Literature Search and Data Extraction

To target studies that would have assessed both the extent of the invasion and the performance of the native community, we limited our search to publications that addressed the vulnerability or resistance of the native community to biological invasions. Thus, in August 26, 2019, we carried two searches in the Web of Science database using the following terms:

We then applied these selection criteria:

After combining the two searches, we extracted information from selected publications on the biophysical features of the system (e.g., location, climate, biome, vegetation type), the type of study (observational or experimental), the variables of the native community measured (biotic or abiotic); identified the driver of vulnerability or resistance to the invasion; and recorded metric of plant performance or community assessment collected on the invasive species, and if available, on the native species. For a full list of variables extracted, see  Supplementary Material 1.

We classified the drivers of vulnerability identified in the publications into six different categories; some were further classified in several subcategories (Table 1). We are aware that some of these categories overlap. For example, biotic resistance, or lack thereof, can be due to different types of biotic interactions (Levine et al. 2003). In our data, we denoted as “biotic resistance” those observations in which competition between native and invasive plants was the proposed mechanism for vulnerability or resistance. Propagule availability as an invasion mechanism may reflect excess propagules from invasive plants and/or lack of native propagules. Although invasive species propagule pressure is not a feature of the native community, an excess of invasive propagules over native propagules overlaps with the concept of priority effects driving vulnerability or resistance (Dickson et al. 2012; Stuble and Souza 2016). For plant performance or community assessment data for both invasive and native plants, we gathered information on the metrics measured (Table 1); these included abundance (either density or cover), biomass (total or aboveground), individual plant growth, recruitment (seed production, establishment), individual survival, and richness (number of species). We also recorded sample size, mean response value, and variability around mean response (SD, SE, or variance). Values from figures were extracted using the Web Plot Digitizer online application ( http://arohatgi.info/WebPlotDigitizer/app). A flowchart of the publication selection process is provided in  Supplementary Material 2.

Table 1.

Drivers of vulnerability and plant performance and community assessment metrics.

Data Analysis

Effect size, that is, differences in plant performance between communities with high and low invasion, was calculated as:

By this formulation, ES of invasive plants (ES_invasive) is always positive. We then compared it with native ES (ES_native): change in native performance under the invasion. Specifically, we wanted to: (1) investigate conditions under which native performance increases, decreases, or does not change; and (2) quantify the magnitude in native performance change (ES_native) with respect to the magnitude of increase in invasive performance (ES_invasive).

Because a substantial portion of the observations, 10%, did not report variance associated with mean performance, instead of using standard metrics (e.g., Hedges' g), we ran a simulation to estimate ES, treating missing variances as latent variables to be estimated as a function of the largest ES variance calculated from studies with reported variances (see following section). Sample size was also considered in these estimations by weighing variances by their sample size (Gurevitch and Hedges 2001). See  Supplementary Material 3 for simulation code.

We carried out extensive exploratory data analysis to assess whether any of the variables gathered (e.g., climate, latitude, type of study) contributed to the observed variability in ES, but none did. To address our research questions, we then used multilevel mixed-effects models to analyze ES as a function of the driver of vulnerability and the nested subcategories within each driver (Table 1), with publication as a random effect. To see whether differences between the invasive and native communities depended on the metric used (e.g., abundance, growth, survival), we also analyzed data for each plant performance or community assessment metric. Additional analyses for each biome–driver combination and vegetation type–driver combination were also done, in this case without including random effects (some categories were represented by only one publication).

We used a hierarchical Bayesian approach to be able to incorporate missing variances as latent variables (Ibáñez et al. 2019). Missing variances were estimated by sampling from normal distributions (limited to be positive) with an SD of 1, with mean being the largest variance among observations with reported variance; this is the most conservative, lowest bias, imputation method to deal with missing variances (Batson and Burton 2016). Parameters of the mixed-effects models were all estimated from noninformative prior distributions (code for these analyses can be found in  Supplementary Material 3).

Because a large portion of our observations (40%) referred to disturbance as a main mechanism of vulnerability to invasion, we carried out an additional analysis to better asses the role of disturbance on vulnerability or resistance to plant invasion. We performed an analysis that compared sites with and without disturbance. For this analysis, differences in plant performance (i.e., ES) were estimated as:

In our records, high invasion was not always associated with disturbance; thus ES estimation here is different from the analyses described earlier. We ran a similar hierarchical model to the one described above, but in this case as a function of the six disturbance categories and the subcategories that we identified in the data (Table 1). Effect size calculations and analyses were carried out in OpenBUGS (Thomas et al. 2006). Effect size posterior estimates that did not include zero in their 95% credible intervals were considered statistically significant. Effect sizes with 95% credible intervals that did not overlap were considered significantly different from one another.

Assessing Vulnerability to Invasion

The two instances in which the native community experienced a significant decline, while invasive plants benefited significantly, were studies in which propagule availability and lack of biotic resistance (or higher competitive ability of the invasive plants) were identified as the drivers of vulnerability to invasion (Figure 1). In all other cases, the native community response to the invasion was neutral or positive (i.e., under herbivory; Figure 1). There were also several instances of native and invasive performances differing, with an overall pattern of higher effects sizes among invasive plants (Figure 1).

In the field of biological invasions, propagule availability refers to both the number of propagules and the rate of arrival (Simberloff 2009). High propagule pressure from introduced species has been strongly associated with their spread and abundance (e.g., Catford et al. 2011; Ibáñez et al. 2009). Our results link higher levels of invasive propagules to a reduction in native plant performance and thus to higher native plant community vulnerability to invasion (Figure 1). Furthermore, these results could simultaneously reflect low native seed abundance (Schuster et al. 2018; Vilà and Ibáñez 2011; von Holle and Simberloff 2005). This finding underscores the importance of legacy (what is left) and priority (what arrives first) effects during plant establishment, particularly after disturbance (Corbin and D'Antonio 2012; Uricchio et al. 2019). If invasive propagules are the most abundant, any removal of vegetation will likely result in reinvasion (Pearson et al. 2016; Prior et al. 2018). Thus, management practices that decrease invasive propagules (e.g., removal before seeding), while at the same time increasing native propagules (e.g., via seeding or planting), may have particularly successful results (Reinhardt Adams and Galatowitsch 2008).

Lack of biotic resistance or high competitive ability of invasive plants is frequently associated with successful invasions (Carboni et al. 2018; Vilà and Weiner 2004). Numerous studies have documented a variety of plant traits conferring invasive plants an advantage over the invaded native communities. For example, high total and specific leaf area (Allred et al. 2010), high germination rates (Deschenes et al. 2019), specific mechanism of nitrogen acquisition (Laungani and Knops 2009), and chemical inhibition of native plant photosynthesis (Musil et al. 2009) have all, among others, been identified as features of invasive plants contributing to their invasion success. Also, competition for space and resources during recruitment or the production of allelochemicals by invasive plants may prevent native species from growing populations that could resist the invasion (e.g., Edwards et al. 2019; Esch et al. 2018; Grove et al. 2017).

Still, high invasive competitive ability is not only a function of the invasive itself, but is usually associated with particular features of the ecosystem (Daehler 2003; Metlen et al. 2012); which, if managed, could confer the native community a higher level of resistance to the invasion (Byun et al. 2018). High resource ability and/or low enemy pressure are conditions under which invasive plants become highly competitive (Blumenthal 2006; Burns et al. 2007; Garcia-Serrano et al. 2007). Our review showed that under these conditions, native plants tended to underperform compared with invasive species: native plants marginally benefited from enemy release and from changes in resources, while invasive plants strongly benefited from both increases and decreases of resources and lack of enemies (Figure 1).

Figure 1.

Results for the analysis of invasive and native plant performance or community assessment (effect size [ES] mean + 95% credible interval [CI]), as a function of drivers of vulnerability (darker bars) and their subcategories (white and light gray bars). CIs that do not overlap with zero are considered statistically significant. An asterisk indicates invasive and native species ES 95% CIs do not overlap but are of the same sign (i.e., same direction of change). Two asterisks indicate invasive and native species ES 95% CIs do not overlap and are of different sign (i.e., opposite direction of change). Numbers denote number of observations included.

After the enemy release hypothesis was formulated as a main mechanism of invasion success (Keane and Crawley 2002), several studies questioned its relevance (Agrawal and Kotanen 2003; Beckstead and Parker 2003; Colautti et al. 2004; Maron and Vilà 2001). Our results show that with respect to herbivory, the category for which we have more observations, invasive species strongly benefited from absence of herbivory, while native plant response was neutral (although invasive and native performances were not significantly different). This might be an indication of overall higher palatability among invasive plants, as acquisitive and fast-growing traits are associated with lower plant defenses (Blossey and Nötzold 1995; Blumenthal et al. 2009). Thus, successful invasive species control might involve managing herbivores, for example, by providing access or shelter.

Figure 2.

Results for the analysis of plant performance or community assessment (effect size [ES] mean + 95% credible interval [CI]), invasive and native, as a function of metric used. CIs that do not overlap with zero are considered statistically significant. Asterisks indicate invasive and native species ES 95% CIs do not overlap and are of different sign (i.e., opposite direction of change). Numbers denote number of observations included.

Fluctuation of resources has also been postulated as a major mechanism underlying biological invasions (Davis et al. 2000). Our analysis shows that changes in resources were strongly associated with higher invasive performance, while native plants varied in their response (Figure 1). Although the sample sizes for native species are low in some of these comparisons, native plants were significantly outperformed by invasive plants in instances of decreasing water and of increasing light and nutrients (Figure 1; note there were no observations of native responses to water increase). Increases in light have often been identified as a major driver of invasive plant establishment (e.g., Huebner et al. 2018), even if the abundance of other resources also promotes native plants (Knight et al. 2008). Traits that lead to rapid growth (e.g., high specific leaf area and leaf nutrient content) are most advantageous under high light conditions, and these traits are common among invasive plants (Allred et al. 2010; van Kleunen et al. 2010; Vilà and Weiner 2004). In contrast, there were also a few instances in which a decrease in resources (i.e., water) favored invasive plants over natives (Figure 1), but we lack enough data to generalize as to when this is the case. Together, these results underscore the importance of managing available resources to avoid situations that favor invasive species, for example, maintain resource levels within their natural range of variability.

Vulnerability to Invasion across Metrics, Vegetation Types, and Biomes

Performance and impact of invasive plants were assessed with different metrics across our review. Abundance and biomass were the two most common measurements of invasive performance (68%), while richness and abundance were the prevalent metrics across native communities (45%; Figure 2). When assessing the impact of plant invasions through the lens of the native community, the two metrics that strongly differentiated invasive and native plants were abundance (e.g., cover, density) and richness (i.e., number of species) (Figure 2). Other metrics (i.e., growth, recruitment, survival) were quite variable among native plants (also note the low number of observations). Differences in native biomass between sites with low and high invasions were of the same magnitude as those for invasive species (Figure 2).

These results might indicate that invasive impact on the native community is exerted at the population level through higher density of individuals (or higher cover). Of course, changes in population density come through changes in recruitment, growth, and survival (Harper 1977). In our review, we had very few observations in these categories, showing a wide range of responses, which might explain the lack of significant differences. Biomass, for which we had a higher number of observations, is a common metric used in experimental work. Our results indicate that biomass may not be a good measure of native community vulnerability to invasions; that is, on an individual basis, native plants accumulate as much biomass as invasive, plants but at the population level (abundance or fecundity) their performances differ (Figure 2), indicating that invasion success may be more complex than just straight competition for resources (Daehler 2003; Maron and Vilà 2001).

Our results also corroborate other studies that document a decrease in native plant species in communities that are being invaded by introduced plants (e.g., Linders et al. 2019; Powell et al. 2013). Still, there was no correlation (not shown) between number of native species and the size of the effect, contradicting many studies that have found higher levels of invasion in richer native communities (Aguiar et al. 2006). Several studies have associated high native diversity with high levels of invasion (e.g., Long et al. 2009; Peng et al. 2019; Stohlgren et al. 1999, 2003), but such comparisons between systems with inherently high or low levels of diversity make it very difficult to assess the actual effect of invasions on native diversity. Here, our comparisons were done for the same ecosystem, with the only difference being the degree of invasion, and indicated a strong decline in native richness under invasion.

Figure 3.

Effect sizes (ES) on invasive and native plant performance/community assessment across drivers of vulnerability by biome (A) and vegetation type (B). Credible intervals (CI) that do not overlap with zero are considered statistically significant. An asterisk indicates invasive and native species ES 95% CIs do not overlap but are of the same sign (i.e., same direction of change). Two asterisks indicate invasive and native species ES 95% CIs do not overlap and are of different sign (i.e., opposite direction of change). Numbers denote number of observations included.

Our assessment of how native plant communities perform under invasion across vegetation types and biomes is limited by the number of observations (Figure 3); thus, we are cautious not to overinterpret our results. Only Mediterranean and temperate communities were well represented, although not for all mechanisms of vulnerability (Figure 3A). In boreal and temperate areas, native plants seem to fare well under invasions, with natives outperforming invasive plants in temperate areas when disturbance was the mechanism driving the invasion. Still, invasive plants showed higher performance under all other drivers. Tropical and Mediterranean native vegetation tended to experience a considerable drop in performance under high levels of invasive plants (Figure 3A). When looking across vegetation types represented in the data, differences between the native and invasive communities become more significant (Figure 3B). Most of the data came from forests and grasslands, which drove the trends discussed above: native species performance was positive under invasion except when lack of biotic resistance or propagule availability were the drivers of vulnerability (Figure 3B). The only vegetation type diverging from the overall trend was wetlands. Disturbance had a strong detrimental effect on wetland natives, while the effect was positive on natives in all other vegetation types (Figure 3B). Wetland habitats are highly susceptible to invasion (Sobrino et al. 2002), they act as “landscape sinks” of residuals, that is, areas where loose soil and plant material from other systems accumulate, and under disturbance, alterations in their hydrology and nutrient levels create conditions for invasive plants to succeed (Zedler and Kercher 2004). Surprisingly, in this vegetation type, invasions attributed to lack of biotic resistance were not associated with lower native performance, as was the case in all other vegetation types (Figure 3B). Because most of these observations, 9 out of 12, were related to species richness, information on other native performance metrics might result in a different outcome.

Figure 4.

Effect of disturbance in invasive and native plant performance/community assessment for each disturbance type (darker bars) and subcategories (lighter bars). Credible intervals (CI) that do not overlap with zero are considered statistically significant. Asterisks indicate invasive and native species ES 95% CIs do not overlap and are of different sign (i.e., opposite direction of change). Numbers denote number of observations included.

Disturbance and Vulnerability to Plant Invasion

Disturbances—natural, anthropogenic, or as result of management—are common among plant communities and in many cases provide optimal conditions for invasive species establishment and spread (Jauni et al. 2015; Lembrechts et al. 2016). Resource fluctuations usually follow disturbances (Jentsch and White 2019); then, species better adapted to rapidly use those resources, like many invasive plants, are likely to outcompete later arrivals (Dickson et al 2012; Radford 2013). As a result, native community capacity to respond to a disturbance will largely determine its vulnerability to invasion.

In this analysis, we found mostly neutral to positive responses of native plants to disturbance. Natives were only outperformed by invasive species when disturbance was caused by human activities (e.g., pollution, edge effect, trampling, hiking; Figure 1). However, our meta-analysis included observations wherein the lack of disturbance was associated with higher dominance of invasives (16% of the records). Thus, to better assess the role of disturbance in invasion, we specifically compared plant performance of native and invasive plants in communities that had experience a disturbance event (Figure 4). Overall, plant performance after disturbance was quite variable and tended to be positive for both invasive and native plants (Figure 4). Our assessment of the effect of disturbance on plant species performance was based only on the publications selected specifically to address questions of vulnerability and resistance to invasion. Our search terms were not targeted to select all papers that report disturbances, or their lack, during invasions. Thus, our results and discussion are limited to this search.

In the case of fire, we only had observations for native plants for human-initiated burns (e.g., prescribed burns), none from wildfire. Here, native species tended to have a negative response to burning, although this was not statistically significant. We found a similar trend when the disturbance was removal of vegetation via thinning. Both burning and removal of vegetation (by using herbicides, cutting, hand pulling) are the most common invasive plant removal management practices (Kettenring and Reinhart Adams 2011). However, as our results show, these practices could have unintended consequences for the native community. Extensive use of herbicides to reduce invasive plants, for example, can negatively affect native plants as well (Flory and Clay 2009; Rinella et al. 2009). Removal without further management may result in profound changes to the ecosystem that the native community is not adapted to handle (Zavaleta et al. 2001). Furthermore, because removal of vegetation is a disturbance on its own, this practice may only be effective in reducing invasions if natives are not affected and the availability of invader propagules is low (Firn et al. 2008). Only disturbances that promote native plants will be associated with resistance to invasion (Chance et al. 2019).

Recommendations

Risk assessments of plant invasions rarely account for native plant performance (Daehler 2003; Maron and Vilà 2001); but this information could provide additional insights for management aimed at decreasing vulnerability, or increasing resistance, to plant invasions. Informed by results from our analyses, we have developed four key recommendations that could be followed to minimize vulnerability to plant invasions: