as a result of higher C:N ratios (Hartley et al. 1997) and increased concentrations of secondary plant metabolites (Bryant et al. 1983, Bryant 1987, but see Reichardt et al. 1991). Consequently, use of large gaps may peak initially as a result of high herb abundance but should decline rapidly as tree regeneration displaces herbaceous vegetation and declines in nutritional quality.

Recent work by Kuijper et al. (2009) provides partial support for this hypothesis. They observed that visitation rates of harvest openings (236–2,803 m²) by a broad suite of forest ungulates (European bison, Bison bonasus; red deer, Cervus elaphus; roe deer, Cervus capreolus; moose, Alces alces; and wild boar, Sus scrofa) in the Białowieża Primeval Forest in Poland increased with gap openness (Kuijper et al. 2009). Use and associated browsing pressure on planted trees, however, did not appear to be deterred by lower nutritional quality in more open environments (Kuijper et al. 2009). In temperate forested regions, white-tailed deer use large forest openings (≥ 8 ha) if they are within their predisturbance home range (Campbell et al. 2004) and provide forage complementary to adjacent older forests (Johnson et al. 1995). Exceptions have been observed (Campbell et al. 2006), and in northern latitudes there is some evidence that deer may avoid openings with high forage availability during the winter because deep snow increases metabolic costs and/or perceived predation risk (Rieucau et al. 2007). White-tailed deer use of smaller openings, such as those associated with natural disturbances and uneven-aged management, has received less study (Holladay et al. 2006, Holmes and Webster 2011, Kern et al. 2012).

It is widely appreciated that ungulates have the potential to influence vegetation community composition and structure through their foraging behavior (Augustine and McNaughton 1998, Russell et al. 2001, Rooney and Waller 2003, Horsley et al. 2003, Côté et al. 2004, Hester et al. 2006). Chronically high levels of herbivory have been associated with biological impoverishment of forest understories (Rooney et al. 2004) and the formation of alternate stable states (Stromayer and Warren 1997, Augustine et al. 1998). Nevertheless, ungulate effects on forest ecosystems can be surprisingly complex, at times producing seemingly paradoxical results (Heckel et al. 2010, Royo et al. 2010, Rutherford and Schmitz 2010, Jensen et al. 2011). Unexpected results may be caused by unanticipated feedbacks between direct and indirect effects of herbivory, context-dependent foraging behavior, and variability in initial conditions (Schmitz 2008). For example, Heckel et al. (2010) observed concomitant declines in stature and reproductive output of both palatable and unpalatable species in response to white-tailed deer browsing of palatable species in Pennsylvania, USA, which they attributed to deer-mediated soil quality declines (e.g., soil compaction). Conversely, Jensen et al. (2011) observed greater herb-layer cover and heterogeneity with increasing winter use of relict eastern hemlock (Tsuga canadensis (L.) Carr.) stands by deer in northern Michigan, USA. High winter deer use was associated with greater soil resource heterogeneity (Jensen et al. 2011).

Soil-mediated effects have been well documented for migratory grazers (e.g., McNaughton et al. 1997, Frank and Groffman 1998, Augustine and Frank 2001, Singer and Schoenecker 2003) but have only recently begun to be explored empirically for semimigratory forest ungulates such as white-tailed deer (Jensen et al. 2011, Bressette et al. 2012, Murray et al. 2013). Unlike their grassland counterparts, forest ungulates are broadly perceived as decelerating N cycling and availability (Pastor et al. 1993, Ritchie et al. 1998, Carline et al. 2005). The mechanism underlying this relationship is that selective foraging of legumes (Ritchie et al. 1998, Knops et al. 2000, but see McNeil and Cushman 2005) and broadleaves (Pastor et al. 1993) shifts vegetative community composition toward plant species that produce lower-nutrient, recalcitrant litter (Pastor et al. 2006). Inputs of nitrogenous wastes are more than offset by changes in vegetation tissue quality (Pastor et al. 2006). However, as noted by Ritchie et al. (1998), forest ungulates may accelerate nutrient cycling if major forages have high tissue N, which allows them to maintain efficient use of another limiting resource, such as light or water, or they are able to tolerate herbivory. Alternately, Seagle (2003) proposed that the movement of forest ungulates between patches of high (e.g., cropland) and low (e.g., forest) N availability could result in a “spatial subsidy.” We hypothesize that concentrated foraging in hemlock-northern hardwood canopy gaps produces a spatial subsidy (unidirectional movement of resources between habitats) with cascading effects on the composition and developmental trajectory of gap vegetation, which is contingent on gap size.

