Short-Term Response of Native Flora to the Removal of Non-Native Shrubs in Mixed-Hardwood Forests of Indiana , USA

While negative impacts of invasive species on native communities are well documented, less is known about how these communities respond to the removal of established populations of invasive species. With regard to invasive shrubs, studies examining native community response to removal at scales greater than experimental plots are lacking. We examined short-term effects of removing Lonicera maackii (Amur honeysuckle) and other non-native shrubs on native plant taxa in six mixed-hardwood forests. Each study site contained two 0.64 ha sample areas—an area where all non-native shrubs were removed and a reference area where no treatment was implemented. We sampled vegetation in the spring and summer before and after non-native shrubs were removed. Cover and diversity of native species, and densities of native woody seedlings, increased after shrub removal. However, we also observed significant increases in L. maackii seedling densities and Alliaria petiolata (garlic mustard) cover in removal areas. Changes in reference areas were less pronounced and mostly non-significant. Our results suggest that removing non-native shrubs allows short-term recovery of native communities across a range of invasion intensities. However, OPEN ACCESS Forests 2015, 6 1879 successful restoration will likely depend on renewed competition with invasive species that re-colonize treatment areas, the influence of herbivores, and subsequent control efforts.


Introduction
As land managers with limited resources strive to counteract the negative effects of invasive species, context-dependent information about the invader and the invaded ecosystem becomes increasing critical.Specifically, an understanding of effective control strategies, spatial and temporal processes, and biotic and abiotic factors contributing to successful invasion are all important when prioritizing management actions.In the case of invasive plants, restoration efforts often focus upon the removal of invasives with the goal of allowing native species to re-establish.Evaluating the effectiveness of these efforts depends upon scientifically-valid before-and-after studies as opposed to relying on anecdotal evidence of suspected impacts [1].While some investigators have examined the effects of invasive plant management, a more comprehensive body of work is needed given the gap between scientific research and management programs [2].
Within forest ecosystems, woody invasive shrubs are particularly problematic because their longevity and persistence in the understory exerts considerable influence over herbaceous-layer processes, including forest regeneration [3].These species are often able to establish in forest understories in the absence of overstory disturbance [4].For example, Lonicera maackii (Amur honeysuckle), an invasive shrub introduced from Asia [5], has become a serious management concern in forests across much of eastern North America.Numerous studies have shown that this shrub negatively affects native plant communities by reducing the survival, reproduction, and growth of native species [6][7][8][9][10], inducing apparent competition [11], and driving decreased abundance and diversity of native flora [12].The control of this and other invasive shrubs has become a high priority for restoration, highlighting the need to evaluate the response of native plants to the removal of this widespread species.
Although the effects of L. maackii removal on native vegetation have been examined [13][14][15][16], most of these studies have employed small experimental removals; only two studies [15,17] treated areas larger than 40 m 2 .However, because invasions are typically treated at the forest or woodlot scale, examinations of larger-scale removal provides a more accurate evaluation of real-world environmental and dispersal conditions.For example, research has shown that immature L. maackii seedlings are typically clustered around mature shrubs [18], suggesting that residual shrubs that remain next to small treatment plots may hasten recolonization by serving as local sources of propagules or as perch sites for birds.Additionally, invasive shrub removal affects the abundance of mice (Peromyscus spp.; [19,20]), whose foraging for seeds has been shown to negatively affect the recruitment of tree seedling [21].The size of a removal area relative to the home range of an individual mouse could affect protective cover and foraging behavior.In addition, plant community diversity can vary spatially due to a multitude of biotic and abiotic factors [22]; therefore, pre-and post-treatment data are critical to understanding treatment effects.However, with the exception of [23], studies have lacked pre-removal data.
The primary objective of this study was to determine the response of native and non-native herbaceous and woody plants to the removal of L. maackii and other non-native shrubs across a gradient of invasion intensities and overstory compositions.Specifically, we examined before-and-after changes in the species diversity of native spring and summer herbaceous flora and woody seedlings in 0.64 ha areas where L. maackii and other non-native shrubs were removed and reference areas where non-native shrubs were left intact.In removal areas, we used a combination of mechanical and chemical treatments and removed all slash so that aboveground biomass of woody invaders was absent.We hypothesized that species diversity of native herbaceous plants and tree seedlings would increase following removal of non-native shrubs.We predicted a particularly dramatic increase for the vernal flora (herbaceous plants that flower primarily in the spring) given that L. maackii expands its leaves earlier than overstory trees, thus, directly competing with vernal herbs [6,24,25].We also hypothesized that the cover of Alliaria petiolata (garlic mustard) would increase following removal treatments given the presence of this species at our study areas prior to removing non-native shrubs.

Study Area
We collected data in six mature, second-growth mixed-hardwood forests in central Indiana (39°20′ N to 40°26′ N, and 86°57′ W to 87°26′ W; Figure 1 5) Ross Biological Reserve (Ross); and (6) Pursell Woodlot (Pursell).Study sites were on glacial landforms with out-wash and alluvial parent materials resulting in soil textures and drainage classes ranging from very poorly drained loams to excessively drained sandy loams (Table 1).Deer hunting was not allowed at four of our study sites; Fowler Park, Hawthorne Park, RR, and Ross.Sites were generally level with slopes of 5% or less.
Canopies across all study sites consisted of deciduous trees with ≥85% canopy closure.Lonicera maackii comprised >88% of non-native shrubs/ha across all sites and the last major canopy disturbance pre-dated L. maackii invasion.Our six study sites were chosen to represent a range of overstory species composition and a gradient of L. maackii density (Table 1).Honeysuckle density (stems ≥1.37 m height) ranged from 1042 stems•ha −1 at Ross to 3135 stems•ha −1 at FNR Farm (Table 1).

Experimental Design
We sampled vegetation in two, 80-m × 80-m areas (extending from the forest edge into interior) at each site.We placed one boundary of a given sample area along the forest edge, with the other three boundaries located ≥10 m from any edge.In one of the areas at each study site (chosen randomly), we removed L. maackii and all other non-native shrubs (hereafter removal area).While L. maackii was the dominant invasive species at all sites, we also removed all other non-native shrubs, which represents a more realistic management practice.In the other sample area, non-native shrubs were left intact (hereafter reference area).From November 2010 through March 2011, within removal areas we hand-pulled or cut all non-native shrubs at ground level using a gas-powered clearing saw and treated stumps with herbicide (20% Garlon 4 ® -triclopyr, 1% Stalker ® -imazapyr, and 79% Ax-it ® -basal oil).Large slash (diameter ≥ 5 cm) was removed from each site.Slash <5 cm diameter was scattered on the forest floor, but noticeable piles were not created.The purpose of removing the large slash was to create a condition reflective of future condition in the most effective restoration programs, where the largest individuals are cut during the first entry, re-sprouts and new seedlings are killed in successive years, and the slash from the first entry decays, eventually resulting in little or no invasive shrub biomass.

