Comparing the plant diversity between artificial forest and nature growth forest in a giant panda habitat

Artificial restoration is an important way to restore forests, but little is known about its effect on the habitat restoration of the giant panda. In the present study, we investigated the characteristics of artificial forest in the Wanglang Nature Reserve to determine whether through succession it has formed a suitable habitat for the giant panda. We compared artificial forest characteristics with those of natural habitat used by the giant panda. We found that the dominant tree species in artificial forest differed from those in the natural habitat. The artificial forest had lower plant species richness and diversity in the tree and shrub layers than did the latter, and its community structure was characterized by smaller tree and bamboo sizes, and fewer and lower bamboo clumps, but more trees and larger shrub sizes. The typical community collocation of artificial forest was a “Picea asperata + no-bamboo” model, which differs starkly from the giant panda’s natural habitat. After several years of restoration, the artificial forest has failed to become a suitable habitat for the giant panda. Therefore, a simple way of planting individual trees cannot restore giant panda habitat; instead, habitat restoration should be based on the habitat requirements of the giant panda.

of the earliest nature reserves established for protecting giant panda 16 . Our main objective was to investigate the characteristics of the artificial forest to determine whether through succession it has successfully approximated suitable giant panda habitat. We were more interested in the tree and bamboo vegetation of the forest because they are the key components of giant panda habitat 17 , and thus, they affect the habitat selection of the giant panda 18 . We hope this timely field study could provide an important reference tool for practices to restore habitat of the giant panda and its co-occurring wild animals.

Results
Species composition. 32 woody plant species (belonging to 25 genera and 15 families) were recorded in the artificial forest plots, and 32 woody plant species (belonging to 24 genera and 15 families) were recorded in the giant panda habitat plots.
14 tree species (belonging to 11 genera and 10 families) were recorded in the artificial forest plots, whereas 23 tree species (belonging to 19 genera and 13 families) were recorded in the giant panda habitat plots. 8 tree species were common to these two types of plots, for which the similarity coefficient was 0.43. The dominant tree species in the artificial forest plots was Picea asperata, which differed from the dominant ones in the giant panda habitat plots (Table 1). 24 shrub species (belonging to 20 genera and 10 families) were recorded in the artificial forest plots, and 28 shrub species (belonging to 21 genera and 14 families) were recorded in the giant panda habitat plots. 14 shrub species were common to these two types of plots, for which the similarity coefficient was 0.54. Philadelphus incanus was the dominant shrub species in both the artificial forest plots and giant panda habitat plots (Table 1).
19 regeneration species (belonging to 16 genera and 11 families) were recorded in the artificial forest plots, and 16 regeneration species (belonging to 15 genera and 10 families) were recorded in giant panda habitat plots. 10 regeneration species were common to these two types of plots, for which the similarity coefficient was 0.57. Lonicera microphylla, Ribes glaciale, and Berberis sichuanica were the dominant regeneration species in both the artificial forest plots and giant panda habitat plots (Table 1).
Furthermore, the same bamboo species (Fargesia denudata) was recorded in artificial forest plots and giant panda habitat plots.
Species richness and diversity. The species richness and Shannon-Wiener indexes of tree and shrub in artificial forest plots were all significantly lower than those of the giant panda habitat plots (all P-values < 0.05, Table 2). However, for the regeneration species, these two variables were not significantly different (P-values > 0.05, Table 2). Community structure. 6 of the 10 variables differed significantly between artificial forest plots and giant panda habitat plots: tree number, tree size, shrub size, bamboo clump number, bamboo clump height, and bamboo size (all P-values < 0.05, Table 3). As compared to the giant panda habitat, the artificial forest was characterized by smaller tree size and bamboo size, fewer and lower bamboo clumps, but more trees and a larger shrub size (Table 3).

Community collocation.
In the artificial forest plots, Picea asperata was the important constructive species, as it appeared in 91.7% (11 of 12) of the artificial forest plots, whereas in the giant panda habitat plots, Abies faxoniana and Betula albosinensis were the important constructive species, as they appeared in 66.7% (8 of 12) and 41.7% (5 of 12) of these plots, respectively (Table 4). Meanwhile, 75.0% (9 of 12) of artificial forest plots had no bamboo in them, whereas every giant panda habitat plot sampled showed the presence of bamboo (Table 4).

Discussion
Many factors can potentially affect habitat selection by animals 19 , such as food availability, water resources, quality shelter, places to hide, etc. Previously, we have found that the giant panda did not use the artificial forest habitat in the Wanglang Nature Reserve 15 . In this study, we still did not find the entity of giant panda and any trace of giant panda activity in the artificial forests and artificial forest plots, but we found a significant difference between the artificial forest and giant panda habitat in terms of plant species composition, community structure, and community collocation, which together may explain the giant panda's lack of use of artificial forest in this area.
From the perspective of species composition, artificial forest had moderate similarity with the giant panda habitat in different layers after several years of succession. However, as the main planted tree species in Wanglang, Picea asperata was still the dominant species in the tree and shrub layers of artificial forest, which was different from the giant panda habitat, indicating that no substantial change has yet occurred in the tree and shrub layers of artificial forest. Furthermore, the richness and diversity of tree and shrub in artificial forest were all lower than those of giant panda habitat, suggesting that the species composition of artificial forest is simpler, which may have been driven by the planting of a single tree species. Thus, in spite of many years of succession, from this perspective, artificial forest is still very different from giant panda habitat. In addition, it is worth noting that the richness and diversity of regeneration in artificial forest were all had no significant differences with those of natural giant panda habitat, which may indicate that the artificial forest could on its way to restore naturally to the natural forest after a relatively longer time.
From the perspective of community structure, artificial forest differed significantly from giant panda habitat, except in the regeneration layer. Planting individual trees at a high density is the typical approach to afforestation; thus, it is no surprise that the tree number in artificial forest was more than twice that found in giant panda habitat. Trees with a larger size are more likely to provide suitable habitat for the giant panda; for example, both branches and under-the-tree environments are used as resting sites by the giant panda 16 . Although more trees   were grown in the artificial forest, their size was significantly thinner than those trees growing in the giant panda habitat; to support pandas, so those thinner trees need further growth. Furthermore, the growth of shrubs in artificial forest was better than in the giant panda habitat plots, most likely because interspecific competition was weaker in the former where bamboo was small or absent. Bamboo is nearly the sole food resource of the giant panda, which reportedly can eat approximately 23-38 kg bamboo shoots per day 20 ; therefore, an environment with sufficient bamboo is essential to qualify as giant panda habitat. However, in the present study, we found that bamboo was scarce in artificial forest, and even where it was present, its growth was very poor. Therefore, it is difficult to imagine how the giant panda can effectively use this kind of habitat, one that clearly cannot meet its food requirements. "Tree-bamboo-giant panda" is an organic whole 17 , in that the typical giant panda habitat is usually described as consisting of tall trees and dense bamboo. It follows that tree and bamboo should be considered the important vegetation components of giant panda habitat. For the trees, the main constructive species of the artificial forest plots was Picea asperata, which differed completely from the constructive species of local giant panda habitat, such as Abies faxoniana, Betula albosinensis, Picea purpurea, etc. For the bamboo, 75.0% of artificial forest had no bamboo. For the three artificial forest plots that had bamboo, two of them were planted artificially and only one had regenerated naturally, and the number of bamboo clumps was only six. Presence of little or no bamboo does not characterise a suitable giant panda habitat. Thus, the current community collocation of artificial forest was mainly characterized by a "Picea asperata + no-bamboo" model (Table 4). Although the population structure of Picea asperata suggests that it is in decline, individuals with DBH at 5-20 cm in tree size still occupied >70% of the total population (Fig. 1). Hence, the situation of Picea asperata growing in artificial forests as the constructive species may not change in the short term. Furthermore, considering that the recovery of bamboo generally takes approximately 20-30 years 11 , it can be inferred that no major changes are likely to occur in artificial forest under natural conditions in the near future.
In this study, we found several striking differences between artificial forest and giant panda habitat in terms of their species composition, richness and diversity, community structure, and community collocation. Based on these results, we conclude that after several years of succession, the existing artificial forest in Wanglang does not provide a suitable habitat for the giant panda. Furthermore, according to our previous experience and the current status of artificial forest, we predict that no substantial changes would occur in the artificial forest in the near future. Consequently, a simple form of tree-planting cannot restore the giant panda habitat; instead, habitat restoration practices should be based on the habitat requirements of the giant panda.

Methods
Study area. The field work was carried out in the Wanglang National Nature Reserve, Pingwu County, Sichuan Province, China (103°50′-104°58′E, 32°49′-33°02′N). As an important giant panda habitat in the north Minshan Mountains 21 , Wanglang was established in 1963, covering an area of 32,297 hm 2 4 at an elevation ranging from 2,320 m to 4,891 m. The annual average precipitation at this nature reserve is approximately 862.5 mm, the lowest mean air temperature reached is −6.1 °C in January and the highest reached is 12.7 °C in July 22 . Fargesia denudata is the main bamboo species in this area.  To investigate the characteristics of natural giant panda habitat, we first selected 12 typical distribution areas of the giant panda in Wanglang Nature Reserve, according to previous studies 23 . In each area, one 20 m × 20 m plot that at least contained giant panda faeces was established to represent the giant panda habitat. The investigation method and time of sampling of these giant panda habitat plots were identical those used for the artificial forest plots.
A total of 24 plots were sampled for analysis (12 artificial forest plots and 12 giant panda habitat plots) (Fig. 2).

Data analysis.
To describe the characteristics of the general plant species composition of artificial forest and giant panda habitat, we first counted the total number of family, genus, and species of woody plants and bamboo in all plots of artificial forest and giant panda habitat, respectively. Subsequently, we counted the species composition and calculated the similarity coefficient of different vegetation layers (tree, shrub and regeneration layers). Lastly, we compared the dominant species-those having an importance value exceeding 0.05-composition of different layers between the two types of plots. To identify differences in plant species richness and diversity between the artificial forest and the giant panda habitat, we first counted the number of species and then calculated the Shannon-Wiener indexes of each artificial forest plot and giant panda habitat plot for the different layers. Then, for each layer, we compared the means of these two variables using the one-way ANOVA when statistical assumptions were met (the data were normally distributed, the variances were homogeneous) or Mann-Whitney U test when statistical assumptions were not met.
To detect the differences in plant community structure between the artificial forest and the giant panda habitat, we first defined and calculated 10 variables related to community structure according to the plot-level data (see Table 5). Subsequently, we used either the ANOVA or Mann-Whitney U test to separately compare the means of these variables between artificial forest plots and giant panda habitat plots.
To analyse the community collocation of artificial forest and giant panda habitat, we compared their constructive plant species from each plot; these were defined as species with a single or cumulative importance value  Table 5. Definition of the variables used in this study.
exceeding 0.50 in the tree layer of a plot. We also examined the presence of bamboo in each artificial forest plot and giant panda habitat plot.
In this study, we used the Sorensen index as the similarity coefficient 24 : (1) where a and b represent the number of species in artificial forest plots and giant panda habitat plots, respectively, c stands for the number of species common to both artificial forest plots and giant panda habitat plots. We used the Shannon-Wiener index as a measure of species diversity 24 : where P i stands for the proportion of the number of the i th species in each plot. We used the relative basal area, the mean values of relative density and relative frequency, and the mean values of relative density and relative height to represent, respectively, the importance values of species in the tree layer, shrub layer, and regeneration layer.
The statistical significance level used in this study was set at 0.05.