How does species rarity influence the functional diversity and resilience along the successional pathway in heterogeneous karst forests?

Abstract

Global climate change and human activities have increased the risk of rare species loss due to their sensitivity to disturbances. Considering their dominance in hyper-diverse communities, the contribution pathways and extent of rare species loss to functional diversity and resilience should be clarified. Therefore, we established 30 forest dynamic plots in heterogeneous degraded karst forests and measured the plant functional traits along the successional pathway. The correlations between trait uniqueness and species rarity were quantified, and the effects of rare and common species losses on functional diversity and trait networks were simulated. The correlations between trait uniqueness and species rarity showed that rare species tended to occupy marginal functional space positions in the mid- and late successional stages. In addition, rare species mainly supported functional redundancy in the early successional stage and significantly impacted functional resilience. In contrast, the unique traits supported by rare species increased along the successional pathway, thus increasing functional diversity and resilience in the late successional stage. This study highlights the crucial role of rare species in the functional diversity and resilience of degraded karst forests, warranting increased attention to rare species during biodiversity conservation and ecological management in heterogeneous, vulnerable ecosystems.

1. Introduction

Global biodiversity is declining at unprecedented rates [1], necessitating prompt conservation efforts. Previous studies demonstrated the dominance of rare species within hyper-diverse communities, which exhibited lower abundance and limited distributions [2]. Since rare species are more vulnerable to natural and anthropogenic disturbances, they face higher risks of extinction and species loss within the community [3]. A growing body of evidence supports the more important role of rare species in maintaining ecosystem services and biodiversity [4,5]. However, the contributions of rare species to ecosystem functioning [6] or resilience [7] remain poorly understood compared with common species.

Researchers have demonstrated that rare species dominate at least some ecosystem processes because they often occupy marginal functional space within the community and capture infrequent resources [5,8,9]. Such unique patterns result in their distinctive contributions to ecosystem functioning. Moreover, the differential resource utilization between rare and common species enhances community resource utilization and ecosystem productivity [10–12]. However, the contribution pathways and extent of rare species to ecosystem functioning require further discussion [3–5,13]. Since functional diversity often serves as a key driver of ecosystem function, distinguishing the effects of common and rare species on plant functional diversity may effectively reveal species–functioning relationships [14]. Several studies emphasized the greater contributions of rare species to functional trait divergence and functional diversity, despite low abundances, especially when considering functional diversity in conservation and restoration decisions [3,4,6,7,15]. However, functional diversity alone often proves insufficient to reveal the disproportionate roles of rare species, specifically their complex relationships with functional traits and ecosystem functioning [16], due to their high susceptibility to loss, which potentially disrupts interactions among species [3,17–19]. By contrast, network analysis may effectively quantify the unique roles of rare species in mediating these relationships within a community [4,20]. The locations and connections with other species in the network can represent the roles of a species in functional diversity and resilience [16,21]. Previous studies suggested that losing rare species with unique traits may have a greater impact on functional diversity, while losing rare species with redundant traits may have a greater impact on functional resilience [3,4,6]. Nonetheless, evidence of the impact of rare species loss on ecosystem functioning resilience is still lacking.

Species rarity can predict extinction risks as rare species are sensitive to environmental disturbances [22,23]. Thus, studying species rarity patterns along temporal or spatial gradients can provide critical information on ecosystem changes [13]. However, the roles of various species can covary along spatio-temporal gradients, confusing the species–functioning relationship [23]. Ubiquitous disturbance regimes can cause varying degrees of vegetation degradation and alter community species composition [24–26]. Post disturbances, the disturbed forest exhibits a successional pathway, generally tending to have richer species compositions, more complex structures and more stable ecosystem functioning over time [27–29]. The roles of rare and common species may also transform with altered species composition, thus varying the proportion of rare species along the successional pathway. With constant species loss and turnover along the successional pathway, however, the interaction between organisms and the environment may alter their similarities and differences [30], and some ecologists attributed such alterations to plant functional traits [28,31]. The strong habitat filtering in the early successional stage limits the range of traits [28], resulting in trait convergence between rare and common species, thereby contributing to functional redundancy [6]. In the succession process, the alleviated environmental conditions allow the establishment of more species, and the severe limiting similarity gradually dominates the community assembly rules, thus increasing functional evenness [31]. Other ecologically similar species possessing unique traits to adapt to distinct niches may replace rare species facing competitive disadvantages, which can increase functional diversity [32]. Nonetheless, functional diversity may decrease due to the increased functional redundancy under greater species richness [33,34]. Therefore, the mechanisms driving the variations in functional diversity and resilience along the successional pathway remain unclear, as is the impact of rare species on these two matrices. Furthermore, species resource acquisition strategies reflect the sensitivity of plant functional traits to abiotic environmental changes, which varies across vegetation types [35,36]. Nutrient limitations in early succession may prompt species to adopt acquisitive strategies, whereas dispersal limitation in late succession can restrict colonization despite favourable abiotic conditions [37,38]. Whether a rare species possesses unique traits during succession is significantly related to its resource acquisition strategies [39]. Therefore, studying the variations in the traits and adaption strategies of rare species highly sensitive to environmental changes under heterogeneous conditions with relatively longer environmental gradients may effectively reveal the underlying mechanisms.

The karst region in Southwest China is ecologically fragile, where severe stone desertification significantly constrains the regional ecological environment [40]. Poor water retention, low soil fertility and discontinuous soil cover often result in a unique combination of temporary water shortages and extreme micro-habitat heterogeneity in karst forests [41,42]. Rare karst forest species, more susceptible to environmental changes, may exhibit different resource utilization patterns and adaptive strategies [23,43,44]. Compared with the subtropical evergreen broad-leaved forests at the same latitude, the subtropical karst evergreen and deciduous broad-leaved mixed forest has richer species and hyper-diverse microhabitats, possibly leading to a large proportion of rare species within the community [44]. Recent studies revealed the unique species composition, functional traits, ecological strategies, spatial community structure and successional pathway of karst forests compared with subtropical evergreen broad-leaved forests [26,34,45,46]. However, it is still unclear how the presence and loss of rare species affect functional diversity and resilience in such heterogeneous forests with higher environmental gradients. Rare species may support functional redundancy and resilience due to trait convergence under stronger habitat filtering in the early successional stage. Meanwhile, they could largely impact functional diversity due to trait uniqueness under stronger habitat heterogeneity in the late successional stage. Thus, we established forest dynamic plots (FDPs) in heterogeneous degraded karst forests and measured the plant functional traits along the successional pathway. By simulating the impact of rare species loss on functional diversity and resilience, we aimed to verify the following hypotheses: (1) rare species possess unique traits and occupy infrequent niche spaces; (2) rare species loss has a greater impact on functional diversity than common species; and (3) loss of rare species located at centre positions of the trait network could have severe impacts on functional resilience.

2. Material and methods

(a). Study area and data collection

The study area is the Maolan National Nature Reserve (107°52′ E to 108°05′ E and 25°09′ N to 25°20′ N) in Southeast China. Adjoining Mulun National Nature Reserve, it collectively forms the world’s best-preserved and most contiguous karst primary forest [44]. The area features a typical subtropical climate, with an annual temperature of 15.3℃ and annual rainfall of 1752.5 mm. Due to poor economic conditions and low ecological protection awareness, random selective cutting for firewood and grazing were the main disturbances within the reserve in the past [47], which were relatively mild and usually caused degradation of the shrub-tree mixed forests. The disturbances ceased since the establishment of the natural reserve and the relocation of settlements in 1984 [47]. The disturbed forest has undergone secondary succession and recovered to a relatively stable community with subtropical evergreen and deciduous broad-leaved mixed forest, mainly structured by Celtis sinensis, Euonymus alatus and Zelkova serrata.

To investigate the contributions of plant species to functional diversity and resilience during three successional stages, we established 10 FDPs (30 m × 30 m) in each successional stage within the natural reserve from June to September 2022 according to the standard handbook of FDP establishment [48]. The shrub-tree mixed forest (SC) was considered the early successional stage, in which the last random selective cutting for firewood occurred around 20 years ago in 2004, and the vegetation propagule was not destroyed. The secondary forest (SG) was considered the mid-successional stage, in which the same human interference (random selective cutting for firewood) last happened around 40 years ago in 1984. The old-growth forest (OG) was considered the late successional stage, in which no human disturbance occurred for over 100 years based on previous field investigations and interviews. Throughout the succession, the same regional species pool was shared, and no human interference occurred (see electronic supplementary material, figure S1 for geographic locations of the study area and sampling sites). The coordinates of each FDP were confirmed using a Real-Time Kinematic (RTK) positioning device (DJ-Innovations Ins, Shenzhen, China), as were the vertex coordinates of the 10 m × 10 m grids within them. The FDPs were spaced at least 50 m apart to minimize the confounding effects of spatial autocorrelation and ecotone. In each FDP, all woody individuals with a diameter at breast height (DBH) ≥1 cm were tagged, located, measured and identified to species.

We selected nine plant leaf functional traits to determine functional diversity, namely, specific leaf area (SLA, cm² g⁻¹), leaf dry matter content (LDMC, g g⁻¹), leaf nitrogen concentration (LNC, g kg⁻¹), leaf phosphorus concentration (LPC, g kg⁻¹), leaf carbon concentration (LCC, g kg⁻¹), leaf thickness (LT, mm), leaf carbon/phosphorus ratio (leaf C:P), leaf carbon/nitrogen ratio (C:N) and leaf nitrogen/phosphorus ratio (N:P). These functional traits have been demonstrated to reflect the performance of individual plants in adapting to environmental conditions [49,50]. Moreover, these traits encompass competitive and cooperative relationships with co-occurring organisms and associate closely with ecosystem functions like nutrient cycling, productivity, carbon storage and soil fertility [43,51].

According to the standardized plant functional trait measurement handbooks worldwide [52], at least three individuals (all individuals if the abundance was below three) of each tree species were randomly selected from within each FDP as the sample. At least three to five fully expanded, healthy and fully exposed to light leaves were collected from each individual in a sample [52]. SLA is a key trait reflecting the species’ resource acquisition strategies. A high-speed photographic apparatus (GK821, Deli, China) was employed to photograph the leaves, and ImageJ (National Institutes of Health, USA; https://imagej.nih.gov/ij) and the ‘LeafArea’ package in R [53] were used to calculate the leaf area (cm²). An electronic balance was used to weigh the fresh weight of the leaves. Then, the leaves were placed in envelopes and dried at 80℃ for at least 48 h in an oven (YT-700, YETO, China) before dry weight measurement. SLA and LDMC were calculated based on the measured leaf area, fresh weight and dry weight. LT was determined using an electronic digital vernier caliper. LNC and LCC were determined using elemental analysers (UNICUBE trace, Elementar, Langenselbold, Germany). Finally, LPC was determined spectrophotometrically at 700 nm using a continuous flow automated analyser (AA3, Bran + Luebbe, Germany).

(b). Statistical analyses

(i). Species rarity quantification along the successional pathway in degraded karst forests

We quantified species rarity using a fuzzy clustering algorithm based on species abundance and frequency [13]. Briefly, the algorithm calculates the dissimilarity among species using the Gower distance via the ‘daisy’ function and performs fuzzy clustering analysis using the ‘fanny’ function in the ‘cluster’ package to classify species into two clusters (common versus rare species). We set the number of clusters (K) to 2 based on the requirement $K\ge n/2-1$ (where n is the number of species), necessitating n ≥ 6. Additionally, fuzzy membership degree (ranging from 0 to 1) of each species in the rare species cluster was calculated, representing the species rarity index in the community [13]. A rarity value closer to 1 indicates that the species is rarer, while a value closer to 0 indicates it is more common (see electronic supplementary material, figure S2 for species rarity quantification). We performed all calculations using the ‘fuzzyq’ function in the ‘FuzzyQ’ package [13].

(ii). Species positions in the functional trait space along the successional pathway in degraded karst forests

The functional trait space shows the distribution and abundance of species functional traits in a multivariate space [54]. When building the functional trait space, we performed Z-score standardization to preprocess species average functional trait data, thus ensuring comparability across traits of different scales. Then, we conducted principal component analysis (PCA), and the first two principal components (PCs) were retained for subsequent analyses since they captured most of the variance [54]. We used PC scores to perform kernel density estimation, constructing a functional space based on trait probability density. In this study, we set the probability threshold to 0.95. Arrows in the functional space represent original variables, with their direction indicating the correlation between the original variables and principal components, and their length representing the variance explained by the original data for the principal components. To understand where rare species are located in the functional space, we use the mean position of the species in the functional trait space as the centroid to calculate its Euclidean distance (electronic supplementary material, equation S1) to each species. Additionally, we performed linear regression to observe whether species rarity influenced the distance of each species to the centroid of the functional trait space.

(iii). Impacts of species loss on functional diversity along the successional pathway in degraded karst forests

We calculated the functional distances between species based on the Gower distance (electronic supplementary material, equation S2) [55], and quantified the functional space quality by the deviation between the original trait distances and the spatial distance after dimensionality reduction, revealing that higher quality in a five-axis functional space was exhibited in all three successional stages. Three complementary indices of functional diversity were conducted in the framework based on five-axis functional space: functional richness, functional specialization and functional originality. Functional richness is the convex-hull volume of the functional space filled by all species within communities during successional stages, indicating the range of trait combinations [56]. Functional originality is the average distance between one observation and all others, reflecting the degrees of uniqueness of species traits in the community [3]. Functional specialization is the mean Euclidean distance between each species and the centroid of the functional space, reflecting the adaptation strategy of the species to specific environmental conditions or ecological niches [3]. To reveal the impact of rare and common species loss on functional diversity, we separately simulated the impact of three species loss scenarios on three functional diversities. We quantified functional richness loss by sequentially removing species in the descending order of rarity (i.e. rarest first) and calculating changes in the convex-hull volume of the trait space using the ‘convhulln’ function in ‘geometry’. For functional specialization, we calculated species’ Euclidean distances to the multivariate mean of the trait space, standardized the distance values of each species and tracked changes in the mean distance after each removal. We evaluated functional originality using nearest neighbour distance, calculated as the minimum Euclidean distance to any other species in the trait space. We standardized these distances by the maximum observed nearest neighbour distance, and their mean values were tracked across removal sequences. Control simulations of species loss from common to rare were conducted by reversing the removal order [3]. In the random species loss scenario, we shuffled the order of species loss 999 times while keeping their trait values constant. After each removal, the impact on the three functional diversities was calculated. We computed the median of the null model matrix to reflect data centrality, and the 0.025 and 0.975 quantiles were used to construct the null model confidence interval. By comparing the observed values with null models, it is possible to distinguish the impacts of losing rare and common species on functional diversity.

(iv). Impacts of species loss on the functional trait network along the successional pathway in degraded karst forests

Researchers have widely used network analysis to estimate microbial community stability [57–59], yet they less commonly use it to quantify the variations in functional trait networks in plant communities [60]. Ecologists have constructed functional trait networks with plant inter-trait relationships as the edges and plant traits as the nodes to capture and visualize the relationships among plant functional traits [20,60,61]. Such functional trait networks proved effective in assessing the importance of a specific trait, the interdependency of multiple traits and the overall phenotypic integration [18], but showed limitations in reflecting the impacts of species attributes on the networks. Setting the species as nodes and trait similarities as positive connections offers a greater capacity to assess species contributions to functional resilience [19,62]. Thus, we calculated the correlation coefficient matrix of the mean values of all species functional traits within the communities across the succession stages, and used the absolute values of the Pearson correlation coefficients to describe the correlation strength. To avoid false correlation, we set the significance threshold of the correlation coefficient as R² > 0.6 and p < 0.05. We constructed adjacency matrices by assigning 1 to significant correlations and 0 to non-significant ones, thereby defining network edges between species nodes. We established an undirected network by ignoring the directivity between nodes and calculated a series of topological coefficients to represent the complexity of the network (see electronic supplementary material, table S1 for topological coefficient quantification). Lower values of diameter and average path length, combined with higher values of mean clustering coefficient, network density and average degree, indicate higher network complexity. Additionally, the number of edges and nodes should be incorporated for a comprehensive evaluation. We calculated the Zi-Pi values of the network. Zi (within-module connectivity) measured the connection density of a node with other nodes in its module, while Pi (among-module connectivity) quantified the distribution of connections between a node and nodes in other modules [63]. In our study, we identified nodes with Zi < 2.5 and Pi > 0.62 as keystone nodes, which linked different modules. We considered these modules as species groups with similar functional traits or niches within the community. Thus, these keystone nodes probably contribute to maintaining the structural resilience of the plant community while also reflecting trait correlations between the keystone species and multiple functional groups. Besides, we simulated the loss of 50% of the rarest or most common species separately. By comparing network stability following the species loss, we can reveal how the loss of rare and common species affects functional trait networks.

