https://orcid.org/0000-0002-7971-3593
Abstract
Invasive plants are often sold as ornamental garden plants, but how often are species sold in the same locations as where they are invasive? To answer this question, we assessed the geography of ornamental plant sales in the United States in relation to existing invasions and potential invasion risk with climate change. Using a data set of 672 nurseries selling 89 invasive plants, we estimated the distance between nursery sales and invasive species’ observed distributions. We also used species range maps to identify nursery sales within current potential invaded ranges, as well as within species’ future potential ranges given climate change. Half of the species were sold by a nursery within 21 kilometers of an observed record of invasion. Under the current climate, horticulture could be seeding invasions for 73 of the 89 species studied, and horticulture could facilitate climate-driven range expansion of 25 of 89 species. Our results illustrate risks associated with horticultural introductions of invasive species, providing guidance for regulatory and educational interventions that reduce the spread of ongoing and future invasions.
The negative impacts of plant invasions are well documented and span levels of biological organization, from altering soil microbial communities to changing ecosystem function (Vilà et al. 2011, IPCC 2022). Despite efforts to limit the spread of potential invaders, nonnative species introductions are unlikely to decrease in the coming decades (Seebens et al. 2015, 2017), with common pathways for species movement, such as the pet and plant trade, predicted to contribute to novel invasions worldwide (Early et al. 2016). In the United States alone, more than half of invasive plants are grown and sold as ornamental garden plants despite being identified as invasive somewhere in the country (Beaury et al. 2021b). In some cases, these invaders may be sold in regions where they have minimal risk of escaping cultivation because of unsuitable environmental conditions in the surrounding natural areas. However, climate change is shifting these environmental conditions, resulting in changes to invasive species distributions across the United States (Allen and Bradley 2016, Evans et al. 2022). To prevent ornamental plantings from seeding future invasions, it is important that we understand how horticulture sales relate to the geography of plant invasions and that we respond with appropriate federal and state regulations (Gordon et al. 2016).
The US horticultural industry is composed of more than 50,000 registered growers and plant dealers (Knuth et al. 2021), as well as online marketplaces for purchasing plants via websites such as Amazon and eBay (Humair et al. 2015, Beaury et al. 2021b). The growth of this industry has already been linked to increased invasion risk (Drew et al. 2010, Chapman et al. 2017, Van Kleunen et al. 2018), and more opportunities to purchase plants online could indicate a growing capacity for horticulture to move plants over long distances. If so, nursery sales might contribute to invasions near retail nurseries, as well as in disjunct geographies. Within the United States, horticultural sales are primarily regulated by state governments, which often restrict the sale of a specified list of invasive plants within their state’s borders (McCubbins et al. 2013, Beaury et al. 2021b). However, many US states’ regulations are outdated, biased toward a few agricultural weeds, or are irregularly enforced because of a lack of capacity (McCubbins et al. 2013, Quinn et al. 2013). As a result, the identity of regulated plants varies considerably across US states (Peters and Lodge 2009, Lakoba et al. 2020, Beaury et al. 2021a), and it is common for a state's prohibited species to be offered for sale legally in a neighboring state (Beaury et al. 2021b). Surveys of the horticultural industry suggest that nurseries are willing to comply with regulations (Burt et al. 2007, Coats et al. 2011), but inconsistencies in regulations exacerbate the potential that people unwittingly grow, sell, and plant ornamental invaders (Beaury et al. 2021b). Regulations also tend to be reactive; prohibited plant lists rarely include plants in anticipation of their spread to new areas (Lakoba et al. 2020, Beaury et al. 2021a), which is a missed opportunity to reduce the likelihood of ornamental escapes.
The reactive nature of the regulatory landscape also leaves the United States vulnerable to climate-driven invasions of ornamental plants (Lakoba et al. 2020). Nurseries primarily sell nonnative species (Harris et al. 2009, Bayón and Vilà 2019), and the more frequently a species is sold, the more likely it is to invade (Dehnen-Schmutz et al. 2007, Fertakos et al. 2023). Furthermore, many imported nonnative plants are adapted to warmer climates (Van der Veken et al. 2008, Bradley et al. 2012). As temperatures warm with climate change, natural areas could become more suitable for warm-adapted nonnative species (Haeuser et al. 2018). For species already known to be invasive in the United States, distribution models predict that climate change will cause major shifts in the hotspots of their invasive presence and abundance (Allen and Bradley 2016, Evans et al. 2022). Horticulture could already be facilitating this movement (e.g., Haeuser et al. 2018 naturalization risk in Europe), but the potential for this to occur has not yet been quantified in the United States.
To address this gap, we synthesized existing spatial data to assess the potential for horticultural sales to contribute to plant invasions. We first quantified the geographic extent across which US plant nurseries distribute any ornamental plant. We then compared this extent with horticultural sales of species defined as invasive in the United States: For 89 ornamental invasive species, we quantified the geographic proximity between horticultural sales and species’ observed invasive distributions, potential invaded ranges under current climate, and predicted future ranges given climate change. Many factors determine invasive species distributions, and our intent is not to establish a causal link between a horticultural sale and a species’s invasion. Rather, we aimed to use the overlap between horticultural sales and species’ invasions to assess potential risks associated with horticultural introductions, including identifying cases in which horticultural sales are unlikely to contribute to invasive species spread. Characterizing the geography of horticultural sales of invasive species can therefore be used to guide proactive regulation and outreach to prevent the spread of invasive species (Haeuser et al. 2018).
Data sources and approach
We focused analyses on 89 ornamental plant species that are defined as invasive in the contiguous United States. Our focal list included 64 species defined as invasive by either the federal noxious weed list or a US state prohibited plant list (Beaury et al. 2021a) and 25 species described as invasive by the Invasive Plant Atlas (www.invasiveplantatlas.org); invasive status was verified using the CABI Compendium Invasive Species (CABI 2022) and the Global Biodiversity Information Facility (GBIF 2022). We defined species as ornamentals if they were sold by four or more plant nurseries from the 1330 nurseries identified as selling invasive plants in the lower 48 US states (Beaury et al. 2021b); on average, the species were sold by 35 nurseries in this data set. Finally, we only included species that had sufficient distribution data to conduct spatial analyses (described below).
Quantifying the geographic extent of horticultural sales
We first identified how far an average US plant nursery distributes any ornamental plant in the contiguous United States, which provided a range of distances for which we might expect a plant to be planted in relation to a horticultural sale. We computed these distances for mail-order and retail nurseries (described below), hypothesizing that mail-order nurseries expand the range of distances across which plants could be distributed. We then used these distances to assess whether nursery sales could contribute propagule pressure to invasions, given how often observed invasions fell within the median and mean nursery sale radius.
To compute distances, we used Dave's Garden (https://davesgarden.com), which is an online resource for gardening that tracks plant trade across thousands of retail and mail-order nurseries in the United States. Dave's Garden provides geotagged customer reviews of each nursery, which we used to approximate each nursery's retail and shipping radius. We extracted nursery locations and customer reviews for nurseries specializing in retail sales (nurseries in the “Go Gardening” resource) and nurseries specializing in online mail-orders (nurseries in the “Garden Watchdog” resource). We extracted nursery names, addresses, reviewer locations, and total number of reviews per nursery using the rvest package for web scraping in R (Wickham 2022). Reviewers’ locations were listed by US town and state and may not coincide with where a plant is planted. Given these uncertainties, we focused on the range, mean, and median distances between reviewers and nurseries, which provides a coarse approximation of how far an ornamental plant might be distributed relative to a nursery's exact location.
The nursery addresses and reviewer locations were geolocated on geocod.io (www.geocod.io). Using the sf package in R (Pebesma 2018), we calculated the Euclidean distance between each nursery and a reviewer’s location. For retail nurseries, we calculated the distances between 3083 reviews of 1998 nurseries. For mail-order nurseries, we focused on a subset of 361 nurseries that were found to sell US invasive species according to Beaury and colleagues (2021b). Focusing on this subset increased our confidence in quantifying a potential range of distances across which invasive species could be distributed and provided a robust sample size of 22,617 reviews of the 361 nurseries.
Observed distributions
We compiled point locations for species invasions from five sources of plant abundance data, including data from EDDMapS, CalFlora, iMap Invasives, VegBank, Florida Invasives, and the Standardized Plant Community with Introduced Status Database (Petri et al. 2023; all sources are referenced and described in supplemental table S1). These data include hundreds of reports from natural resource practitioners, as well as standardized vegetation surveys and state data repositories, together representing more than 500,000 point locations, each associated with a species identification, a measure of percentage cover, and geographic coordinates located in the lower 48 United States. We only included abundance data (rather than presence observations) to ensure that species had established populations, which included point locations where species were recorded at at least 5% cover (Bradley 2016, Jarnevich et al. 2021, O'Neill et al. 2021). Each of the 89 species had a minimum of 100 geolocated observations of invasion, with an average of 1500 points per species (the final data set included 65,536 point locations across species). By only including abundance data, we provide a highly conservative estimate of observed distributions.
Using these point locations, we measured the distance between each observation of invasion and the nearest nursery selling that species on the basis of a subset of all nurseries compiled by Beaury and colleagues (2021b; n = 672 nurseries selling the 89 ornamental invasive species).
Predicted invaded ranges under current and future climate
Because of data gaps, the available point location data only capture a subset of the areas that have already been invaded, and it is likely that many invasive species in the United States are still spreading to fill their ecological niches (Bradley et al. 2015, Allen and Bradley 2016, McMahon et al. 2021). We were therefore interested in whether horticultural sales overlapped with each species’s potential invaded range, hypothesizing that nurseries might be at risk of facilitating invasive species range infilling or range expansion with climate change.
Each species’s potential invaded range was predicted using a rigorous ensemble distribution modeling framework (Evans et al. 2022) focused on modeling the range where species could achieve at least 5% cover. Following methodology developed by Young and colleagues (2020) and Engelstad and colleagues (2022), this modeling framework uses a species’s observed distribution to predict suitable habitat for invasion from an ensemble of five modeling algorithms and eight environmental predictor variables (primarily quantifying temperature and precipitation patterns). Environmental variables were selected on the basis of their explanatory power in a previous modeling effort mapping more than 50 invasive plant distributions (Engelstad et al. 2022) and given the availability of future climate projections (Abatzoglou et al. 2018). Models used a targeted background sampling approach and were validated using testing data and a tenfold cross validation approach (Young et al. 2020, Engelstad et al. 2022). All models performed well on the basis of AUC-ROC validation statistics (supplemental material; Evans et al. 2022). Additional model details are provided in the supplemental material, as well as in Evans and colleagues (2022), Young and colleagues (2020), and Engelstad and colleagues (2022). Specifically, Evans and colleagues (2022) describe the geography of invasive plant distributions, shifts and hotspots of invasion with climate change, and model uncertainties or limitations. The resulting suitability maps for each species are publicly available on FigShare (https://doi.org/10.6084/m9.figshare.21395001.v1.).
From the continuous suitability maps created and provided by Evans and colleagues (2022), we identified each species’s potential invaded range as the areas with high suitability for invasion given majority model agreement (at least 70% of the models predicted high suitability). This resulted in a binary range map for each species, with areas categorized as suitable or unsuitable for invasion at at least 5% cover (i.e., abundant invasions).
Evans and colleagues (2022) repeated the ensemble distribution modeling approach described above to map suitability for invasion under a scenario of 2 degrees Celsius of warming (representative concentration pathways 4.5–6.0, in which emissions either stabilize or slowly decline by 2100; IPCC 2022). Using these maps (see the supplemental material), we asked whether nurseries were located within areas predicted to be highly suitable for abundant invasions under future climate change (at least 70% of models predicted high suitability given climate change).
Geographic overlap between horticulture and plant invasions
US nurseries were reviewed by customers anywhere from less than 0.2 kilometers (km) to 4500 km away from the nursery location, representing proximities of from down the road to transcontinental (figure 1). Across nurseries, the median distance of a customer review was 255 km, with a mean of 778 km. Retail nurseries tended to have a smaller sale radius, with customer reviews at a median distance of 21 km (and a mean of 161 km). Customer reviews of mail-order nurseries were within a median distance of 1064 km (and a mean of 1395 km). Distribution distances were generally variable, but even nurseries specializing in retail sales could be distributing plants across large distances (figure 1). Although this is a coarse estimate, these distances provide a mean and median expectation for the distance between a nursery sale and where an ornamental plant may have been planted.
Distribution of distances between nurseries and geolocated customer reviews of businesses specializing in retail (n = 3083 reviews of 1998 nurseries) versus mail-order (n = 22,617 reviews of 361 nurseries) sales. Median values are 21 km and 1064 km, respectively.
We then asked whether the observed distributions of the 89 ornamental invasive species were within the median retail and mail-order radius (21 km and 1064 km) of a nursery where that species is sold. There was considerable variation in the distances between nursery sales and the species’ observed distributions (supplemental figure S2): The median distance between a nursery and a species’s invasion record was 647 km (with a mean of 681 km). However, 49 of the 89 ornamental invasives were sold within 21 km of an observed record of invasion, and 88 of 89 species were sold within 1064 km of a record of invasion. These geographic proximities suggest that for most species, nurseries are within a reasonable distance in which they could provide propagule pressure to existing invasions, particularly for the 49 species sold within a retail sale radius.
The closest proximity between an invasion and a horticultural sale was measured at 0.69 km, and the furthest was at 4249 km (the median distance was 427 km). Coincidentally, both the minimum and maximum distances were associated with invasions of Foeniculum vulgare (fennel; figure 2). This species is a regulated noxious weed in Washington state and considered invasive elsewhere in the United States, including coastal California, where most of the abundant populations were observed (the USDA PLANTS Database reports F. vulgare as introduced to 35 of the lower 48 United States; https://plants.sc.egov.usda.gov/home). It is possible that horticultural sales of fennel are cultivated varieties that lack invasive traits (Skinner 2022), but this has not been verified by other sources (Cal-IPC 2022). In general, noninvasive varieties could reduce invasion risk (Deng et al. 2020, Trueblood et al. 2010), but most evidence suggests that cultivars are able to genetically mix with invasive varieties and can revert to their original traits over time (Knight et al. 2011, Datta et al. 2020). Therefore, F. vulgare illustrates the potential for horticultural sales to provide propagule pressure to invasions in specific regions of the United States but also highlights cases in which horticulture is distributing F. vulgare in areas where this species has not been observed to be invasive (e.g., the northeast United States in figure 2).
Recorded invasions (observations of populations at least 5% cover) and nursery sales of invasive fennel (Foeniculum vulgare) within the contiguous United States, highlighting both the close proximity between nursery sales and invasions as well as nursery sales that are far outside of the range of areas where species invade.
We next compared nursery sales with the species’ predicted invaded ranges under current climate. Most species (73 of 89) were sold by at least one nursery within their potential invaded range, indicating a widespread risk of range infilling. We found that 28 species had more than 10 nurseries within their potential range, and three species had more than 50 nurseries within their potential range: Cichorium intybus (chicory, prohibited in Colorado and New Mexico), Daucus carota (wild carrot, prohibited in Iowa, Kansas, Michigan, Minnesota, and Washington; figure 3d), and Vinca minor (common periwinkle). Vinca minor is not currently prohibited by the federal noxious weed list or any state prohibited list in the United States, but it is described as highly invasive by the CABI Compendium Invasive Species (CABI 2022).
Examples of nursery sales within current and future ranges of four ornamental invasive plants with high risk of spreading because of horticulture: (a) Albizia julibrissan, (b) Pyrus calleryana, (c) Ficus carica, and (d) Daucas carota.
Nursery sales could also facilitate the range expansion of ornamental invaders, given the rapidity of climate change (IPCC 2022). For 65 of 89 species, at least one nursery was located in an area predicted to be suitable for invasion under future climate change (e.g., figure 3b). For 40 of these species, this occurred because a nursery was already located in an invader's range, and this area will remain suitable for invasion under future climate change. Therefore, the risk of nursery-assisted expansion is already high under current climate and continues with climate change.
For the other 25 species, a nursery was located in a currently unsuitable area that will become more suitable for invasion as temperatures warm (e.g., figure 3). For these ornamental invaders, we have an opportunity to proactively limit their spread by increasing education and outreach to horticulture and by adding species to climate watch lists (box 1). Although this provides a promising starting point for intervention, this analysis only examined about 10% of the larger pool of US ornamental plants known to be invasive and sampled only a subset of the nurseries that could be selling invasive species in the United States (Beaury et al. 2021b). We are therefore likely underestimating the extent to which ornamental plant sales could contribute to current and future invasions.
Discussion and solutions
The horticultural industry is the primary source of invasive plants into the United States (Lehan et al. 2013), and hundreds of plant species identified as invasive somewhere in the United States are still available for sale through mail-order and retail nurseries (Beaury et al. 2021b). Many factors determine which species are sold and where, such as climatic suitability (Haeuser et al. 2018), cost and ease of propagation (Dehnen-Schmutz et al. 2007, Drew et al. 2010), and market competition (Drew et al. 2010). But prior to this study, it was unclear how often US nurseries sell ornamental plants in the same geographies where they are known to invade and affect native ecosystems. We highlighted instances in which nurseries fell well outside of areas where species are likely to invade (e.g., figures 2 and 3), but we also found substantial evidence that invasive species are sold as ornamentals near existing invasions and within the geographic range of areas susceptible to invasion. Furthermore, we show that nursery sales could facilitate climate-driven range expansion of 25 of the 89 high-impact invasive species analyzed in the present article. Horticulture therefore presents a major risk of spreading invasive plants under both current and future climates, and the proximities we highlighted can inform interventions that focus on high risk of propagule pressure from trade.
In the United States, horticultural sales are primarily regulated by state governments, but inconsistencies across regulations preclude effective management of ornamental invaders at regional and continental scales. We found that the geographic distribution of horticultural sales extends far beyond the state level (figure 1), particularly for mail-order nurseries. If nursery trade is not regulated in a manner consistent with the scope of horticultural trade, new invasions could arise. Regulating online plant trade is an emerging challenge for invasive species management, but geographic analyses such as those discussed in the present article (and those conducted for other regions; e.g., Haeuser et al. 2018) can inform where and how to strengthen regulations. For example, proactively regulating the 73 species sold within their potential ranges under current climate, and the 25 species that we identified as likely to be facilitated by horticulture under future climate, could prevent further ornamental escapes (box 1). In addition, increasing awareness of where invaders’ potential ranges are located could help nurseries avoid shipping ornamental species to locations where they could become invasive (box 1).
Like most sectors, horticulture must adapt to a changing climate, but this will come at an ecological cost if horticultural practices continue to include nonnative plants (Haeuser et al. 2018, Van Kleunen et al. 2018). Previous studies suggest that most nursery stock is nonnative (Harris et al. 2009, Bayón and Vilà 2019, Fertakos et al. 2023), but this may be changing as invasive species awareness increases and gardening trends shift to favor native species (e.g., the growing popularity of native pollinator gardens). Native plants have lower invasion risk (Simberloff et al. 2012) and provide benefits to native biodiversity (Fusco et al. 2018). For example, landscapes dominated by native plants support a higher diversity and abundance of native birds, bees, and other pollinating insects (Tallamy and Shropshire 2009, Burghardt et al. 2010, Narango et al. 2018).
Gardening with warm-adapted native and neonative (range-expanding native) plants can also be a viable climate adaptation strategy (Bradley et al. 2023). Native species are 40 times less likely to spread rapidly than are introduced species (Simberloff et al. 2012), but appropriate caution is still needed when expanding the ranges of native species in response to climate change (i.e., neonative species; Brodie et al. 2021). For example, ornamental neonative species could expand and outcompete other species in natural areas if they are weedy or common in their native ranges or if they alter ecosystem structure, chemistry, or function (Wallingford et al. 2020). Additional challenges include breeding easy-to-grow native plants that maintain the ecological benefits of wild types (e.g., breeding for new flower shapes can reduce pollination services; Corbet et al. 2001, Ricker et al. 2019) and cultivating enough native nursery stock to meet the rise in demand (Beckwith et al. 2022, Rihn et al. 2022). But studies have shown that consumers are willing to pay more for native plants (Yue et al. 2011), and ecologically, it is clear that native plants must be part of the solution if we are to reduce invasion risk and adapt to climate change.