Data Collection
At each site, vegetation data were collected within fixed-area plots located along transects.Each removal or reference area contained three transects (spaced 20 m apart) extending from the forest edge into the interior.Along each transect, we placed four, 40-m 2 circular plots and four 2-m × 2-m square quadrats.Circular plots were spaced 20 m apart along each transect with the first plot 5-10 m from the forest edge.Each quadrat was randomly placed to the upper right, upper left, lower right, or lower left within the circular plot and oriented parallel to the transect.This design resulted in 12 circular plots and 12 quadrats per sample area.Table 1.Dominant cohort species (woody stems ≥ 10 cm dbh), soil types, and mean (± 1 SE) densities of L. maackii (stems ≥ 1.37 m height; defined as a group of basal stems originating from the same rootstock) in six mixed hardwood forests in Indiana.Overstory and L. maackii data were collected as described in [17].Overstory species listed in descending order of importance value (IV = [relative basal area + relative density]/2).Soil information was obtained from the USDA Web Soil Survey [27].In the circular plots, we recorded the number of woody individuals by species in the sapling layer (stems < 10 cm diameter at breast height (dbh) and ≥1.37 m tall).An individual was defined as a single stem or a clump of stems originating from the same point at ground level.In each quadrat, data were recorded in lower (≤1 m) and upper (1.01-5 m) strata.In the lower stratum, we recorded percent cover of native and non-native herbaceous and woody species, coarse woody debris (CWD; midpoint diameter ≥10 cm), fine woody debris (FWD; midpoint diameter <10 cm), dead leaves/dead herbaceous stems, and bare soil.We also recorded number of seedlings and shrubs (<1.37 m tall).In the upper stratum, we recorded percent cover of native and non-native trees, shrubs, and vines.All percent cover estimates were based upon categories modified from [28]: 0%-1%, 1.1%-2%, 2.1%-5%, 5.1%-10%, 10.1%-25%, 25.1%-50%, 50.1%-75%, 75.1%-95%, and >95%.Cover classes were estimated by a single observer to reduce bias.

Study Site
Quadrats in removal and reference areas were sampled in the spring (April) and summer (July-August) before (2010) and after removal (2011).Seedlings and shrub densities were tallied during the summer of each year.Circular plots were sampled in September of both years.Taxa not identifiable to species were categorized into multi-species groups or identified to genus.Nomenclature follows the United States Department of Agriculture Plants Database [29].

Data Analyses
For herbaceous-layer vegetation and substrate variables, we calculated percent cover by taxon using cover-class midpoints.Mean percent cover was also calculated by stratum, sample area (removal, reference), study site, season (spring, summer), and year (2010, 2011).For woody saplings, and for seedlings and shrubs, we calculated mean stems/ha by taxon within a plot, as well by sample area, study site, and year.
Quadrat-level taxonomic richness (S, number of taxa), Pielou's Evenness Index (J'), and Shannon's Diversity Index (H') were calculated based on the percent covers of native herbaceous and woody plants in the lower stratum.We also calculated mean H', S, and J' by sample area, study site, season, and year.For summer data, any native taxa recorded in the quadrats were considered.However, spring data calculations only included herbaceous species that flower primarily between March and June [30].
We used Wilcoxon signed rank tests [31] to examine differences between 2010 and 2011 in removal and reference areas for the following response variables: density of L. maackii shrubs in the sapling layer, density of L. maackii shrubs <1.37 m tall, percent cover of Alliaria petiolata in spring and summer, spring and summer native S, J', and H', density of native seedlings (species were combined as a group), percent cover of vegetation groups based on nativity and growth habit, and percent cover of substrate variables (CWD, FWD, dead leaves/herbaceous stems, bare soil).The USDA Plants Database [29] was used to create the following groups based upon nativity and growth habit: native forbs, native grasses, native sedges, native ferns, native vines, native trees and shrubs, non-native forbs, non-native grasses, non-native vines, and non-native shrubs.Important assumptions of the Wilcoxon signed rank test are (1) data are paired and pairs come from the same population; (2) each pair is chosen randomly and is independent of another pair; and (3) data need not be normally distributed, but must be ordinal.For data that do not adhere to parametric assumptions of normality (we used histograms to determine that data and differences were not normally distributed), such as those in this and many other ecological studies, it is more appropriate to use non-parametric procedures instead of parametric procedures, such as paired t-tests or ANOVA [32].The program R was used to perform Wilcoxon signed rank tests [33].We used the R package vegan [34] to calculate H', S, and J'.
Because of the large number tests performed, we used a graphically-sharpened procedure based upon control of the false discovery rate [FDR]; [35,36] to adjust p-values [37].Because they retain statistical power while keeping the proportion of false discoveries small relative to all significant results [37,38], multiple comparison techniques based upon FDR have become more widely used in ecological studies as an alternative to traditional controls of family-wise error rate.P-values of reference and removal areas were pooled separately for analysis and adjustment.

Response of Native Flora and Alliaria petiolata
Between 2010 and 2011, changes in cover of native vegetation groups in the lower-vertical stratum were generally greater and more positive in removal than reference areas in spring (Table 2) and summer (Table 3).In spring, cover of forbs in the lower stratum in reference areas did not change significantly (20.6 ± 6.5 in 2010 vs. 22.6 ± 6.9 in 2011; p = 0.22), but increased from 17.2 ± 3.8 to 31.0 ± 6.3 in removal areas (p = 0.04; Table 2).Summer forb cover was largely unchanged (20.9 ± 6.6 in 2010, 20.1 ± 6.3 in 2011; p = 0.78) in reference areas, but increased from 22.7 ± 5.2 to 37.8 ± 6.8 in removal areas (p = 0.04; Table 3).Changes in the upper stratum for native trees/shrubs and vines were less pronounced than those in the lower stratum (Table 2).

Table 2.
Mean percent cover (±1 SE) and Wilcoxon signed rank test results for native and non-native plants (grouped according to growth habit) and substrate variables in reference and removal areas, during the spring seasons of 2010 and 2011 at six mixed-hardwood forests in central Indiana.Although mean values are presented, Wilcoxon p values were calculated to examine significance of differences between 2010 and 2011, based on ranks.For each group/substrate variable, mean values were calculated from the six study sites (n = 6 for each species in a given treatment type, year, and season).Plant data were collected in the lower stratum (≤1 m) and upper stratum (1.01-5 m) whereas substrate data were collected only in the lower stratum.p-values were adjusted for multiple comparisons with a graphically-sharpened procedure to control the false discovery rate [35].Adjustments were made separately for reference and removal areas.T = trace (<0.1% cover).Native H', S, and J' generally increased from 2010 to 2011 in both spring and summer, but changes were greater in removal areas (Figures 3 and 4).During spring, H' (0.61 ± 0.41 in 2010 vs. 0.55 ± 0. 15 in 2011; p = 0.17), S (2.46 ± 0.48 vs. 2.68 ± 0.52; p = 0.17), and J' (0.54 ± 0.10 vs. 0.58 ± 0.11; p = 0.19) changed little in reference areas.In removal areas during spring, H' increased from 0.45 ± 0.13 in 2010 to 0.88 ± 0.10 in 2011 (p = 0.04).Species richness increased from 1.86 ± 0.37 to 3.43 ± 0.44 (p = 0.04), and J' increased from 0.43 ± 0.11 to 0.71 ± 0.03 (p = 0.04).Trends in the summer were similar to those of spring; H', S, and J' changed little in reference areas (Figure 4), but generally increased in removal areas; H' increased from 1.68 ± 0.19 in 2010 to 2.08 ±0.13 in 2011 (p = 0.04), S increased from 8.43 ± 1.47 to 13.71 ±1.04 (p = 0.04), but J' did not change significantly (p = 0.20).At all study sites, native seedlings as a group increased in response to the removal of non-native shrubs, whereas in reference areas, seedling densities changed little (Figure 5).In removal areas, seedling density increased from 12,847 ± 4361 stems•ha

b d c a
Alliaria petiolata cover also increased following the removal of non-native shrubs (Figure 7), but this varied with pre-treatment percent cover of the species across study sites.Alliaria petiolata was not detected in either sample area at RR or Hawthorn before or after removal treatments and was not detected in the reference area during spring or summer at Fowler (Figure 7).Alliaria petiolata did occur in all other sample areas, and changes in mean percent cover from 2010 to 2011 were greater in removal areas than reference areas (Figure 7).Cover in reference areas did not change significantly in spring (p = 0.77) or summer (p = 0.54).Conversely, in removal areas during the spring, cover increased from1.We also observed changes in some substrate variables.Mean cover of bare soil decreased in removal areas during both spring and summer (Tables 2 and 3).Conversely, mean percent cover of FWD increased from 2010 to 2011 in removal areas (p = 0.04 for both seasons; Tables 1 and 2).

Discussion
Our results suggest that removing Lonicera maackii and other non-native shrubs allows the recovery of native herbaceous and woody plant communities, at least in the short term.Removal areas exhibited significant increases in the cover and diversity of native species in both spring and summer, as well as increased densities of native seedlings, compared to reference areas.Furthermore, we observed significant decreases in the percent cover of bare soil in removal areas in spring and summer but no significant changes in the cover of dead leaves/dead herbaceous stems, indicating that growing space was being filled by native vegetation.These subsequent increases in native flora were documented at all six study sites, which represented a wide range of pre-treatment L. maackii densities.Thus, our results suggest that even at sites with invasions in the advanced expansion or saturation phases [3], control efforts can allow the recovery of native taxa.The depauperate ground layer (Figure 6a) that occurs during the saturation stage could lead a manager to conclude that the area is completely void of native vegetation with little hope for recovery in response to restoration efforts.However, removing L. maackii and other non-native shrubs at our sites caused dramatic increases in the abundance of native flora, particularly vernal herbs.Based upon a review of the literature, our study is the first to identify before-and-after effects across a range of forest compositions and invasion intensities in forests that contained little to no aboveground biomass from non-native shrubs following removal treatments.As such, our study adds to a limited, but growing, body of work that documents the recovery potential of post-treatment forests (e.g., [15,23]).
The major environmental change resulting from removing non-native shrubs was likely an increase in light availability near the forest floor.Increased light may have played a particularly important role for the recovery of native vernal herbs [6,9].Because L. maackii expands its leaves earlier in the year than other woody plants in its invaded range [5], it competes with spring-flowering herbs that utilize light available prior to canopy leaf out [39].This response is illustrated by the dramatic increases in H' and S we observed for vernal herbs after removal.
For herbaceous-layer taxa that increased in post-removal cover, dispersal and re-establishment undoubtedly occurred via a combination of sexual and vegetative reproduction [40].The most dramatic increases were observed for taxa that exhibited high fecundity and effective dispersal mechanisms.For example, during the summer, generalist species such as P. pumila, Sanicula spp., and P. quinquefolia (Appendix A1) exhibited the greatest increases in cover.While this response by generalists is not surprising, their cover was quite low prior to removal, suggesting that invasions by L. maackii and other shrubs suppressed even the most resilient native herbs.Although L. maackii removal also allowed woody seedlings to establish, subsequent canopy disturbance will be necessary for most of these seedlings to recruit into the overstory.
Following removal of invasive shrubs, long-term recovery of native taxa will be largely influenced by competition with non-native species and herbivory by white-tailed deer (Odocoileus virginianus).As originally hypothesized, we observed a post-harvest influx of L. maackii seedlings and garlic mustard, indicating that undesirable outcomes may also result from control efforts.Given that L. maackii seeds are readily dispersed by birds [41], white-tailed deer [42], and possibly mice [43], effective management will require the continual control of new seedlings in treatment areas.Alliaria petiolata, like native taxa, responded to the increased light levels and possibly the soil disturbance associated with cutting non-native shrubs and dragging slash [44].Increased cover of A. petiolata following removal of L. maackii has been observed elsewhere when the species co-occur [13,15,16].
It is widely acknowledged that overabundant deer populations can cause declines in native taxa through intense herbivory [45], and help perpetuate the dominance of invasive species by preferentially browsing on native plants [46,47].While we observed a strong positive response of native plant species to L. maackii removal, this increased herbaceous cover, in combination with the removal of a thick shrub layer that reduced access, may lead to increased herbivory of native species and increased dominance of A. petiolata [16,48,49].This negative interaction between deer herbivory and invasive plant species could be particularly problematic following shrub removal in natural areas where hunting is prohibited.

Conclusions
Our results suggest that implementing control of invasive shrubs may allow for the recovery of native plant species, even in forests with heavy densities resulting from long-term invasions.However, managers implementing invasive control in forests should understand that removal treatments may result in increased cover of both native and exotic plant species.Given the increased density of L. maackii seedlings after removal, as well as increased A. petiolata cover, long-term control efforts will be necessary to sustain the recovery of native communities.Long-term efforts to control invasive plants, such as L. maackii, should balance reclaiming heavily invaded areas with preventing areas with more recent invasions from reaching the saturation phase of population growth [18].

Figure 1 .
Figure 1.Location of study sites within Indiana, USA. 1 = Hawthorn Park, 2 = Fowler Park, 3 = Rifle Range, 4 = FNR Farm, 5 = Ross Biological Reserve, 6 = Pursell Woodlot.Study sites represented by a single symbol were too close together to separate at a state-wide scale.Shading on map represents forest canopy cover as derived from the National Land Cover Database [26].

Figure 2 .
Figure 2. Mean Lonicera maackii sapling-layer densities (shrubs/ha for shrubs ≥ 1.37 m tall) in reference areas (a) and removal areas (b) and seedlings/ha (shrubs < 1.37 m tall) in reference areas (c) and removal areas (d), in 2010 and 2011, at six mixed-hardwood forests in central Indiana.Error bars are ± 1 SE.

Figure 3 .
Figure 3. Mean Shannon's Diversity Index (H') in reference areas (a) and removal areas (b); mean taxonomic richness (S) in reference areas (c) and removal areas (d), and mean Pielou's Evenness Index (J') in reference areas (e) and removal areas (f) during the spring seasons of 2010 and 2011 at six mixed-hardwood forests in central Indiana.Error bars are ±1 SE.Indices were based on percent cover data from native herbaceous species that flower primarily during the spring.

Figure 4 .
Figure 4. Mean Shannon's Diversity Index (H') in reference areas (a) and removal areas (b); mean taxonomic richness (S) in reference areas (c) and removal areas (d); and mean Pielou's Evenness Index (J') in reference areas (e) and removal areas (f) during the summer seasons of 2010 and 2011 at six mixed-hardwood forests in central Indiana.Error bars are ± 1 SE.Indices were based on percent covers of all native herbaceous and woody taxa encountered during the summer.

Figure 6 .
Figure 6.(a) Lonicera maackii thicket showing no native plant regeneration on the forest floor; (b) Trillium recurvatum (bloody butcher); (c) Podophyllum peltatum, and (d) Alliaria petiolata displayed positive responses to the removal of non-native shrubs.All photos were taken at Purdue University, Department of Forestry and Natural Resources Farm (FNR Farm), by Michael Jenkins.

Figure 7 .
Figure 7. Mean percent cover (±1 SE) of Alliaria petiolata during the spring in reference areas (a) and removal areas (b); and during the summer in reference areas (c) and removal areas (d); in 2010 and 2011, at six mixed-hardwood forests in central Indiana.

Table 3 .
[35] percent cover (±1 SE) and Wilcoxon signed rank test results for native and non-native plants (grouped according to growth habit) and substrate variables in reference and removal areas, during the summer seasons of 2010 and 2011 at six mixed-hardwood forests in central Indiana.Although mean values are presented, Wilcoxon p values were calculated to examine significance of differences between 2010 and 2011, based on ranks.For each group/substrate variable, mean values were calculated from the six study sites (n = 6 for each species in a given treatment type, year, and season).Plant data were collected in the lower stratum (≤1 m) and upper stratum (1.01-5 m) whereas substrate data were collected only in the lower stratum.p-valueswereadjusted for multiple comparisons with a graphically-sharpened procedure to control the false discovery rate[35].were made separately for reference and removal areas.T = trace (<0.1% cover).