Relationship Between Forest Strata Structure and Regeneration in Subtropical Evergreen Broad-Leaved Forest


 Background: Regeneration is an extremely important and complex ecological process, which is disturbed by many factors. The current stand structure has an important influence on regeneration. The aim of this study is to provide theoretical reference for improving the regeneration capacity subtropical evergreen broad-leaved forest and formulating management measures of regeneration restoration.Methods: A permanent plot of 100m × 100m was set up in the evergreen broad-leaved forest of Tianmu Mountain National Nature Reserve, Zhejiang Province, China. The plot was divided into 25 survey units of 20m × 20m by the adjacent grid survey method, and all the trees in the plot were investigated. The tree height, DBH, crown width, density, species richness index, aggregation index, competition index and mingling of each forest stratum were used as the stand structure index. The tree height, DBH, crown width, density and species richness index of regeneration trees were used as regeneration indicators. Redundancy analysis (RDA) was used to explore the relationship between forest strata structure and regeneration of evergreen broad-leaved forest. Results: In the whole stand, DBH, tree species richness index and crown width were the main structure factors affecting regeneration. In the upper forest stratum, the tree height was the main structure factor affecting regeneration. In the middle forest stratum, the tree species richness index and crown width were the main factors affecting regeneration. In the lower forest stratum, crown width, competition index, tree height and tree species richness index were the main factors affecting regeneration. The effects of tree species richness index and crown width on regeneration in the whole stand were mainly reflected in the middle and lower forest strata in each forest stratum. Conclusions: The influencing order of each forest stratum structure on regeneration was: lower forest stratum > middle forest stratum > upper forest stratum. Different regeneration indicators had different responses to the main stand structure indices, while the young tree height and DBH, and the tree species diversity and density of regeneration trees were most affected by the main stand structure indices. In order to promote the regeneration of evergreen broad-leaved forest in the future, different management measures should be taken for different forest strata, and the threshold value of each index should be controlled.


Introduction 31
Subtropical evergreen broad-leaved forest is one of the typical forest types in China, which plays an important 32 role in protecting the environment and maintaining the global carbon balance and the sustainable development of 33 human beings (Cao et al., 2010). However, due to the people's unreasonable development and utilization of forest 34 resources in the early stage, the area of evergreen broad-leaved forest had been continuously reduced, resulting in 35 the degradation of ecological functions and other problems (Song et al., 2005;Zhuo and Zheng, 2019). Natural regeneration is the main way of forest resources reproduction, which is particularly important for the restoration 37 and protection of evergreen broad-leaved forest, and has great research and protection significance (Shi et al., 2014; 4 neighborhood trees and regeneration, but most of their research used the artificially regeneration trees as object divided into 25 survey units of 20m × 20m by the adjacent grid survey method (Fig.1). Each grid was used as the 96 investigation unit to measure all the trees in the plot, record each tree species, Coordinates (x, y, z), DBH, tree  Trees with DBH > 5cm were defined as large trees, and DBH, tree height, crown width, density and diversity 116 index were selected to describe the characteristics of stand non-spatial structure.

117
The tree species diversity index is used to describe the proportion of species to individuals in a biological 118 community. The tree species richness index was calculated as (Liu et al., 2011): where, S is tree species richness index, m is the number of species in each grid, and M is the total number of 121 individuals of all species in the plot.

Stand spatial structure indices 123
Mingling, competition index and aggregation index were selected to describe the characteristics of stand 124 spatial structure, and Voronoi diagram based on the relationship of neighborhood trees was used to calculate the 7 the object tree are regarded as neighborhood trees (Tang et al., 2007). To eliminate the edge effect, the 127 eight-neighborhood method was used to edge correction of the plot.
where, Mi is the mingling of the object tree i; Mci is the complete mingling of the object tree i; Ni is the number of 134 neighborhood trees; Vij is a discrete variable, Vij = 1 when the neighborhood tree j and the object tree i have 135 different tree species, otherwise Vij = 0; ci is the number of different tree species in pairs of neighborhood trees in 136 the spatial structure unit i; Di is the Simpson diversity index of the tree species in the spatial structure unit i.

137
Competition index is used to describe the competitive relationship among trees within a forest. The Hegyi 138 competition index (hereinafter referred to as competition index) based on Voronoi diagram was calculated as 139 (Hegyi, 1974): where, CIi is the competition index of object tree i, Lij is the distance between object tree i and neighborhood tree j,

8
Aggregation index is used to describe the spatial distribution patterns in forest, and is defined as the 147 proportion of the average distance between object trees and their nearest neighborhood trees to the expected 148 average distance under a random tree distribution pattern (Clark and Evans, 1954). It was calculated as: Where, R is the aggregation index, N is the number of trees in the plot, F is the plot area, and ri is the distance from 151 the object tree i to its nearest neighborhood tree.

Regeneration indicators 153
The trees with DBH < 5cm were defined as regeneration trees. According to the tree height and DBH, the 154 regeneration trees were divided into three grades: seedlings, saplings and young trees. Seedlings: H ≤ 1.5m, DBH 155 < 1cm; Saplings: H ≤ 1.5m, DBH ≥ 1cm; young trees: H > 1.5m, DBH < 5cm (Tang et al., 2006). The DBH, tree 156 height, crown width, species richness index and number of regeneration trees were selected as regeneration 157 indicators. The species richness index was calculated by using Eq. (1).

Data analyses 159
The software IBM SPSS Statistics 20 was used to analyze the differences of different forest strata structure.  stand structure indices were taken as environmental variables, and regeneration indicators as species variables, the 165 relationship between them was analyzed using the software Canoco 5. Firstly, in order to select a suitable model 9 maximum gradient of the four axes was less than or equal to 3, the linear model was used; when the maximum original data. Variance inflation factor (VIF) was used to test the multicollinearity between variables, and the 171 variance inflation factor was less than 20, which indicated that there was no multicollinearity among the stand 172 structure indices. The most significant stand structure indices affecting regeneration were screened out through 173 interactive forward selection. Finally, the specific relationship between the most significant stand structure indices

Results 176
3.1. Differences in different forest strata structure 177 Stand structure characteristics of different forest strata are shown in Table 1. There were significant 178 differences among the different forest stratum structure index except the aggregation index (p < 0.05). The 179 mingling has significant differences between upper forest stratum and lower forest stratum, and no significant 180 differences between middle forest stratum and other forest strata. The species richness index has significant 181 differences between upper forest stratum and other forest strata, and no significant differences between middle 182 forest stratum and lower forest stratum. With the rise of forest strata, the mingling increased, and the competition 183 index and species richness index decreased. Therefore, it was reasonable to divide into three forest strata in this 184 subtropical evergreen broad-leaved forest.

Effect of whole stand structure on regeneration 189
The results of RDA showed that 49.22% of the regeneration variation was explained by the whole stand richness index and crown width had the most significant effect on the regeneration, which the explained variations 194 of regeneration were 18.4%, 9.4%, 7.2%, accounting for about 71.11% of the explained variation of the 8 whole 195 stand structure indices (Table 2).  showed an increasing trend, and young tree density and species richness index showed a unimodal distribution.

217
The young tree density and species richness index kept a high response value when the species richness index of 218 whole stand was between 4 and 5 (Fig. 3B). With the increase of the crown width in the whole stand, young tree 219 DBH and height showed a single valley distribution, while sapling density and species richness index showed a 220 unimodal distribution. Young tree DBH and height maintained a small response value when the crown width was 221 between 4m and 5.5m. The sapling density and species richness index maintained a high response value when the 222 crown width of whole stand was between 4.5m and 6m (Fig. 3C). The results of RDA showed that 37.76% of the regeneration variation can be explained by the upper forest 228 stratum structure index with 19.83% being explained by the first axis and 10.87% being explained by the second 229 axis. It can be seen that RDA can better explain the relationship between upper forest stratum structure index and 230 regeneration. The interactive forward selection results of upper forest stratum showed that the tree height was the 231 most significant structure factor affecting regeneration, and the interpretation rate of regeneration was 13.9%,

232
which accounts for about 36.81% of the total interpretive ability of the 8 upper forest stratum structure indices 233 (

237
According to the ordination diagram of RDA (Fig. 4)  The redundancy analysis results of middle forest stratum structure and regeneration is shown in Table 4.