To test this hypothesis, we investigated the influence of white-tailed deer use of 12 artificial canopy gaps in a hemlock-northern hardwood forest on nitrogen availability and ground-layer vegetation dynamics. Gaps were created to provide a range of opening sizes (~ 36–450 m²), and arrays of subplots and exclosures were established within each opening. Our specific hypotheses were (1) deer use will be negatively associated with canopy gap size, (2) deer use of gaps results in a spatial subsidy, demonstrated by greater N availability in deer access plots than exclosures, and (3) successional change in gap ground-layer vegetation is influenced by cascading direct and indirect effects of herbivory.

Methods

Study Site

The study was conducted in a hemlock-northern hardwood forest at Michigan Tech.'s Ford Center Research Forest near Alberta, Michigan (46° 37′N, 88° 29′W). Dominant tree species included sugar maple (Acer saccharum Marsh.), eastern hemlock, and yellow birch (Betula alleghaniensis Britt.), which comprised 42, 34, and 12% of stand basal area, respectively (James Schmierer, Forester, Ford Center Research Forest, Michigan Technological University, pers. comm., Apr. 19, 2012). Other minor tree species included balsam fir (Abies balsamea L.), red maple (Acer rubrum L.), white spruce (Picea glauca (Moench) Voss), black cherry (Prunus serotina Ehrh.), basswood (Tilia americana L.), and American elm (Ulmus americana L.). Soils were generally moderately well drained Kallio cobbly silt loams (coarse-loamy, mixed, superactive, frigid Oxyaquic Fragiorthods), 1–20% slopes (Berndt 1988). Regional white-tailed deer densities vary seasonally but typically range from 6.5 deer km⁻² in the spring to 9.3 deer km⁻² in the fall (Mayhew 2003). During high snow years, overwintering densities in hemlock stands can be substantially higher as a result of conditional migration to areas of dense conifer cover (Verme 1973, Sabine et al. 2002, Witt et al. 2012). Local climate is moderated by Lake Superior. Average daily temperature ranges from −9.8° C in winter to 17.4° C in summer (Berndt 1988). Annual precipitation averages 87.4 cm, with mean seasonal snowfall of 382.5 cm (Berndt 1988).

Experimental Design

During the winter of 2002/03, 20 artificial canopy gaps were created. Three size classes of gaps were formed: small (50–150 m², n = 7), medium (151–250 m², n = 7), and large (251–450 m², n = 6). All trees > 1 m tall and associated logging slash were removed from the openings. Four to twelve 1 × 1 m sample plots were established at randomly selected locations within each gap and monumented. In 2005, one to three plots within each gap were chosen at random and enclosed in 1.52 m tall wire fencing to exclude white-tailed deer. Exclosures were circular and approximately 2 m in diameter. The mesh size (~15 cm²) was small enough to exclude deer but large enough to permit access by smaller herbivores (e.g., snowshoe hares, Lepus americanus; and microtine rodents). Ground-layer vegetation on all sample plots was censused during the summer of 2007. Initial results can be found in Holmes and Webster (2010, 2011). For the current study, we selected at random 12 gaps that contained at least two exclosures and provided a gradient in gap size for a detailed assessment of vegetation change and N availability.

Logging slash was removed so that deer exclosures could be constructed and the deer effect isolated (Holmes and Webster 2010). High levels of logging slash and/or fine and coarse woody debris retention following natural gap formation can provide a measure of protection for tree regeneration and herbaceous plants from forest ungulates (Smit et al. 2012, but see Fredericksen et al. 1998). Consequently, our results may not be directly comparable to natural canopy gaps or silvicultural treatments that retain or concentrate slash in openings.