Although substantial evidence links horticultural trade with species invasions (Dehnen-Schmutz et al. 2007, Drew et al. 2010, Van Kleunen et al. 2018, Beaury et al. 2021b), our correlative results do not necessarily suggest a causal link between a horticultural sale and an observation of a species’s invasion. There are many drivers of invasion in addition to horticultural introductions, and it remains a challenge to identify how invasive populations move across a landscape following initial introduction. Biases in distribution data and uncertainty in distribution model output also create challenges for prioritizing management on the basis of invasion risk (Sofaer et al. 2019). For example, the models presented by Evans and colleagues (2022) meet most criteria for informing management decision-making (see Sofaer et al.’s 2019 table 1b and 1c; quality location data, predictive relevance, algorithm choice, etc.). But the range maps are a first attempt of predicting climate change impacts on invasive plant distributions (i.e., the models are not yet iterative and are difficult to verify for accuracy; Sofaer et al. 2019) and therefore must be interpreted with caution. Furthermore, we focused on 89 species with data sufficient for this analysis using a conservative cutoff that excluded any point location that lacked abundance information or where species were observed at low cover (Evans et al. 2022). These 89 species represent a small subset of the approximately 800 ornamental plants invading the United States (Beaury et al. 2021b), such that we are likely further underestimating risks from horticulture given the limitations of the analysis.
As more distribution data become available, we may be able to identify additional potential invaders from ornamental escapes and work with horticulture to prevent novel introductions. For example, in a less data-intensive analysis, Allen and colleagues (2022) identified 24 high-impact ornamental invasive plants projected to range shift into the Northeast with climate change (box 1), nine of which overlap with the species analyzed in the present article. Allen and colleagues (2022) created this resource to inform horticultural professionals about range-shifting invasive species and to support community members in planting alternatives to high-risk invaders. Similarly, Haeuser and colleagues (2018) explored invasion risk associated with species that are already present in European horticulture but have yet to naturalize, potentially because of an invasion debt associated with climate change. Proactively identifying these risks, and other resources or infrastructure that incentivizes people to make ecologically mindful decisions could have a major impact on what the ecosystems of the future look like, especially in light of climate change (Bradley et al. 2023).
Conclusions
We show that current nursery practices are at risk of seeding existing and future invasions, and the reach of horticulture extends far beyond state borders. State-centric and reactive regulations are therefore unlikely to reduce invasive plant spread across the United States. Preventing invasive species spread and range expansion ultimately requires broadening the scale at which we regulate and manage ornamental invaders. This includes using geographic and predictive analyses to inform invasive species regulations and climate watch lists (box 1), sharing information about high-risk ornamental invaders across states and regions, and working with horticulture and community members to reduce the escape of ornamental species into natural areas.
Acknowledgments
Thank you to members of the Princeton Levine Lab and M.F. Michael Nelson for feedback. Additional thanks to the US Geological Survey and the INHABIT team for their contributions to the distribution models used to map species ranges. This work was supported by the U.S. Geological Survey, Northeast Climate Adaptation Science Center Award Number G19AC00091, and the U.S. Geological Survey, Northeast Climate Adaptation Science Center Award Number G21AC10233-01.
Author contributions
Evelyn M. Beaury (Conceptualization, Data curation, Formal analysis, Methodology, Validation, Visualization, Writing – original draft, Writing – review & editing), Jenica M. Allen (Conceptualization, Visualization, Writing – original draft, Writing – review & editing), Annette E. Evans (Conceptualization, Data curation, Formal analysis, Writing – review & editing), Matthew E. Fertakos (Conceptualization, Data curation, Visualization, Writing – review & editing), William G. Pfadenhauer (Conceptualization, Data curation, Visualization, Writing – review & editing), and Bethany A. Bradley (Conceptualization, Visualization, Writing – original draft, Writing – review & editing).
Author Biography
Evelyn Beaury ([email protected]) is a Postdoctoral Research Associate at Princeton University, in Princeton, New Jersey, in the United States. Bethany Bradley is a professor, Jenica Allen is an adjunct assistant professor, Annette Evans is a postdoc, Matthew Fertakos and William Pfadenhauer are both PhD candidates at the University of Massachusetts Amherst, in Amherst Massachusetts, in the United States.
