Browse Preference and Browsing Intensity of White-Tailed Deer (Odocoileus virginianus) in Allegheny High Plateau Riparian Forests, USA
Abstract
Decades of chronic browsing by overabundant white-tailed deer (Odocoileus virginianus Zimmerman) has strongly influenced forest pattern and process on the Allegheny High Plateau Ecoregion of northwestern Pennsylvania, USA. Previous research has found that riparian forests contain the greatest herbaceous plant species richness of regional plant communities but little is known about the impacts of deer browsing on the structure and composition of the herbaceous layer (all vascular plants < 1 m tall) of Allegheny High Plateau riparian forests. We examined browse preference and browsing intensity by white-tailed deer on the herbaceous layer of five riparian forest study sites in the Allegheny National Forest during the summer growing season (July, September). Browsing intensity was low to moderate and differed significantly among sites and sample periods. Deer selectively foraged on a few preferred plant species during certain sampling periods, particularly Aster divaricatus, A. prenanthoides, Chelone glabra, Impatiens capensis, Pilea pumila, Polygonum virginianum and Ranunculus hispidis. We found that plant species richness and composition, and browsing intensity by white-tailed deer, are highly variable across riparian forests of the region. In order to assess or predict deer browsing impacts to regional riparian forests, we suggest that riparian sites be studied individually, perhaps on a watershed basis, as the surrounding landscape and available habitat may influence deer densities and foraging activity in an individualistic manner.
Introduction
Decades of chronic browsing by overabundant white-tailed deer (Odocoileus virginianus Zimmerman) has strongly influenced forest pattern and process on the Allegheny High Plateau Ecoregion of Pennsylvania [1]. Although white-tailed deer were severely depleted by overhunting in the early 1900s, the extensive forest cutting that occurred on the Allegheny High Plateau between 1880 and 1920 stimulated woody plant re-growth that provided copious browse. Abundant food, in conjunction with effective game protection laws and few natural predators, allowed populations to irrupt and reach a peak of about a million deer statewide by 1940 [2].
In the eastern United States, chronic over-browsing by white-tailed deer has diminished regeneration and height growth of many tree and shrub species, affecting forest composition and slowing the establishment of new, productive stands [3]. On the Allegheny High Plateau, Marquis [4] found that deer browsing caused 87% of forest regeneration failures that occurred outside of deer-exclusion fences. Limited regeneration of trees due to deer browsing has led to direct and indirect changes in species composition and structure in virtually all forest strata [2 - 3]. For example, deer-induced shifts in forest composition can cause microclimatic changes and alterations in light quality and quantity that can affect a wide range of plant species [3].
Another potential impact of chronic over-browsing by white-tailed deer is the creation of alternate stable states in plant communities [4 - 5]. Stromayer and Warren [4] defined an alternate stable state as “a stable condition in an ecological community at a different stage than what would be predicted, based on prevailing ecological and successional conditions”. Numerous studies have shown that deer preferentially forage on certain herbs and tree seedlings [3], diminishing them to such an extent that less palatable plants invade and dominate sites. On Pennsylvania’s Allegheny High Plateau for example, heavy browsing of blackberry (Rubus spp.) by deer has facilitated invasion of non-palatable ferns and grasses, particularly hayscented fern (Dennstaedtia punctilobula), New York fern (Thelypteris noveboracensis) and short husk grass (Brachyelytrum erectum), which can dominate the herbaceous layer of forests, diminishing plant diversity and interfering with forest regeneration [6].
Our research has shown that riparian or streamside forests of Pennsylvania’s Allegheny High Plateau support among the greatest herbaceous plant species richness of regional plant communities [see 7,8 and references therein]. Little information is presently available concerning the impact of deer browsing on riparian forests of the Allegheny High Plateau, particularly on the herbaceous layer stratum (all vascular plants <1 m tall) where summer plant species richness is often high. The goal of this study was to provide baseline information on the impacts of white-tailed deer browsing on the herbaceous layer of riparian forests of the Allegheny High Plateau of northwestern Pennsylvania during the summer growing season. Our specific objectives were to: 1) document browsing intensity of white-tailed deer on the herbaceous layer of riparian forests; and 2) determine browsing preferences of white-tailed deer in riparian forests.
Methods
Study area
This study was conducted in the Allegheny National Forest (ANF) (41º 45’ N; 79º 00’ W), northwestern Pennsylvania, USA. ANF lies within the Allegheny High Plateau Ecoregion, a non-glaciated landscape typified by relatively flat to gently rolling plateaus dissected by deeply eroded stream valleys [9]. Plateau elevations range from 500 to 700 m above sea level (asl); stream valley floors generally range from 300 to 400 m asl [10]. Summers are cool with an average temperature of 18.9oC [11]. Precipitation is distributed fairly evenly throughout the year and averages between 100 and 110 cm [10].
Field Sites and Data Collection
Five study sites - all part of a long-term riparian forest dynamics project - were used in this study [see 7,12,13 for descriptions of the study sites]. Four permanently marked 50 m transects were installed in the floodplain at each site, parallel to the stream active channel. The location of transects, and the distance between them, varied due to constrained valley morphology. Transects were separated by a minimum distance of 10 m and typically alternated in placement along each side of the stream for a distance of 200 to 500 m. Five permanently marked 1-m2 plots were established along each of the transects in a stratified random manner; thus each site had a total of 20 1-m2 vegetation plots that were repeatedly monitored during the study.
Sampling of riparian herbaceous layer vegetation occurred in July and September 1999. Within each 1-m2 plot, total percent ground cover and percent cover by species were estimated visually. For each plant species, the number of browsed or non-browsed browseable units (BUs) was tallied in each plot. The categories of BUs we used are the same as those used by Williams et al. [7] who modified the browsing assessment protocol of Anderson [14]. BUs having a jagged and torn appearance were scored as browsed by deer; those appearing cleanly-cut were scored as browsed by lagomorphs such as eastern cottontail (Sylvilagus floridanus J.A. Allen) or snowshoe hare (Lepus americanus Erxleben) [7]. We did not tally reproductive units of plants in browse plots since many herbaceous layer plants in Allegheny Plateau riparian forests produce flowers and fruit sporadically in response to soil moisture and light conditions [7]. We recognize that some inherent biases may exist in our observational assessment of browsing. For example, deer may preferentially browse plant species that are more apparent because of growth form or relative height. Likewise, deer may consume so much of a plant that any evidence of browsing is lacking.
Data Analysis
The hierarchical agglomerative classification analysis UPGMA (unweighted pair-groups method using arithmetic averages) was used to examine the compositional similarity of the herbaceous layer at each study site [15]. Presence/absence data was used in classification analysis, and Sorensen’s index - a binary measure - was chosen as the similarity measure. Since outliers may negatively impact cluster analysis [e.g., 16], only species present in ≥ 5 plots at a site were used in analysis. Results were summarized in a dendrogram in which floras of similar species composition are grouped most closely. UPGMA was done using MVSP version 3.14 [17]. Importance values were calculated for herbaceous layer plant species (relative frequency + relative cover/2) by site and sample period to provide a compositional summary of the vegetation [15]. Importance values were averaged within UPGMA community types and between sample periods to represent the species composition of the flora during the study. Nomenclature follows Gleason and Cronquist [18].
One-way analysis of variance (ANOVA) was used to examine differences in total BUs, percent of BUs browsed, percent total cover, and species richness (number of species/m2) across sites using individual plot values within sample periods. Tukey’s multiple range test was used to distinguish means when ANOVAs were significant. Repeated measures analysis of variance (RANOVA) was used to examine differences in total BUs, percent of BUs browsed, species richness, and total cover between sample periods. Data were transformed [arcsin transformation for percentage data; log (x + 0.5) for continuous data; square root (x + 0.5) for count data] prior to analysis to meet normality assumptions of parametric tests [19] and back-transformed for tabular presentation. Fisher’s exact test was used to determine whether individual plant species were browsed to a greater or lesser extent than the population as a whole by site and date [19]. Statistical analyses were done using SYSTAT version 7.0 [20].
Significance was accepted at p < 0.05 for all analyses.
Results
Herbaceous layer attributes
The UPGMA dendrogram showed a distinct separation of the herbaceous layer floras of the five study sites into two main clusters or community types (Fig. 1). The Pumpjack and Waid Run sites formed one UPGMA community type; the Irwin Run, Sibbald Run, and Turnup Run sites formed the other. Based on importance values, the Pumpjack - Waid Run community type was strongly dominated by the ferns Thelypteris noveboracensis and Dryopteris intermedia (Table 1). In contrast, the Irwin Run - Sibbald Run - Turnup Run community type had a more equitable species distribution and was not strongly dominated by one or a few species. The forbs Aster prenanthoides, A. divaricatus, and Tiarella cordifolia, and the grass Glyceria melicaria, were among the more common species in the herbaceous layer of the Irwin Run - Sibbald Run - Turnup Run community type.
The total number of species encountered in the herbaceous layer flora at the five riparian sites ranged from a low of 28 species at Pumpjack in September to a high of 67 species at Sibbald Run in July (Table 2). Mean overall floral richness across study sites was similar between the July (48.6 + 6.6 SE species) and September sample periods (46.6 + 6.8 species SE). Mean species richness at the plot level (number of species/m2) varied significantly among sites (RANOVA, F=4.512, df=4,15, p=0.014) but not between sample periods (F=2.802, df=1,15, p=0.115). The site by sample period interaction was not significant (F=0.584, df= 4,15, p=0.679). Within sample periods, species richness at the plot level varied significantly across sites during both sample periods (ANOVA, July: F=5.326, df=4,15, p=0.007; September: F=3.540, df=4,15, p=0.032).
Percent total groundcover did not vary significantly across sites (RANOVA, F=1.358, df=4,15, p=0.295) and was marginally non-significant between sample periods (F=4.073, df=1,15, p=0.062). The site by sample period interaction was not significant (F=0.640, df=4,15, p=0.642). Within sample periods, percent total cover did not vary significantly across sites in either the July or September sample periods (ANOVA, July: F=1.194, df=4,15, p=0.353; September: F=1.184, df=4,15, p=0.358). Total cover ranged from a low of 29.2% at Waid Run in September to a high of 57.5% at Irwin Run in September (Table 2).
Density of total BUs varied significantly across sites (RANOVA, F=5.15, df=4,94, p=0.0001) and between sample periods (F=81.182, df=3,11, p=0.0001). The site by sample period interaction was also significant (F=11.95, df=3,12, p=0.0001). Total BUs ranged from a high of 179.6 units/m2 at Pumpjack to a low of 74.4 units/m2 at Waid Run, both in July (Table 3). Density of BUs typically decreased between the July to September sample periods, probably due to the senescence of relatively abundant annual species such as Impatiens capensis and Pilea pumila.
Browsing intensity
Browsing intensity was generally low to moderate across study sites during both sample periods. The percentage of available units browsed varied significantly across sites (RANOVA, F=3.412, df=4,15, p=0.036) and sample periods (F=5.009, df=1,15, p=0.041). The site by sample period interaction was not significant (F=1.463, df=4,15, p=0.262). Percent browsing ranged from a high of 34.0% at Irwin Run during July to a low of 13.6 % at Sibbald Run in September (Table 3). Mean overall percentage of the total flora browsed across study sites was higher in July (24.5 + 3.0% SE) than in September (17.9 + 1.7% SE).
Browse preference
Forty-four plant taxa sustained some level of browsing by white-tailed deer across study sites and between sample periods (Table 4). Because some browsed plants occurred at very low densities, calculation of p-values for Fisher’s exact test was suspect for these species due to sparsely fitted cells [20]. Thus, only 17 taxa occurred at densities that could be used in meaningful analysis across study sites and sample periods. Seven species were browsed in significantly greater proportion (Fisher’s exact test, p<0.05) than the population of herbaceous layer plants as a whole within at least one sample period and study site: Aster divaricatus, A. prenanthoides, Chelone glabra, Impatiens capensis, Pilea pumila, Polygonum virginianum and Ranunculus hispidis. Five taxa were browsed in significantly lower proportion (Fisher’s exact test, p<0.05) than the population of herbaceous layer plants as a whole over at least one sample period and site: Clematis virginiana, Galium spp., Oxalis stricta, Viola spp., and Acer rubrum. The UPGMA community type that included the Irwin Run, Sibbald Run and Turnup Run study sites had by far the greatest number of occurrences (11) in which certain plant species were browsed in significant proportion. In contrast, there was only a single occurrence of significant browsing on Pilea pumila during September at Waid Run for the UPGMA community type that included also the Pumpjack site.
Discussion
Based on past studies that documented white-tailed deer overabundance and heavy browsing in Allegheny High Plateau forests [1 - 6], we expected that browsing intensity in the herbaceous layer of riparian forests would be greater than what we observed. Most previous studies found sufficient evidence to suggest that deer had a negative impact on the structure of regional forest ecosystems. For instance, it has been suggested that excessive deer browsing in regional forests has resulted in the creation of black cherry monocultures, reduced height growth of hardwood trees and shrubs, and degradation of understory habitat and cover for wildlife [1,6,21].
Since riparian forests of the region support relatively rich plant communities, we expected that deer would focus their foraging in these resource-rich areas. Studies suggest that if provided the opportunity, deer typically prefer a diverse assemblage of plants on which to forage [22 - 23]. The benefits of foraging on a diverse array of plants for deer are many, including enhanced weight maintenance during winter, increased fawn survival, greater opportunity to find and select more nutritious foods, and dilution of compounds which inhibit digestion [22 - 23]. Thus, communities with a greater diversity of plants may provide nutritional or energetic benefits that should be preferred by deer. The results of our study provide some evidence that deer may forage preferentially in more species-rich herbaceous layers of Allegheny High Plateau riparian forests. For example, we generally observed greater browsing intensity by deer at our study sites with a more species-rich flora, particularly at Irwin Run, Sibbald Run and Turnup Run. Moreover, these sites also supported a greater number of preferred browse species such as Aster divaricatus, A. prenanthoides and Chelone glabra than those study sites with a less diverse flora (e.g., Pumpjack and Waid Run).
Ferns (e.g., Dryopteris intermedia, Thelypteris noveboracensis) and grasses (e.g., Brachyelytrum erectum, Glyceria melicaria), unpalatable to deer, were among the dominant herbaceous layer plants found at most of our riparian study sites, especially the Pumpjack and Waid Run sites. Abundance of Rubus spp. was also very low across study sites. Deer exclosure and enclosure studies have shown a shift from high amounts of Rubus spp. to grass and fern dominated vegetation in areas with high deer densities [1 - 6]. These studies found that interference by ferns and grasses, brought about by high deer densities, negatively influenced the establishment of hardwood reproduction. Ferns and grasses reduce light levels below their canopies [24] and may also produce potential allelopathic substances that affect germination and growth of plant species [25 - 26]. Also, the litter layer below ferns is often deeper than in areas free of ferns, which may prevent seeds from reaching moist mineral soil necessary for germination and seedling establishment [24]. Furthermore, a dense fern stratum can provide habitat and shelter for seed and seedling predators, which can reduce densities of seedlings [e.g., 27]. It may be that high cover of ferns and grasses, perhaps induced by deer browsing, suppressed more palatable herbaceous plants at some of our study sites, particularly Pumpjack and Waid Run.
A shift in dominant vegetation types exemplifies the notion of alternate stable states in plant communities [4 - 5]. Stromayer and Warren [4] suggested that in order for alternate stable states to occur, there must be a suppression of plant species over a long period of time. Furthermore, Westoby et al. [28] concluded that for a transition in stable states to occur in a community, a combination of climatic or management actions are usually required. On the Allegheny High Plateau, the combination of historical timber harvest followed by heavy deer browsing is an example of management actions having effects over long periods of time. Thus, the production of alternate stable states in regional plant communities is likely [6]. However, unequivocal evidence for deer-induced alternate stable states at our riparian study sites is difficult to attain, as long-term monitoring is required to document compositional and structural changes [4]. In particular, we lack historical data on plant species composition at our study sites to establish a baseline for deer browsing impact. Deer exclosure studies could also help identify a shift in stable states, but to our knowledge none have yet specifically targeted riparian areas. If alternate stable states exist for the herbaceous layer vegetation of Allegheny High Plateau riparian forests, this may explain, in part, the low levels of deer browsing observed at some of our study sites, particularly the fern-dominated Pumpjack and Waid Run sites.
The configuration of the surrounding landscapes about our riparian study sites may have had an influence on deer foraging activity. Our review of aerial photographs of the study areas revealed a noticeable patchiness of the forested landscape. Habitat patches included tornado blowdowns, clearcuts and thinnings, agricultural fields, residential areas, and oil and gas fields. For example, within a half-kilometer of the Pumpjack site there is a recent clearcut with fresh browse and heavy Rubus cover, and oil and gas wells also perforate the area. Sibbald Run is in close proximity to a large tornado blowdown as well as several hayfields. Browse-rich patches are known to cause deer congregations due to availability of browse and forage [1]. Landscape patchiness creates a network of edge habitat preferred by deer [29]. The fact that recent forest harvesting practices have excluded riparian zones means that the immediate riparian areas themselves may have less induced edge habitat than adjacent managed uplands. The combination of landscape characteristics mentioned above may result in upland areas that are significantly more attractive to deer. Further research at varied spatial scales is needed to assess the influence of landscape diversity on deer foraging activity in Allegheny High Plateau forests [e.g., 30].
In conclusion, the results of our study show that herbaceous layer plant species richness and composition, and browsing intensity by white-tailed deer, are highly variable across riparian forests of the Allegheny High Plateau. Thus, in order to assess deer use of regional riparian forests, it appears that each riparian system must be studied individually [e.g., 31], perhaps on a watershed basis, as adjacent habitats within the landscape matrix may influence deer density and activity in certain riparian areas in an individualistic manner.
Acknowledgments
We thank Beth Brokaw, Rachael Cook, Sharon Krock, April Moore, and Amanda Saul for field assistance. Comments of anonymous reviewers improved the manuscript. This study was supported in part a participating agreement between the USDA Forest Service, Allegheny National Forest, and Clarion University of Pennsylvania.