Canopy disturbance and gap partitioning promote the persistence of a pioneer tree population in a near‐climax temperate forest of the Qinling Mountains, China

Abstract An unresolved question of temperate forests is how pioneer tree species persist in mature forests. In order to understand the responsible mechanisms, we investigated a near‐climax mixed temperate forest dominated by Betula albosinensis in the Qinling Mountains of China. Through establishing four 50 m × 50 m plots, we examined the canopy disturbance characteristics and its effects on tree recruitments. We further test the intra‐ and interspecific effects on the recruitment of B. albosinensis. The obtained data demonstrated canopy disturbance was frequent but most small‐sized. The canopy gaps are caused mainly by adult B. albosinensis by snapping. The regeneration of coexistent tree species shows a distinct preference for gap size. B. albosinensis were clumped at the juvenile stage and small scales. B. albosinensis juveniles were positively associated with B. utilis juveniles and negatively associated with the conspecific and B. utilis large trees. In addition, B. albosinensis juveniles showed negative associations with contemporary other tree species. Our results suggested that canopy disturbance caused by canopy trees and gap partitioning among the coexistent tree species are important for the persistence of the mixed forest. As a main gapmaker, B. albosinensis appear to develop a self‐perpetuating life‐history trait and allow them to persist.

However, the correspondence between disturbance regimes and species-specific life-history traits is not well understood.
Hardwood-conifer mixed forests are widespread in warm temperate subalpine forests of China. Particularly in the subalpine range of the Qinling Mountains, the near-climax hardwood-conifer mixed forests, extending over a wide elevation range over 700 m, constitute one of the most important forest ecosystems of central China (Dang, Zhang, Zhang, Jiang, & Zhang, 2010;Qin et al., 2011;Wang, Franklin, Ren, & Ouellette, 2006). Intriguingly, most of these mixed forests are dominated by a pioneer tree species, Betula albosinensis Burk. The analysis of soil sporopollenin indicates that B. albosinensis have existed in these forests at least from the Late Quaternary (Zhu, 1991). Therefore, many historical records and investigation reports in the Qinling Mountains listed the forest as a climax forest (Yue, Dang, & Gu, 2000;Zhu, 1991). To date, however, the ecological mechanism responsible for the long-term persistence of B. albosinensis in these mature forests has not been studied.
In the present work, therefore, we seek to understand the mechanism explaining the long-term persistence of pioneer B. albosinensis in the canopy of mature mixed forest. First, we documented the characteristics of canopy disturbance and the regeneration patterns of canopy tree species in the mixed forest. Secondly, we examined how the conspecific and heterospecific adults affect the recruitment of B. albosinensis. Finally, we discuss the possible mechanism responsible for the persistence of B. albosinensis based on the given disturbance regime.
Historically, the B. albosinensis forests in the Qinling Mountains were selectively cut several times during the period 1790-1870 (Dang et al., 2010;Zhang, 1989). In the past century, however, human activities are F I G U R E 1 Location of study area in Shaanxi Province, China rare due to the relatively difficult accessibility (Liu, Tang, Dai, Tang, & Cui, 2002). Therefore, the forest exhibits major natural disturbances (e.g., winds, floods, snowstorms; Dang et al., 2010) without anthropogenic disturbances. In mature B. albosinensis forest, most canopy B. albosinensis exceed 0.5 m in DBH (diameter at breast height). The largest tree recorded even exceeds 1 m in DBH and is nearly 100 years of age (calculated by equations between DBH and age).

| Field methods
The spatial locations of trees provide signatory information for inferring inter-and intraspecific relationships in field conditions (Gray & He, 2009); therefore, analyzing the spatial patterns of trees is an effective method of investigating inter-and intraspecific interactions of tree species in the absence of growth data (Browning, Franklin, Archer, Gillan, & Guertin, 2014;Guo et al., 2015). In the present study, we sampled four stands to represent the typical floristic composition of B. albosinensis forest along the altitude (Table 1). Each stand was sampled with a 50 m × 50 m plot. Each plot was established on the consistent slop and microtopography to avoid the confounding effects of environmental heterogeneity on tree survival.
Within each plot, only canopy tree species with more than 30 individuals were investigated to analyze spatial pattern. For each investigated tree species, all living stems greater than 5 cm in DBH were measured and mapped. Each tree species were divided into three growth stages according to DBH: juvenile (J) (5 cm ≤ DBH < 10 cm), medium (M) (10 cm ≤ DBH ≤ 25 cm), and large (L) (DBH > 25 cm).
In addition, canopy gaps within four plots were investigated.
We considered gaps as all canopy openings with an area exceeding 20 m 2 that appeared to have been formed by the death of one or more trees. The actual gap size was calculated with elliptical method based on the length (longest distance within the gap) and width (perpendicular to the length) for each gap (Runkle, 1981(Runkle, , 1982. At each canopy gap, the gap-forming species (gapmakers) were identified and measured. At the same time, the gapmakers were categorized as: standing dead, snapped, uprooted, and branch fallen following Nakashizuka (1989). In addition, we recorded the regeneration of trees inside the expanded gap by direct counting seedlings (DBH < 5 cm). The expanded gap was defined as the actual gap plus the adjacent area extending to the bases of canopy trees surrounding the canopy gap; therefore, its benefit is that it includes area directly and indirectly influenced by the canopy opening (Runkle, 1982).

| Data analyses
The O-ring statistic was used to analyze the spatial patterns and associations of trees in each plot. Based on Ripley's K (Ripley, 1981) and mark correlation functions (Stoyan & Stoyan, 1994), the O-ring statistics replaces the circles used for Ripley's K with rings and uses the mean number of neighbors in a ring of radius r and ring width around an individual, thus isolating specific distance classes (Wiegand & Moloney, 2004). The O-ring statistics characterizes patterns by the frequency of points co-occurring at a given distance, so it can analyze the spatial patterns derived from ecological processes easily and intuitively (Schurr, Bossdorf, Milton, & Schumacher, 2004;Wiegand & Moloney, 2004). The O-ring statistics includes both univariate and bivariate statistics (Getzin et al., 2006). The univariate statistics are used to analyze the spatial pattern of one object, while the bivariate statistics are used to analyze the spatial association between two objects (pattern 1 and pattern 2). Following the notation by Wiegand and Moloney (2004), the bivariate O-ring statistic O 12 (r) is calculated as: where n 1 is the number of points of pattern 1; R 1 , w i (r) the ring with radius r and ring width w centered in the ith point of pattern 1; In our study, univariate O-ring statistic was used to analyze the spatial patterns of B. albosinensis at the different growth stage, respectively, and bivariate O-ring statistic was used to examine the intraspecific and interspecific associations in spatial distribution. Different null models were selected for various analyses because the use of an inappropriate null model may lead to a misinterpretation of spatial patterns (Goreaud & Pélissier, 2003). For the univariate O-ring statistic, complete spatial randomness was selected as the null model because there was no broad-scale heterogeneity in the stand caused by exogenous factors. For the bivariate O-ring statistic, data were analyzed using an antecedent-condition null model that keep the older stages fixed and randomize the earlier ones because the different growth stages of trees were not achieved at the same time but in sequence.

| Population structure
A total of 5 canopy tree species were investigated within the sampled mixed forests (Table 1). B. albosinensis dominated all the plots with the highest density, accounting for 47%, 56%, 42%, and 43%,

| Gap formation and regeneration
A total of 65 gaps were measured, covering 32.7% of the total investigated area. Plot II showed a highest gap density and area, followed by Plot I, Plot III, and Plot IV (Table 1). Most gaps are small-sized in the forest, and only, two gaps exceeded 100 m 2 , which were found in Plots I and II (Figure 4). Within the 65 gaps, a total of 131 gapmakers were identified (Table 1). The mechanism of gap formation was mainly tree snapping which accounted for 36% of total (4.0%) were infrequent gapmakers in their Plot I. In addition, 6.3% of gapmakers were highly decomposed and unidentified.
A total of 870 tree seedlings were found in the gaps. The total and species-specific seedling density showed significant relationships with gap size ( Figure 5). Seedling density of total and P.
armandii both showed a parabola trend with the increased gap size.
Seedling density of B. albosinensis and B. utilis increased significantly with gap size. Q. liaotungensis and A. fargesii decreased significantly with gap size.

| Spatial patterns and relationships
The spatial pattern of B. albosinensis varied along the life cycle In addition, some notable interactions were found between B. albosinensis juveniles with the nonconspecific trees (

| Disturbance and regeneration dynamics
In our study, a total of 65 canopy gaps were identified within the total of 1.0 ha forest area, indicating that canopy disturbance is abundant in the present forests. However, most gaps were <100 m 2 , which was consistent with other reports in temperate forests that small gaps dominate natural disturbance regimes (Fukui, Hirao, Murakami, & Hirakawa, 2011;Sipe & Bazzaz, 1994). In the gap formation, tree snapping was the most common mechanism, which is consistent with the mechanism of canopy gap formation in mixed hardwood-conifer forests in Foping National Natural Reserve of the Qinling Mountains where 40% of gaps are tree snap types (Wang et al., 2006). In the mixed B. albosinensis forests, seasonal wind and snow are the main natural disturbance types (Dang et al., 2010;Qin et al., 2011). Most canopy trees are susceptible to wind events and snow accumulation; therefore, tree snapping is most frequent after these disturbance events. Besides tree snapping, we found that uprooting was more F I G U R E 6 Spatial distribution pattern of B. albosinensis population at different growth stages in the four Plots of the mixed forest. The solid line indicates ring statistic, O(r); the gray-filled area indicates the upper and lower limits of the 99% confidence envelope of the null model. "J," "M," and "L" represent juvenile, medium-sized, and large-sized trees, respectively common in Plots I and II, and standing death was more common in the Plots III and IV. The difference might be attributed to the change in slope between the Plots. In our study, Plots I and II were located in steep sites, whereas Plots III and IV had gentle slopes (see Table 1).
Steep topography develops shallow soils and confronts intense seasonal wind and flood, which may increase the likelihood of tree uprooting. In contrast, the trees growing on gentle slopes may seldom uproot, but they are prone to become standing death due to disease in the poorly-drained soils (Sefidi, Mohadjer, Mosandl, & Copenheaver, 2011). In addition, Betula trees are the most common gapmakers in the mixed forest; in particular, B. albosinensis accounted for 48% of the gapmakers. The high mortality rate of Betula trees also has been found in the other hardwood-conifer forests of China (Taylor, Jinyan, & ShiQiang, 2004;Wang et al., 2006). As pioneer trees, Betula trees grow rapidly and develop fragile stems and large crown (Lusk, 1995).
The way Betula grows makes them susceptible to local disturbances, such as seasonal wind and snow accumulation. Therefore, canopy gaps created by Betula trees were prevalent in the present forest.
In our study, seedling density was largest in the gaps of medium size, suggesting medium-sized gap may be more important for the regeneration of most tree species. However, most tree species showed species-specific regeneration pattern in relation to gap size ( Figure 5). The recruits of B. albosinensis and B. utilis increased significantly with gap size, which is consistent to the previous findings that F I G U R E 7 Spatial association among the different growth stages of B. albosinensis population in the four Plots of the mixed forest. The solid line indicates ring statistic, O(r); the gray-filled area indicates the upper and lower limits of the 99% confidence envelope of the null model. "J," "M," and "L" represent juvenile, medium-sized, and large-sized trees, respectively an environment with much light and exposed soil are important for the establishment and survival of Betula sp. (Nakamura, 1985;Taylor & Qin, 1988;Seiwa & Kikuzawa, 1996;Guo et al., 2013). Betula sp.
has life-history traits that are associated with large gap specialists, such as shade intolerance, long-distance seed dispersal, and fastgrowing seedlings, which contribute to their regeneration in largesized gaps. In our study, therefore, Plots I and II have more abundant B. albosinensis seedlings than Plots III and IV since Plots I and II have more large-sized gaps (Figure 4). In contrast, densities of Q. liaotungensis and A. fargesii seedlings decreased significantly with gap size. Previous research suggested that Quercus seedling response to light availability varies among species (Sevillano, Short, Grant, & O'Reilly, 2016;Truax, Lambert, & Gagnon, 2000). Nevertheless, most researchers have found positive effects of increased light on Q.
liaotungensis seedling establishment. For example, Li and Ma (2003) found that seedling recruitment of Q. liaotungensis was significantly better in gaps than in shaded understory. However, Q. liaotungensis seed fall occurs under and near the crown of the parent tree, which does not allow them to reach large-sized gaps, although the environment is favorable for establishment (Mi & Hou, 2009). Thus, Q.
liaotungensis seedlings may be limited to the small-sized gaps due to limited dispersal. It can explain the decline of Q. liaotungensis seedling along the increased gap size. Compared with Q. liaotungensis, A. fargesii is considered to be more shade tolerant and have lower light requirements for regeneration than hardwood trees although it also depends on gaps to keep the population sustainable and stable (Dang et al., 2010). Fu, Liu, and Xiong (2010)  armandii is less shade tolerant (Lan, Lei, & An, 2005;Yu et al., 2014).
In addition, moist soil is also necessary for the establishment of P. armandii seedling (Gao, 1991). Soil moisture is variable with gap size. In general, intermediate-sized gaps can maintain wetter soil than small-and large-sized gaps because the total input of rainfall in small gap is less due to the interception loss and the large gap suffer increased evaporation and transpiration (Gray, Spies, & Easter, 2002;Ochiai, Okuda, & Sato, 1994). Therefore, the moist soil of the medium-sized gap might be possible to influence on the regeneration of P. armandii. Note: "+" indicates positive association, "−" indicates negative association, and "r" indicates random relationship; "J," "M," and "L" represent juvenile, medium-sized, and large-sized trees, respectively. TA B L E 2 Spatial associations of B. albosinensis juveniles with coexistent tree species in the four Plots of the mixed forest