262
The species richness index of the middle forest stratum had a larger positive effect on seedling species 14 forest stratum had a larger positive effect on sapling species richness index and density, and had a larger negative 265 effect on young tree height, and sapling crown width and height (Fig. 6).

268
When the species richness index of the middle forest stratum was between 1 and 1.5, the young tree species 269 richness index first increased after reaching the minimum value, the increasing rate of seedling density slowed 270 down, and the increasing rate of young tree density accelerated. When the species richness index of the middle 271 forest stratum increased to between 3 and 3.5, the young tree species richness index reached the maximum value, 272 and the increasing rate of seedling density began to accelerate, the change of young tree density tended to be stable 273 (Fig. 7A). With the increase of crown width in middle forest stratum, sapling crown width and young tree height 274 showed a single valley distribution, the young tree density and species richness index showed a unimodal 275 distribution. When the crown width of middle forest stratum was between 6m to 7m, the young tree height and 276 sapling crown width reached the minimum value, and the sapling density and species richness diversity reached 277 the maximum value (Fig. 7B).
15  The redundancy analysis results of lower forest stratum structure and regeneration is shown in Table 5.

282
49.58% of the regeneration variation can be explained by the four axes, 43.78% of the regeneration variation can 283 be explained by the first two axes with 30.09% being explained by the first axis and 13.69% being explained by 284 the second axis. Therefore, the first two axes provided the optimal explanation for the variation in both lower 285 forest stratum structure index and regeneration. From the forward selection results of lower forest stratum, the 286 most significant structure factors affecting regeneration were: crown width, competition index, tree height and 287 species richness index, which the explained variation of regeneration were 11.2%, 10.8%, 9.5%, 7.2%, accounting 288 for 78.06% of the total explained variation of the 8 stand structure indices.

301
The specific effect of the main lower forest stratum structure indices on regeneration is shown in Fig. 9. When 302 the crown width of the lower forest stratum was between 2.0m and 3.2m, the seeding density and species richness 303 index, and sapling density and species richness index had a single valley distribution, and sapling density and 304 species richness index reached the minimum value. When the crown width of the lower forest stratum was between 305 3.2m and 4.5m, the seeding density and species richness index, and sapling density and species richness index had 306 a unimodal distribution, and sapling density and species richness index reached the maximum value (Fig. 9A).

307
With the increase of competition index in the lower forest stratum, the seedling and sapling density and species 308 richness index had a single valley distribution, the young tree density and species richness index showed 309 downtrend. The seedling and sapling density and species richness index reached the minimum value when the the lower forest stratum, the sapling density showed a decreasing trend, young tree DBH and sapling species 312 richness index showed a single valley distribution, and young tree height showed an increasing trend. The sapling 313 species richness index reached the minimum value when the tree height in the lower forest stratum was between 314 5m and 5.3m. The young tree DBH reached the minimum value when the tree height in the lower forest stratum 315 was between 4.8m and 5m (Fig. 9C). When the tree species richness index in the lower forest stratum was between 316 1.0 and 2.5, the young tree density showed an increasing trend, the seedling density and species richness index 317 showed a unimodal distribution, and young tree species richness index showed a single valley distribution. When 318 the tree species richness index in the lower forest stratum was between 2.5 and 4.0, the young tree density showed 319 a steady trend, the seedling density and species richness index showed a single valley distribution, young tree 320 species richness index showed a unimodal distribution, which species richness index reached the maximum value 321 (Fig. 9D). . It is generally believed that the larger DBH and crown width of the forest, the older 333 stand age and the more mature seed trees in the forest, which can provide enough provenance for regeneration.

334
Some researches showed crown width plays a role of shading and shelter for regeneration, and affects the growth

Conclusions 385
In this paper, redundancy analysis was used to study the relationship between different forest strata structure 386 and regeneration, it can not only independently determine the contribution and explanation of each stand structure 387 variable (Liu et al., 2011), but also reduce the number of stand structure variables that can effectively explain the 388 regeneration variation. It can be seen from the ordination diagram of RDA that although the competition index of 389 the whole stand and the crown width of the upper forest stratum had no significant effect on the regeneration, they 390 had strongly correlated with sapling density and richness index. Hence the effect of non-significant stand structure