Short‐term decline of Castanopsis fargesii adult trees promotes conspecific seedling regeneration: The complete process from seed production to seedling establishment

Abstract Declining forests usually face uncertain regeneration dynamics and recovery trajectories, which are challenging to forest management. In this study, we investigated the decline pattern of Castanopsis fargesii and examined the effects on conspecific seedling regeneration. We found that 61.45% of adult individuals were in decline and the smaller DBH size classes of trees (10–40 cm) had a greater probability of decline. Most of the intermediate decline (94.52%) and nondecline individuals (95.23%) did not worsen, and the crowns of 21.91% of the intermediate decline trees were recovered during 2013–2018. Adult tree decline had a negative effect on seed production (mean mature seed density of nondecline, intermediate decline, and high decline individuals was 167.3, 63.3, and 2.1 seeds/m2, respectively), but no effect on key seed traits. The seed survival rate of declining trees was greater than that of nondeclining trees at both the seed production and seed dispersal stages. The seed to seedling transition rates in canopy gaps, decline habitats, and nondecline habitats were 7.94%, 9.47%, and 109.24%, respectively. The survival rate and height growth of newly germinated seedlings were positively correlated with the light condition, which was notably accelerated in the canopy gaps. Taken together, these results indicate that the reduction in seed production of some adult trees had a weakly negative effect on new seedling recruitment, while the improved environmental condition after the decline significantly enhanced the survival and growth of both advanced and new germinated seedlings. Looking at the overall life history, the short‐term defoliation and mortality of some C. fargesii adult trees can be regarded as a natural forest disturbance that favors conspecific seedling regeneration. High‐intensity management measures would be unnecessary in cases of an emerging intermediate decline in this forest.


