Mechanism of Terrestrial Plant Community Assembly under Different Intensities of Anthropogenic Disturbance in Dianchi Lakeside

Abstract

A lakeside is a functional transition zone that connects the lake aquatic ecosystem and the land ecosystem. Understanding the community assembly mechanism is crucial for regional ecological restoration, habitat management, and biodiversity conservation. However, research on the terrestrial plant community assembly in lakesides under anthropogenic disturbance is still lacking. The present study used phylogeny and functional traits to assess the community assembly of three habitat types with different anthropogenic disturbances in Dianchi lakeside. The factors that influenced the community assembly were also explored. Results indicated that the phylogenetic signals of all the examined functional traits of the dominant species were weak, suggesting that the traits were convergent. The community phylogenetic and functional structures of the different habitat types showed random patterns. Thus, the assembly of terrestrial plant communities in the three habitat types was driven by competitive exclusion and neutral processes in Dianchi lakeside. The trait trade-off strategies of species in the different habitats varied with the different habitat types. Anthropogenic disturbance played an important role in the process of community assembly. The present study provides a scientific basis for the assessment and management of ecological restoration in Dianchi lakeside and other plateau lakes and enriches the knowledge on the community assembly mechanism of disturbed plant communities.

1. Introduction

A plant community is a collection of species within a certain range of time and space, and it is made up of individuals and populations of plants [1]. The composition of species in a community is influenced by diffusion limitation, environmental constraints, and internal interactions [2]. Exploring the mechanism of community assembly can help us maintain species diversity and diagnose the success of regional ecological restoration [3].

At present, niche theory and neutral theory are the main hypotheses to explain community structures and dynamics [4,5]. According to niche theory, community assembly is a process in which species from regional species banks are selected into local communities, and environmental filtering and biological interaction in this process are regarded as a multi-nested filter [6]. On the one hand, environmental filtering filters out species that have adapted to the habitat, so different species are expected to have convergent traits [7]; on the other hand, competitive exclusion reduces the overlap range of traits between species in similar niches, thus allowing numerous species to coexist [8]. Two opposing ecological drivers, namely, environmental filtering and competitive exclusion, work together to maintain species diversity in communities [9]. Neutral theory emphasizes the role of stochastic factors in community assembly, and the community structure is affected by diffusion limitation [10]. Some scholars have attempted to integrate niche theory and neutral theory when analyzing the mechanism of community assembly. Although different models have been proposed, a consensus exists on how to clearly measure the relative contribution of different ecological processes to community assembly [11].

Disturbance is a common factor that affects community assembly. Studies have shown that human activities have a large impact on ecosystem maintenance [12,13]. Disturbance can change the availability of resources and open up new space by removing some species, thus providing opportunities for invasive species, and altering the succession process and structure of the community [14,15]. In addition, disturbance as a filter affects the community species composition and structure, while different types of disturbance have various effects on the community structure [15,16]. Several studies have shown that community restoration after human disturbance is dominated by deterministic ecological processes. Management legacy will continue to influence the succession of the community in the future, and the relative contribution of environmental filtering and competitive exclusion will change with such succession [17,18,19,20].

Dianchi Lake, which is located in Kunming, the capital city of Yunnan Province, belongs to the Jinshajiang River system of the Yangtze River Basin. The lake is listed as the sixth largest freshwater lake in China [21]. Previous studies focused on combating eutrophication in Dianchi Lake [22,23,24]. Recently, an intensive effort for the ecological restoration of the lake was initiated by the local government. However, the effects have not been promising [21]. A lakeside is a functional transition zone that connects the lake aquatic ecosystem and the land ecosystem. Lakeside ecological restoration is crucial for combating eutrophication in plateau freshwater lakes [25]. The lakeside of Dianchi is adjacent to the city and towns with dense populations and under high-frequency human activities. Thus, the ecological restoration process of Dianchi lakeside is influenced by exotic human disturbance and changes in land use patterns. Although the species compositions and flora of spermatophytes in Dianchi lakeside have been studied [26], no report is available on the assembly of plant communities in Dianchi lakeside during regional ecological restoration. This study is the first to explore the mechanism of terrestrial plant community assembly in Dianchi lakeside under certain degrees of anthropogenic disturbance during ecological restoration. The following questions were put forward. (a) What are the characteristics of communities and invasiveness degree of species under different degrees of anthropogenic disturbance? (b) What are the ecological adaptation strategies of plants at the community level under different degrees of anthropogenic disturbance? (c) Which ecological process dominates the assembly of terrestrial plant communities in Dianchi lakeside?

2. Materials and Methods

2.1. Study Area and Sampling Sites

The study was conducted in the lakeside of Dianchi Lake (102°36′ E–102°47′ E, 24°40′ N–25°02′ N). Dianchi Lake is located in the middle of Yunnan Province, southwest China, with an average water-level height of 1887.4 m. The north and east of the lake make up the urban region of Kunming City. The shoreline of the lake is 163 km, and the length is 40 km from north to south and 7 km from east to west [26,27]. Dianchi Lake watershed has a subtropical monsoon climate, with an average annual temperature of 14.7 ℃ and distinct dry and wet seasons [28]. The rainy season is from May to October, the dry season is from November to April, and the mean annual precipitation is 1000 mm [28,29]. Dianchi lakeside is adjacent to Kunming City with high-frequency human activities. The lakeside environment has been seriously damaged and polluted by industries and agriculture. Moreover, the continuous human disturbance and severe alien invasion species still affect the fragile ecosystem in the fragmented habitat of Dianchi lakeside [27,30].

From April 2019 to August 2020, the plant communities in 10 sample sites along Dianchi Lake were investigated in detail (Figure S1). These sample sites of 20 × 50 m plots were divided into three habitat types, namely, mountain land (ML), abandoned farmland (AF), and abandoned wetland (AW), in accordance with the degree of human disturbance, land use pattern, and habitat type (

Table 1. Condition of 10 typical plots in the three habitat types along Dianchi lakeside.

Plots	Habitat Types	Human Disturbance	Duration of Ecological Restoration
DW	Mountain land	Weak	≥10 a
HYC	Mountain land	Weak	5~10 a
LNW	Mountain land	Weak	≥10 a
WLC	Abandoned farmland	Medium	5~10 a
XHG	Abandoned farmland	Medium	5~10 a
SP	Abandoned farmland	Medium	5~10 a
FB	Abandoned farmland	Medium	5~10 a
HF	Abandoned wetland	Strong	5~10 a
XH	Abandoned wetland	Strong	5~10 a
PLJ	Abandoned wetland	Strong	5~10 a

Note: “a” means years.

). Each plot was divided into 10 quadrats for different plant growth: ten 100 m² (10 × 10 m) quadrats for woody plants and ten 1 m² (1 × 1 m) quadrats for herbaceous plants. All plants in the plots were identified, and the number of individuals and coverage of species in each quadrat were recorded. We identified the alien invasive plants in the plots in accordance with the “List of Alien Invasive Species of China,” which was published by the Ministry of Ecology and Environment of the People’s Republic of China [31]. Importance values (IV) were used to indicate the dominance of invasive herbs in the herbaceous layer of a community. IV is the average of the relative covering (RC), relative abundance (RA), and relative frequency (RF) of species in a community. The relevant formulas are given below [32].

RCi = (Covering of species i)/(Total covering of all species),

(1)

RAi = (Abundance of species i)/(Total abundance of all species),

(2)

RFi = (Frequency of species i)/(Total frequency of all species)

(3)

IV = (RCi + RAi+ RFi)/3,

(4)

We performed the Kruskal–Wallis rank sum test to identify the difference in invasive plants’ IV between the three habitat types by using the statistical software SPSS (version 23).

2.2. Soil Sampling and Measurement

We collected three soil samples per plot at a depth of 10 cm from the surface [4,5]. From the sides and middle of a plot, two soil portions were collected each and mixed together as one sample. We measured the following five key chemical and physical variables: soil pH, soil organic carbon (SOC), soil total nitrogen (STN), soil total phosphorus (STP), and soil total potassium (STK). Soil pH was measured with an acidity meter (PHS-3C, Rex, China). SOC was measured by a method where the organic components of soil are oxidized by K₂Cr₂O₇ and titrated by FeSO₄ [33]. STN was measured using an automatic Kjeldahl apparatus (SKD-200, Peiou, China). The molybdenum–antimony anti-colorimetric method was applied to measure STP with a UV-visible single-beam spectrophotometer (752 type, Jinghua, China). STK was measured via flame atomic absorption spectrophotometry.

2.3. Plant Trait Selection and Measurement

From July to August of 2019 and 2020, the functional trait data of the dominant species in the plots were collected in accordance with the handbook for standardized measurement of plant functional traits [34]. The dominant species in the plots were determined by their coverage in the community. Ten individuals of each dominant species were randomly selected, and their petiolate leaves that were fully expanded and healthy were collected. Simultaneously, five leaves per individual of herbaceous plants and 10 leaves per individual of woody plants were obtained. We selected the following eight leaf functional traits to measure: leaf thickness (LT), leaf area (LA), specific leaf area (SLA), leaf dry matter content (LDMC), leaf chlorophyll content (Chl), leaf organic carbon concentration (LOC), leaf total nitrogen concentration (LTN), and leaf total nitrogen concentration (LTP). LT was measured with a Vernier caliper (accuracy of 0.02 mm). LA was measured with a laser area meter (CI-203; CID Bio-Science, Camas, WA, USA). We measured the saturated fresh mass (FM) of the leaves on the day of picking, dried the leaves at 60 °C for 72 h, and measured the dry mass (DM). SLA was calculated as LA divided by DM, and LDMC was calculated as DM divided by FM. The LOC, LTN, and LTP were measured using the same method as that used for SOC, TC, and TP. Chl was measured immediately after the leaves were picked by using a handheld chlorophyll meter (Dualex Scientific+, FORCE-A, Orsay, France).

2.4. Phylogenetic Tree and Phylogenetic Signal Tests

In accordance with the method of Webb [35], a phylogenetic tree (Figure 1) containing 86 species from the 10 plots in Dianchi lakeside was constructed using the online tool Phylomatic (http://phylodiversity.net/phylomatic/ (accessed on 24 September 2020)) (Figure S2). The phylogenetic branching length was estimated by Zanne’s phylogenetic tree, which contains 31,749 species of terrestrial plants [36]. To quantify the evolutionary patterns of functional traits, the phylogenetic signal was tested using Blomberg’s method for continuous traits [37]. The phylogenetic signal was estimated by the statistic K, which is the specific value of observed values compared with the values of the null model. K = 1 indicates a random phylogenetic signal of functional traits, and the evolutionary characteristic of traits conforms to the Brownian evolution model. K > 1 suggests that traits are more conservative than the expectation of the Brownian evolution model, namely, closely related species have more similar traits than the expectation of the random model. K < 1 suggests that traits are more convergent than the expectation of the Brownian evolution model, that is, closely related species are not more similar in traits than the expectation of the random model. Significance was tested through 999 randomizations that randomized the trait distribution on the tips of the phylogenetic tree. The community weighted mean (CWM) of functional traits at the species level was used to test the phylogenetic signal [38], and the trait data were log10 transformed. The formula for CWM is as follows [39]:

$$\overline{t_i}=\frac{{{\displaystyle \sum }}_{j=1}^{P }{a}_{ij}{t}_{ij}}{{{\displaystyle \sum }}_{j=1}^{P }{a}_{ij}}$$

(5)

where $\overline{t_i}$ is the CWM of species i, P is the number of plots, a_ij is the abundance of species i in plot j, and t_ij is the trait value of species i in plot j. The test on the phylogenetic signal was performed using the multiPhylosignal functions in the R package “picante” [40].

2.5. Phylogenetic Community Structure Tests

We used the net relatedness index (NRI) to indicate the phylogenetic structure of a plant community. NRI is a standardized effect size of the mean pairwise distance (MPD) on the phylogenetic tree for all species in each plot. The equation is as follows:

NRI = −1 × [MPDobserved − mean (MPDrandomized)]/sd (MPDrandomized)

(6)

where MPDobserved is the observed value of the mean pairwise distance between species in a community, MPDrandomized is generated through a null model that randomly shuffles the species on the tips of the phylogeny in a local community, mean (MPDrandomized) is the mean of 999 random MPDrandomized, and sd (MPDrandomized) is the standard deviation of the random distribution. When NRI values deviate significantly from the null model, positive NRI values indicate a phylogenetically clustered structure of communities, and negative NRI values suggest a divergent phylogenetic structure. When no significant difference exists between NRI and the null model, the phylogenetic structure of the community is random. In this study, the calculation of NRI was implemented with the R package “picante” [40]. The significant deviations of NRI from the expectation of the null model in the different habitats were assessed through Student’s t-test, which was implemented with the statistical software SPSS (version 23).

2.6. Functional Community Structure Tests

For the analysis of the phylogenetic structure based on phylogeny, we used a functional trait dendrogram to examine the community functional structure. The trait dendrogram was constructed from the Euclidean distance matrix of three PCA axes of eight functional traits. Then, we used the standardized effect size of pairwise distance (S.E.S PW) in the trait dendrogram to indicate the functional structure of the plant community. The equation is as follows:

S.E.S PW = −1 × [PWobserved − mean (PWrandomized)]/sd (PWrandomized),

(7)

where PWobserved is the observed value of the mean pairwise distance between species in the trait dendrogram of a community, mean (PWrandomized) is the mean of 999 random PWrandomized generated through a null model similar to the analysis of phylogeny relatedness, and sd (PWrandomized) is the standard deviation of the random distribution. When S.E.S PW values deviate significantly from the null model, positive S.E.S PW values indicate a clustered functional structure of communities, and negative S.E.S PW values suggest a divergent functional structure. When no significant difference exists between S.E.S PW and the null model, the functional structure of the community is random. In this study, the calculation of S.E.S PW was implemented with the R package “picante” [40]. The significant deviations of S.E.S PW from the expectation of the null model in the different habitats were assessed by Student’s t-test, which was implemented with the statistical software SPSS (version 23).

2.7. Trade-Off Analysis of Leaf Economic Spectrum Traits

We selected SLA, LDMC, and LTP as leaf economic spectrum (LES) traits [41,42]. The CWM of functional traits at the community level was used for the analysis, and the formula for CWM is as follows [39]:

$$\overline{p_i}=\frac{{{\displaystyle \sum }}_{j=1}^{s }{a}_{ij}{t}_{ij}}{{{\displaystyle \sum }}_{j=1}^{s }{a}_{ij}}$$

(8)

where $\overline{p_j}$ is the CWM of plot j, S is the number of species, a_ij is the abundance of species i in plot j, and t_ij is the trait value of species i in plot j.

Principal component analysis (PCA) was employed to analyze the trade-off of LES traits, which reflects the ecological strategies of plants at the community level in different habitats [43]. Meanwhile, a Pearson correlation analysis was performed to explore the relationship between the first principal component (PC1) and three LES traits and soil factors in the whole habitats. The difference in LES traits in the three habitats was tested through ANOVA, and the data on the three traits were log10 transformed. PCA was performed with the R package “FactoMineR” [44].

3. Results

3.1. Community Characteristics in the Three Habitat Types

The ML habitat in this study had less cultivated vegetation and few species of trees and shrubs. The dominate species of this habitat are Coriaria nepalensis Wall., Pyrus pashia Buch.-Ham. ex D. Don, Eragrostis pilosa (L.) Beauv., Imperata cylindrica (L.) P. Beauv. In addition, some alien invasive plants, such as Ageratina adenophora (Spreng.) R. M. King and Bidens pilosa L., were the dominate species distributed in this habitat. The plant communities in habitat AF were affected by human activities due to large amounts of farmland having been abandoned and gradually replaced by artificial forests in the lakeside. In these artificial forests, often, only a single tree species and many weeds, which mainly include grasses and invasive herbaceous plants, are located. The dominate species in the communities of this habitat are Populus yunnanensis Dode, Celtis tetrandra Roxb., Malvastrum coromandelianum (L.) Gurcke, Imperata cylindrica. The alien invasive plants of Bidens Pilosa, Ageratina adenophora were also largely distributed in this habitat. The community of habitat AW is also subjected to strong human disturbance. Trees and shrubs that were cultivated by humans were the major species in the community of this habitat. The dominate species in the communities of this habitat are Koelreuteria paniculata Laxm., Photinia glomerata Rehder & E.H.Wilson, Liquidambar formosana Hance, Osmanthus fragrans (Thunb.)Lour. In addition, some invasive herbaceous plants such as Alternanthera philoxeroides (Mart.) Griseb., Bidens Pilosa, Ageratina adenophorum were colonized in patches with good light transmittance and margins of communities.

Several alien invasive plants were found in every plot of the three habitat types, and they were all herbs. The major invasive plants were A. adenophora, Bidens pilosa, Alternanthera philoxeroides, Ipomoea purpurea, and Erigeron sumatrensis. In most plots, alien invasive species were dominant in the herbaceous layer of the communities. In addition, the dominance of invasive herbs increased with the increase in human disturbance intensity in the different habitat types. A significant difference in IV was observed between ML and AW (Figure 1).

3.2. Phylogenetic Signal of Traits

For all functional traits, the statistic K was less than 1, suggesting that the phylogenetic signal was weaker than the expectation under a Brownian motion model of trait evolution. In addition, the P value from the permutation test was more than 0.05 for six traits, except for LT and LOC (

Table 2. Phylogenetic signal of functional traits of dominant species in the study plots in Dianchi lakeside.

Functional Traits	K	P
SLA	0.073	0.371
LDMC	0.209	0.090
LA	0.111	0.501
LT	0.509 *	0.017
LTN	0.109	0.164
LTP	0.130	0.095
Chl	0.062	0.401
LOC	0.408 *	0.009

Note: “*” means significant phylogenetic signal.

), indicating that most traits had no statistically significant phylogenetic signal. This result implies that the functional traits of the species in Dianchi lakeside may be more influenced by the external environment than by phylogeny.

3.3. Community Phylogenetic Structure

The NRI value of ML was less than zero, and the NRI values of AF and AW were more than zero. With the increase in human disturbance intensity in the different habitat types, the NRI value tended to rise (Figure 2a). However, the results of the Student’s t-test showed that the NRI values of the three habitat types did not deviate significantly from the null expectation (

Table 3. Student’s t-test on the phylogenetic (a) and functional (b) structures differing from the null expectation in the three habitat types.

	NRI			S.E.S PW
Habitat Types	Mean	SE	p	Mean	SE	p
ML	−0.78	1.057	0.620	0.62	0.193	0.370
AF	0.26	0.255	0.671	0.87	0.304	0.151
AW	0.79	0.280	0.194	−0.75	0.320	0.154

Note: ML: mountain land, AF: abandoned farmland, AW: abandoned wetland.

). This finding indicates that the phylogenetic structures of plant communities in the three habitat types in Dianchi lakeside were random.

3.4. Community Functional Structure

The results of the Student’s t-test on S.E.S PW were similar to the results on NRI, suggesting that the community functional structures of the three habitat types were also random (Table 3). However, the S.E.S PW values of ML and AF were more than zero, whereas the S.E.S PW value of AW was less than zero (Figure 2b).

3.5. Ecological Strategy of Plants in the Three Habitat Types

The PCA results for three LES traits of the different habitat types and the overall level of all habitats at the community level are shown in Figure 3. The first PCA axis (PC1) explained 88.3%, 98.6%, and 57.1% of the variation in LES traits in the three habitat types. PC1 explained 88.3% of the variation in LES traits in the overall habitats. SLA and LDMC had a negative relationship in the three habitat types and in the overall habitat. Therefore, PC1 was used as the trade-off axis of LES traits at the community level.

The ANOVA results for LES traits between the different habitats showed that the SLA and LTP of AF were the largest, but only LTP was significantly larger than AW. The LDMC of ML was significantly greater than that of AF, but no significant difference was found between ML and AW (Figure 4).

The LES traits and PC1 were not significantly correlated with most soil factors according to the results of the Pearson correlation analysis. However, LDMC was significantly negatively correlated with soil pH, and a significantly positive correlation was observed between PC1 and STK (Figure 5).

3.6. Relationship between Community Structure and Soil Factors

The Pearson correlation analysis revealed that the phylogenetic and functional structures of the communities were not significantly related to the five soil factors, and no significant correlation was found between NRI and S.E.S PW. However, NRI was significantly positively correlated with the IV of alien invasive plants (Figure 5).

4. Discussion

4.1. Phylogenetic Signal and Trait Convergence in Different Habitat Types

Anthropogenic disturbance, as a filter, affects community species composition mainly through phenotypic traits. Species with the same tolerance to disturbance have similar traits; in other words, species subjected to the same disturbance pressures show a trend toward functional convergence [45]. In the present study, the test on the phylogenetic signal for plant functional traits in Dianchi lakeside showed that the statistical K values of all traits were less than 1, that is, distant species may have more similar traits than expected under a Brownian motion model of trait evolution. Furthermore, for most traits (except LT and LOC), no significant phylogenetic signal was observed. This result implies that the functional traits of dominant species in the 10 plots of Dianchi lakeside were convergent overall. Similar ecological processes have been observed in other studies. For example, extant research has shown that herbaceous species present similar adaptations to the selection of shade tolerance in traits at low elevations, and the functional convergence of herbaceous species has been observed in habitats with high soil fertility [5]. Moreover, 10 functional traits of 180 species in the meadows and pastures of North Adriatic Karst grasslands, which are semi-natural habitats, showed a convergent pattern of trait evolution in a previous study, and harsh soil conditions were hypothesized to act as a selection pressure [46]. Thus, species that have similar adaptations to disturbance are selected from the regional species pool into the local community.

4.2. Disturbance and Invasion Shaped the Phylogenetic Structure of a Community

Anthropogenic disturbance always affects the phylogenetic structure of a community. The phylogenetic structure of a disturbed habitat is more clustered than what is expected because the abiotic filter of the environment increases after the dominant species is removed [3,15]. Moreover, the legacy of disturbance gradually decreases with succession, and the phylogenetic structure changes from clumping to divergence [13,17]. However, phylogeny and disturbance do not always have a relationship, and the hypothesis that disturbance filters species with similar adaptive traits, thereby filtering phylogenies, has a premise that traits related to disturbance and environment adaptation are phylogenetically conserved [15,16]; that is, closely related species are highly similar in terms of ecological adaptation traits [35]. A similar tendency is that disturbance causes a convergent phylogenetic structure when traits with related competition are less phylogenetically conserved [15]. Zhang et al. [16] found that phylogenetic diversity does not correlate with the anthropogenic disturbance gradient, although phylogenetic clustering occurs in different disturbance types. Disturbed habitats are often vulnerable to species invasion [47,48]. Invasion often leads to phylogenetic convergence of the community structure [49], and it may be related to the fact that invasive plants with general adaptability are widely distributed and often belong to species-rich families [50].

Our results showed that the NRI of the three habitat types had no significant difference from the expectation of the null model, suggesting that competitive exclusion may lead to a random phylogenetic structure in the case of functional traits that are overall convergent [9]. However, we observed a trend where phylogenetic clumping increased with increasing human disturbance intensity and species invasion. Anthropogenic disturbance may have several effects on the assembly of plant communities in Dianchi lakeside. Human activities directly affected the species composition in the community; for instance, most of the tree and shrub species in the plots were planted artificially. Meanwhile, weeds were often removed artificially in the communities of AF and AW habitats, so the community succession of the understory remained at the early stages. Furthermore, the removal of some species by mowing provided opportunities for the colonization of alien species during the long-term dynamic process of the community; alien invasive plants have become a common species composition of the plant community in the lakeside [30]. Some invasive species from Asteraceae and grasses from Poaceae, which have strong adaptability to disturbance and phosphorus-rich soil, were the dominant species of plant communities in the different habitat types. Therefore, the increase in related species led to a trend of phylogenetic clustering.

4.3. Assembly Processes of Communities Reflected by Functional Traits

The community functional structure is shaped by disturbance directly or indirectly [12,13,51]. The distribution pattern of traits in response to the environment exhibits clustering under disturbance stress, and the functional traits related to competition show an even pattern [52]. Recent studies have explored the drivers of community assembly under disturbance based on functional traits, and they found that the community functional structure of lowland rainforests exhibits clustering under strong logging disturbance; in addition, abiotic filtering dominates the process of community assembly [3]. Land use history and management affect community succession, and community restoration is a deterministic process [18,20]. The high trait convergence under intense land use disturbance and the high trait divergence under low disturbance indicate that environment filtering and competitive exclusion drive the community assembly process in different habitats [20]. However, different traits are always affected by different drivers (e.g., management, soil fertility, and climate) [53] and show different distribution patterns of traits. The patterns of trait distributions can be random in early succession, and stochastic factors, such as colonization stochasticity and dispersal, influence the assembly of communities [19]. The functional structure of the community in Dianchi lakeside also showed a random pattern, and the phylogenetic structure suggested that neutral processes drove community assembly. Management and land use history may have played an important role in such community assembly. Anthropogenic disturbance directly altered the species composition, and frequent restarts of succession in the herbaceous layers caused by management provided opportunities for random colonization of native and invasive species.

Trade-off between functional traits is common among species [41]. The leaf economic spectrum exists not only at the global scale [42], but also at the species and community levels [43]. The relationship between traits is affected by disturbance, and the functional trait spectrum shifts from conservation to rapid resource acquisition along the disturbance gradient [12]. In this study, species with resource-conservative strategies (high LDMC and low SLA and LTP) dominated the plant communities in the ML habitat. By contrast, species with resource-acquisitive strategies (high SLA and LTP and low LDMC) dominated the plant communities in the AF habitat. The species in the AW habitat were in the middle of the trade-off axis. The strategy of trait trade-off also existed at the community level in the overall habitats. Meanwhile, LDMC and PC1 that represented the axis of trade-off were significantly correlated with soil pH and STK, respectively. These results suggest that a common trade-off of functional traits occurred at the community level, and the ecological strategies of species based on trait trade-off in the different habitats were significantly affected by several external factors, such as disturbance and soil properties. Subsequently, the community assembly processes were also influenced.

5. Conclusions

The phylogenetic and trait-based tests we conducted in Dianchi Lakeside reinforce the importance of disturbance in understanding the complex mechanisms driving community assembly. The community phylogenetic and functional structures of the different habitat types showed random patterns. The functional traits of dominate species in the communities under different levels of disturbance were overall convergent, and competitive exclusion may have resulted in the random phylogenetic structure. However, it may also be a combination of disturbance as a habitat filter and biotic interactions between invasive and non-invasive species, due to the phylogenetic structure changing from divergence to convergence along the gradient of disturbance and species invasion. Thus, the assembly of terrestrial plant communities in the three habitat types was driven by competitive exclusion and neutral processes in Dianchi lakeside. No significant correlation was found between phylogenetic and functional structures, and the variation and distribution of traits had no obvious relationship with phylogeny under the pattern of convergent evolution of traits. This implies that the different habitat types caused by anthropogenic disturbance obviously affect trait trade-off strategies of plant species in Dianchi lakeside. In summary, the present study proposed that anthropogenic disturbance played an important role and strongly affected the process of community assembly. The results enrich the body of knowledge on plant community ecology and provide new insights into the assembly mechanism of disturbed plant communities. The study also offers a scientific basis for the assessment and management of ecological restoration in Dianchi lakeside and other plateau lakes.