Understory Structure and Vascular Plant Diversity in Naturally Regenerated Deciduous Forests and Spruce Plantations on Similar Clear-cuts: Implications for Forest Regeneration Strategy Selection

The active effect of natural regeneration on understory vegetation and diversity on clear-cut forestlands, in contrast to conifer reforestation, is still controversial. Here we investigated differences in understory vegetation by comparing naturally regenerated deciduous forests (NR) and reforested spruce plantations (SP) aged 20–40 years on 12 similar clear-cuts of subalpine old-growth spruce-fir forests from the eastern Tibetan Plateau. We found that 283 of the 334 vascular plant species recorded were present in NR plots, while only 264 species occurred in SP plots. This was consistent with richer species, higher cover, and stem (or shoot) density of tree seedlings, shrubs, and ferns in the NR plots than in the SP plots. Moreover, understory plant diversity was limited under dense canopy cover, which occurred more frequently in the SP plots. Our findings implied that natural deciduous tree regeneration could better preserve understory vegetation and biodiversity than spruce reforestation after clear-cutting. This result further informed practices to reduce tree canopy cover for spruce plantations or to integrate natural regeneration and reforestation for clear-cuts in order to promote understory vegetation and species diversity conservation.


Introduction
Species composition, diversity, and structure of understory vegetation are keys to providing complex structure and conserving indigenous floras within forests [1,2].The understory can provide habitat and food for faunal communities [3], and act as a driver of nutrient cycling [4], stand productivity [4,5], and forest regeneration and succession [6][7][8].Thus, the understory community and biodiversity are focal objectives for sustainable forest management, effective forest biodiversity conservation, and successful forest restoration [4,6,9].However, the effects of different strategic applications for forest regeneration on understory vegetation and species diversity on clear-cuts remains controversial, advocating further study [9][10][11][12].It is a challenge for forest managers to promote forest regeneration, while conserving indigenous biodiversity in a large area of clear-cut forestlands.
Natural regeneration and conifer reforestation on clear-cuts are two major regeneration strategies that have been long employed in the northwestern Sichuan Province, China [13], and globally [14].Natural regeneration without artificial reforestation often depends on remnant vegetation, its seed pool, and dispersals surrounding vegetation and involves the synchronous development of both native trees and other plant forms together with their abiotic environment.It usually leads to a mix of tree species and unevenly-aged individuals, which exhibit connectivity among their components and are self-organized into hierarchies and cycles [15,16].In contrast to natural regeneration, artificial reforestation schemes are designed with targets for establishing overstory structure and satisfying production demands, and they often repress vegetation with the potential to hinder target tree growth.They are also designed to achieve a stand of individuals that are even-aged and regularly spaced.Thus, artificial reforestation efforts usually do not include activities that are conducive to developing understory biodiversity [14].This is true depending on the extent to which a site is prepared for planting, which has not been addressed [11,12,17] up to now.A substantial body of research has compared species composition and diversity between coniferous plantations and naturally regenerated forests or secondary forests worldwide (for reviews see Brockerhoff et al., 2008;Bremer & Farley, 2010) [11,12].However, results vary and are even contradictory.Reforested plantations might have similar, or significantly lower or higher, vascular plant species richness and diversity of understory vegetation in comparison with naturally regenerated forests [10][11][12][18][19][20][21].Consequently, the effects remain unpredictable and differ according to the manner and intensity of disturbance from different regeneration pathways [11,12].
The conflicting results presented in the literature regarding species richness and understory vegetation structure are due to several inconsistent factors used when making comparisons: different historical origins of previous vegetation [12], distinct initial site condition [22,23], different successional stage or forest age [16,24,25], target tree identity and mixture [10,11], and site degradation intensity and management [11,12,14,22].To reach reliable conclusions on the effects of different regeneration methods on understory vascular plant composition and diversity requires controlling variability among study sites.Moreover, it is also necessary to consider the sampling design and the accompanying statistical methodology.During sampling design and the investigation of species composition, different scales (stand, plot, and quadrats) have long been widely used by researchers.Recently, systematical sampling has become more popular [25,26].Quadrats are often nested plots, and plots are often nested stands during the systematical sample process.However, these data are usually not independent [27,28] and may also be non-normal.These are troubling issues for many researchers who are used to applying independent tests in their studies.The generalized linear mixed-effect models (GLMMs) is an extension of generalized linear models (GLMs), including random effects to deal with correlated data structures, in particular, with clustered structures [28,29].This model also provides a more flexible approach for non-normal data [30].However, it has not been widely used in studies that utilize nested data to explore differences in understory vascular plant structure and biodiversity between naturally and artificially regenerated forests [29,31].
In the current study, we aimed to evaluate the understory community structure and plant composition over the same regeneration period in two stand types: naturally regenerated deciduous forest and planted spruce forest, originating from similar clear-cuts of old-growth spruce-fir forests in the eastern Tibetan Plateau.GLMMs were applied to explore the difference of understory vegetation between the two regenerating forests.We addressed three questions: (1) Which regeneration strategies result in higher vascular plant diversity in the forest understory: natural regeneration or spruce reforestation?(2) How do the vascular plant groups (tree seedlings, shrubs, ferns, forbs, and graminoids) differ in species composition, richness, and community structure between the understory of the two forests?(3) What are the differences in structures of the overstory and understory and their relationships between the two forests?We hypothesized that: (H1) The naturally regenerated forests would host higher species diversity compared to the planted spruce forests, with woody plant diversity being the key driver of the diversity difference; and (H2) the tree canopy structure of the two forests disparately influences the structure and species diversity of the understory.

Study Area
This study was conducted in the Aba Tibetan and Qiang Autonomous Prefecture (30°35′ N-34°19′ N, 100°30′ E-104°27′ E), which is an area of approximately 80,000 km 2 in the northwestern Sichuan Province, China.This region is located in the northeastern Hengduan Mountains region, a famous biodiversity hotspot known in China and worldwide.It is also part of the Southwestern National Forest Region in China [13].The forested elevations range from 2400 m to 3900 m, and the climate is temperate with an annual rainfall of 800-1000 mm and a mean annual temperature of 6-10 °C.The frost-free period in this region is less than 100 days.Mountain brown soil (luvisols) is the major soil type [32,33].
The old-growth coniferous forests are dominated by one to three species of firs (Abies spp.), spruces (Picea spp.), or larches (Larix spp.).They are widely distributed in the subalpine region and harbor a very high biodiversity with over 6000 species of vascular plants [13].There is large-scale clear-cut logging in the primary coniferous forests beginning in the 1960s and ending with the Natural Forest Protection Program in 1998.As a result of forest harvest, there are large numbers of clear-cuts with patch sizes of 3-10 ha throughout the national forestland in the studied regions.Most clear-cuts were reforested with a single native species, spruce (Picea asperata Mast.), in accordance with the manual [32].Four-year-old spruce seedlings were planted at an initial density of at least 3300 stems per hectare.Once or twice in the initial two to five years following, planting management activities, including weeding and cutting shrubs, were employed to reduce competition and promote target seedling growth.After these initial treatments, no further management was applied.During this period, some clear-cut patches left to regeneration naturally succeeded towards deciduous broad-leaved forests dominated by Betula albo-sinensis Burk., B. platyphylla Suk., Acer mono Maxim., A. maximowiczii Pax, A. davidii subsp.grosseri (Pax) P. C. de Jong, Populus davidiana Dode, Sorbus koehneana Schneid, S. setschwanensis Schneid, and S. hupehensis Schneid [13,33].As a result, the study area is a mosaic of spruce plantations of various ages and naturally regenerated forest patches [13].The naturally regenerated deciduous broad-leaved forest accounts for approximately 25% of the forested area in the region, and the reforested spruce forest accounts for >40% [13].The large area of secondary forest allowed us to select paired stands of the same ages, one reforested and the other resulting from natural regeneration, to explore the difference of understory structure and vascular plant diversity between the two forest regeneration strategies.