| INTRODUC TI ON
Widespread tree defoliation and mortality rates have increased considerably in recent decades (Allen et al., 2010;Cohen et al., 2016;Jung, Blaschke, & Oßwald, 2000;van Mantgem et al., 2009; Vilà-Cabrera, Martínez-Vilalta, Galiano, & Retana, 2013;Vindstad, Jepsen, Ek, Pepi, & Ims, 2019). Although the underlying causes for this phenomenon are not known for certain, they are frequently associated with global environmental changes, such as more intense and frequent droughts, higher temperatures, outbreaks of invasive pests and pathogens, or the interactions among these factors (Carnicer et al., 2011;Heitzman, Grell, Spetich, & Starkey, 2007;Kayes & Tinker, 2012;Klutsch et al., 2009;Kurz et al., 2008;Loo, 2009;Thomas, Blank, & Hartmann, 2002;Vindstad et al., 2019). These factors share a common feature in that they tend to induce species selection and individual removal, leading to changes in the species composition of the community in the short term in conifer-or hardwood-dominated forests (Aynekulu et al., 2011;Collins, Rhoades, Hubbard, & Battaglia, 2011). The management of declining forests has attracted the attention of scientists, forestry practitioners, and policymakers; however, our current understanding of management strategies for declining forests is still limited.
Tree defoliation and mortality can induce a series of changes in local environmental conditions (e.g., higher solar radiation at the forest floor, more extreme temperatures, and accelerated nutrient fluxes in the soil) that alter the probability of establishment of conspecific tree seedlings (Ibáñez et al., 2017;Kayes & Tinker, 2012;Redmond & Barger, 2013). For example, an increase in radiation levels and drought stress in the gaps that open after tree death sometimes preclude the establishment of late-successional shade-tolerant species, and indirectly favor pioneer, drought-tolerant species that can lead to changes in the forest structure (Diskin, Rocca, Nelson, Aoki, & Romme, 2011;Ibáñez et al., 2015;Redmond & Barger, 2013).
Moreover, the trajectories of recovery after drought-or insect-driven tree mortality may depend on advanced regeneration being established prior to the disturbance (Collins et al., 2011;Kayes & Tinker, 2012;Redmond & Barger, 2013). For example, tree dieback in mature forests can release suppressed saplings of shade-tolerant species, which allows the late-successional species to continue dominating the stands and indirectly limiting the establishment of light-demanding pioneer species that would otherwise be typical of the disturbed sites (DeRose & Long, 2010). These examples illustrate the complexities inherent to the postmortality regeneration dynamics, where species with different niches and seedlings at different age stages tend to respond differently (Galiano, Martínez-Vilalta, Eugenio, Granzow-de la Cerda, & Lloret, 2013).
New seedling recruitment dynamics is a little studied aspect of declining forests, despite the knowledge being important for forest management (McDowell, Ryan, Zeppel, & Tissue, 2013). New seedling recruitment is a multiphase process involving seed production, seed dispersal, seed germination, seedling emergence, and early survival of seedlings (Wang & Smith, 2002;Yang, Huang, Qian, & Fukuda, 2015). The accurate quantification of the impact of canopy tree decline at each stage of the seedling recruitment process is significant for predicting the long-term dynamics and management of declining forests. Although earlier studies have shown that adult tree defoliation can cause a reduced seed production due to a smaller gain in carbohydrate (Nakajima & Ishida, 2014;Palacio, Hoch, Sala, Körner, & Millard, 2014;Vilà-Cabrera, Martínez-Vilalta, & Retana, 2014), little is known about the possible effects on seed traits (e.g., seed mass and nutrient content) that play decisive roles in seed survival and germination and seedling early competition (Lebrija-Trejos, Reich, Hernández, & Wright, 2016). At the seed dispersal stage, predation and fungal infections are the main seed-killing agents that determine the success of the seed to seedling transition (Tomita, Hirabuki, & Seiwa, 2002;Yang et al., 2015). Both of these factors are negatively density-dependent and easily influenced by local environmental conditions (e.g., soil moisture and canopy openness) postdecline (Yang et al., 2015). Nevertheless, we have scant knowledge of the effects of tree defoliation and the subsequent changing environmental conditions on the fate of seeds at the seed dispersal stage. Accordingly, studies need to be conducted to reveal the underlying ecological mechanisms involved in the process of new seedling input postforest decline.
Subtropical evergreen broad-leaved forests are recognized as an important global vegetation formation type that contributes to the sustainable development of subtropical regions (Song & Da, 2016).
In recent years, cases of Castanopsis fargesii (one of the most widely distributed dominant species in evergreen broad-leaved forests) decline have been seen along the Yangtze River. The decline of C. fargesii dominated forests would pose a great threat to the integrity of the forests and would further influence the supply of ecosystem services. Although C. fargesii is recognized as a late-successional and shade-tolerant species (Cornelissen, 1993;Yang et al., 2015), its regeneration depends on the understory light condition. Since 2008, a serious decline of C. fargesii has been found in the Jinyun Mountain National Nature Reserve in the middle and upper reaches of the Yangtze River. Sano et al. (2014) and Takahashi et al. (2014) examined pathogenic bacterial infections as possible causes of C. fargesii adult tree decline. In the current study, we address the pattern of decline of C. fargesii adult trees, and assess the effects of the adult tree decline on seedling establishment, including seed production, seed fate at the seed dispersal stage, seedling emergency, and seedling survival and growth. We tested the following three predictions: (a) The decline of C. fargesii has a negative effect on seed production but has no effect on seed traits. (b) The seeds of declining adult trees have a higher survival rate at the seed production stage and seed dispersal stage, which would increase the seed to seedling transition rate. (c) Improved light conditions for establishment beneath the canopies of declining trees contribute to both newly germinated seedlings and advanced seedling survival and growth. By testing these predictions, we can confirm whether or not the defoliation and mortality of C. fargesii adult trees can be identified as short-term natural disturbances that would benefit short-term conspecific seedling regeneration.

| Study site and study species
We conducted this study on Jinyun Mountain (29°49′N, 106°20′E), a national nature reserve in Chongqing in Southwestern China. The site has a subtropical monsoon climate with an annual mean temperature of 13.6°C, a mean January temperature of 3.1°C, and a mean August temperature of 24.3°C. The annual mean precipitation and relative humidity are 1,611.8 mm and 87%, respectively . Vegetation on Jinyun Mountain is composed of typical middle-subtropical evergreen forests, which are codominated by C. fargesii, Machilus pingii, Symplocos setchuensis, and Castanopsis carlesii var. spinulosa (Yang et al., , 2015. Adult C. fargesii trees can attain 30 m in height. Mean seed mass of C. fargesii is about 0.55 g and ranged from 0.1 to 1.8 g (Yang et al., 2015). Seeds are primary dispersed by gravity and secondary dispersed by rodents such as squirrels and mice (Du, Guo, Gao, & Ma, 2007). The fruits ripen from October to November, and the peak seed rain occurs between October and November (Yang et al., 2015). The emergence of seedlings usually starts in July and August and ends in late November.

