Characteristics and factors influencing the natural regeneration of Larix principis-rupprechtii seedlings in northern China

Study site

This study was conducted in the Guandi Mountains, Jiaocheng county, Lvliang country, Shanxi Province (111°22′–111°33′E, 37°45′–37°55′N) (Yang et al., 2017). The annual, minimum, and maximum temperatures are 4.3 °C, –29.1 °C, and 17.7 °C, respectively. The average, minimum, and maximum atmospheric pressure is 936.6 hPa, 936.4 hPa, and 936.8 hPa, respectively. The average water vapor pressure is 5.3 hPa. The average annual precipitation is 822 mm, and evaporation is 1,268 mm. The altitude is between 1,500 and 2,831 m. The area has a temperate continental monsoon climate. The soil is mainly mountain cinnamon soil, cinnamon soil, and brown soil (Yang et al., 2017).

The forests are made up of dominant plants (L. principis-rupprechtii, Pinus tabuliformis, Picea wilsonii, Picea meyeri, Picea asperata, and Populus simonii), shrubs (Acer tataricum subsp. ginnala, Rosa multiflora, Spiraea salicifolia, Clematis florida, and Berberis amurensis), and grass (Fragaria orientalis, Geranium wilfordii, Chrysanthemum chanetii, Lathyrus humilis, Rubia cordifolia, Bupleurum smithii, and Thalictrum aquilegiifolium).

Experimental design

We investigated L. principis-rupprechtii seedling regeneration in the Guandi Mountains in a research area of 2 hm² in August 2020. The forest area was a plantation that did not have artificial intervention or management. Six fixed 40 ×40 m plots were selected (Fig. 1). The regeneration degree was evaluated (high = number of regenerated seedlings >5,000 tree/ha, medium = number of regenerated seedlings between 3,000 and 5,000 tree/ha, and low = number of regenerated seedlings between 2,500 and 3,000 tree/ha). In order to efficiently identify useful methods to stimulate natural regeneration, we studied the sampling plots with high regeneration densities, and then analyzed the effects of the naturally regenerated seedlings.

Data collection

Field measurements were carried out in six plots between August 1 and August 30, 2020. After the plots were laid out, we measured their land parameters, including aspect, slope, elevation, longitude, latitude, and canopy density (Table 1). Simultaneously, all L. principis-rupprechtii individuals were mapped in a plane-coordinate system. The coordinates of each L. principis-rupprechtii individual was measured using a tape and compass. We recorded each adult tree’s growth parameters, including height, diameter, breast height, and crown. The distances between L. principis-rupprechtii seedlings and the adult trees were recorded. Two categories were established: seedlings (0.3 m <height <2.5 m) and saplings (height ≥ 2.5 m). Five sampling points (1 ×1 m) in each plot were selected for herb species investigation using a X shape sampling method. Each herb species’ height, number, and coverage was recorded before bagging and weighing. The significant value was adopted to measure diversity indices such as the Simpson index, Shannon–Wiener index, and Pielou index. The shrub species across five quadrats (5 ×5 m) were then investigated and recorded using X-shape sampling. Five 30 ×30 cm litter sampling points near each herb and shrub sampling plot were additionally investigated. After the stones were removed, the undecomposed and semi-decomposed litter was classified, bagged, weighed, and labeled for further laboratory analysis. Five litter samples and five herb samples were collected from each plot.

Spatial pattern statistical analysis

Our point pattern analysis illustrated the spatial relationship and described the effect of distance between the L. principis-rupprechtii adult trees and seedlings. The null model identified potential differences in this spatial distribution. Pair-correlation function g(r) combined with Ripley’s K function showed the spatial relationship of different species. The random labeling null model was utilized to complete spatial randomness (CSR). The expected number of points at distance d and a short analysis area, K(d), were calculated by: (1)${\displaystyle \mu \mathrm{K}\left(\mathrm{d}\right)=\sum \sum \frac{{\mathrm{w}}_{\mathrm{ij}}\left(\mathrm{d}\right)}{\mathrm{n}}}$

where µis the density of regenerated seedlings per plot, n is the number of trees, and w_ij(d) is a weight function that considers the edge effect and represents the reciprocal of the number of circles centered on i (the radius is the distance from i to j) falling into the area (Barbeito et al., 2008).

The pair-correlation function g(r) assessed the critical range of regenerated clusters, which was more efficient compared to the K function that removed the scale dependence of K(d). The pair-correlation was the derivate of the K function, calculated by: (2)${\displaystyle \mathrm{g}\left(\mathrm{r}\right)={\mathrm{K}}^{\prime }\left(\mathrm{r}\right)/2\pi \mathrm{r}}$

where r is the spatial scale, and dot is the sampling circle with the target trees. g(r)>1, g(r) =1, and g(r)<1 indicated the pattern aggregation, randomness, and regularity, respectively, on the scale r.

The spatial relationship between seedlings and trees was analyzed using the g₁₂(r) function: (3)${\displaystyle {\mathrm{g}}_{12}\left(\mathrm{r}\right)=\mathrm{d}{\mathrm{K}}_{12}\left(\mathrm{r}\right)/\left(2\pi \mathrm{r}\cdot \mathrm{d}\left(\mathrm{r}\right)\right)}$

where g₁₂(r) =1, g₁₂(r)>1, and g₁₂(r)<1 indicate insignificant, positive, and negative correlations, respectively, between two different species on the scale r. These species were selected and combined within the two defined categories (seedlings and adult trees) to analyze their relationship.

The spatial relationship between the regenerated seedlings and adult trees was examined using CSR. Random labeling is not commonly used in sample plots. Consequently, we adopted a case-control design with L. principis-rupprechtii seedlings as pattern 1 (the control pattern) and L. principis-rupprechtii adult trees as pattern 2 (case). g₁₂(r) provided insight into the spatial distribution of L. principis-rupprechtii seedlings around adult trees. At a given scale, if g₁₂(r) was above the confidence envelope, L. principis-rupprechtii seedlings were positively associated with L. principis-rupprechtii adult trees on the scale r. However, if g₁₂(r) was below the confidence envelope, that showed a negative association between seedlings and trees.

All analyses were conducted using the Programita software package and Origin 2021 software (Wiegand & Moloney, 2004). The scale r represented spatial scales within 20 m (calculated up to half of the shortest side of the sampling plots). We used a 1 m cell scale and calculated the statistics up to a scale r of 20 m in view of plot size. One hundred and ninety-nine Monte Carlo simulations with a significance level of 99% were used to determine whether the species distribution pattern and correlation were significant (Wiegand & Moloney, 2004).

Spatial distribution patterns of L. principis-rupprechtii seedling populations across different sampling plots

Only seedlings of the same age and adult trees were found in the sampling plots, and no saplings were found. We observed that L. principis-rupprechtii seedlings and L. principis-rupprechtii adult trees were randomly distributed (I–VI) (Fig. 2). There was low spatial heterogeneity between seedlings and trees. We additionally explored L. principis-rupprechtii seedling spatial patterns across six sampling plots, and the g(r) function analysis results showed that the spatial distribution patterns changed with the spatial scale (Fig. 3). L. principis-rupprechtii seedlings in plot I showed aggregated distribution at a scale of 0–17.5 m, and random distribution at a scale of 17.5–20 m (Fig. 3–I). L. principis-rupprechtii seedlings in plot II showed aggregated distribution at a scale of 0–15 m (Fig. 3–II). L. principis-rupprechtii seedlings in plot III showed aggregated distribution at a scale of 0–9 m, random distribution at a scale of 10–12.5 m, and regular distribution at a scale of 12.5–20 m (Fig. 3–III). L. principis-rupprechtii seedlings in plot IV showed aggregated distribution at scales of 0–5 m and 7.8–17.5 m, and random distribution at a scale of 5–7.8 m (Fig. 3–IV). L. principis-rupprechtii seedlings in plot V showed aggregated distribution at scales of 0–9.5 m and 10–19 m (Fig. 3–V). L. principis-rupprechtii seedlings in plot VI showed aggregated distribution at a scale of 0–5 m and random distribution at a scale of 5–20 m (Fig. 3–VI). Generally, L. principis-rupprechtii seedlings showed aggregated distribution at small scales and random or regular distribution with increased spatial scale. The scatter distribution of populations in the sampling plots (Fig. 3) also showed that L. principis-rupprechtii seedlings in different sample plots were clustered in the Guandi Mountains.

Bivariate spatial analysis of L. principis-rupprechtii seedlings and adult trees in different sampling plots

Our spatial correlation analysis of L. principis-rupprechtii adult trees and seedlings showed that their spatial correlations (Fig. 4) were different across these six sampling plots in the Guandi Mountains. In sample plot I, L. principis-rupprechtii seedlings and L. principis-rupprechtii adult trees were positively correlated at a scale of 0–1 m, and showed weakly positive correlation at scales of 3.5–6.5 m and 10–15 m. Furthermore, there was no correlation at the other scales (Fig. 4–I). In sample plot II, there was a slightly negative correlation at a scale greater than 19 m. L. principis-rupprechtii seedlings and adult trees showed a negative correlation at a scale of 17.5–18.5 m and no correlation in any other scales (Fig. 4–II). In sample plot III, there was a negative correlation at scales of 0–1.5 m, 5–6.5 m, and 13–15.2 m, and no correlation in other scales (Fig. 4–III). In sample plot IV, there was a positive correlation at scales of 1.5–2.8 m and 8.2–10 m, and there was no correlation at scales of 0–1.5 m, 2.5–8.2 m, and 10–20 m between L. principis-rupprechtii seedlings and L. principis-rupprechtii adult trees (Fig. 4–IV). In sample plot V, there was a negative correlation at a scale of 10–11 m, and there was no correlation at scales of 0–10 m and 11–20 m (Fig. 4–V). In sample plot VI, there was a positive correlation between L. principis-rupprechtii seedlings and L. principis-rupprechtii adult trees at scales of 2.5–4 m and 8–9 m (Fig. 4–VI), and there was no correlation at other scales.

Effect of understory herb diversity and canopy density on L. principis-rupprechtii seedlings

In these sampling plots, six environment variables (Table 2) could explain 90.2% (PC1: 65.1%; PC2: 25.1%) of the changes found in the number of L. principis-rupprechtii seedlings. The number of L. principis-rupprechtii seedlings was negatively correlated with the Pielou, Shannon-Wiener, and Simpson indices, and the thickness of the semi-decomposed and undecomposed layers (Fig. 5) was weakly correlated with canopy density. Overall, the understory herb diversity and the thickness of the litter layer significantly affected the number of regenerated L. principis-rupprechtii seedlings.

The effect of L. principis-rupprechtii seedlings in different sampling plots

Previous studies have shown that the seedling stage is the most important and sensitive stage during forest regeneration (Deborah & David, 1984; Kaoru et al., 1989; Wright et al., 2005), particularly in forests with limited resources and space (Frost & Rydin, 2000). We found that the number of L. principis-rupprechtii seedlings was mainly determined by reproductive characteristics, as larger populations could produce more high-quality seeds. Natural germinated seedlings are formed when seeds are distributed under a large forest gap with favorable resources (Fig. 2), and a large number of seedlings are distributed in the community (Harms et al., 2000). The demands of L. principis-rupprechtii seedlings for light, soil nutrients, and water increase with their growth. L. principis-rupprechtii is a typical heliophilous species and their seedlings only grow successfully when occupying a large forest gap. Limited environmental resources and space can increase self-thinning in gradually-matured seedlings (Condit et al., 2000). Therefore, to achieve stable population growth, the living environment of L. principis-rupprechtii seedlings must be improved.

Species spatial distribution patterns change with scale (Condit et al., 2000; Peter & Chesson, 2000). The distribution pattern of a species at a small scale is mainly determined by seedling replenishment, while the distribution pattern at a large scale is mainly determined by habitat heterogeneity (Bai et al., 2012; Shen et al., 2013). In this study, L. principis-rupprechtii seedlings across all sample plots showed aggregated distribution at a scale of 0–5 m, and the distribution pattern changed from aggregated distribution to random or regular distribution as the scale increased, which is consistent with the findings of Zhang & Meng (2007). The difference in distribution patterns at the large and small scales was caused by the aggregated distribution at a small scale, which was beneficial for L. principis-rupprechtii seedlings to shelter each other, resist the adverse environment, and improve the survival rate. The random or regular distribution at a large scale allowed the L. principis-rupprechtii seedlings to avoid intraspecific competition (e.g., for light, water, or other resources) caused by habitat heterogeneity (Jian et al., 2007). During the field investigation, we also found that L. principis-rupprechtii seedlings were mostly distributed under the forest gap, and there were more seedlings above the litter layer, especially in the thinner thickness of the litter layer. On the other hand, the decreasing number of trees created good conditions for the mass reproduction and growth of young individuals. Additionally, when the root system extended to a certain degree, the self-thinning effect changed its distribution pattern from an aggregate distribution to a random or regular distribution, and L. principis-rupprechtii adult trees were mainly in random or regular distributions, which was consistent with the findings of Jian et al. (2007) and Nathan (2006).

Spatial correlation analysis of L. principis-rupprechtii seedlings and adult trees

The spatial correlation of individuals of different species within a population shows the spatial distribution and functional relationships among individuals at a particular time (Shields et al., 2014). This relationship is the result of the long-term interactions between trees and seedlings, and between the population and environment, and is the measure by which the current situation and population development trend can be predicted. A positive spatial correlation reflects the mutually beneficial ecological relationship within the population while a negative spatial correlation confirms a mutually exclusive ecological relationship within the population (Zhang et al., 2012). Previous studies have shown that the spatial distribution characteristics of adult trees affect the spatial correlation of tree growth (Getzin et al., 2008a; Walder & Walder, 2008). When trees are aggregated, growth reflects an obvious negative correlation, and negative correlation intensity increases with aggregation intensity. Law et al. (2010) found that the spatial correlation pattern of tree growth and the spatial distribution pattern of trees are independent to a certain extent. In this study, we observed no or a weak negative correlation between adult trees and seedlings with the increase of scale (except at a small scale), which may be because adult trees have a lower demand for soil water, light, and other resources and do not compete with seedlings (Caquet et al., 2010; Chen & Cao, 2014). As a result, there is a large distribution density of individual seedlings around adult trees. The spatial correlations between L. principis-rupprechtii seedlings and adult trees in plots V and VI, I and IV, and II and III were similar (Fig. 4), and illustrated the effect of trees on seedlings. We also found that there were few seedlings around trees in some sampling plots (Fig. 2), which was probably caused by limited seed diffusion (Seidler & Plotkin, 2006). Because L. principis-rupprechtii seeds have seed wings and rely on the wind to spread, most seeds are scattered at a small scale to form seedlings. Additionally, there is less wind in plots at high elevations, which causes scattered distribution of L. principis-rupprechtii seeds in these areas. Additionally, the effect of L. principis-rupprechtii adult trees on seedlings gradually decreased at a large scale, which shows that the competition between trees and seedlings tends to be gentle and seeds are unable to disperse over a long range. On the other hand, however, this shows the joint action of the bio-ecological characteristics of the population. The findings in this paper are consistent with those of Zhang et al. (2012) and Wang et al. (2017).

We also found that individuals with the same tree age were in aggregated distribution. This indicates positive spatial correlation, interdependency, and consistency in the selection of habitats (Andersen, 1992) because these individuals are at similar development stages and shelter each other, improving the survival rate. The non-correlation between L. principis-rupprechtii adult trees and seedlings at a large scale indicates that the individuals do not interfere with each other in the early development stage. With an increase in scale, the correlation between L. principis-rupprechtii adult trees and seedlings became uncorrelated or negatively correlated, indicating that the increasing number of seedlings may lead to growth retardation or death. This explains why there are only a small number of seedlings around L. principis-rupprechtii adult trees in some geographical locations (Getzin et al., 2008b), which validates our hypothesis (1). Additionally, a few L. principis-rupprechtii adult trees produced a better habitat for seedlings, which allowed them to make full use of limited resources in order to grow. Overall, the distance between L. principis-rupprechtii seedlings and adult trees affected the natural regeneration of seedlings in the Guandi Mountains. The distance (1.5–4 m) between seedlings and adult trees can be applied to promote the natural regeneration of seedlings. Therefore, the effectiveness and practicality of seedling transplantation using this distance between seedlings and adult trees should be considered in further research.

The effect of distribution between L. principis-rupprechtii adult trees and seedlings

Habitat heterogeneity is considered the most significant factor that leads to seedling aggregation (Getzin et al., 2008b). Herb diversity potentially impacts species composition and seedling spatial patterns, and significantly reduces community productivity. We found that herb diversity negatively affected the number of L. principis-rupprechtii seedlings, which decreased the number of L. principis-rupprechtii seedlings (Fig. 5 and Table 2), validating our hypothesis (2). Herbs use most of their biomass to produce large and long leaves, increasing the thickness of the litter layer, while reducing the available light, inhibiting the growth of co-occurring species, and impeding the regeneration of seedlings.

An increasing number of herbs increases the competition for soil nutrients among seedlings, affecting species composition. Shading and mechanical hindering (e.g., in twine production) impede seed germination and seedling regeneration and can lead to seedling death. With an increase in herbaceous coverage, the number of seedlings greatly decreased, indicating that herbaceous coverage is the main factor affecting the regenerate degrees of seedlings. The scale of 1.5–4 m was consistent with the range of canopy cover in adult trees. Insufficient light can impede the growth of herbs. A shady environment is optimal for the regeneration of seedlings at an early stage, which further verifies our previous conclusion. Decreased thickness of the semi-decomposed and undecomposed layers lead to a large number of nutrients being decomposed and transferred to the soil to promote the growth of seedlings. According to the results of this study, understory litter is formed at the early stage of regeneration, and the semi-decomposed and undecomposed layers are relatively thin at the scale of 1.5–4 m (Malik et al., 2020). The semi-decomposed and undecomposed litter layers showed a negative effect on natural regeneration across the six sampling plots. Seeds germinated into seedlings in the litter layer, but the seed radicles in the thicker litter layer could not reach the soil and obtain sufficient nutrients for growth, leading to the death of many naturally regenerated seedlings in the early development stage (Nakagawa et al., 2001; Nakagawa, Kurahashi & Hogetsu, 2003). The litter layer was thick around adult trees at a scale of 1 m. However, L. principis-rupprechtii’s thick litter layer had an auto-toxic effect on seed germination and seedling growth (Liira, Sepp & Kohv, 2011), even though the thick semi-decomposed and undecomposed layers can preserve heat, water, and other resources. Generally, the litter thickness in these plots created an ideal stand for regenerated L. principis-rupprechtii seedlings. Thinner and thicker litter layers should be managed in forests. However, the correlation between canopy density and L. principis-rupprechtii seedlings was weak, which was consistent with the findings of Wenjun & Xi (2020) and verifed our hypothesis (3).