Field Techniques

Gap area was estimated in 2005 and 2011 based on an eight-sided polygon anchored on a rebar stake at gap center. Eight radii were measured from gap center (45° intervals) to the crown edge of a main canopy tree bordering the opening (Webster and Lorimer 2002). Winter deer use was quantified through pellet and browse counts. Pellet counts were conducted in May 2011. To locate all deer pellet piles, two observers used a system of concurrent rechecks (see also Witt et al. 2012). To limit our window of observation to deer use the previous winter, only pellet piles on top of the previous year's leaf litter were counted. Number of pellet piles per gap was converted to pellet piles per ha based on the 2011 gap area measurement, which defined the spatial extent of our search. Browse counts were conducted in late June and early July 2011. Depending on availability, up to 10 sugar maple saplings ranging in height from 0.5 to 2 m were randomly selected within each gap. Given its ubiquity and tolerance to herbivory, this species is frequently used to generate a browse index for this forest type (Anderson and Loucks 1979, Frelich and Lorimer 1985, Rooney et al. 2000). For each sapling, numbers of unbrowsed twigs and twigs browsed the previous winter were recorded.

During the summer of 2011, ground-layer vegetation (< 1 m tall) was sampled in each 1 × 1 m sample plot (n = 48). Percent cover was estimated ocularly by a single observer. Species nomenclature follows USDA Plants database.¹ To compare light availability between gaps, a hemispherical canopy photograph was taken at a height of 1 m in the center of each gap in late June 2011. Direct and diffuse under-canopy radiation (mol m⁻² day⁻¹) was estimated using the computer software WinSCANOPY (2005).

Available Nitrogen

Soil reactive N (NH₄⁺ and NO₃⁻) availability was measured in each 1 m² plot by burying three ion exchange resin bags 10 cm beneath the soil surface. Each resin bag contained 7.4 mL (approx. 5.0 g) of Dowex Marathon MR-3 hydrogen and hydroxide form resin beads (Sigma-Aldrich, St. Louis, MO, USA) in a nylon-lycra bag (Giblin et al. 1994, Harpole and Tilman 2007). Available soil nitrogen was measured in two consecutive stages: (1) during snowmelt and the beginning of N uptake by plants (Mar. 19–Apr. 29, 2011) and (2) following the emergence of understory vegetation and canopy leaves (Apr. 29–May 25, 2011). Hereafter, the first stage is referred to as April N and the second stage as May N. To minimize soil disturbance, bags were inserted at a 45° angle and any disturbed snow or leaf litter was replaced after burial. Ions were extracted from the resin beads using 25 mL of 2 M KCl per resin bag. Concentrations of ammonium and nitrate were determined colorimetrically using a Perstorp 3550 EnviroFlow System (Rapid Flow Analyzer). Nitrogen ion availability is expressed as mg [ion] L⁻¹ g⁻¹ resin, based on the postextraction dry weight of resin beads.

Statistical Analyses

Trends in deer use, browse, and available reactive N with canopy gap area were evaluated using linear regression. Regression assumptions of normality and homogeneity of error terms were evaluated using plots of standardized residuals versus fitted values and normal probability plots (Neter et al. 1996). Natural log transformations were used when necessary to meet assumptions. One-tailed paired t-tests were used to test the hypothesis that exclosures contained less reactive N than control plots (i.e., deer-access plots). Regression analyses and paired-t-tests were conducted in Minitab (2010).

Shifts in plant community composition along environmental gradients from 2007 to 2011 were investigated using nonmetric multidimensional scaling ordination as implemented in PC-ORD version 5 (McCune and Mefford 2006). Ground-layer data for 2007 were derived from Holmes and Webster (2011). Because our interest was in describing patterns of community change rather than diversity, rare species (present on less than 5% of sample plots) were removed prior to analysis to reduce noise and enhance the detection of relationships between community structure and measured environmental gradients (McCune and Grace 2002). A total of 49 species were retained in the main matrix. Seven environmental variables were included in the secondary matrix (Table 1). These variables were selected to describe hypothesized relationships between available reactive N, canopy gap size, and ground-layer species composition. The ordination was run in Auto Pilot mode, which uses Sørenson's distance measure and a random starting configuration (McCune and Mefford 2011). To determine whether vegetation communities were more heterogeneous at the plot level within deer exclosures or control plots, we used a permutational analysis of multivariate dispersions (PERMDISP2, Anderson 2004). This test compares the dispersion (mean distance to group centroid) of groups (plots) within the ordination space.

Correlations between 2011 environmental variables and nonmetric multi-dimensional scaling ordination axes. The final ordination was three-dimensional, with a stress of 16.5 and cumulative r² of 0.754.

Table 1.

Correlations between 2011 environmental variables and nonmetric multi-dimensional scaling ordination axes. The final ordination was three-dimensional, with a stress of 16.5 and cumulative r² of 0.754.

^a Under canopy radiation (mol m⁻² d⁻¹).

Results

Influence of Gap Area on Winter Deer Use

Deer use, as described by number of fecal pellet groups per hectare, displayed a significant negative association with gap area (P = 0.048; Figure 1A). Use of small gaps (< 100 m²) was particularly high, resulting in 816 to 3,437 pellet groups ha⁻¹ since leaf-fall. The mean pellet density across the 12 study gaps was 893.0 ± 282.8 pellet groups ha⁻¹. Browse rate on sugar maple saplings within the reach of wintering deer was highly variable between openings (Figure 2B). We found only weak statistical evidence that browse rate declined with increasing opening area (P = 0.102; Figure 1B).

Figure 1.

Relationship between white-tailed deer fecal pellet groups ha⁻¹. (A) ln(pellet groups) = 7.46 – 0.00625 Gap area, F_{1, 11} = 5.06, P = 0.048, R² = 0.336) and percent of last year's growth browsed for 10 randomly selected sugar maple saplings per gap that were within the reach of wintering deer. (B) Browse = 79.1 – 11.2 Gap area, F_{1, 10} = 3.24, P = 0.102, R² = 0.245). Solid lines denote fitted regression lines and dashed lines denote 95% confidence intervals for the prediction.

Figure 2.

Mean (± 1 SE) available reactive nitrogen (mg L⁻¹ g⁻¹ resin) within deer exclosures and control (deer access) plots in (A). April, (B). May, and (C). April + May. Significant paired t-test results are indicated by an asterisk (P < 0.05). Corresponding difference between exclosure and control plots (exclosure – control) within individual gaps are provided in panels D, E, and F in order of increasing gap size.

Availability of Reactive Nitrogen

Exclosures contained significantly lower levels of available reactive N (NH₄⁺ and NO₃) during and shortly after snowmelt (April) than control plots (P = 0.036; Figure 2A). During the May sampling period, no significant differences were observed (P = 0.449; Figure 2B). Cumulatively (April + May), exclosures contained slightly less available reactive N than controls, but the difference was not statistically significant (P = 0.109; Figure 2C). The greatest April and April + May reactive N differences between exclosures and controls were observed in larger gaps (Figure 2D–F). These latter relationships were coincident with significant positive associations (P < 0.05) between available reactive N in both April and April + May and gap size for both exclosures and controls (Figure 3). There were no differences in slope or intercept between exclosures and controls for April N (P = 0.453 and P = 0.916, respectively) or April + May N (P = 0.198 and P = 0.505, respectively).

Figure 3.

Relationship between mean available reactive nitrogen (mg L⁻¹ g⁻¹ resin) within deer exclosures (barred circle) and controls (open circle) and canopy gap area (m²) in (A). April, (B). May, and (C). April + May. No significant differences in slope or intercept were observed for exclosures and control plots (P > 0.198). Regression results are as follows: April N = 0.334 + 0.00263 Gap area, F_{1, 23} = 17.06, P < 0.001, R² = 0.437; April + May N = 1.84 + 0.0108 Gap area, F_{1, 23} = 12.99, P = 0.002, R² = 0.371. Solid lines denote fitted regression lines and dashed lines denote 95% confidence intervals for the prediction.

Composition and Development of Gap Ground-Layer Vegetation

The final nonmetric multidimensional scaling ordination solution was three-dimensional with a stress of 16.5 and instability criterion of 0.00027 after 500 iterations. The cumulative r² was 0.745, with axis 3 explaining the most variation (r² = 0.303) followed by axes 1 (r² = 0.288) and 2 (r² = 0.164). Among the environmental variables, April and April + May N displayed the strongest positive correlations with axis 1 (Table 1). April, May, and April + May N also displayed strong positive correlations with axis 3 (Table 1). Measures of gap area and available light exhibited the strongest positive correlations with axis 1 (Table 1).

Of the 49 species tested, American red raspberry (Rubus idaeus L.) displayed the strongest correlation to axis 3 (r = 0.767; Table 2). The next strongest correlation with axis 3 was for spinulose wood fern (Dryopteris carthusiana (Vill.) H.P. Fuchs; r = 0.476). Three of the five fern species included in the ordination displayed positive correlations with axis 3 (Table 2). All but one of nonnative species included in the ordination had neutral or negative correlations to axis 3 (Table 2). The exception was the robust annual, brittlestem hempnettle (Galeopsis tetrahit L.), which had a somewhat strong positive correlation with this axis (r = 0.242). The strongest negative correlation with axis 3 was exhibited by the nonnative tall hawkweed (Hieracium piloselloides Vill.; r = −0.362) followed by the native calico aster (Symphyotrichum lateriflorum (L.) A. & D. Löve; r = −0.326; Table 2).

Pearson's correlations between species abundance and nonmetric multidimensional scaling ordination axes. See Table 1 for ordination diagnostics. Correlations ≥ 0.2 are denoted with bold text. Only species that occurred on > 5% of sample plots were included in the ordination.

Table 2.

Pearson's correlations between species abundance and nonmetric multidimensional scaling ordination axes. See Table 1 for ordination diagnostics. Correlations ≥ 0.2 are denoted with bold text. Only species that occurred on > 5% of sample plots were included in the ordination.

^a Nonnative.

The two species displaying the strongest correlations with axis 1 were nodding sedge (Carex gynandra Schwein.; r = 0.485) and American red raspberry (r = 0.386; Table 2). Shade-tolerant, woodland species exhibited predominantly negative associations with axis 1 (Table 2). The liliaceous species bluebead lily (Clintonia borealis (Ait.) Raf.) and Canada mayflower (Maianthemum canadense Desf.) exhibited two of the strongest negative correlations with axis 1 (r = −0.479 and −0.472, respectively; Table 2). Three out of the four native liliaceous species included in the ordination were negatively correlated with axis 1. Starflower (Trientalis borealis Raf.), a shade-tolerant member of the primrose family, also exhibited a strong negative correlation with axis 1 (r = −0.479). Sugar maple saplings in the ground-layer (≤ 1 m height) had the strongest negative association with axis 1 (r = −0.652).

Successional trajectories of gap plant communities accessible to deer (control) were divergent from those of exclosures (Figure 4). Movement in the ordination space between 2007 and 2011 control plots suggests a general upward shift along axis 3 and a wider, but generally positive distribution, along axis 1. Exclosure species composition displayed a more subtle upward shift along axis 3 and an increasingly negative association with axis 1. Given the strong positive correlation between available reactive N and these axes (Table 1; Figure 4E), they likely represent a gradient in available N. Axis 1 was also positively correlated with gap area and available light (Table 1). Consequently, species composition in control plots appeared to correspond more with high N availability than it did in exclosures, where composition was more associated with lower reactive N and light levels (Figure 4).

Figure 4.

Nonmetric multidimensional scaling ordination of canopy gap, ground-layer vegetation in a hemlock-northern hardwood forest. Open circles indicating the mean locations of control plots by gap in species space in (A). 2007 and (B). 2011 are scaled based on gap area in 2011. Barred circles indicate the mean location of deer exclosures by gap in (C). 2005 and (D). 2011 are also scaled based on 2011 gap area. Joint plot vectors for April + May N, gap area, and under canopy radiation 1 m aboveground surface are presented as solid lines (E). Note differences in scale between panels A–D, and E, vectors are scaled based on r² values and range from zero to one. For a complete listing of environmental variables, see Table 1. Correlations between ordination axes and plant species are provided in Table 2.

A permutational test of multivariate dispersions (PERMDISP) indicated that species composition was significantly more heterogeneous within exclosure than control (deer-access) plots (t = 2.48, P = 0.021). The mean (± 1 SE) distances from group centroid, based on Bray–Curtis dissimilarities, were 54.4 ± 1.5 and 48.9 ± 1.6 for exclosures and controls, respectively.

Discussion

Forage Maturation Hypothesis

Our results suggest that the forage maturation hypothesis may be useful for describing white-tailed deer use of forest canopy gaps at northern latitudes. Nine years after gap creation, winter deer use (as indexed by deer pellet groups) was greatest on a per unit area basis in small canopy gaps and declined significantly with increasing gap area. A number of mechanisms may favor foraging in small gaps, especially as gaps age. First, even very small canopy gaps can increase leader and lateral branch growth rates of shade-tolerant advance regeneration by an order of magnitude over closed canopy conditions (Canham 1988). Sugar maple, the dominant sapling species in the gaps examined (Holmes and Webster 2010), displays a strongly asymptotic growth response to increasing light availability and opening size (Canham 1988, Ellsworth and Reich 1992, Webster and Lorimer 2002) and is a preferred forage species for deer in the understory of hemlock stands (Anderson and Loucks 1979, Witt and Webster 2010). Second, standing woody biomass is generally lower for several years in small (< 80 m²) versus larger gaps (> 80 m²; Webster and Lorimer 2002), resulting in an intermediate level of biomass as compared to closed canopy areas and large gaps. Third, in coniferous forests snow depths around and within small openings are lower as a result of canopy interception and sublimation, which would favor movement and foraging by wintering deer (Euler and Thurston 1980, Morrison et al. 2003). Finally, small canopy gaps may delay maturation of woody browse as a result of declining light levels associated with lateral gap closure (Runkle and Yetter 1987, Klingsporn et al. 2012) and slower growth rates of less shade-tolerant species (Phillips and Shure 1990, Dale et al. 1995). Furthermore, concentrated use of forest openings by deer has been shown to effectively delay maturation though repeated cropping of woody stems (Pedersen and Wallis 2004, Royo and Carson 2006, Tremblay et al. 2007).

While it has been suggested that saplings growing under higher light conditions should be better defended against herbivory (Bryant 1987) and have lower tissue quality (e.g., higher C:N ratio; Hartley et al. 1997) than those growing under lower light levels, tissue quality results from a previous study in these gaps were inconclusive (Holmes and Webster 2010). This may explain, at least in part, why browse rate was only weakly associated with gap area. Browse rate would also be expected to decline with increasing opening size if saplings in larger gaps were able to escape herbivory more quickly than those in small gaps, which would reduce both time until maturity and forage availability. Some evidence of this was provided at our study site by Holmes and Webster (2010) who observed that saplings (the majority of which were sugar maple) in larger gaps (> 151 m²) were taller on average than those found in small gaps (< 150 m²). While containing fewer tall saplings, small gaps did contain more seedlings and small saplings (≤ 50 cm tall) than larger openings (Holmes and Webster 2010). Consequently, small gaps appear to provide greater biomass of accessible woody browse, whose maturation can be more easily delayed by repeated cropping. As a result, through time small gaps may actually provide larger foraging patches (area of available biomass) in which foraging ungulates would be expected to spend more time (Shipley and Spalinger 1995).

Evidence of a Spatial N Subsidy

Two lines of evidence support the potential of wintering deer to produce a spatial N subsidy in coniferous forest canopy gaps. First, mean observed pellet group densities in our canopy gaps would result in localized nitrogenous waste deposition rates commensurate with atmospheric deposition in the region. Second, significantly lower levels of available reactive N were observed in deer exclosures shortly after snowmelt than were observed for plots accessible to deer.

Using the allometric approach described by Jensen et al. (2011) to estimate total N released per excretion event, our highest level of use (3,437 pellet groups ha⁻¹ since leaf-fall) would produce a nitrogenous waste (urine and feces) deposition rate of 18.5 kg N ha⁻¹ winter⁻¹. Applying the same allometry to mean use (893.0 ± 282.8 pellet groups ha⁻¹) results in a potential deposition rate of 4.8 ± 1.5 kg N ha⁻¹ winter⁻¹. Total dry and wet atmospheric N deposition in the region from 2009–2011 averaged approximately 5 kg N ha⁻¹ yr⁻¹.² Consequently, canopy gaps in conifer forests where deer aggregate in the winter may be receiving additional ungulate-associated N deposition rates on par with and/or in excess of atmospheric deposition.

Seagle (2003) proposed that white-tailed deer foraging in multiuse landscapes could produce substantial spatial N subsidies. Recently, Abbas et al. (2012) showed that roe deer produce spatial subsidies of N and P to forest patches in an agricultural region of France. Effect strength was sensitive to the amount of forest area and how deer used the landscape to feed and defecate (Abbas et al. 2012). Given the forested nature of our study system, a substantial subsidy may seem unlikely. However, conditional migration in response to snow depth by white-tailed deer in northern portions of their range (Verme 1973, Moen 1976, Van Deelen et al. 1998) can produce high localized overwintering densities in conifer yarding complexes (≥ 25 deer km⁻²; Van Deelen et al. 1996) and relict hemlock stands in particular (≥ 83 deer km⁻²; Jensen et al. 2011, Witt et al. 2012). Our results suggest that deer use within these stands is further concentrated by interactions between foraging behavior, canopy gap dynamics, and local variability in snow depth (see Rongstad and Tester 1969, Moen 1976, Morrison et al. 2003). Furthermore, nitrogenous waste produced by wintering deer represents a transfer of resources from summer and autumn ranges to winter range as a result of catabolism of stored fat and muscle associated with winter malnutrition (Bahnak et al. 1979). Consequently, our results corroborate the highly heterogeneous nature of forest ungulate effects on forest biogeochemistry (Abbas et al. 2012, Murray et al. 2013).