| Intra-and interspecific relationships
Our results showed that, with a few exceptions (e.g., clustered distribution of large-sized trees at scales of 4-10 m in plot I), there was an evident shift in spatial distribution from aggregation to randomness along the life cycle for B. albosinensis ( Figure 6). The nonrandom mortality driven by intraspecific competition is also observed in other temperate forests (Duncan, 1991;Szwagrzyk & Czerwczak, 1993), supporting that density-dependent mortality is a prevailing mechanism in the dynamics of temperate tree populations. In addition, we also found a negatively spatial relationship of juvenile versus large B. albosinensis (Figure 7), suggesting the recruitment of B.
albosinensis was affected by conspecific large trees. It is inconsistent with the findings of Hou et al. (2004)  Spatial associations among species are indicative of interspecific interaction in the past (Rejmánek & Lepš, 1996 Besides B. utilis, the regeneration of tree species shows a distinct preference for gaps of different size (see Figure 5), suggesting that they have different microsites for regeneration. Hence, differential patterns of regeneration among these species may be responsible for their negative associations in regenerating site.

| What responsible for the persistence of B. albosinensis
Recruitment limitation is a key factor affecting the persistence of B.
albosinensis in the closed forests over the long term; it is therefore generally considered to be an unstable early-successional species.
However, the viewpoint is based on the inherent life-history characteristics without considering of variable local environmental condition. The direction of community succession depends on the comprehensive influence of species life history and specific abiotic condition, and environmental condition contributes more at some time (Clements, 1936;Connell & Slatyer, 1977). In the mixed forests of present study, although the current recruitment of B. albosinensis was deficient (indicated by the current unimodal pattern in size distribution), the long residence time of B. albosinensis (as inferred from the large size of these trees) in the canopy and its frequency as a gapmaker seems to create conditions that favor its own regeneration. In addition, the spatial segregation of seedlings of different tree species along the gap size gradient suggests the partitioning of gaps by size may play an important role in maintaining the stability of the mixed canopy. Gap partitioning is considered to be an important mechanism in maintaining the coexistence of multiple tree species since it can decrease the interspecific competition (Nakashizuka, 2001). In mature forest, interspecific competition in the seedling stage is crucial to the survival and growth of the pioneer species seedlings. Negative associations in the seedling stage of B. albosinensis with coexistent species suggest gap partitioning can help them avoiding the interspecific competition, which may also contribute a lot to the long-time persistence of B. albosinensis. Therefore, our results support the idea that long life spans, coupled with differences in seedling shade tolerance, and the creation of canopy gaps of different sizes by canopy trees are important factors for the nonequilibrium persistence of tree species mixtures (Gutiérrez et al., 2008;Lusk & Smith, 1998). Canopy disturbance and gap partitioning may redirect the fate of being replaced by B. albosinensis and allow them to persist in the mixed mature forest.

ACK N OWLED G M ENTS
We are grateful to Lingtong Kong, Di Kang, and Junjie Yang for their efforts in field works and the logistical support from Mt. Taibai National Nature Reserve. We also thank Xiaosi Guo for his help in species identification in the field. This study was financially supported by National Natural Science Foundation of China (31600337) and the Key Program of the National Basic Sci-Tech Special Fund of China (2007FY110800).

CO N FLI C T O F I NTE R E S T
No potential conflict of interest was reported by the authors.

AUTH O R CO NTR I B UTI O N S
YG and MY conceived and designed the experiments; YG and PZ conducted field work and analyzed the data; YG wrote the manuscript.

DATA ACCE SS I B I LIT Y
Sampling locations and gap investigational data: