Plant invasions are associated with lower community stability at multiple spatial scales.

Save

Plant invasions are threatening the long-term sustainability of ecosystem functions. Yet, how plant invasions impact the temporal stability of community cover across spatial scales remains poorly understood, hindering effective invasion management. Using the National Ecological Observatory Network dataset, we examine how plant invasions influence multi-scale stability of total community abundance (used total community cover) via altering community functional diversity, as well as growth and dispersal strategies.

We find that invasion-associated losses of functional α diversity and functional shifts toward acquisitive strategies, reduce species asynchrony and species stability, respectively. These changes ultimately lead to lower temporal stability at the local scale (α stability). Meanwhile, invasion-associated functional shifts toward acquisitive seed dispersal strategies weaken functional β diversity, which reduces asynchrony community dynamics among local communities (spatial asynchrony).

These concurrent declines in α stability and spatial asynchrony collectively destabilize metacommunity-scale stability. Our findings reveal that plant invasions may trigger destabilizing effects across spatial scales, associated with changes in functional diversity as well as growth and dispersal strategies, underscoring the need for a multi-scale invasion management framework focusing on both local diversity and the connectivity among local communities.

Introduction

Since the Anthropocene, plant invasions have become a critical environmental challenge in ecosystems worldwide, posing significant threats to biodiversity conservation, ecosystem functions, socioeconomic development, and human well-being1–4. A key concern is whether invasions undermine the long-term sustainability of community abundance and ecosystem functions (e.g., productivity and biomass)5–7. This sustainability is often assessed through temporal stability, usually quantified as the ratio of the temporal mean to the temporal standard deviation of community abundance and ecosystem functions, where a higher value indicates lower fluctuation over time8.

However, current research at the relatively small scale (usually 1 m2) reveals conflicting findings: while plant invasions may weaken the temporal stability of total community cover or biomass by reducing diversity7, they could potentially enhance ecosystem stability when invasive species with higher inherent stability dominate communities9,10. Invasive plants tend to exert stronger negative effects on ecosystem functions at broad spatial scales than at small scales1,11, largely because their strong dispersal capacity enables rapid regional spread and overcomes constraints from native competitors or local resource availability11,12. However, how such invasions influence temporal stability across multiple spatial scales remains largely unexplored, as most existing studies have focused on single-scale assessments7,9,10, presenting a substantial barrier for across-scale invasion management.

The multiscale stability framework offers a hierarchical approach to understanding the temporal stability across spatial scales, which conceptualizes metacommunities as networks of interconnected local communities, linked through species dispersal processes13–15. Within this framework, metacommunity-level temporal stability (γ stability) is decomposed into two components: the average temporal stability of local communities (α stability) and the asynchronous community dynamics among local communities (spatial asynchrony). At the local level, α stability can be further partitioned into the mean temporal stability at the species level (species stability) and asynchronous population dynamics among species (species asynchrony)13,16.

Recent applications of this framework have advanced our understanding of how shifts in community structure (e.g., diversity and community composition) interact with environmental variability to shape the temporal stability of ecosystem functions across spatial scales17–19. Specifically, as functional strategies mediate temporal dynamics of species and communities under climatic flucations20,21, functional α diversity often enhances species asynchrony by increasing the trait dissimilarity between co-occurring species22, while higher functional β diversity can promote spatial asynchrony by facilitating trait differences in species composition across local communiies19, both contributing to higher temporal stability across spatial scales17. Nevertheless, the application of this framework remains limited in elucidating how plant invasions influence temporal stability across multiple scales through their impacts on community diversity and composition.

Within local communities, lower functional α diversity associated with plant invasions could reduce species asynchrony (Fig. 1A). It is well established that plant invasions drive functional α diversity loss across multiple ecosystems, including grasslands and forests4,23,24.

On the one hand, invasive species may induce competitive exclusion of native species with slow growth strategies, due to their high competitiveness, stemming from stronger resource acquisition capabilities, allelopathic inhibition, or higher tolerance to neighboring competition than native species25–28. On the other hand, invasive plants could facilitate further invasion and exacerbate functional diversity decline through habitat modification. Specifically, invasive plants drive shifts in nutrient stocks29, cycling rates23, and soil chemistry30 through their litterfall and root leachates24, creating microenvironmental conditions that selectively exclude native species with specific functional strategy adapted to their original habitats24,29.

In addition, root exudates from invasive plants may reshape soil microbial community composition, disrupting critical mutualistic associations that native species depend upon for survival and reproduction24,31,32. Given the positive correlation between species asynchrony and functional α diversity22,33, we hypothesize that plant invasions could weaken species asynchrony by lowering functional α diversity. However, invasive plants may also occupy distinct ecological niches with native species and introduce novel trait combinations, potentially even temporarily increases functional α diversity34,35, thereby introducing substantial uncertainty into predictions of how invasions will affect species asynchrony.Fig.

1A conceptual diagram showing the research hypotheses.A invasion-associated redecued functional α diversity weakens asynchronous population dynamics among species (i.e., species asynchrony); (B) invasion-associated functional shifts in plant communities towards acquisitive strategies reduce temporal stability at the species level (i.e., species stability); (C) invasion-associated reduced functional β diversity weakens asynchronous community dynamics among local communities (i.e., spatial asynchrony); (D) plant invasions may threathen temporal stability at the metacommunity level (i.e., γ stability). These patterns depict the changes in stability components along the gradient of increasing invasion pressure (from left to right) in each subgraph. ( + ) and (-) represent positive and negative correlations, respectively, e.g., functional α diversity is positive related to species asynchrony.

The plant images were designed by Freepik - Magnific.com.

Within local communities, plant invasions could be associated with functional shifts in plant communities, potentially altering species stability (Fig. 1B). Invasive plants typically exhibit acquisitive strategies, characterized by enhanced photosynthetic capacity (e.g., higher leaf nitrogen and phosphorus content), efficient resource acquisition (e.g., higher specific leaf area), and rapid reproduction (e.g., smaller seed dry mass)36–39.

Compared to invasive species in similar environments (e.g., climatic conditions) and successional stages, native species typically adopt conservative strategies, demonstrating greater structural and hydraulic safety (e.g., higher leaf dry matter content) and reproductive reliability (e.g., larger seed mass with germination advantages)36–38,40. As invasive plants displace natives and dominate communities through competitive exclusion and allelopathic suppression, their increasing dominance drives a functional shift in local communities from conservative to acquisitive strategies36,41. Given that acquisitive strategies typically develop at the expense of stress tolerance, potentially compromising species stability38,42.

We therefore hypothesize that invasion-induced functional shifts toward acquisitive strategies could destabilize communities. However, we also acknowledge that trait shifts may occur independently of invasion43, and that alternative pathways may exist—such as highly productive invasive plants gaining dominance through competive superiority—which could increase biomass production and potentially enhance species stability9,10. These mechanisms could operate simultaneously, with their opposing effects potentially offsetting one another, thereby complicating predictions of how invasions influence species stability and obscuring the relationship between biomass gains and ecosystem stability.

At the metacommunity scale, lower functional differences among local communities (i.e., lower functional β diversity) associated with plant invasions could diminish spatial asynchrony (Fig. 1C). This aligns with the well-established global pattern of functional β diversity loss associated with plant invasions44,45.

The underlying mechanisms can be attributed to two interconnected processes: first, invasive species with greater dispersal capabilities tend to colonize multiple local communities, thereby decreasing species turnover of all species across these communities37,46,47. Given that invasive plants generally exhibit acquisitive strategies36,37, their widespread establishment (i.e., lower species turnover) could reduce functional differences among local communities, i.e., lower functional β diversity. Second, invasive plants could displace native species through interspecific interactions (e.g., competition and allelopathic inhibition), particularly those with conservative strategies due to their limited competitive abilities and lower population densities48.

This displacement process reduces the regional species pool, consequently promoting the homogenization of local community compositions and functional strategy23,26,45. While studies have linked biotic and functional homogenization to destabilized ecosystem functions via reduced spatial asynchrony17–19, whether plant invasions threaten temporal stability by promoting biotic homogenization remains unexplored. However, functional β diversity could also increases or remain unchanged with plant invasion in some contexts, such as when invasive species introduce novel traits49, filter out functionally redundant natives50, or differentially filter native species across communities51, thereby introducing substantial uncertainty into predictions of how invasions will affect spatial asynchrony.

Here, our results show that plant invasions are consistently associated with declines in temporal stability across spatial scales, underscoring the need for a multiscale management approach for invasive species that simultaneously maintains local diversity and landscape connectivity. Specifically, using the terrestrial plant communities survey dataset from the National Ecological Observatory Network (NEON), spanning from 2013 to 202252, we investigate the impacts of plant invasions on the temporal stability of total community abundance (summed species cover) across multiple spatial scales. The dataset consisted of 43 terrestrial sites distributed across the United States (Supplementary Fig.

1 and Supplementary Table 1), with each site comprising 2 to 33 standardized sampling plots (metacommunity, 400 m2). Each sampling plot contained six subplots (Supplementary Fig. 2; local community, 1 m2)53.

We tested the following hypotheses: (1) invasion-associated reduced functional α diversity reduces species asynchrony (Fig. 1A); (2) invasion-associated functional shifts in plant communities towards acquisitive strategies reduce species stability (Fig. 1B); (3) invasion-associated reduced functional β diversity weakens spatial asynchrony (Fig.

1C); (4) plant invasions may threaten γ stability (Fig. 1D). Methodologically, we quantified functional α diversity (functional dispersion)54, β diversity (functional dissimilarity)55, and community‑weighted mean traits using TRY trait data based on all species (invasive and non-invasive plants), and calculated stability components within a multiscale framework13.

These were tested with linear mixed effects56 and structural equation models57.

Results

Negative plant invasions-temporal stability relationships at multiple spatial scales

We found that both α and γ stability (i.e., temporal stability at the local and metacommunity scale) were negatively associated (Fig. 2A, B; P < 0.001 for α stability and P = 0.007 for γ stability), while spatial asynchrony (i.e., asynchronous community dynamics among local communities) remained consistent with invasive cover (Fig. 2C; P = 0.677).

Following the decomposition of α stability into species stability (i.e., temporal stability at the species level) and species asynchrony (i.e., asynchronous population dynamics among species), both species asynchrony and species stability were negatively associated with invasive cover (Fig. 2D and Fig. 2E; P < 0.001 for both species asynchrony and species stability).

Sensitivity analysis yielded similar results, when considering the potential confounding effects of directional cover change on temporal stability (Supplementary Table 2), and confounding variables (Supplementary Fig. 3; observational frequency, initial survey year, and final survey year).Fig. 2Trends associated with invasive cover and temporal stability across spatial scales.A α stability, (B) spatial asynchrony, (C) γ stability, (D) species asynchrony, (E) species stability.

Two-sided t test with no adjustment was used to test whether there is a significant difference in slope from 0 (α stability: estimate [± SE] = − 0.47 [0.13], t892 = − 3.45, P = 5 × 10−4; spatial asynchrony: estimate [± SE] = 0.03 [0.07], t892 = 0.42, P = 0.677; γ stability: estimate [± SE] = − 0.85 [0.31], t892 = − 2.72, P = 0.007; species asynchrony: estimate [± SE] = −0.37 [0.10], t892 = − 3.71, P = 2 × 10−4; species stability: estimate [± SE] = −0.05 [0.01], t892 = − 7.03, P = 3 × 10_12). Conditional R2 and p-value are shown. Dots are the values predicted by partial regressions with each explanatory variable.

The lines and shades are the mean and 95% confidence intervals. Source data are provided as a Source Data file.

Negative plant invasions-diversity relationships at multiple spatial scales

Shannon α diversity, Bray-Curtis distance among subplots, and Shannon γ diversity were negatively associated with invasive cover (Fig. 3A–C; all P < 0.001). Functional α, β, and γ diversity were also negatively associated with invasive cover (Fig.

3D–F; P < 0.001 for functional α and γ diversity, and P = 0.018 for functional β diversity). The richness of native plants were negatively associated with, but the richness of invasive plants were positively associated with invasive cover, resulting in a negative relationship between richness with invasive cover at the metacommunity level (Supplementary Table 3). Community-weighted trait means (CWM) of TraitPC1, which was positively associated with leaf dry matter content (LDMC) and negatively associated with specific leaf area (SLA), nitrogen, and phosphorus content per leaf mass (Nmass and Pmass), negatively related to invasive cover (Supplementary Fig.

4A, B). CWM of TraitPC2, which was positively associated with seed dry mass, also negatively related to invasive cover (Supplementary Fig. 4C).

Higher CWMPC1 and CWMPC2 represented conservative growth and seed dispersal strategies, and lower CWMPC1 and CWMPC2 represented acquisitive growth and seed dispersal strategies (Supplementary Table 4). Meanwhile, CWMPC1 and CWMPC2 of native species negatively related to invasive cover (Supplementary Fig. 4D, E).

Changes in CWMPC1 and CWMPC2 of native species between two consecutive censuses (△CWMPC1 and △CWMPC2) were also negatively related to invasive cover (Supplementary Fig. 5). Functional nestedness positively, but functional turnover of native species among subplots negatively related to invasive cover, resulting in a positive relationship between functional β diversity of native species with invasive cover (Supplementary Fig.

6).Fig. 3Trends associated with invasive cover and Shannon diversity components.A Shannon α diversity. B Bray-Curtis distance.

C Shannon γ diversity and functional diversity compoents (D functional α diversity; (E): functional β diversity; (F): functional γ diversity). Two-sided t test with no adjustment was used to test whether there is a significant difference in slope from 0 (Shannon α diversity: estimate [±SE] = − 0.51 [0.07], t892 = − 6.50, P = 1 × 10−10; Bray-Curtis distance: estimate [± SE] = − 0.14 [0.02], t892 = − 5.52, P = 4 × 10−8; Shannon γ diversity: estimate [±SE] = − 0.68 [0.10], t892 = − 6.57, P = 8 × 10−11; functional α diversity: estimate [± SE] = − 0.31 [0.06], t892 = − 4.91, P = 1 × 10−6; functional β diversity: estimate [± SE] = − 0.05 [0.02], t892 = − 2.36, P = 0.018; functional γ diversity: estimate [± SE] = − 0.29 [0.08], t892 = − 3.48, P = 5 × 10−4). Conditional R2 and p-value are shown.

Dots are the values predicted by partial regressions with each explanatory variable. The lines and shades are the mean and 95% confidence intervals. Source data are provided as a Source Data file.

Decreasing diversity associated with plant invasions reduces temporal stability

Our results showed consistent positive diversity-stability relationships at multiple spatial scales. α stability increased with both Shannon and functional α diversity (Fig. 4A and Supplementary Fig. 7A).

Spatial asynchrony increased with both Bray-Curtis distance among subplots and functional β diversity (Fig. 4B and Supplementary Fig. 7B). γ stability increased with Shannon γ diversity but remained consistent with functional γ diversity (Fig.

4C and Supplementary Fig. 7D). Following the decomposition of local stability into species stability and species asynchrony, species stability decreased with both Shannon and functional α diversity (Fig.

4E and Supplementary Fig. 7E). Species asynchrony increased with both Shannon and functional α diversity (Fig.

4F and Supplementary Fig. 7F). Species stability increased with both CWMPC1 and CWMPC2 (Supplementary Fig.

8A, B). Species asynchrony decreased with both CWMPC1 and CWMPC2 (Supplementary Fig. 8C, D). α stability increased with CWMPC1 and marginally increased with with CWMPC2 (Supplementary Figs.

8E, 7F).Fig. 4Trends associated with functional diversity-stability relationships.A α stability, (B): spatial asynchrony, (C): γ stability, (D): species stability, and (E): species asynchrony. Dots are the values predicted by partial regressions with each explanatory variable.

Two-sided t test with no adjustment was used to test whether there is a significant difference in slope from 0 (α stability: estimate [± SE] = 0.33 [0.06], t892 = 4.88, P = 1 × 10−6; spatial asynchrony: estimate [± SE] = 0.28 [0.09], t892 = 3.01, P = 0.003; γ stability: estimate [± SE] = − 0.01 [0.11], t892 = − 0.06, P = 0.947; species asynchrony: estimate [±SE] = 0.29 [0.05], t892 = 5.83, P = 7 × 10−9; species stability: estimate [±SE] = − 0.09 [0.02], t892 = − 4.34, P = 1 × 10−5). The lines and shades are the mean and 95% confidence intervals. Source data are provided as a Source Data file.

Our structural equation models show decreasing functional α diversity associated with invasive cover, weakened species asynchrony and α stability (Fig. 5 and Supplementary Fig. 9).

Lower CWMPC1 and CWMPC2 associated with invasive cover further diminished species stability and α stability (Fig. 5). In addition, functional β diversity directly decreased as invasive cover increased (Fig.

5 and Supplementary Fig. 9). The decline in CWMPC2 associated with invasive cover also indirectly contributed to reduced functional β diversity (Fig.

5). Furthermore, decreasing functional β diversity associated with invasive cover weakened spatial asynchrony (Fig. 5).

Decreasing α stability and spatial asynchrony associated with invasive cover jointly drove the decline in γ stability (Fig. 5 and Supplementary Fig. 9).

Sensitivity analysis yielded similar results, when considering the potential spatial autocorrelation (Supplementary Fig. 10), the potential effects of Shannon diversity (Supplementary Fig. 11), missing observation within the time series (Supplementary Fig.

12), time lags of plant invasion and its effects on temporal stability (Supplementary Fig. 13), fundamental differences between leaf and seed traits (Supplementary Fig. 14), functional diversity alternatives (Supplementary Fig.

15), plant growth form (Supplementary Fig. 16), and introduced species (Supplementary Fig. 17).

Functional α diversity was positively correlated with Shannon α diversity, and functional β diversity was positively correlated with Bray-Curtis distance between local communities (Supplementary Fig. 11). Invasion cover was associated with reduced diversity (both taxonomic and functional), which in turn was linked to lower species asynchrony and spatial asynchrony, ultimately decreasing temporal stability (Supplementary Fig.

11). Meanwhile, invasive cover decreased with community tree coverage, but increased with human pressure (Supplementary Fig. 9).Fig.

5The structural equation modeling shows how changes in functional diversity and strategies with invasive cover regulate temporal stability at multiple spatial scales.The green and yellow arrows indicate significant positive and negative paths, respectively, with their standardized coefficients displayed (*: P < 0.05, **: P < 0.01, and ***: P < 0.001). Conditional R2 is shown in the corresponding boxes. CWMPC1 is positively associated with leaf dry matter content and negatively associated with specific leaf area, nitrogen, and phosphorus content per leaf mass, representing a conservative growth strategy.

CWMPC2 is positively associated with seed dry mass, representing a conservative seed dispersal strategy. The widths of significant paths are scaled by standardized path coefficients. Fisher’s C test was used to test whether significant paths had not been included in the models.

The SEM fits the data well (Fisher’s C = 112.335, df = 106, P = 0.318). The full SEMs with all paths are shown in the supplementary materials (Supplementary Fig. 9).

Source data and full statistical details are provided as a Source Data file.

Discussion

Our findings demonstrate that plant invasions critically undermine the temporal stability of total community abundance across multiple spatial scales. Specifically, we identified a consistent pattern that lower functional α diversity associated with plant invasions reduces asynchronous population dynamics among species (i.e., species asynchrony), and functional shifts from conservative to acquisitive growth strategies associated with plant invasion weakened temporal stability at the species level (i.e., species stability). Meanwhile, functional shifts toward acquisitive seed dispersal strategies associated with plant invasions reduce functional β diversity, which weakens asynchronous community dynamics among local communities (i.e.,spatial asynchrony).

Ultimately, the combined effects of declining temporal stability at the local scale (i.e., α stability) and spatial asynchrony under plant invasions collectively weaken temporal stability at the metacommunity scale (i.e., γ stability).

In line with our first hypothesis, the reduction in functional α diversity associated with intensified plant invasions diminishes species asynchrony within local communities. While some studies suggest that plant invasions may temporarily elevate α diversity through passenger effects—where the proliferation of invasive species transiently compensates for the loss of native diversity and increases local species pool35,58—our findings corroborate the negative plant invasion-diversity relations, likely through mechanisms such as competitive exclusion, allelopathic suppression, and habitat modification23,24,59. The loss of functional α diversity further diminishes species asynchrony, as functionally unique species tend to exhibit divergent responses to climate change, whereas the abundance gain of environmentally adapted species can offset the decline of maladapted species under climate fluctuations in species-rich plant communities, i.e., compensatory effects8,22,33.

Supporting our second hypothesis, functional shifts from conservative to acquisitive growth strategies associated with plant invasions in local communities reduce species stability. In our study, the invasive plants exhibit acquisitive growth strategies (Supplementary Fig. 4), marked by rapid resource acquisition and intense competitive traits36,37,39.

Once they displace native species and establish dominance, this invasion process may drive functional shifts in local communities towards acquisitive strategies of invasive species41,60. Simultaneously, invasive plants modify both abiotic and biotic conditions in ways that selectively favor native species exhibiting acquisitive strategies. For instance, invasive plants act as ecosystem engineers by enhancing soil nitrogen availability through accelerated litter decomposition, creating high-resource environments that may disproportionately favor fast-growing and resource-demanding natives24,48,60,61.

This functional shift towards acquisitive strategies reduces species stability, as fast growth and reproduction often trade off with stress tolerance38,40. For instance, species with acquisitive strategies (e.g., high specific leaf area, leaf nitrogen, and phosphorus content) enhance photosynthetic efficiency and resource uptake for fast growth, but they also exhibit high growth sensitivity to abiotic stress and environmental fluctuations42,62,63. In contrast, conservative species (e.g., those with high leaf dry matter content) with lower growth rates, greater structural investment, and longer life spans maintain lower inter-annual variability, particularly under environmental stress or disturbance20,21,62.

Although the shift toward acquisitive strategies may increase phenotypic plasticity within communities, promoting more asynchronous and individualized population dynamics under environmental fluctuations21,64, it fails to compensate for the inherent instability of species with acquisitive traits, ultimately undermining α stability.

At the local scale, our findings demonstrate that plant invasions may reduce α stability by diminishing both species asynchrony and species stability. This contrasts with previous experiments reporting no destabilizing effects of plant invasions on temporal stability, which were attributed to either the absence of positive diversity-stability relationships or the higher intrinsic stability of invasive species9,10,65. The absence of positive diversity-stability relationships may arise because simplified experimental communities—such as those assembled randomly or maintained under controlled environmental conditions—favor generalists that monopolize resources (e.g., invasive species), thereby amplifying dominant species effects in driving temporal stability9,16,66.

However, in natural communities, co-evolved species interactions and environmental heterogeneity can suppress single-species dominance and promote compensatory effects, resulting in that plant invasions weaken species asynchrony by reducing α diversity67–70. In addition, the positive selection effects for invasive plants, i.e., overproduction due to their greater growth rate and competitive ability, may compensate for the high inherent growth variability of acquisitive invasive plants in controlled experimental environments9,66,71, resulting in higher species stability of invasive plants. However, our findings reveal that in complex natural systems, the functional shifts toward acquisitive strategies associated with plant invasions exacerbate interannual growth variability, which outweighs any stabilizing effects of overproduction, underscoring that single-species dominance may be highly unstable and vulnerable to climate fluctuations.

Consistent with our third hypothesis, the decline in functional β diversity associated with plant invasions weakens spatial asynchrony. On the one hand, plant invasions may drive environmental homogenization through habitat modification. Invasive plants could induce directional environmental shifts (e.g., increased soil nitrogen content and pH) through litterfall and root exudates, which facilitate subsequent invasions and erode heterogeneity between local communities24,72–74.

In addition, their root traits (e.g., lower root C:N ratios and higher specific root lengths) and allelopathic effects could homogenize soil mutualistic networks, including microbial and nematode communities32,75. This environmental homogenization will trigger the selective extinction of native species and increased community nestedness, but it fails to offset the decline in species turnover, ultimately reducing β diversity23,45,76. This is consistent with our results showing increased nestedness and directional functional shift towards acquisitive strategies, rather than random fluctuations, of native species under plant invasions (Supplementary Figs.

4–6). On the other hand, invasive plants could exacerbate biotic homogenization by overcoming dispersal limitations between local communities. Invasive species often possess traits like lightweight seeds adapted for wind dispersal or animal-mediated transport, enabling widespread propagule spread77–79.

The enhanced dispersal facilitates colonization across multiple communities, homogenizing their species and functional composition and reducing functional β diversity39,46. Aligning with previous findings, our results confirm that β diversity promotes spatial asynchrony, i.e., functional and compositional divergence between local communities drives their asynchronous responses to climate fluctuations17,18. Consequently, reducing functional β diversity associated with plant invasions diminishes the spatial buffering effects, thereby undermining temporal stability at broader scales.

While the NEON dataset offers an opportunity for evaluating how plant invasions affect temporal stability across spatial scales, several methodological limitations should be acknowledged. First, we used total community cover as a proxy for ecosystem function (biomass or abundance) to quantify temporal stability. Although previous stability research report consistent patterns using both cover and biomass/abundance metrics80,81, the absence of biomass and abundance data in our study prevented a direct validation of such consistency within our own framework.

Second, our analytical framework emphasized the spatial asynchrony among local communities. However, the asynchrony index in this framework proposed by Wang et al. partly depends on the total number of populations (i.e., the portfolio effect), and it didn’t distinguish between this and the contribution of compensatory dynamics14,82. Recent methodological developments, such as the metapopulation‑based approach proposed by Segrestin & Lepš82,83, shift the focus toward population‑level asynchrony across spatial and organizational scales.

This population‑explicit approach could more directly disentangle mechanisms like shared environmental responses versus local competitive interactions—a promising direction for future research. Third, our use of species‑mean trait values from the TRY database overlooks within‑species trait variation. Intraspecific trait variability may influence competitive outcomes, resource use efficiency, and ultimately, population and community stability84,85.

Fourth, while our structural equation modeling allowed us to infer pathways, the observational nature of our data inherently limits causal interpretation. Therefore, while our observational study offers insights into invasion-stability relationships across complex natural systems, controlled small-scale experiments remain crucial for inferring specific mechanisms under reduced environmental noise. Future investigations addressing these empirical constraints would yield deeper mechanistic understanding of the destabilizing effects of plant invasion and more targeted restoration frameworks.

In conclusion, our findings provide evidence for the destabilizing effects of plant invasions on the temporal stability of community abundance at multiple spatial scales. Invasion-associated losses of α and β diversity, coupled with functional shifts toward acquisitive strategies, reduce species asynchrony, spatial asynchrony, and species stability, respectively. Our results reveal the destabilizing effects of plant invasions: local declines in temporal stability and diversity propagate to broader spatial scales, ultimately undermining temporal stability at the metacommunity scale.

These findings underscore the need for a multi-scale management framework, where effective invasion control simultaneously prioritizes local diversity maintenance and connectivity among local communities to sustain spatial asynchrony essential for climate adaptation86–88. Importantly, our analysis further situates invasion success within a context of direct anthropogenic pressure (Supplementary Fig. S8), linking plant invasions to land‑use change and human disturbance.

Therefore, coordinated conservation action from local to regional scales, focused on mitigating human‑mediated drivers (e.g., urbanization and agricultural disturbance) and preserving functional diversity, is essential to safeguard ecosystem stability in the face of ongoing plant invasions.

Methods

Study area and inventory data

We accessed the ‘Plant presence and percent cover’ dataset from the National Ecological Observatory Network (NEON; available at https://data.neonscience.org) on 7 September 202452. The NEON program established a comprehensive monitoring initiative across 47 strategically selected sites across the United States (Supplementary Fig. 1).

Utilizing a nested sample plot design, it aimed to investigate and quantify the multifaceted impacts of climate change and invasive species on biodiversity53. Specifically, each site comprised 2 to 33 standardized sampling plots (20 m × 20 m, 400 m2), with each plot containing six subplots (1 m × 1 m, 1 m2; Supplementary Fig. 2).

Field crews conducted annual vegetation surveys, identified plant species, and recorded species-level ground cover during the growing season in each subplot between 2013 and 2022. The growth forms of plants are classified into woody, shrub, and herbaceous according to a global growth form database89. To obtain a reliable estimate of temporal stability, we selected plots with at least four annual observation records17.

Finally, our dataset contained 943 plots and 5658 subplots across 43 sites, with 4445 plant species recorded (Supplementary Table 1). Out of 4445 species, there are 3594 herbaceous plants, 506 trees, and 345 shrubs. The analysis included a total of 6163 annual observations across 943 plots, as 398 plots were not monitored in 2019 or 2020 (mean observation frequency: 6.53 ± 1.51 times; Supplementary Table 1).

The mean inter-plot distance within a site is 3.5 ± 4.6 km (mean ± standard deviation), ranging from 0.91 to 12.7 km, with a median of 2.93 km (Supplementary Table 1).

The invasive status of each plant species was classified according to the United States Register of Introduced and Invasive Species (US-RIIS), compiled by the United States Geological Survey90. This database distinguished between non-native species that are merely introduced and those considered invasive based on their documented impacts to local ecosystems or economies. Of the 4445 species recorded in our study, 530 were non-native, comprising 105 introduced and 425 invasive species.

The most abundant invasive species were Poa pratensis, Bromus inermis, Microstegium vimineum, Dactylis glomerata, and Cynodon nlemfuensis (Supplementary Table 1). For each plot, we calculated invasive cover by first determining the proportion of invasive plant cover within each subplot and then averaging these proportions across all six subplots. Among the 943 plots, invasive plants were absent in 25 plots, appeared after plot establishment in 148 plots, and were already present before establishment in 770 plots (Supplementary Table 1 and Supplementary Fig.

18).

Temporal stability components

We defined each 1 m2 subplot as a local community, as this is the standardized sampling unit in the NEON protocol for capturing fine‑scale species interactions and environmental filtering, aligning with established metacommunity theory and has been successfully applied in previous studies to assess α diversity effects on local stability17,91. Each 400-m2 plot comprising six subplots was treated as a metacommunity, as its spatial extent (tens of meters) appropriately captured the typical dispersal range of invasive plants, thereby capturing local population expansion while minimizing strong dispersal limitation78,92. We then calculated the temporal stability at multiscale spatial scales13–15.1\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${{{\rm{species\; stability}}}}=\frac{{\sum }_{i,k}{\mu }_{k(i)}}{{\sum }_{i,k}\sqrt{{v}_{k\left(i\right),k(i)}}}$$\end{document}species stability=∑i,kμk(i)∑i,kvki,k(i)2\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${{{\rm{species\; asynchrony}}}}=\frac{{\sum }_{i,k}\sqrt{{v}_{k\left(i\right),k(i)}}}{{\sum }_{k}\sqrt{{\sum }_{i,j}{v}_{k\left(i\right),k(j)}}}$$\end{document}species asynchrony=∑i,kvki,k(i)∑k∑i,jvki,k(j)3\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${{{\rm{\alpha }}}}\,{{{\rm{stability}}}}={{{\rm{species\; stability}}}}\times {{{\rm{species\; asynchrony}}}}=\frac{{\sum }_{i,k}{\mu }_{k(i)}}{{\sum }_{k}\sqrt{{\sum }_{i,j}{v}_{k\left(i\right),k(j)}}}$$\end{document}αstability=species stability×species asynchrony=∑i,kμk(i)∑k∑i,jvki,k(j)4\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${{{\rm{spatial\; asynchrony}}}}=\frac{{\sum }_{k}\sqrt{{\sum }_{i,j}{v}_{k\left(i\right),k(j)}}}{\sqrt{{\sum }_{i,j,k,l}{v}_{k\left(i\right),l(j)}}}$$\end{document}spatial asynchrony=∑k∑i,jvki,k(j)∑i,j,k,lvki,l(j)5\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${{{\rm{\gamma }}}}\,{{{\rm{stability}}}}={{{\rm{\alpha }}}}\,{{{\rm{stability}}}}\times {{{\rm{spatial\; asynchrony}}}}=\frac{{\sum }_{i,k}{\mu }_{k(i)}}{\sqrt{{\sum }_{i,j,k,l}{v}_{k\left(i\right),l(j)}}}$$\end{document}γstability=αstability×spatial asynchrony=∑i,kμk(i)∑i,j,k,lvki,l(j)where i and j represent species i and j, while k and l represent local community k and l. μk(i) refers to the mean cover of species i in the local community k across all sampled years; νk(i),k(i) refers to the cover variance of species i in the local community k across all sampled years; νk(i),k(j) means the cover covariance between species i and species j in local community k across all sampled years; νk(i),l(j) means the cover covariance between species i in local community k and species j in local community l across all sampled years.

We selected this framework because it offers an ecologically interpretable approach to scaling ecosystem-function stability from local to broader spatial extents, aligns with metacommunity theory that explicitly links α and β diversity to γ scale stability, and enables direct comparison with prior research13,14,17,18.

Functional traits

We selected five functional traits, which were considered to reflect plant resource acquisition strategy and how plants respond to climate change62,93, including leaf dry matter content (LDMC, mg g−1), nitrogen content per leaf mass (Nmass, mg g−1), phosphorus content per leaf mass (Pmass, mg g−1), seed dry mass (SM, g), and specific leaf area (SLA, mm2 mg−1). We obtained species mean trait values from the TRY database94 and then replaced missing trait values with genus-level and family-level means of species observed in our study (Supplementary Table 5). There are 79-153 species without known traits-information congeners, accounting for 1.8% of the total coverage, with missing trait values replaced by mean values from all other species recorded in our study.

We performed a principal component analysis (PCA) based on the above five functional traits (Supplementary Fig. 4 A). TraitPC1, which explained 40.6% of the variation, was positively correlated with LDMC, and negatively correlated with SLA, Nmass, and Pmass, indicating traits associated with conservative growth strategies (Supplementary Table 4)38; TraitPC2, which explained 19.7% of the variation, was positively correlated with SM, indicating traits associated with conservative dispersal strategies (Supplementary Table 4)40.

Quantification of biodiversity and environmental drivers

We calculated the mean distance of each species to the cover-weighted centroid of all species in a multidimensional functional space at the subplot level, representing functional dispersion at the subplot level, which served as our measure of functional α diversity54. We then averaged functional α diversity across six subplots. Similarly, functional dispersion at the plot level was calculated as functional γ diversity.

Functional β diversity was derived as the mean pairwise functional dissimilarity among the six subplots within each plot, computed using “beta” function in the “BAT” package55. This approach partitions total functional dissimilarity into replacement and richness‑difference components (i.e., nestedness and turnover), offering a decomposition‑oriented perspective on trait turnover across local assemblages54,55. Although expressed differently, both functional α and β diversity metrics we used are derived from pairwise functional distances and are consistent within a common trait‑based diversity framework.

We used the community-weighted trait means (CWM) of TraitPC1 and TraitPC2, weighted by species cover, to reflect functional strategies at the plot level (i.e., CWMPC1 and CWMPC2). Meanwhile, we calculated Shannon α and γ diversity weighted by species cover at the subplot and plot level, respectively. β diversity was calculated as the average pairwise Bray-Curtis dissimilarity among six subplots within each plot, weighted by species cover76. These diversity indices were calculated based on all species within the community, including invasive and non-invasive plants.

Considering the dependence of temporal stability at multiple spatial scales on environmental divers, we extracted monthly mean temperature, potential evapotranspiration, and precipitation from a gridded multivariate climate dataset with a resolution of 0.5° × 0.5°95. We then calculated the long-term averages of mean annual temperature (MAT) during the plot investigation period. The climate moisture index (CMI) was equivalent to annual precipitation minus annual potential evapotranspiration and represented climate water availability during the plot investigation period.

To measure the interannual fluctuations of climate during the survey, we also calculated the coefficients of variation of the annual mean temperature (CVT) and climate moisture index (CVM) during the plot investigation period. Soil nutrient availability was derived from the Harmonized World Soil Database (1 km resolution), which includes information on soil texture, organic carbon, pH, and total exchangeable bases96. Human influence was quantified using the terrestrial human modification index at a 1 km resolution, which integrates 13 spatially explicit anthropogenic stressors (e.g., population density, land use intensity, road influence) and models their cumulative impact on natural systems97.

Statistical analyses

To explore how temporal stability and functional diversity at multiple spatial scales change with invasive cover, we fitted the following linear mixed effects model using the ‘nlme’ package56:6\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${{{{\rm{Stability}}}}}_{i}={{{{\rm{\beta }}}}}_{1}\times {{{{\rm{Cover}}}}}_{i}+{{{{\rm{\beta }}}}}_{2}\times {{{{\rm{MAT}}}}}_{i}+{{{{\rm{\beta }}}}}_{3}\times {{{{\rm{CMI}}}}}_{i}+{{{{\rm{\beta }}}}}_{4}\times {{{{\rm{CVT}}}}}_{i}+{{{{\rm{\beta }}}}}_{5}\times {{{{\rm{CVM}}}}}_{i} \\+{{{{\rm{\beta }}}}}_{6}\times {{{{\rm{Soil}}}}}_{i}+{{{{\rm{\beta }}}}}_{7}\times {{{{\rm{Huamn}}}}}_{i}+{{{{\rm{\beta }}}}}_{8}\times {{{{\rm{Tree}}}}}_{i}+{{{\rm{\nu }}}}{{{\rm{k}}}}+{{{{\rm{\pi }}}}}_{{{{\rm{ki}}}}}+{{{\rm{\varepsilon }}}}$$\end{document}Stabilityi=β1×Coveri+β2×MATi+β3×CMIi+β4×CVTi+β5×CVMi+β6×Soili+β7×Huamni+β8×Treei+νk+πki+ε7\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${{{\rm{Diversity}}}}{{{\rm{i}}}}={{{{\rm{\beta }}}}}_{1}\times \,{{{{\rm{Cover}}}}}_{i}+{{{{\rm{\beta }}}}}_{2}\,\times {{{{\rm{MAT}}}}}_{i}+{{{{\rm{\beta }}}}}_{3}\,\times \,{{{{\rm{CMI}}}}}_{i}+{{{{\rm{\beta }}}}}_{4}\times {{{{\rm{CVT}}}}}_{i}+{{{{\rm{\beta }}}}}_{5}\times {{{{\rm{CVM}}}}}_{i} \\+{{{{\rm{\beta }}}}}_{6}\times {{{{\rm{Soil}}}}}_{i}+{{{{\rm{\beta }}}}}_{7}\,\times {{{{\rm{Huamn}}}}}_{i}+{{{{\rm{\beta }}}}}_{8}\times {{{{\rm{Tree}}}}}_{i}+{{{\rm{\nu }}}}{{{\rm{k}}}}+{{{{\rm{\pi }}}}}_{{{{\rm{ki}}}}}+{{{\rm{\varepsilon }}}}$$\end{document}Diversityi=β1×Coveri+β2×MATi+β3×CMIi+β4×CVTi+β5×CVMi+β6×Soili+β7×Huamni+β8×Treei+νk+πki+εwhere i and k is ith plot in kth site; β is the estimated coefficients in the model; Stabilityi means five temporal stability components (γ stability, spatial asynchrony, α stability, species stability, and species asynchrony) in ith plot; Diversityi means three diversity components (functional α, β, and γ diversity) in ith plot; Coveri is the mean invasive cover during the survey in ith plot; MATi, CMIi, CVTi, and CVMi are mean annual temperature, climate moisture index, and the coefficients of variation of the annual mean temperature and climate moisture index during the survey in ith plot, respectively. Soili and Huamni are the soil nutrient availability and human influence in the ith plot, respectively; Treei is the overstory canopy coverage, calculated as the proportion of tree coverage to total vegetation coverage; νk is a random effect accounting for site-specific regional climate and geology; πki is a nested plot-specific random effect accounting for spatial non-independence among plots within the same site; ε is a random error. All environmental predictor variables were scaled for coefficient comparison.

Similarly, we fitted linear mixed models with environmental drivers as covariates to test how temporal stability at multiple spatial scales changes with Shannon diversity, functional diversity, and CWM. Meanwhile, while null model simulations are usually used for distinguishing directional selection from random drift, they are not appropriate for large-scale datasets like ours (4445 plant species) due to computational limitation98. Therefore, as a robust alternative to test whether plant invasion may drive directional functional shifts rather than random fluctuations, we calculated changes in CWMPC1 and CWMPC2 of native species between two consecutive censuses (△CWMPC1 and △CWMPC2), and then fitted linear mixed models with invasive cover as the independent variable and environmental drivers as covariates.

Considering the non-normality of the stability component, we bootstrapped the linear mixed effects model with 1000 iterations. Specifically, we performed 1000 bootstrap resamples (with replacement) of the original data, and for each resample, we refitted the linear mixed-effects model and extracted the coefficient estimates. The 2.5% and 97.5% percentiles across all 1000 estimates were then taken as the 95% confidence interval for each coefficient.

The bivariate relationships tested via linear mixed effects models were used to verify the individual mechanistic links that together constitute the integrated pathways in the structural equation model, as described below.

To examine how invasive cover alters multi-scale temporal stability by affecting functional diversity and strategies, we employed piecewise structural equation modeling (SEM) using the “piecewiseSEM” package57, guided by an a priori theoretical framework (Supplementary Fig. 19). Specifically, to account for our hierarchical sampling design, we specified nested random intercepts (i.e., plot nested within site) in all constituent linear mixed models comprising the SEM.

We began with the conceptual model (Supplementary Fig. 19) and incorporated potential pathways if they significantly improved the model fit (P < 0.05, based on Shipley’s d-separation test). The global model fit was assessed using Fisher’s C statistic (P > 0.05).

Considering the impact of environmental factors and human influence on invasion cover and functional strategies19,99, we also incorporated the direct effects of environmental factors and human influence on invasive cover and functional indicators (FD and CWM) in our structural equation modelings. In addition, standardized path coefficients were computed to facilitate comparisons across coefficients.

To strengthen the robustness of our findings, we performed the following sensitivity analyses: (1) to address the potential confounding effects of directional cover change on temporal stability8, we substituted the variance or covariance of species cover with the variance or covariance of residuals derived from the linear regression of species cover against sampling time in the stability calculations (Supplementary Table 2); (2) to evaluate the influence of temporal series length on temporal stability17, we refitted model (Eq. 6) by including confounding variables (observational frequency, initial survey year, and final survey year) as fixed effects (Supplementary Fig. 3); (3) to account for potential spatial autocorrelation among plots, we incorporated an exponential spatial autocorrelation structure (implemented via the “nlme” package) into all linear mixed models comprising our structural equation model (SEM) and refitted the models (Supplementary Fig.

10)56; (4) considering the potential relationships between taxonomic and functional diversity, we added the Shannon α diversity and Bray–Curtis distance in the structural equation model and refitted it (Supplementary Fig. 11); (5) considering the impact of missing years in the time series on stability estimation, we refitted the structural equation model using 545 plots with continuous annual measurements (Supplementary Fig. 12); (6) considering the potential time lag between colonization by invasive species and the impacts on temporal stability, we refitted the structural equation model using plots with invasive plants absent or already present before plot establishment (Supplementary Fig.

13); (7) to address the potential mismatch between leaf and seed traits, we performed PCA on four leaf traits and separately considered seed dry matter, and refitted the structural equation model (Supplementary Fig. 14); (8) we recalculated functional α and β diversity using the unified diversity framework proposed by Chao et al. based on species pairwise functional distances100, and refitted the structural equation model (Supplementary Fig. 15); (9) considering the potential effects of plant growth form, we recalculated functional α and β diversity by incorporating growth form (woody, shrub, or herbaceous) as three binary (0/1) variables into the functional trait matrix, and refitted the structural equation model (Supplementary Fig.

16); (10) considering the impact of non-invasive non-native plants, we also used introduced cover (total coverage of invasive and introduced plants) as independent variables to refit the structural equation model (Supplementary Fig. 17). These sensitivity analyses consistently yielded results similar to those reported in the main text.

All statistical analyses were conducted using R (version 4.3.1).

Reporting summary

Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.

Supplementary information

Supplementary InformationReporting SummaryTransparent Peer Review file

Source data

Source Data

Related articles

PLOS

Associations of palmar surface lactic acid with hand scaling and in vitro pathogen inactivation: A cross-sectional study

The relationship between palmar surface lactic acid and hand scaling is not well understood. This cross-sectional study examined associations between palmar surface lactic acid, hand scaling, and in vitro anti-pathogenic activity, to generate hypotheses relevant to hand skin condition and pathogen inactivation. In 49 healthy adults, after standardized hand washing and acclimation (20°C, 40% RH), we assessed scaling (visual score), surface lactic acid and stratum corneum ceramides (LC–MS), stratu...

PLOS

MFMNet: Adaptive tri-layer graph pyramid with cross-scale attention router for synergistic drug prediction

Background and objective: Experimental screening of synergistic anticancer drug pairs is time-consuming, while effectively integrating complementary molecular graph representations with cell-line-specific information remains challenging for computational drug-synergy prediction. Methods: We developed MFMNet with a Tri-Layer Graph Pyramid (TGP) comprising GCN, GIN, and GAT branches for extracting complementary molecular graph representations. A Cross-Scale Attention Router (CSAR), implemented as...