Vegetation Recovery in Response to the Exclusion of Grazing by Sika Deer ( Cervus nippon ) in Seminatural Grassland on Mt . Kushigata , Japan

We examined the recovery of vegetation in seminatural grassland in central Japan after eliminating grazing by sika deer (Cervus nippon) by fencing. By 2012, after 5 years of fencing for exclusion of sika deer, the species composition of quadrats within the enclosure reverted to the original species composition in 1981, not browsed by sika deer. Conversely, outside the fence was different from the baseline quadrats in 1981. Iris sanguinea, a prominent flower in the area, recovered within the enclosure, while it continued to decrease with grazing outside the fence. Nevertheless, the I. sanguinea cover had not recovered to the 1981 levels in the enclosure. Fencing can effectively restore vegetation as the species compositionwithin the enclosure gradually reverts to the original vegetation. Preventing grazing in intensively grazed seminatural grassland might lead to different successional pathways. Since I. sanguinea did not recover fully within the enclosure and the species composition differed slightly from the original vegetation, this suggests that the vegetation within the enclosure will change to an alternative state. Therefore, different management is needed to promote the correct succession pathways for ecological restoration, perhaps by enhancing the colonization of target species, to prevent restored sites from giving rise to alternative states.


Introduction
Traditionally, seminatural grasslands in Japan have been managed for grazing cattle and harvesting agricultural materials (e.g., fertilizer; [1]).However, most of these have been abandoned with changes in lifestyle and farming methods [2], as in other countries (e.g., [3]).In Japan, forests cover 78% of the land area [4] and the mild, humid climate promotes the growth of forests [5].Consequently, abandoned grasslands are invaded by trees and are important especially for early successional species and species favoring open habitats (e.g., [6]).So, such seminatural grassland has a high priority to conserve biological diversity.
Recently, the population of sika deer (Cervus nippon) in Japan has increased sharply because of the declining numbers of hunters and their aging, extinction of predator, and so on [7,8].The effects of deer on natural grassland vegetation have been reported worldwide (e.g., [9]).In Japan, deer have had serious effects on natural forests [10,11], plantations [12,13], and grasslands [14,15].Since sika deer prefer open habitats as foraging sites [16], patchy grasslands surrounded by forests are used heavily by sika deer.Consequently, the plant species composition in the grasslands has been altered markedly by grazing sika deer.
On Mt.Kushigata (2053 m a.s.l. at the summit) in central Japan, seminatural grasslands are distributed patchily with plantations and fragmented natural forests [17].In the grassland, the flowers of Iris sanguinea and other meadow species (e.g., Veronicastrum japonicum) had been renowned.However, those flowers have been decreasing since 2000 because of grazing by sika deer.As countermeasures for the grazing, fences were set in the grassland in 2007.Fencing is effective for recovering from herbivore grazing [18,19] and promotes tree regeneration [20,21].The effects of grazing exclusion 2 ISRN Biodiversity from seminatural grasslands on species richness [22] and spatial patterns [23] have been studied.However, undesirable results, in which execution at ecological restoration might lead to alternative states [24], can occasionally occur when fencing is used as a tool for conservation in overgrazed environments [25].Therefore, restoration efforts should be checked by comparing the results with the restoration target [3,4].
This paper presents the vegetation recovery pattern in seminatural grassland in central Japan after preventing grazing with fencing and addresses the following question: can I. sanguinea and a vegetation community recover to the original situation and community before they were affected by sika deer?On Mt.Kushigata, seminatural grasslands are distributed patchily with plantations and fragmented natural forests [17].The origin of the grassland is not clear, but elevations in this region around 2000 m are typically dominated by subalpine coniferous forests composed of Abies and Tsuga species.The grassland was probably the result of human activity, such as mowing or burning.We studied the Hadakayama area of Mt.Kushigata.Hadakayama means "naked mountain, " which indicates that this area has not been covered by forest for a long time.The grasslands in the Hadakayama area are renowned for the flowers of I. sanguinea.Over a 10-year period, however, the number of I. sanguinea flowers has decreased, possibly as a result of the natural succession from grassland to the typical subalpine coniferous forest.Grazing by sika deer was another potential reason.Therefore, fences to exclude sika deer were erected to prevent grazing in 2007.
In October 2007, we established a 20 × 20 m plot and fenced half of it (10 × 20 m) to protect it from deer grazing.The quadrats set by Koma High School were 20 m from the plot.In June 2008, we established 20 1 × 1 m quadrats, 10 inside and 10 outside the fence, and conducted annual surveys from 2008 to 2012 using the Braun-Blanquet method.We counted the number of individuals of I. sanguinea in each quadrat every year.[6] did not list all rare species (i.e., species with low coverage in the quadrats studied).Therefore, we analyzed the species with coverage with a score >1 on the Braun-Blanquet scale.

Analysis. Koma High School
We used nonmetric multidimensional scaling (NMS; [27]) to provide an ecologically interpretable quantification of the compositional differences among original vegetation (1981) and quadrats inside and outside the fence (2008)(2009)(2010)(2011)(2012).NMS applied Sørensen's similarity index to calculate a distance matrix.We used the species cover in each quadrat for NMS after transforming the Braun-Blanquet scale quantitatively.The data transformed to cover values (the midpoints of the cover intervals for each score) were used; that is, scores of 1 to 5 were converted to the values 2.5%, 12.5%, 37.5%, 62.5%, and 87.5%, respectively [26].NMS was performed using PC-ORD [27].
To show the recovery of I. sanguinea, we compared the number of individuals and cover.

Results
The recovery of I. sanguinea inside the fence was good, while outside the fence, it decreased continuously with grazing (Figure 1).The number of flowering I. sanguinea also increased inside the fence to 25 in 2010, 198 in 2011, and 307 in 2012, while no flowers occurred outside the fence (unpublished data, Committee of Conservation of I. sanguinea at Mt. Kushigata).Nevertheless, the I. sanguinea cover had not recovered fully compared to 1981 (Figure 2).
The changes in the number of species inside and outside the fence showed contrasting trends (Figure 3).The number of species increased inside the fence, but not outside it.The species composition differed markedly inside and outside the fence (Figure 4).In 2008, most of the quadrats were located in the upper left position in the NMS diagram.Then, the species composition of the quadrats within the fence shifted to the upper right position in the diagram, where the species composition in 1981 was located (i.e., not browsed by sika deer).Conversely, the species composition of the quadrats outside the fence was shifted to the lower right.Species that occurred in more quadrats inside the fence were Dianthus superbus var.longicalycinus, Phedimus aizoon var.floribundus, Serratula coronata, and Chamerion angustifolium as well as I. sanguinea (Table 1).Angelica pubescens, Veronicastrum japonicum, and I. sanguinea, which were categorized by tall herbs, were only dominant inside the fence.Brachypodium sylvaticum and Ranunculus japonicas were less dominant before grazing but were dominant after exclusion of sika deer.Outside the fence, Artemisia princeps initially dominated, and subsequently the graminoids Stipa pekinensis and B. sylvaticum, which appeared to be unpalatable, dominated.

Disscusion
Exclusion of sika deer by fencing was successfully recovering the cover and number of individuals of I. sanguinea because those outside the fence were continuously low.Thus, effects of grazing by sika deer were continuously severe.In the seminatural grassland of northern Japan, Iris setosa dominated the most at sites grazed by horses, which improved the surface soil characteristics [28].In our study, sika deer grazed on I. sanguinea directly, causing serious damage.Thus, to conserve I. sanguinea, fencing seems necessary under the present circumstances.Since Tamura [29] showed that vegetation recovery, particularly tall herb species, was poor when fencing was delayed, it will be impossible for I. sanguinea to recover fully even if fences are erected now.However, I. sanguinea would not recover fully compared to situation before grazing despite preventing grazing.Fencing appeared to be effective at restoring the vegetation as the species composition in the fence gradually  reverted to the original vegetation.Herbivores often, but not always, increase plant diversity in grasslands [30].Outside the fence, however, since the species composition was altered and the number of species was low, intense grazing pressure by sika deer likely existed.
Rooney and Dress [31] showed that species with relatively lower abundance were more likely to be missing due to browsing than more abundant species.Actually, species with lower abundance in 1981 (e.g., Polygonatum odoratum, Ixeridium dentatum, and Hakonechloa macra) were not recovered even after fencing.By grazing, tall-growing herbs were reduced and lower-growing species were increased [32] and grazing-resisted species were shorter in height than grazingsusceptible species [33].Thus, tall herb species were tending to grazing and hard to recover after grazing.

Conclusions
Preventing grazing after intensive grazing of seminatural grassland might result in different successional pathways being followed and the species composition is slightly different from the original vegetation.This suggests that the vegetation inside the fence will change to an alternative state [24].Galvánek and Lepš [3] showed that the species composition of the restored plots after the reintroduction of mowing was still far from the target composition.Therefore, different management methods are needed to ensure the correct succession pathways are followed for ecological restoration and to enhance colonization of the target species [34], rather than the restored site resulting in an alternative state [24].Thus, other methods to restore the vegetation (e.g., removing of unpalatable recalcitrant species [35]) inside of the fence would be necessary.Moreover, as Wright et al. [36] suggested, complete removal of ungulates may be required for recovery in heavily browsed forest understory vegetation in New Zealand.Hence, the control of sika deer population should be required.

2 )Figure 1 :Figure 2 :
Figure 1: Changes of mean number of individuals of Iris sanguinea in each quadrat.Vertical bars showed standard deviation.

Figure 3 :
Figure 3: Changes of mean number of species that occurred in each quadrat.Vertical bars showed standard deviation.

Figure 4 :
Figure 4: Results of nonmetric multidimensional scaling (NMS) for species that occurred in each quadrat.
2.1.Study Site.The study was conducted on Mt.Kushigata (2053 m a.s.l. at the summit), Yamanashi Prefecture, in the cool-temperate zone of central Japan (35 ∘ 35  N, 138 ∘ 22  E).The mean annual precipitation and temperature at the nearest meteorological station (Oizumi, 867 m a.s.l.) are approximately 1140 mm and 10.7 ∘ C, respectively.The estimated mean annual temperature at the summit of Mt.Kushigata is about 3.5 ∘ C. Snow cover is usually less than 1 m.