Field Investigation and Data Collection
The field investigation occurred in the summers of 2006 and 2007.Forest management records from local forest management centers were used to select suitable sites with paired stands of reforested spruce forest (SP) and naturally regenerated forest (NR).Each pair originated from similar clear-cuts with a same harvested time and with a similar topography.Twelve sites (each with a pair) from three counties with 20-40-year-old stands were selected.We systematically set three plots in the NR and three plots in the SP stands at each site.Evaluation, aspect, and slope were measured for each plot, and the stand age was also recorded according to forest management records.Each plot was the same size, 20 m × 20 m.We further systematically placed nine 2 m × 2 m shrub quadrats to investigate shrubs, and nine 1 m × 1 m herb quadrats were fixed to the upper left corner of each shrub quadrat to investigate herbs in each plot.Overall, 12 stands, including 36 plots, 324 shrub quadrats and 324 herb quadrats from naturally regenerated sites, and another paired 12 stands, also including 36 plots, 324 shrub quadrats, and 324 herb quadrats from reforested spruce plantations, were examined.
We defined tree canopy cover as the proportion of the forest floor covered by the vertical projection of tree crowns and carefully estimated the projected canopy cover and total shrub cover (including tree seedlings) for each quadrant following the method used by Strong (2011) [25].Then, for all shrubs present in a quadrat, we recorded the species name, measured its average height, counted the stems, and estimated the cover.A similar investigation was implemented for each herb quadrat as well.To improve the estimation, a grid (the size of the shrub or herb quadrat) with 20 cells was used to estimate the total cover and that of each species.Specimens of dominant or unknown species for each of the stands were collected in the field and identified in a laboratory using various volumes of Flora Popularis Republicae Sinicae (Chinese version of Flora of China, China) [34].All specimens are stored in the herbarium at the Chengdu Institute of Biology of the Chinese Academy of Sciences.

Statistical Analysis
In the study, we focused on the difference of species composition, structure, and species richness between the naturally regenerated forests and reforested spruce plantations based on the same background, but different forest management strategies.Hence, we checked for differences at the site level (n = 12) of the altitude, aspect class, slope, and time elapsed since clear-cutting between the NR and the SP stands by employed an independent t-test.The two stands did not differ significantly in altitude, aspect, slope, and stand ages (Table A1).
We then compared the difference between the NR and the SP and investigated the effects of canopy and shrub cover on the structure and species richness of the understory vegetation by applying GLMMs analysis based on the following facts: (1) nested data structure was not independent; (2) response variables were not fitted to normal distribution; and (3) there were many zeroes in our data because of the presence or absence of some group or species in each quadrat.During GLMMs analysis, treatment (NR vs. SP) was introduced as a fixed factor and stands (12 vs. 12) as a random factor, with three plots nested in each stand, and nine quadrats nested in each plot.Poisson error distribution, using a log-link function, was recommended for cover, average height, density, and species richness during GLMMs analysis.
Except for directly comparing species composition and total species richness between the NR and the SP forests, we classified all species into five species groups by growth form (tree seedlings, shrubs, ferns, forbs, and graminoids).The graminoids included species from Poaceae, Cyperaecae, and Junacaceae.The difference of each growth form group was analyzed by GLMMs.Poisson error distribution, using a log-link function, was also used for cover, average height, density, and species richness.We compared differentiation in vascular plant richness or abundance between the two forests for species group analysis as well.First, we categorized each species into one of three frequency-tendency distribution groups according to their occurrence tested by GLMMs model.The three identified species groups were: (1) reforestation species group (RES): species exclusively or more frequently found in the SP; (2) natural regeneration species group (NRS): species found exclusively or more frequently in the NR; and (3) generalist species group (GES): species that are recorded synchronously in the NR and SP, but do not show significant differences in occurrence (Tables A2 and A3).We postulated that with a background of the same origins (similar clear-cuts of the same old-growth spruce-fir forests) and regional species pools, the two forests provided different habitats and environments due to two regeneration strategies and, consequently, early stand succession.Thus, vascular plants with higher occurrence frequency in either the NR or the SP could indicate stronger habitat preference.
GLMMs analysis was also utilized to identify the effects of canopy and shrub cover on the structure and species richness of understory vegetation.Structure and species richness of the understory vegetation were selected as dependent variables.Canopy and/or shrub cover was selected as an explanation variable and stand as a random factor, with three plots nested in each stand, and nine quadrats nested in each plot.In both cases, Poisson error distribution with log-link function was selected in GLMMs.

Structural Parameters
The paired (NR vs. SP) stands with similar ages had similar topographical conditions (Table A1).However, they had different structural parameters at the understory both in the nested 2 m × 2 m shrub quadrats and the 1 m × 1 m herbaceous quadrats, except for the herbaceous species richness in the unit of square meters (Table 1).Furthermore, the two stands also displayed various frequency distribution patterns in covers of tree canopy, shrubs, and herbaceous plants (Figure 1).Higher average tree canopy cover was presented in the SP than the NR, both in the shrub quadrats (Table 1; Figure 1a) and the herb quadrats (Table 1).The frequency distributions demonstrated that most quadrats in both forests had higher tree canopy cover and more frequently presented between 50% and 100% (Classes 5 and 6, respectively).However, comparatively, the tree canopy cover was more frequently at class 6 (76%-100%) and less often at class 1 (<1%), 4 (26%-50%), and 5 (51%-75%) in the SP than in the NR (Figure 1a).Significantly less shrub cover both at shrub and herb quadrats, lower stem density, and shorter average height of shrubs at the shrub quadrats presented in the SP than in the NR (Table 1).The frequency distributions also revealed that the SP had more quadrats with shrub cover of less than 5% (Cover Classes 1 and 2), but the NR had more quadrats with cover between 6% and 75% (Cover Classes 3 -5) (Figure 1b).Furthermore, the SP had less cover and shorter average shoot heights, but greater shoot density for herbaceous plants than those of the NR (Table 1).The SP also had higher frequency in herbaceous cover presenting both at the lowest two classes (1 and 2, <5%) and the highest cover class (6, >75%); whereas the higher frequency in the NR was presented at the medium cover class (Classes 3 and 4, 6%-50%) (Figure 1c).Table 1.Structure and species density (mean ± SE) in shrub and herb layers from the naturally regenerated forests (NR) and the reforested spruce forests (SP) originating from similar clear-cuts in the eastern Tibetan Plateau.The difference between the NR and the SP was tested by generalized linear mixed-effects models (GLMMs).Treatment (NR vs. SP) was introduced as a fixed factor and stands (12 vs. 12) as a random factor, with three plots nested in each stand, and nine quadrats nested in each plot.In each case, Poisson error distribution with log-link function was selected in GLMMs.* indicated that shrub included shrub and tree seedlings.

Understory Species Richness and Composition
We recorded a total of 334 vascular plant species in the understories of both the SP and the NR stands (Tables A2 and A3).Fewer species occurred in the SP than in the NR (264 vs. 283 species) (Figure 3).A total of 87 woody plant species, including shrubs, tree seedlings, and saplings less than three meters high were recorded in the understory, but fewer species were in the SP than the NR (62 vs. 82) (Figure 3a).Fifty-seven woody species co-occurred in the two forests, making up a ratio of 69.5% in total woody plant species richness.A total of 247 herbaceous species were found in the two forests investigated, with 202 species in the SP and 200 species in the NR (Figure 3b).Many more herbaceous plant species (155 species, 63.2% of the totality) were commonly recorded in both forest types.natural regeneration species group (NRS), species only present or more frequent in naturally regenerated forests relative to reforested spruce plantations; reforestations species group (RES), species only or more frequent in reforested spruce plantations species relative to natural stands; generalist species group (GES), common in both forests, and with no significant difference in frequency between the two forests.The difference of frequency for each species between the NR and the SP was tested by generalized linear mixed-effects models (GLMMs).Treatment (NR vs. SP) was introduced as a fixed factor and stands (12 vs. 12) as a random factor, with three plots nested in each stand, and nine quadrats nested in each plot.In common species, a binomial error distribution with logit-ling function for presence (1) and absence (0) of each species in each quadrat was selected in GLMMs.
The two forests also differed in species richness within nested 2 m × 2 m shrub quadrats and 1 m × 1 m herbaceous quadrats (Table 1 and Figure 2).The SP had significantly less woody plant species richness, but similar herbaceous species richness in comparison to the NR (Table 1).The frequency distribution of shrub species richness in a unit of four square meters showed that the SP quadrats often had less than three species, whereas, the NR quadrats usually contained more than five species in each four m 2 quadrat (Figure 2a).Comparatively, the frequency distribution of herbaceous species density was similar to that of the NR (Figure 2b).The GLMMS analysis further showed a significant difference in frequency tendency distribution for common species between the NR and the SP stands for both woody species and herbaceous species (Tables A2 and A3).Except for those species existing either in the NR or the SP, there were 20 woody species and 32 herbaceous species with higher frequency in the NR than the SP; in contrast, there were six woody species and 36 herbaceous species whose frequency was higher in the SP than the NR.

Species Groups
Species group analysis showed that higher woody plant species richness, but less herbaceous species, existed in the NRS group than the RES group (45 vs. 11; 77 vs. 83) (Figure 3), demonstrating that more woody species and fewer herbaceous species tended to live in the habitats under the naturally regenerated forests.The understory plant growth forms also displayed some differentiations in structure and species richness between the NR and the SP (Table 2).Three groups (tree seedlings, shrubs, and ferns) always had much higher presence frequency, cover, average height, stem or shoot density, and species richness under the NR than the SP.However, forbs only had higher average height but lower cover, shoot density, species richness per square meter, and total species richness; the graminoids only had slightly higher cover and average height under the NR than the SP.Comparatively, the NR had higher species richness from three growth form groups (tree seedlings, shrubs, and ferns) and less species richness from forbs and graminoids (Table 2).Table 2. Species richness and structural parameter values (mean ± SE) of growth-form species groups and their differences in the naturally regenerated forests (NR) and the reforested spruce forests (SP) originating from similar clear-cuts in the eastern Tibetan Plateau.The difference between the NR and the SP was tested by generalized linear mixed-effects models (GLMMs).Treatment (NR vs. SP) was introduced as a fixed factor and stands (12 vs. 12) as a random factor, with three plots nested in each stand, and nine quadrats nested in each plot.In each case, Poisson error distribution with log-link function was selected in GLMMs, except binomial error distribution with logit-ling function for the presence (1) and absence (0) of each growth form.

Relationships of the Tree Canopy Cover with the Structure and Species Richness of Understory Shrubs
The tree canopy cover in the NR and the SP presented different influences on structures and species richness of understory shrubs (Table 3).The tree canopy cover limited only covers of total shrub layer and tree seedlings under the NR; however under the SP, it significantly hindered not only covers, but also stem density and the average heights of tree seedlings and shrubs.It was noted that the tree canopy cover had no significant influence on species densities of both forests.Comparatively, the SP canopy cover had a more seriously negative influence on shrub assembly structure than that of the NR.

Relationships of the Tree Canopy Cover and Understory Shrubs with the Structures and Species Richness of Understory Herbs
The tree canopy cover in the NR and SP also showed different influences on structures and species richness of understory herbs (Table 4).The NR canopy cover insignificantly limited the herbaceous layer development; however, in contrast, the SP canopy cover significantly hindered herbaceous community development, including covers and shoot densities of totality and various growth form groups.Under the context of the tree canopy cover, in the SP, the shrub cover significantly influenced herbaceous cover and shoot density, fern shoot density, forbs cover and shoot density, and graminoids cover and shoot density, but only significantly affected the graminoids cover in the NR.Comparatively, the tree canopy cover in the SP had a more serious negative influence on herb community development than its shrub cover.Table 3. Results of generalized linear mixed-effects models (GLMMs) for the effect of tree canopy cover on the understory shrub in the naturally regenerated forests (NR) and the reforested spruce forests (SP) on similar clear-cuts in the eastern Tibetan Plateau.Structure and species richness of the understory shrub were selected as dependent variables.Canopy cover was selected as an explanation variable and stand as a random factor, with three plots nested in each stand, and nine quadrats nested in each plot.In both cases, Poisson error distribution with log-link function was selected in GLMMs.

Discussion
The present study highlighted the importance of the reasonable selection of forest regeneration strategies for the development of the understory vegetation structure and in situ conservation of vascular plant biodiversity.Our results clearly showed that implementation of two regeneration strategies on similar clear-cutting sites, the natural regeneration and spruce plantation, produced distinct stand structures of both overstory and understory (Table 1 and Figure 1) and inevitably led to different understory plant composition and diversity (Figures 2 and 3; Tables A2 and A3).

Understory Vascular Plant Species Diversity
We found a high ratio (63%, 212 species of total 334 species) of total vascular plant species co-occurring in two forests.Some important late-successional species, such as Allium cyaneum Regel, Allium ovalifolium Hand.-Mazz, and Abies fabri (Masters) Craib, which are possibly remnants of clear-cuts from the old-growth spruce-fir forests [24], could be preserved within the two forests (Tables A2 and A3).This suggests that forest regeneration, regardless of natural regeneration or conifer reforestation, can effectively promote and conserve some native plants on clear-cuts.This result supports the current insight that reforestation with indigenous trees may play an important role in biodiversity conservation [11,12,18].
The two forests (SP and NR) were both at the early successional stage [13,24] and included not only many pioneer species, but also some late-succession plant species in the understory (Tables A2 and A3), definitely contributing to relatively high species diversity.Thus, our results also support the previous assertion that the successional stage plays an important role in determining biodiversity and composition in the understory [12,16,25].We further found that plant species composition was complicated and rich in the clear-cuts at the early developmental stage, containing not only many shade-intolerant and wind-dispersal species, such as annuals and ruderals, but also several remnant shade-intolerant or shade-tolerant species (Tables A2 and A3), as previously reported elsewhere [11,24,38].Therefore, we confirmed that the initial species compositions and their attributes after clear-cutting are fundamental drivers of understory biodiversity and its response to different regeneration pathways.
However, our results underscored significantly different effects of spruce reforestation and natural regeneration in species composition and diversity.We found that in total the NR had 19 more vascular plant species in the understory than the SP (283 vs. 264), 20 woody plant species more than the SP (25 vs. 5), and only two herbaceous plant species less than the SP (Figure 3).The growth form species group analysis also showed that higher total species richness for tree seedlings (28 vs. 21), shrubs (54 vs. 41) and ferns (22 vs. 12), but less for forbs (159 vs. 167) and graminoids (20 vs. 23) were present in the NR than the SP (Table 2).The findings were also supported both by frequency distribution patterns of species density (Figure 2) and species group analysis (Figure 3).In conclusion, our results definitely indicated that the NR harbored more vascular plant species in the understory than the SP in similar site conditions with the same vegetation origination in the eastern Tibetan Plateau, mostly due to higher species richness of woody plants and ferns.This provided reliable support for the initial hypothesis that natural regeneration with deciduous tree mixture could improve the understory plant diversity preservation better than the spruce reforestation on clear-cuts, because natural regeneration could provide more suitable understory microhabitats to encourage plant settlement and regeneration than spruce reforestation.Our results also revealed the important insight that various growth forms in the understory could respond differently to the regeneration treatments, resulting in the naturally regenerated forests having higher species richness in ferns, shrubs and tree seedlings, but less in forbs and graminoids (Tables 1 and 2).The present result relating to tree seedling demography also supported previous speculations in the eastern Tibetan Plateau that traditional dense single tree reforestation can hinder settlement and natural regeneration of some indigenous pioneer deciduous trees [24].

Structure of Tree Canopy Cover and Understory Vegetation, and Their Correlations
We also found a significant difference in tree canopy cover and understory structure between the reforested spruce plantations and naturally regenerated stands.The SP had higher tree canopy covers than the NR, both in shrub quadrats and herb quadrats (Table 1).The results were further explained by the differences in frequency of size patterns of tree canopy cover, with higher a frequency present in Cover Class 6 (76%-100%) for the SP, but more frequently in Cover Classes 1 (<1%), 4 (26%-50%), and 5 (51%-75%) for the NR (Table 1; Figure 1a).It is clear that monospecific and high density reforestation can be more effective and rapid to establish dense canopy structure than naturally regeneration in the study area [24,33].We further showed that the two forests presented significant differences in the understory vegetation structure (Table 1 and Figure 1).The SP had more undesirable shrub assembly structural features with less shrub cover, smaller stem density and shorter average height when compared to the NR (Table 1).This was also supported by the frequency distributions with more quadrats in shrub cover less than 5% (Cover Classes 1 and 2) in the SP and more quadrats in the cover between 6% and 75% (Cover Classes 3-5) in the NR (Figure 1b).Similarly, the SP also had more disadvantageous herbaceous community structures with less cover, shorter average shoot heights and slightly greater shoot density, in comparison with the NR (Table 1), which can be explained by different frequency distribution patterns with higher frequency in herbaceous cover at the lowest two classes (<5%) and the highest class (>75%) in the SP, and higher frequency at the medium cover class (6% -50%) in the NR (Figure 1c).These results were also identical to the results of the growth form species group analysis (Table 2).
Our results further demonstrated that the important differences in understory vegetation structure between the NR and the SP may be ascribed to their distinct tree canopy cover (Tables 3 and 4).The tree canopy cover in the SP limited the structure of the shrub assembly and herbaceous community more seriously than the NR.Due to higher tree cover, the shrub cover only slightly hindered the cover and shoot density of graminoids (Table 4).Therefore, we identified our hypothesis (H2) that the tree canopy structure of the two forests disparately influences the structure and species diversity of the understory.Tree canopy closure for reforested spruce forests usually requires 8-14 years from time of cultivation [24], which is faster than the naturally regenerated deciduous forests with 18-20 years in the focal region of the eastern Tibetan Plateau [33], meaning that faster and stronger sunlight restriction in the SP hinders the understory plant growth and, accordingly, vegetation development more in this region than in the NR.Therefore, compared to the NR, the SP always had more quadrats with a shrub and herbaceous cover of less than 5% (Figure 1b,c) and a lower woody plant species richness (<four species per quadrat) (Figure 2a).Furthermore, the tree canopy cover still differed even after canopy closure (Table 1; Figure 1a), and it continued to hinder the understory community development and biodiversity at the early successional stage (Tables 3 and 4; Figure 2).Strong (2011) also found that poplar diameter or stem densities and spruce size in the forest canopy layer could explain three-fourths of the variation in understory species abundance in the boreal forests [25].This finding further illustrated that the dense tree canopy more significantly limited the organizational structure of the understory vegetation in the SP stands in comparison with the NR natural sites (Tables 1, 3 and 4), which inevitably influenced the understory plant composition and biodiversity [18,39].Therefore, reducing the tree canopy cover in the dense spruce plantation by earlier thinning can be a reasonable management choice to promote understory development and in-situ plant diversity conservation.
It should be noted that the disturbance regime during reforestation has long been considered to influence plant settlement and development in the early stages [4,40,41].Reforestation practice comprises a series of activities, including site preparation, pit digging, seedling planting, initial weeding, and subsequent seedling tending and trampling, which can also directly influence the remnant understory vegetation community [24,39] Natural regeneration, on the other hand, has no further anthropogenic disturbance after clear-cutting.Moreover, the reforestation management activities expose the soil surface by reducing ground vegetation [41].The engineering activities during reforestation on clear-cuts also destroy habitat and transform the microclimate, so that its conditions are less favorable for the establishment and growth of remnant shade-tolerant plants.Consequently, many reforested microhabitats were altered into -more hostile environments‖ for some shade-tolerant species (e.g., orchids), while settling opportunities for pioneers and disturbance species were enhanced [39].In the initial years, the weeding and tending measures also continued to restrain population growth and the reproduction of high shrubs and large herbs, and indirectly drove some shade-tolerant plants into decline or led them to disappear, such as Kingdonia uniflora I. B. Balfour & W. W. Smith, the red-list protection herb endemic to China, Paris polyphylla Smith, orchids (Listera puberula var.maculata S.C. Chen & Y. B. Luo and Platanthera chlorantha F. Maekawa), and so on (Table A3).Therefore, because of human-made activities on clear-cuts, the spruce reforestation severely restricted the shrub community development and obviously increased invasions by pioneer annuals and ruderals.Meanwhile, however, it was harmful to those remnant species populations sensitive to habitat alteration [3,12].Thus we suggestthat to reduce the initial planting density of target trees during reforestation design was also a fundamental measure to decrease damage to initial ground vegetation and to allow the combination of reforestation and natural regeneration.

Conclusions and Implications
Regeneration strategies are critical for consequent forest succession, biodiversity conservation, and timber production on clear-cuts.However, their effects in deciding the understory vegetation and biodiversity are continually controversial and currently not well-known [11,12,14].We implemented the current study to compare the understory structure and vascular plant diversity between the naturally regenerated deciduous forest and the reforested spruce plantation with similar age, following the same clear-cut logging of old-growth spruce-fir forests in the eastern Tibetan Plateau.We tried to explore the effects of two regeneration strategies on the understory structure and plant diversity, natural regeneration, and spruce reforestation.We found that the naturally regenerated forest harbored richer vascular plant species, featuring more species of tree seedlings, shrubs and ferns, but similar forbs and graminoids in comparison to the reforested spruce forest.Furthermore, the naturally regenerated deciduous stands had less tree cover, but more desirable understory vegetation structure than the reforested spruce stands.Comparatively, the tree canopy cover more seriously hindered the understory structure development in the spruce plantation than in the naturally regenerated deciduous forest.Our findings comprehensively suggest that forest regeneration alternatives have distinct effects on the understory plant community and biodiversity, mostly due to initial disturbances and subsequent tree canopy attributes.It is implied that, relative to the coniferous reforestation, natural regeneration is better for the preservation of indigenous plant diversity and the understory vegetation at the early forest succession stage (20-40-years of age).The present study highlights the importance of regeneration strategy selection in biodiversity preservation, which has been neglected during in forest restoration on large areas of degraded forestlands worldwide.Given that conifer plantations are increasing in China and other biomes [11,14], it is urgent to modify the current reforestation management prescription for the promotion of the stand structure, the understory vegetation, and biodiversity preservation.Therefore, we recommend choosing the natural regeneration strategy on clear-cuts in the eastern Tibetan Plateau to better improve indigenous plant diversity conservation in the early successional forests, because this region with high elevation environmental fragility and importance in ecological and biodiversity conservation has been acknowledged as a key area aiming at ecological preservation and biodiversity conservation in the China National Region Development Strategy.However, due to the greater stand productivity in the spruce plantation [13], if we aim at striking a balance between biodiversity conservation and timber productivity, integrating natural regeneration and artificial reforestation into the local regeneration prescription would be a better choice.Such a mixed approach should greatly decrease the initial spruce seedling planting density for reducing reforestation disturbances and improving the proportion of mixed deciduous tree canopy by natural regeneration on clear-cuts.Furthermore, for current large areas of dense spruce plantation forests, we propose the timely implementation of reasonable selective thinning or the creation of artificial gaps to maintain the heterogeneous crown structure and to improve understory development and biodiversity conservation.Table A2.Woody plant species composition of the naturally regenerated forests (NR) and the artificial reforested spruce plantation (SP) originating from similar clear-cuts in the eastern Tibetan Plateau.The difference of frequency between the NR and the SP was tested by generalized linear mixed-effects models (GLMMs).Treatment (NR vs. SP) was introduced as fixed factor and stands (12 vs. 12) as random factor, three plots nested in each stand, and nine quadrats nested in each plot.In each case, binomial error distribution with logit-ling function for presence (1) and absence (0) was selected in GLMMs.Frequency tendency distribution (FTD): natural regeneration species (NRS), species only present or more frequent in naturally regeneration forests relative to reforested spruce plantations; reforestation species (RES), species only or more frequent in reforested spruce plantations species relative to natural stands; GES, generalist species.Growth-form: S, shrub; T, tree seedlings.

Figure 2 .
Figure 2. Frequency distribution of species density for woody plant species in four m 2 quadrats (a) and for total herbaceous species in one m 2 quadrats (b) within the naturally regenerated forests (NR) and reforested spruce forests (SP) originating from similar clear-cuts in the eastern Tibetan Plateau.

Figure 3 .
Figure 3. Species richness of totality and frequency tendency distribution species groups occurring under the naturally regenerated forests (NR) and the reforested spruce forest (SP) originating from similar clear-cuts in the eastern Tibetan Plateau.(a) woody plant species classification; and (b) herbaceous species classification.Frequency tendency distribution:natural regeneration species group (NRS), species only present or more frequent in naturally regenerated forests relative to reforested spruce plantations; reforestations species group (RES), species only or more frequent in reforested spruce plantations species relative to natural stands; generalist species group (GES), common in both forests, and with no significant difference in frequency between the two forests.The difference of frequency for each species between the NR and the SP was tested by generalized linear mixed-effects models (GLMMs).Treatment (NR vs. SP) was introduced as a fixed factor and stands (12 vs. 12) as a random factor, with three plots nested in each stand, and nine quadrats nested in each plot.In common species, a binomial error distribution with logit-ling function for presence (1) and absence (0) of each species in each quadrat was selected in GLMMs.

Table 4 .
Results of generalized linear mixed-effects models (GLMMs) for the effects on the understory herbaceous layer by covers of the tree canopy and shrub in the naturally regenerated forests (NR) and the reforested spruce forests (SP) on similar clear-cuts in the eastern Tibetan Plateau.Structure and species density of herbaceous layer were selected as dependent variables.Tree canopy and shrub cover were selected as explanation variables and stand as a random factor, with three plots nested in each stand, and nine quadrats nested in each plot.In both cases, Poisson error distributions with log-link function were selected in GLMMs.

Table A3 .
(1)baceous plant species composition of the naturally regenerated forests (NR) and the artificial reforested spruce plantation (SP) originating from similar clear-cuts in the eastern Tibetan Plateau.The difference of frequency between the NR and the SP was tested by generalized linear mixed-effects models (GLMMs).Treatment (NR vs. SP) was introduced as fixed factor and stands (12 vs. 12) as random factor, three plots nested in each stand, and nine quadrats nested in each plot.In each case, binomial error distribution with logit-ling function for presence(1)and absence (0) was selected in GLMMs.Frequency tendency distribution (FTD): NRS, species only present or more frequent in naturally regeneration forests relative to reforested spruce plantations; RES, species only or more frequent in reforested spruce plantations species relative to natural stands; GES, generalist species.Growth-form: FB, forbs; FN, fern; GM, graminoids.