The second line of evidence of a spatial subsidy is provided by higher levels of available reactive N outside of deer exclosures soon after snowmelt. The greatest differences were observed in larger gaps, which contained more reactive N. Concentrations of available N tend to increase with increasing opening size as a result of reduced fine root density and microbial assimilation (Parsons et al. 1994, Prescott et al. 2003). Small gaps are more likely to be bridged by roots of gap border trees, which draw down available N (Parsons et al. 1994). Declining overstory litter inputs with increasing gap area reduce the availability of labile C, which leads to a reduction in N assimilation by microbes (Hart et al. 1994, Prescott et al. 2003). Consequently, ungulate derived inputs should be more readily apparent in larger openings even if less intensely used by foraging ungulates than smaller gaps where N would be expected to be more rapidly taken up by tree roots and assimilated by microbes. Total available reactive N outside of exclosures in our sample gaps was greater than levels observed for a similar period in heavily used closed canopy deeryards (4.45 versus 4.09 mg L⁻¹ g⁻¹ resin), while within exclosure levels were nearly identical (3.77 versus 3.79 mg L⁻¹ g⁻¹ resin; Murray et al. 2013).

Our results suggest that the duration of the subsidy is brief. During the May sampling period, no consistent differences were observed between exclosures and controls. An advantage of our approach was that ion exchange resins simulate the uptake of plant-available N by roots and microbes (Binkley and Vitousek 1989, Qian and Schoenau 2002). Consequently, they integrate transient fluxes in plant-available N over time (in our case two months of continuous monitoring during a period of rapid change). Some other commonly employed techniques better capture mineral and mineralizable N (Binkley and Vitousek 1989), the latter of which would be more sensitive to changes in litter quality associated with selective herbivory (Pastor et al. 2006). For example, Harrison and Bardgett (2004) assessed the influence of red deer on soil nitrogen availability in regenerating forests in the central Scottish Highlands using direct assays and incubations of bulk soil samples taken at three times during the growing season. They found that dissolved inorganic and organic N as well as mineralizable N were lower in browsed verses unbrowsed areas (Harrison and Bardgett 2004). Bressette et al. (2012), on the other hand, assessed white-tailed deer impacts on soil N (NO₃⁻ and total N) in northern Virginia based on a single bulked sample from each of their plots. At two out of three sites, NO₃⁻ and total N were greater outside of exclosures, which the authors attributed to lower uptake by the depauperate vegetation in deer access plots (Bressette et al. 2012). Most of the N associated with winter deer waste comes from urinary N (Jensen et al. 2011), which has a patchy spatial distribution and is readily available to plants. Therefore, instantaneous samples of extractable N from soil cores or ex-situ incubations would be unlikely to capture this flux. For example, while pellet groups were abundant in our sample gaps, few occurred within our deer access plots. Consequently, the signal detected by our ion exchange resins was likely a diluted pulse associated with urinary N coming out of suspension in the snowpack and moving laterally through the soil profile. Incubations could capture mineralization of fecal pellet N, but as noted by Pastor et al. (2006, p. 301) fecal pellets of herbivores feeding on low N forages like twigs are “mainly sawdust” and recalcitrant to mineralization.

Another previously unacknowledged factor is the timing of ungulate-associated N deposition in relation to deer body condition and foraging behavior. In response to deep snow on their winter range, migratory white-tailed deer reduce forage consumption and movement in an effort to conserve energy (Ozoga and Verme 1970, Moen 1976, Dumont et al. 2005) and avoid predation (Nelson and Mech 1981). White-tailed deer on low nutrition winter diets may lose 11–28% of their body weight in response to reduced forage consumption and changes in metabolic activity (Silver et al. 1969). Catabolism of fat and muscle tissue during starvation increases urinary N (Bahnak et al. 1979) up to five times above background levels (DelGiudice and Seal 1988). Consequently, unlike other forest ungulates and deer in more southern latitudes, behavioral and metabolic constraints on northern white-tailed deer populations during winter may facilitate a pulsed resource subsidy in coniferous forests (Jensen et al. 2011).

Cascading Effects on Gap Ground-Layer Plant Communities

Pulsed resource subsidies, defined as “rare, brief, and intense episodes of increased resource availability in space and time,” can have persistent effects on plant community structure and composition (Yang et al. 2008). Our nonmetric multidimensional scaling ordination results suggest that the potentially substantial but brief influx of plant-available N may influence the developmental trajectory of gap plant communities. Locations of control plots in the ordination space displayed an increasingly positive correspondence with axes associated with high availability of reactive N and light. Exclosure plots, on the other hand, displayed an increasingly negative association with axis 1 over time, which was most strongly correlated with April N and light availability. As expected, nitrophilic genera (e.g., Rubus) displayed strong positive associations with gradients in available N and light (Jobidon 1993). Negative associations between shade-tolerant herbs and environmental gradients associated with greater N availability were consistent with trends observed in atmospheric N deposition and fertilization studies (Gilliam 2006, Strengbom and Nordin 2008, Talhelm et al. 2013). Given the small size of our exclosures and limited range of light environments provided by our experimental gaps, the degree of differentiation observed within individual gaps is noteworthy (see also Holmes and Webster 2011).

Herbivory at our study site has homogenized plant communities across a range of gap sizes (Holmes and Webster 2011). This is likely a result of the combined influence of winter and summer use of these stands by migratory and resident deer. In areas with stronger seasonal gradients in ungulate use, the herbaceous layer may be relatively insulated from direct effects of herbivory (Jensen et al. 2011, Murray et al. 2013). Under these conditions, ground-layer communities may respond to the heterogeneous distribution of pulsed resources by increasing in cover and/or heterogeneity (de Mazancourt et al. 1998, Jensen et al. 2011, Murray et al. 2013). Forest ground-layer communities subjected to continuously high rates of consumption, on the other hand, tend to become increasingly homogenous and species poor (Rooney et al. 2004, Wiegmann and Waller 2006, Thiemann et al. 2009). Additionally, the concentrated use of gaps for foraging as compared to closed canopy environments (Kuijper et al. 2009) may negate positive fertilization effects for summer herbs and/or decrease fine scale heterogeneity in resource availability. Consequently, the spatial patterning of ungulate-excreted resource pulses in canopy gaps clearly warrants further investigation.

Conclusion

Our results suggest that nitrogenous waste excretion by aggregations of wintering deer may represent an important but underappreciated pulsed resource in forest ecosystems. The magnitude of this pulse is likely influenced by behavioral and metabolic responses to winter severity and the spatial patterning of winter and summer habitats. Nitrogen pulses resulting from concentrated foraging and defecation in canopy gaps may produce persistent shifts in plant community composition. Given the importance of canopy disturbances and gaps to the perpetuation of forest ecosystems, localized and/or heterogeneous impacts may be magnified as forests turn over (Nuttle et al. 2011). A better understanding of these relationships is needed to identify when and where forest ungulates may have a meaningful impact on the biogeochemistry and developmental trajectory of forests.

Endnotes

1.

Please visit the USDA Plant database at http://plants.usda.gov.

2.

For more information, please visit http://epa.gov/castnet.

Acknowledgments: We thank Miranda Aho, Brandon Bal, Dan Cerney, Fay Dearing, Jennifer Eikenberry, Krista Fischler, Kayla Griffith, Dan Hutchinson, Melissa Jarvi, Ben Jensen, Joanna Rogers, and Amber Voight for assistance with field and lab work. Janet Marr and John Hribljan assisted with identification of plant specimens. James Schmierer and students in the FERM program assisted with the creation of canopy gaps. We are grateful to Stacie A. Holmes and her field assistants for contributions to earlier phases of the gap project. Helpful comments on an earlier draft of this manuscript were provided by two anonymous reviewers. Financial support was provided by the McIntire–Stennis Cooperative Forestry Research Program and Ecosystem Science Center and School of Forest Resources and Environmental Sciences at Michigan Technological University.