All analyses were performed using ‘FuzzyQ’, ‘funspace’, ‘ggraph’, ‘mFD’, ‘ape’, ‘ggClusterNet’, ‘igraph’, ‘geometry’ and ‘vegan’ packages in R 4.3.2 [53].

3. Results

(a). Species positions in functional trait space along the successional pathway in degraded karst forests

The functional traits exhibited consistent correlations in the three successional stages. Specifically, specific leaf area, leaf nitrogen concentration, leaf phosphorus concentration and leaf nitrogen-to-phosphorus ratio had positive correlations, and leaf carbon concentration, leaf dry matter content, leaf thickness, leaf carbon-to-phosphorus ratio and leaf carbon-to-nitrogen ratio were positively correlated. Furthermore, leaf nitrogen concentration, leaf carbon-to-nitrogen ratio and leaf nitrogen-to-phosphorus ratio contributed the most to the primary axis of the functional trait space during the early successional stage and mid-successional stage (figure 1A,B); in contrast, the leaf nitrogen concentration, leaf carbon-to-nitrogen ratio and specific leaf area were the primary contributors in the late successional stage (figure 1C, electronic supplementary material, table S2). We also observed a significant positive correlation between the rarity index of species and their distance from the centroid in the functional trait space in the mid-successional and late successional stages (figure 1D), indicating that rare species occupied a unique trait space.

Figure 1.

The functional trait spaces along the successional pathway in degraded karst forests. SC represents the early successional stage (A), SG represents the mid-successional stage (B), OG represents the late successional stage (C) and the relationships between the functional spatial centroid distance of species and their species rarity index (D). In panels A–C, points represent species in the functional space where the greener the colour of the point, the lower the species rarity is, while the bluer the colour of points, the higher the species rarity is. The background gradient from yellow to red represents the functional trait space. The redder colour indicates the higher trait probability density. SLA is for specific leaf area; LDMC is for leaf dry matter content; LNC is for leaf nitrogen concentration; LPC is for leaf phosphorus concentration; LCC is for leaf carbon concentration and LT is for leaf thickness; C:N is for leaf carbon-to-nitrogen ratio; C:P is for leaf carbon-to-phosphorus ratio; N:P is for leaf nitrogen-to-phosphorus ratio. In panel D, the circles are the values of species, with the solid and dashed lines representing the significant and non-significant results of the correlation analysis. NS, non-significant; *p < 0.05; **p < 0.01.

(b). Impacts of species loss on functional diversity along the successional pathway in degraded karst forests

The comparison between observations and null models, demonstrated that rare species loss influenced functional diversity more noticeably than common species loss. Furthermore, functional richness underwent an accelerated decline when species were lost from the rarest to the most common, compared with random species loss. The mid-successional stage exhibited the highest susceptibility among the successional phases, with significantly lower functional richness when 13.9% of rarest species were lost than in the random loss scenario. Notably, random species loss exhibited significantly low functional richness when 16.5% of the rarest species were lost in the late successional stage. In contrast, no significant difference was observed in this factor across most loss levels during the early successional stage. Compared with random species loss, functional richness decreased by 40.5% in the late successional stage and 27.9% in the mid-successional stage when 25% of the rarest species were lost. In contrast, the relative decline in the early successional stage was 13.7%. The loss of the rarest species diminished functional originality and specialization, whereas the loss of the most common species obtained the opposite results. In the case of approximately 50% of the rarest species lost in the early successional stage, the functional specialization was significantly lower than in the random species loss scenario. Additionally, functional specialization dropped by 7.7% in the late successional stage and 5.1% in the mid-successional stage when 50% of the rarest species were lost compared with random species loss. The corresponding reduction in the early successional stage represented 3.1%. The trends of functional originality and specialization coincided across the three stages (figure 2).

Figure 2.

Impact of species extinction on the functional diversity (functional richness, specialization, originality) along a successional pathway. The sequential loss of the rarest species (blue solid line) is compared with a scenario where the most common species are lost first (green dashed line), with a null scenario simulating a random sequential extinction (orange line indicates the median of this scenario among 999 replicates and the 95% confidence interval as the shaded area). SC, early successional stage; SG, mid-successional stage; OG, late successional stage.

(c). Impacts of species loss on the functional trait network along the successional pathway in degraded karst forests

The functional trait network structure in the early successional stage was more complex than in the rest. Handeliodendron bodinieri, Tirpitzia sinensis and Ligustrum lucidum, among others, constituted the keystone species within the network, while Callicarpa giraldii, Bridelia retusa and Xylosma longifolia primarily contributed to the connections in the late successional stage (electronic supplementary material, figure S5). In the mid-successional stage, the functional trait network exhibited relatively low structural complexity, with no identified keystone species (electronic supplementary material, figure S5). The loss of the rarest species in the early successional stage significantly diminished the complexity and connectivity of the functional trait network (electronic supplementary material, figures S5D, S5E; figure 3A). Similar reductions were found due to losing the most common species in the mid-successional stage (electronic supplementary material, figures S5F, S5G; figure 3B). In the late successional stage, functional trait network parameters decreased slightly when the rarest species were removed compared with the loss of the most common species (electronic supplementary material, figures S5H, S5I; figure 3C).

Figure 3.

Network topological properties along a successional pathway. The topological properties are for the early successional stage (A), the mid-successional stage (B) and the late successional stage (C).

4. Discussion

(a). Species positions in functional trait space along the successional pathway in degraded karst forests

Previous research has suggested that variations in plant functional traits were indicative of the various strategies employed to adapt to resource availability [51,54]. Our results unveiled that the dominant plant functional traits changed over successional stages, suggesting that the plant strategies varied along the successional pathway. Consistent with Leitão et al. [3], the current research indicated that rare species tended to occupy marginal positions within the functional space. Our findings also revealed a significant positive correlation between species rarity and their distances to the centroid within the functional space during the mid-successional and late successional stages (figure 1D). Rare species were less competitive regarding environmental adaption and resource utilization compared with common species [23]. Thus, common species with similar strategies may occupy rare species’ potential niches [32]. Consequently, they had a narrow distribution and occupied marginal functional niches in the community [2,64].

We also found that the proportion of rare species with unique functional traits increased significantly along the successional pathway (figure 1; electronic supplementary material, figure S2). Changes in the community assembly rules might contribute to such patterns. The convergence of traits between rare and common species occurred in the early successional stage due to enhanced habitat filtering, leading to fewer rare species possessing unique functional traits and a narrow functional trait space [6,28]. Conversely, the increasing species richness in the late successional stage can expand the functional trait space [34,65], heightening the occurrence possibility of unique functional traits [6]. Moreover, the limiting similarity became dominant with the increase in soil fertility along the successional pathway in the karst forests, enabling rare species to develop more unique traits to adapt to specific microhabitats and avoid competition [31,32,66,67]. The stable environmental conditions in the late successional stages facilitate the growth of large trees. However, this type of stability also could intensify resource competition, driving the competitive exclusion of species without unique functional traits [68,69], which resulted in rare species with large size and unique traits being more prevalent, as they could persist under highly competitive pressure compared with smaller individuals without such traits [39,70,71]. Furthermore, the abiotic environment typically affects the adaptive strategies of plants [50]. Some studies focusing on tropical dry forests showed that the dominant traits shifted from conservative strategies to acquisitive strategies [50]. Due to temporary water shortage and extreme micro-habitat heterogeneity in karst areas, plants develop conservative traits to enhance drought tolerance, maximize photoprotection and mitigate water loss in the early succession (figure 1; electronic supplementary material, figure S4, table S2) [42,72,73]. Several studies suggested that rare species, which have unique functional traits and occupy the marginal niche or position, may develop additional traits contributing to resource acquisition and growth [7,39]. However, belowground traits and reproductive traits might further shape the observed patterns in functional uniqueness, which were not measured in this study. Root traits have comparable impacts on ecosystem functioning to aboveground traits, and they might explain the part of the variation that is independent of aboveground traits [74]. Similarly, the relationships among seed dispersal distance, seed mass and plant height vary significantly across dispersal modes [75]. Thus, belowground traits and reproductive traits might lead to confusing impacts on the functional roles of rare and common species.

These results support hypothesis 1 that the proportion of rare species with unique traits increases along the successional pathway in the karst region, possibly due to transition in species’ adaptation strategies induced by biotic and abiotic factors. However, species with lower abundance preferentially establish and survive relative to those with higher abundance, leading to increased abundance of rare species during successional stages, which may also increase the probability of rare species with unique traits [76].

(b). Impacts of species loss on functional diversity along the successional pathway in degraded karst forests

The redundancy hypothesis posited that the ecosystem functioning might be buffered due to species redundancy, where the loss of specific species can be compensated by those performing similar roles [77]. By comparison, losing species with unique functional traits might severely threaten ecosystem functioning due to their irreplaceableness [7]. This research found that functional diversity experienced a more pronounced decline under the rarest species loss compared with random species loss (figure 2). This observation indicated that losing rare species disproportionately impacted functional diversity. In contrast, losing common species might not lead to such a significant effect. Mouillot et al. [4] revealed that the distinctive traits afforded by rare species were susceptible, and the species that perform the most vulnerable functions were typically rarer than expected. This paper revealed that rare species exhibited significantly unique functional traits but were susceptible to extinction, indicating a heightened risk of losing functional resilience in karst forests [4]. Given that these species are typically clustered around the centroid of the functional space (figures 1 and 2), the loss of common species could decrease such clustering, thereby increasing functional originality and functional specialization.

Furthermore, losing the rarest species had the most significant impact on functional diversity during the mid-successional stage and the least effect on functional diversity in the early successional stage (figure 2). It is speculated that the impact of species loss on functional diversity is related to the species abundance distribution (SAD); SAD reflects the dominance of species and the proportion of rare species, making it effective to determine the trade-off between functional diversity and functional redundancy [78,79]. The species dominance was greatest in the early successional stage, followed by the mid-successional stage, and the late successional stage exhibited the lowest degree (electronic supplementary material, figure S3). Wu et al. [34] found that the SAD negatively impacted differences in species richness from the early successional to mid-successional stage. Their finding indicated that although the large proportion of rare species did not have a significant advantage in terms of abundance, its influence on species richness was minor. Rare species might be restricted to specific niches and possess similar functional traits to other co-occurring species (figure 1D) due to their lower abundance. As a result, their loss hardly influenced functional diversity in the early successional stage [15,34]. In addition, Mouillot et al. [80] reported that the number of unique species saturated with increasing species richness reinforced the redundancy of the most common traits rather than providing expected insurance. Therefore, even in the more species-rich mid-successional stage, rare species still hold great relevance despite a high probability of functional redundancy [4,34]. Our study also identified a threshold for rare species loss, which could significantly impact functional diversity in the late successional stage, and this effect was more pronounced than that in other stages (figure 2). It was indicated that the sub-rarest species had more unique traits than the rarest species at the late successional stage. High immigration rates contribute to a stochastic assembly in communities with high species richness, potentially leading to temporary species coexistence [81]. Specifically, the occurrence of the rarest species may cause niche overlap briefly, resulting in functional redundancy [82]. In contrast, the sub-rarest species are highly adaptable to specific environments within the community and differ from common species in resource utilization. Moreover, we used null models to simulate the random loss of species. However, this simulation method has limitations. Many studies suggested that the non-random loss of species might be typical since the loss of species depends on their attributes (e.g. body size, rarity and sensitivities to environmental change [83–85]). This validated the scenario where the rarest species were lost first in our simulation. Especially, karst regions are characterized by high environmental heterogeneity and a large proportion of rare species, resulting in a higher likelihood of non-random loss in rare species due to their low adaptability [16,21].

Our results supported hypothesis 2 that rare species was integral to ecosystem functioning due to their unique and irreplaceable functional traits. Moreover, their loss had a significantly disproportionate impact on functional diversity, especially in the late successional stage. Inconsistent with previous studies, our results indicated that communities might become unstable along the successional pathway in degraded karst forests, mainly due to the increasing proportion of rare species with unique traits in the late successional community [40].

(c). Impacts of species loss on the functional trait network along the successional pathway in degraded karst forests

Functional trait network can reflect the interaction, dependency and functional redundancy among species (for example, the positive edges represent the similar traits between species) as well as the resilience of the network [19,62]. We simulated the effects of losing 50% of the rarest and the most common species on the functional trait network along the successional pathway in degraded karst forests. The results showed that the loss of the rarest species delivered a greater impact on network connectivity and complexity than losing the most common species at the early successional stage (figure 3A). Rare species exhibited more positive connection edges in the early successional stage (electronic supplementary material, figure S6). This presence suggested resembling functional traits to other species, contributing to a more stable functional trait network [62]. According to the stress-gradient hypothesis, the facilitative interactions among plants dominate under harsh conditions, such as in the early successional stage [86,87]. Enhanced environmental filtering leads to trait convergence regardless of limiting similarity, and further improvement results in higher occurrence of positive species interactions in the early successional stage [34,65,88]. Moreover, functional redundancy is often observed in dominant species, contributing to the maintenance of community resilience post-disturbances [89,90]. The relatively high species abundance in the early successional stage may also be responsible for the functional redundancy of more rare species (figure 3A). However, many rare species act as the keystone species within the network, linking multiple nodes in the early successional stage (electronic supplementary material, figure S5A). Losing keystone species significantly degrades functional resilience (figure 3A) [91–93]. Thus, the contribution of rare species to maintaining functional resilience is limited. The rare species loss slightly impacted the functional trait networks in the mid-successional stage (figure 3B). This does not fully align with hypothesis 3. The reason is that rare species with unique traits may act as isolated nodes in the network due to their low correlation with other species in this stage. Similarly, rare species connected most positive edges in the late successional stage (electronic supplementary material, figure S6), and their loss differed minorly from that of common species. This may be attributed to the rarest species, which maintain the functional trait network resilience by connecting more positive edges. In addition, the sub-rare species possess the most unique traits, thereby buffering the impact of rare species loss on the resilience of the functional trait network.

Overall, our results provide clear evidence that rare species performed different functional roles across successional stages. Rare species contributed to functional redundancy and enhanced community functional resilience in the early successional stages. This resilience might experience disruption when species loss reaches a certain level. Based on the above analysis, we should focus on the functional roles played by rare species across different successional stages.

5. Conclusion

This study simulated the effect of the loss of rare and common species on the functional diversity and resilience along the successional pathway in degraded karst forests. We found that the impact of rare species loss did not simply change linearly throughout the succession. Specifically, rare species primarily supported functional redundancy in the early successional stage. In this case, their loss minimally influenced the functional diversity while having a significant effect on functional resilience. In the mid- and late successional stages, losing rare species significantly affected functional diversity, and the impact on functional resilience was reduced. This contrasting observation indicated that rare species had unique and irreplaceable ecological functions in the later successional stages. Consequently, the impact on functional diversity and network resilience could be buffered. When we lost a certain proportion of rare species, this event disrupted functional diversity more significantly. These results highlight the crucial role of rare species in the heterogeneous degraded karst forests, i.e. rare species are highly vulnerable to disturbances due to their pronounced functional uniqueness. This study underscores that protecting rare species warrants special attention in biodiversity conservation. However, we did not consider belowground traits and reproductive traits, which might confound our results. Future research should integrate multidimensional traits (reproductive, growth, defensive, etc.) when quantifying plant functional community structure.

Ethics

This work did not require ethical approval from a human subject or animal welfare committee.

Data accessibility

All raw data and code used in this manuscript can be found from Dryad [94].

Supplementary material is available online [95].

Declaration of AI use

We have not used AI-assisted technologies in creating this article.

Authors’ contributions

L.M.: conceptualization, data curation, formal analysis, investigation, methodology, validation, visualization, writing—original draft, writing—review and editing; S.J.: conceptualization, formal analysis, methodology, writing—review and editing; M.S.: conceptualization, formal analysis, funding acquisition, project administration, supervision; D.C.: conceptualization, funding acquisition, investigation, methodology, project administration, supervision; G.Z.: funding acquisition, project administration, supervision; Q.L.: funding acquisition, project administration, supervision; Y.D.: conceptualization, formal analysis, methodology, supervision, writing—review and editing; H.X.: conceptualization, formal analysis, methodology, supervision, writing—review and editing; L.Z.: conceptualization, data curation, formal analysis, funding acquisition, investigation, methodology, project administration, resources, supervision, validation, writing—original draft, writing—review and editing.

All authors gave final approval for publication and agreed to be held accountable for the work performed therein.

Conflict of interest declaration

We declare we have no competing interests.

Funding

This work was supported by the National Natural Science Fund (32360380, 32360278, 32460377), the Guizhou Provincial Key Technology R&D Program (General (2023)111) and the Guizhou University scientific Research and Innovation Team Project ((2023) 07).