| Fieldwork
In September 2013, we established a 0.5 ha (50 m × 100 m) core monitoring plot and eight 30 m × 30 m sampling plots that were dominated by C. fargesii to monitor the long-term dynamics of declining forests ( Figure 1a). Detailed information about the core plot and eight sample plots is summarized in Table 1. This sampling design was adopted because evergreen broad-leaved forests generally occur in scattered patches, making it difficult to find a large area of continuously distributed C. fargesii forests at our study site. Monitoring was done to assess: in all plots, we determined the vigor levels based on the percentage of crown defoliation, using a semi-quantitative scale (Linares & Camarero, 2012). Six tree vigor levels were defined for this study:

| Survey of advanced seedlings
In September 2013, we set up 30 circular advanced seedling (existing before our study) subplots (radius = 4. In December 2013, all advanced seedlings (H < 1.3 m, existing before this study) in these subplots were investigated by determining their ages (by counting the bud scars on the seedlings) and measuring the heights. We defined three seedling stages based on the ages of the advanced seedlings in each circular subplot: small seedling, age ≤5 years; intermediate seedling, 5-10 years; large seedling, age >10 years (Yang et al., 2015). The annual ring analysis showed that the decline of C. fargesii adult trees and the formation of canopy gaps in the study plots were earlier than 2000.
Consequently, most seedlings in the CGA habitats were established after the decline.

| Survey of seed production and seed traits
We investigated seed production of the 30 selected C. fargesii adult trees using seed traps. The seed traps were made of a funnel of polyethylene cloth (1-mm mesh), with a receiving area of 0.5 m 2 (1 m × 0.5 m) and a height of 1 m above the ground (Yang et al., 2015).
Considering the terrain, crown shape, and vine coverage, we set up two or three seed traps in each subplot at a distance of 1.5 m from each adult tree (72 seed traps in total, Figure 1). We fully considered the distance between two selected adult trees to ensure that a certain distance (>3 m) was present between the seed traps and the adjacent adult tree canopies. In addition, C. fargesii seeds are about 0.5 g, and the horizontal dispersal distance is very close by seed rain.
Therefore, the risk that seeds in the seed traps had come from adjacent adult trees was very low.
The seeds that accumulated in the traps were collected at 1-week intervals from October 2014 to January 2015 (until no seeds accumulated for more than 2 weeks). The collected seeds were air-dried and sorted. We categorized the collected seeds as mature seeds (intact, without any injury), immature seeds (the embryo and cotyledons were incompletely developed), vertebrate-attacked seeds (mainly eaten by squirrels before falling), and insect-attacked seeds (suffered predation by invertebrate moth larvae with obvious wormholes on the seed surface) (Figure 2), and counted the total number of seeds in each category.
We combined the mature seeds of each adult tree collected every week. Among the combined seeds, we randomly selected 30 intact mature seeds from each of the 11 nondeclining adult trees and including fresh weight (seed mass), nutritional (starch), and defensive (cellulose and tannin) material properties. We did not measure the seed traits of high declining adult trees, because the collected seeds were too few to sample. Starch content was estimated using the optical rotation method following the national standard (NY/T11-1985) for cereals. Cellulose content was estimated using the acid and alkali washing method following the national standard (NY/T13-1986) for cereals (Wang, Zhang, Zhang, Li, & Yi, 2016). The tannic acid content was determined using the spectrophotometry method following the national standard (GB/T15686-2008) for sorghum (Wang & Chen, 2012).

| Survey of new seedling recruitment and environmental conditions
In April 2015 (before seed germination), we set up a 0.5 m 2 (1 m × 0.5 m) seed bank quadrat at the right side, close to the seed trap to investigate the seed fate at the seed dispersal stage (a total of 72 seed bank quadrats; Figure 1b). We collected the humus and topsoil of each soil seed bank quadrat and then separated and sorted C. fargesii seeds from the soil samples. The collected seeds were categorized into four groups after a careful examination: (a) viable seeds (already had an elongated radicle or healthy embryos and cotyledons); (b) fungi-infected seeds (killed by fungal infection, where the embryo and/or cotyledons were covered with hyphae and decayed); (c) desiccated seeds (killed by desiccation, where the roots and/or cotyledons were withered and brown), and (d) rodent-attacked seeds (mainly killed by rodents after falling) ( Figure 2). This classification was according to Tomita et al. (2002) and Yang et al. (2015).
In December 2015 (after seedling emergence), we established a 0.5 m 2 (1 m × 0.5 m) new seedling quadrat at the left side, close to each of the 72 seed traps (Figure 1b). We censused the newly germinated seedlings (hereafter, "new seedlings") within each seedling quadrat and measured their stem heights. In December, 2016, we revisited the seedling quadrats to investigate seedling survival and stem height.
Within each seedling quadrat, we measured the light conditions and soil conditions in 2016. We measured the photosynthetic photon flux density (PPFD) at 1.3 m above the forest floor using a LightScout

| Data analysis
The seed densities (seeds/m 2 ) of each seed trap were calculated as x i /y i × 100%, where x i was the total seeds collected from a seed trap, and y i was the area of a seed trap. The seed density of an F I G U R E 2 Seed fate pathways of C. fargesii in seed to seedling transition process adult tree was calculated as the mean value of all seed traps set under the tree.
We used a one-way ANOVA to test differences in seed production (seed density) and seed traits for the declining and nondeclining trees. We used Tukey's HSD to test differences in stem height and density of the advanced and newly germinated seedlings at each stage across different habitats. Before the analysis, all data were tested for normal distribution using the Shapiro-Wilk test and homogeneity was tested using the Breusch-Pagan test. Data for seed production were log-transformed to accomplish normality and homoscedasticity assumptions.
We tested the effects of the environmental conditions of the seedling quadrats on 2-year survival and stem height of newly germinated seedlings using a generalized linear mixed model (GLMM).
For the 2-year survival of seedlings, we used a binomial family, logit link model with living or dead seedlings (0/1) as the response variable, environmental conditions as the fixed effects, and tree ID as the random effect. For stem height of seedlings, we used a Gaussian distribution, log-link model with seedling stem height of seedlings as the response variable, environmental conditions as the fixed effects, and tree ID as the random effect. We used an odds ratio regression to determine the effect of each environmental condition on both seedling 2-year survival and height growth (Liang et al., 2016). An odds ratio (significance >1) indicated that environmental conditions increased the seedling survival and growth, and an odds ratio (significance <1) indicated that environmental conditions caused seedling mortality and decreased seedling growth.

| Seed production and seed traits
The density of mature seeds, immature seeds, vertebrate-attacked seeds, and insect-attacked seeds of declining adult trees was significantly lower than nondeclining trees ( Table 2). The proportion of immature seeds, insect-attacked seeds, and vertebrateattacked seeds of declining adult trees was lower than that of F I G U R E 3 (a) Distribution of percentage decline individuals across DBH size for sapling and adult trees and (b) comparisons of tree abundance in the three decline levels of adult trees. In (a), the numbers above each bar refer to the sample size in each group nondeclining trees (Table 2). No significant differences were seen in seed mass or starch content of seeds produced by declining or nondeclining adult trees (Figure 5a,b). However, the tannin content and cellulose content of seeds produced by declining adult trees were significantly lower than that produced by nondeclining trees (Figure 5c,d).

| Seed fate at seed dispersal stage
At the seed dispersal stage, the density of rodent-attacked seeds and fungi-infected seeds in NDE habitats were significantly higher than DE habitats and CGA habitats (df = 27, F = 7.24, p = .0030, Figure 6e; df = 27, F = 8.58, p = .0013; Figure 6d). The proportion of rodentattacked seeds in the NDE habitats was significantly higher than in the CGA and DE habitats (df = 27, F = 11.2, p < .01; Figure 6e). The proportion of desiccated seeds in the CGA habitats was higher than that of the NED and DE habitats (df = 27, F = 9.71, p < .01; Figure 6f).
The density of viable seeds in the CGA habitats was 3.24 seed/m 2 , which was 1.5-times that of mature seeds (Figure 6b). The density of new seedlings in the NDE habitats, DE habitats, and CGA habitats was 13.29 seedlings/m 2 , 6.01 seedlings/m 2 , and 2.29 seedlings/m 2 , respectively. In addition, the seed to seedling transition rate in NDE, DE, and CGA habitats was 7.94%, 9.47%, and 109.24%, respectively.

TA B L E 2
Comparison of seed density of the four types of seeds collected from seed traps from the three levels of declining adult trees There were more viable seeds than mature seeds, and the transition rate was higher than 100% in CGA habitats because some seeds were transported from other habitats (such as NDE habitats with high seed density) to CGA habitats by seed dispersers and seeds in CGA habitats had a higher survival rate.

| Survival and growth of new germinated seedlings
In 2015, the survival rate of new seedlings in the CGA habitat was significantly higher than that of the NDE and DE habitats. In 2016, the survival rates of new seedlings in the CGA and DE habitats were significantly higher than that of the NDE habitats (Figure 7a). While stem heights did not differ among the three habitats in 2015, stem height of new seedlings in the CGA was higher than that of the DE habitats by 2016, which was higher than that of the NDE habitats

| Distribution pattern of advanced seedlings and saplings
We recorded 1,247 advanced seedlings ( Younger offspring, such as small seedlings, were most abundant in the NDE habitats. Older recruits, such as the saplings, were more abundant in the CGA habitats than in the NDE habitats (Figure 8a).

Mean stem heights of intermediate seedlings and large seedlings in
the DE and CGA habitats were also significantly larger than those in the NDE habitats (Figure 8b).

| Pattern of decline of Castanopsis fargesii
In past decades, studies that addressed forest decline patterns were

| Effects on seed production
At the seed rain stage, negative effects were seen from the decline in adult trees on seed production, similar to the situation of Pinus

TA B L E 3
Comparison of environmental variables for the three types of habitat and the effects of environmental variables on seedling survival and stem height of new germinated seedlings in 2016, based on the GLMM analysis limits seed production (Palacio et al., 2014). The seed abortion rate (proportion of immature seeds) of declining trees, however, was relatively lower than that of nondeclining trees, suggesting that they had a higher fruiting efficiency. In addition, the proportion of vertebrateattacked and insect-attacked seeds of declining trees was lower than that of nondeclining trees, which suggests that their seeds had less survival pressure (e.g., vertebrate attack and insect attack) at the seed production stage. As a result, despite the reduction in absolute number of seeds, the intermediate decline individuals were still able to maintain a considerable seed production (over 60 seed/m 2 ).
However, the sharp reduction in seed production of high decline trees (2.1 seeds/m 2 ) could have an adverse effect on the seedling regeneration in canopy gaps, at least during the seed rain stage.
Negative effects of adult tree decline on seed mass and nutritional traits were not found, suggesting that seed quality from the declining seeds did not decrease. Interestingly, the content of defensive materials (e.g., tannin and cellulose) of the seeds produced by declining trees was significantly lower than that of the nondeclining adult trees. Thus, not even the reduced defensive material content had a negative effect on seed survival at the seed production and seed dispersal stages ( Figures 5 and 6). By combining these findings, except for severely declining individuals, the declining adult trees can still produce large numbers of mature seeds. More importantly, their key traits (i.e., potential germination capacity and competitiveness of seedlings) were not significantly reduced. In summary, no seed limitation was found in the declining adult trees. These results support Prediction 1.

| Effects on new seedling recruitment
At the seed dispersal stage, rodent attack was the main seed-killing agent, causing the death of most mature seeds in both nondeclining and declining habitats (Figure 6e). Rodent attack was a density-dependent seed-killing agent that caused seeds to face a higher risk of loss in vitality in the NDE habitats due to the higher seed density (Tomita et al., 2002;Zhu, Comita, Hubbell, & Ma, 2015). In contrast, seeds in the DE and CGA habitats had a higher survival rate due to their relatively low seed densities (Figure 7). Interestingly, a considerable number of seeds were transported into the CGA habitats by rodents, resulting in more viable seeds in the CGA habitats, compared to the input of mature seeds by seed rain (Figure 6b). Our recent research (unpublished seed dispersal experiment using 1,500 labeled seeds) also showed that about 5% of the total seeds under the canopies of the adult trees have been transported into forest gap habitats. These seed dispersals by rodents played an important role on seedling recruitment in the CGA habitats. At the seed germination stage, no significant difference was seen in the germination rate of viable seeds in the three types of habitats, indicating that the adult tree decline did not affect seed germination. Considering the whole seedling recruitment process, the seed to seedling transition rate in the DE and CGA habitats was significantly higher than that of the NDE habitats (9.47% and 109.24% vs. 7.94%; Figure 6d).
In summary, the higher seed to seedling transition rate in the NE and CGA habitats, to a great extent, reduced the negative effects of seed production on new seedling recruitment. These results support Prediction 2.

| Effects on seedling establishment
The survival and growth of new seedlings play an important role in seedling regeneration and community dynamics (Zhu et al., 2015).  Figure 7). In addition, an abundance of advanced seedlings and saplings existed in the community. The younger recruits (e.g., small seedlings and medium seedlings) were concentrated in the NDE habitats, while the older recruits such as large seedlings and saplings were concentrated in the CGA and DE habitats (Figure 8a). Thus, the abundance of small seedlings does not mean a higher efficiency of seedling establishment in the NDE habitats since most of the seedlings could not survive due to the unfavorable understory environment (Yang et al., 2015;Zhu et al., 2015).
In contrast, the few seedlings in the declining habitats, especially in the forest gaps, were more likely to occupy spaces generated by the death of adult trees due to the better understory environment. In total, even though the tree species is shade-tolerant and late-successional, the improved light conditions that follow a decline in C. fargesii can indeed promote the survival and growth of its seedlings. The higher seedling survival and growth rate in the CGA habitats further reduced the negative impact of drastic seed production reduction in adult trees decline on seedling regeneration. These results support Prediction 3.

| CON CLUS I ON AND MANAG EMENT IMPLIC ATIONS
Successful regeneration postdecline was determined by multiple factors, such as decline intensity, seed production capacity, new seedling recruitment, seedling survival, and sapling growth (Brown & Allen-Diaz, 2009;DeRose & Long, 2010;Ibáñez et al., 2015;Nakajima & Ishida, 2014). Regeneration failure can occur at any stage, causing the long-term alteration of community species composition and forest structure (Figure 9). The prediction of regeneration trajectories of declining species is not an easy task. In this study, we found that C. fargesii decline has some special characteristics, compared to that of Pinus and hardwood species-dominated forests, namely: (a) The decline of C. fargesii adult trees did not dramatically change the forest stand structure in the short term; (b) negative effects of adult tree decline were seen in seed production but no effects were seen on seed key traits; (c) although few seeds were seen in canopy gaps, the seed to seedling transition rate was relatively high due to the effective seed dispersal; (d) newly germinated seedlings in the canopy gaps and decline habitats had a higher survival rate and relative growth rate; and (e) a large number of well-grown advanced seedlings were distributed in the forest gaps. These positive phenomena suggested that the short-term defoliation and mortality of C. fargesii adult trees could be seen as a natural disturbance that promoted conspecific seedling regeneration. Hence, we predict that the canopy gaps occurring after adult tree mortality and decline will be occupied by conspecific individuals. Nevertheless, longerterm monitoring is necessary to confirm that these trends are maintained through time or if other mechanisms might prevent C. fargesii regeneration.
The evergreen broad-leaved forests were near-climax and codominated by some late-successional species such as C. fargesii, Machilus pingii, Schima superba, and C. carlesii var. spinulosa (Wang, Kent, & Fang, 2017). Moreover, the understory environmental condition of these forests was characterized by high humidity and low light intensity. The most stressful environmental condition that limited the establishment of seedlings occurred at the seedling stage.
The regeneration and maintenance of these dominant tree species depended mostly on improvements in the local environmental conditions, such as fine-scale disturbances and the creation of small gaps. In general, seed abundance and seed germination as limiting factors were not present for these species, even though some individuals were declining. There were abundant small seedlings under the crown of healthy adult trees. When the adult trees were in decline, these pre-established seedlings will be released due to the improved environmental conditions. Thus, we regard the intermediate decline of the evergreen broad-leaved forest as a forest disturbance that can promote the regeneration of its dominant species. We also believe that high-intensity management measures would be unnecessary in cases of an emerging intermediate decline in this forest.
Furtherly, the management of this mature forest could benefit from the creation of small gaps by cutting some early-successional trees to improve the late-successional species regeneration.

ACK N OWLED G M ENTS
This study was supported by Chongqing Technology Innovation and Application Demonstration Major Theme Special Project (cstc2018jszx-zdyfxmX0007) to YY, and the graduate research and innovation foundation of Chongqing, China (Grant No. CYB18038) to LH.

CO N FLI C T S O F I NTE R E S T
The authors declare no conflicts of interest.

DATA AVA I L A B I L I T Y S TAT E M E N T
The data that support the findings of this study are available at the Dryad Digital Repository (https://doi.org/10.5061/dryad.dncjs xkx4).