INTRODUCTION

Invasive species are well known to wreak substantial ecological harm to biological communities (e.g., Pyšek et al. 2012; Gallardo et al. 2015). Attempts to lessen the impact of invasive species by eradicating or controlling them can be a major cost for protected areas managers (Pimentel et al. 2005). Numerous strategies are available to combat nonnative species (Ruiz and Carleton 2003; Hulme 2009). Because of the great cost and often futility of eradicating nonnative species once they become established, managers have increasingly attempted to follow an “early detection and rapid response” (EDRR) strategy to prevent wide and extensive outbreaks (Reaser et al. 2020a). EDRR comprises several components, including target analysis (determining the effort needed to detect an invasion), detection, identification, reporting, risk screening, feasibility screening, and response (Reaser et al. 2020a).

Watch lists, defined as lists of invasive species to be prioritized for surveillance, reporting, and other responses that may reduce the risk of impact to valued resources, can serve as an important tool for the target analysis component of EDRR (Frey 2017; Reaser et al. 2020b). For this project, we consider watch lists as lists of species that have not been recorded in the area under management but may arrive and present management challenges in the future. Watch lists can help managers prioritize the types of habitats where ongoing surveillance for nonnative species takes place and to guide the selection of survey techniques employed to search for newly arrived nonnative species. Watch lists can also help managers determine the equipment, training, and other needs to facilitate rapid control responses in the event of an arrival of a new nonnative species. For example, the equipment needed to deter an invasive aquatic plant is distinct from that used to control an invading insect herbivore.

By their nature, watch lists cannot usually be developed empirically from survey data. In most cases, they require expert knowledge of the nonnative species occurring in the region surrounding a managed area. However, many managers lack sufficient resources and staff capacity to develop watch lists. State floras can help to identify nonnative plant species, but they are typically updated too infrequently (compared to the rate of arrival of nonnative species) to serve as the sole source of input into a watch list. For example, the flora for New York was last published in 2017 (Werier 2017), updating the previous 2003 list (Mitchell and Tucker 2003). During that 14 y period, 263 nonnative species were added (Werier 2017). Also, for large and/or ecologically heterogenous states, state-level records of nonnative species are less useful because of the uncertainty of the proximity of records to a particular managed area (and therefore the likelihood of invasion into the managed area).

Citizen science (also known as “community science”) schemes and the access they provide to broadscale observations across species’ ranges present an opportunity to create provisional watch lists in the absence of expert knowledge about regional biota. With a list of nonnative species occurring in a country or region, it should be possible to mine databases of georeferenced citizen science observations to develop watch lists for managed areas. To be successful, such an exercise should take place in an area where the citizen science scheme is in broad use (i.e., at least hundreds of observations of nonnative species each year), the species of interest are reliably identifiable by citizen scientists, and the data are made available soon after observations take place. Although citizen science observations are inherently incomplete and the level of incompleteness is unknown, they nonetheless have the potential to provide useful information in the absence of more rigorous survey data.

Here we demonstrate a novel use of observation data made available through iNaturalist (iNaturalist 2020) to develop provisional watch lists of nonnative plants for US national park units with limited budgets. iNaturalist enables users to share georeferenced images and other media of plants and animals on a mobile-friendly platform. We describe and make available a series of scripts that draw from a master list of nonnative species to extract records of these species occurring within a defined buffer of a national park unit, but that are not yet recorded in the park. We summarize the results of this exercise, describe the experience of an independent team implementing the process, suggest ways to refine the output to make it more useful for managers, and discuss taxonomic considerations, sourcing of lists of nonnative species, and challenges that arose. Our method can be adopted to develop provisional watch lists for both plants and animals in managed areas worldwide.

METHODS

Derivation of Watch Lists

To derive provisional watch lists for each unit, we identified all species with records inside the 160 km zone but not inside the unit's boundaries. Because exceedingly long watch lists can be unwieldy and perhaps less helpful to resource-limited land managers, we arbitrarily truncated each list at 100 species, prioritizing the species with the greatest number of records in the buffer. In the case of a tie between the 100th most recorded species and other species with the same number of records, we included the additional species, yielding lists slightly longer than 100 species in some cases. The final watch lists include both the list of species and the corresponding number of records within 160 km of the unit.

Figure 1.

Locations of US National Park Service (NPS) units included in the study. Unit codes as in Table 1. Colors separate NPS administrative units used to stratify the units addressed in the study.

RESULTS

Among the 36 park units, there were 0–119 (mean 27.4, SD 27.1) nonnative species recorded within their boundaries and 0–875 (mean 417.3, SD 279.1) nonnative species outside their boundaries but within the 160 km buffer (Table 1). The number of iNaturalist records for nonnative species recorded both within park units and in the buffers varied across several orders of magnitude (Table 1). The full watch lists generated for each unit are presented in  Table S1 (307_naar-41-04-08_s01.docx), with species listed in decreasing order by the number of records. Sample output maps are shown in  Figures S1 (307_naar-41-04-08_s01.docx) and  S2 (307_naar-41-04-08_s01.docx).

Figure 2 shows the locations of iNaturalist records for 2 nonnative species occurring in the buffer area around Fort Vancouver National Historic Site located in southern Washington. In this example, Cakile maritima Scop. has numerous records within the buffer but is restricted to the coast and therefore likely does not pose a threat to Fort Vancouver. Sorbus aucuparia L., on the other hand, has been recorded throughout the terrestrial area surrounding Fort Vancouver and therefore may pose a true threat.

NPS personnel were able to implement the process successfully for all established units with natural areas. They identified small changes needed in the script to improve the workflow and results, such as adding the analysis date to the output maps. We subsequently revised the script accordingly. For Jean Lafitte National Historical Park, the team supplemented the USDA nonnative species list with local lists of nonnative species for a more complete search for nonnative species. Care must be taken when doing this as duplicate species can appear on the draft watch list when taxonomic synonyms occur on the different nonnative species input lists. A botanist reviewed the list for Jean Lafitte, (1) identifying several species on the draft watch list that had been recorded in the park but not in iNaturalist and (2) noting some important invasive species in the region of the park that were detected by iNaturalist participants but not frequently enough to appear in the top 100 species in the draft watch list.

DISCUSSION

This study demonstrates the potential utility of citizen science data for automating the development of provisional watch lists to contribute to EDRR programs at areas managed for biodiversity protection. These lists can serve as draft versions of watch lists that could be refined using a variety of other tools as discussed below. These initial lists can be quickly generated and do not require any botanical expertise.

Table 1.

Species counts and numbers of records of nonnative plants found within US National Park Service units or in a 160 km buffer surrounding the unit (and not in the unit itself). Abbreviations: NHP, National Historical Park; NHR, National Historical Reserve; NHS, National Historic Site; NM, National Monument; NP, National Park; NRA, National Recreation Area.

The provisional watch lists created with citizen science data can be improved in several ways, all of which refine the list to species posing the greatest risk. First, as demonstrated in Figure 2, the distribution of species on the list can provide clues about the degree to which they threaten to disperse and establish in the managed area. In this example, a coastal species was highlighted as likely to be unthreatening. In other contexts, species restricted to elevations that do not exist in the managed area or those in landforms such as riparian areas that are not found in the protected area can be eliminated from the watch list.

Incorporating botanical knowledge can also aid in interpretation of species on provisional watch lists. Species that grow in microhabitats that do not occur in the managed area could be eliminated. A botanist should also be able to eliminate ornamental or crop species that cannot become established on their own. For example, the nonnative plant recorded most abundantly in the buffer around Fort Vancouver was Digitalis purpurea L. The images accompanying some of these records suggest that, although many are likely growing as weeds, others were likely planted as ornamentals (e.g., iNaturalist observation 25874403). Also, as the Jean Lafitte experience demonstrated, a botanist familiar with a park can identify species on a draft watch list that are already known from the park but not detected by citizen scientists.

Figure 2.

Map showing the locations of iNaturalist records of two nonnative plants occurring within a 160 km buffer around Fort Vancouver National Historic Site (FOVA) in southern Washington. Shading depicts topography.

Additionally, adding information about the invasive potential of species in the region where the managed area is located could help prioritize species on the list (Frey 2017). Several methods are available to assess the potential invasiveness of nonnative species (Morse et al. 2004; Colautti et al. 2014), and databases with the results of these assessments could be compared with the provisional watch list to prioritize known invasive species. Similarly, watch list species that appear on lists published by state invasive plant/species councils or exotic pest plant councils could also be prioritized. For example, the Washington Invasive Species Council lists 25 priority invasive plants and plant groups ( https://invasivespecies.wa.gov, accessed 29 Oct 2020). Four of these species (Alliaria petiolata (M. Bieb.) Cavara & Grandee, Buddleja davidii Franch., Lythrum salicaria L., Polygonum cuspidatum Siebold & Zucc.) appear on our provisional watch list for Fort Vancouver and should be prioritized for surveys within the site. Four additional species, Myriophyllum spicatum L., Centaurea nigra L., Centaurea diffusa Lam., Centaurea stoebe L., were also reported in the buffer, but were not among the 100 most common species and therefore did not enter the default watch list. Given the documented threats they pose to natural ecological communities, these four species should be elevated to appear in the watch list. Similarly, the Jean Lafitte exercise showed that important invasive species that deserve watch list status may not be among the 100 most commonly reported species for the buffer area. This process of risk evaluation is fundamental in developing watch lists for natural areas with high public scrutiny (Reaser et al. 2020b).

Although a watch list drawn up by a local expert is often desirable, there are benefits of following the more automated method we present here. Most importantly, the automated method can be repeated regularly to keep abreast of recently documented nonnative species. A newly compiled list can be compared with one compiled a year previously to identify newly detected and possibly newly arrived nonnative species. For example, the nonnative Cichorium intybus L. was not documented by iNaturalist observers in the buffer of Fort Vancouver until 2020 and therefore would not have appeared on an automated watch list generated in 2019.

Except in the case of taxonomic mismatches, the automated method will comprehensively list all species recorded in the observational database used that fulfill the criteria for inclusion on a provisional watch list. On the other hand, experts may be aware of arrivals of nonnative species that have been reported elsewhere. Experts can also filter noninvasive species or those without available habitat on the target managed area more efficiently than can an algorithm. For Fort Vancouver, prioritizing the default watch list by number of records in the buffer missed four important species.

Like many research applications, using citizen science observations for compiling a watch list has advantages and disadvantages. The benefit, of course, is free access to ongoing, often large numbers and geographically distributed observations of nonnative species. Biases in what and where citizen scientists report are well known (Kosmala et al. 2016). Biases that could affect the development of watch lists include avoidance of sites that are inaccessible, far from home, or in areas with low species richness (Mair and Ruete 2016; Tiago et al. 2017). However, many invasive nonnative plants are weedy species that occur along roads or trails, areas that citizen scientists are documented to favor (Mair and Ruete 2016).

Good coverage by citizen science observers of the buffer around managed areas is a prerequisite to using their data to develop watch lists. We found abundant iNaturalist records of nonnative plants in the buffers around most parks, with over 1000 records of nonnative plants at 29 (81%) of the 36 units examined. Fort Vancouver, located near the Vancouver, Washington-Portland, Oregon, metropolitan area with a sizeable pool of potential citizen scientists, had 27,379 records of nonnative plants in the buffer area. This large number of observations confers confidence that the provisional watch list likely contains many important invasive species threatening the site. Even in poorly surveyed areas, the available citizen science data can provide an initial, albeit incomplete, list of nonnative species that potentially could invade a managed area. Note though, that in remote areas such as Gates of the Arctic National Park and Preserve, the lack of invasive species records is most likely due to the lack of observations; citizen science cannot aid in watch list development in these regions. Managers can encourage additional reporting of nonnative species through a targeted project in iNaturalist (which notifies observers to specific needs for observations) and by adjusting their iNaturalist subscriptions to be alerted via email whenever a watch list species is observed within the park boundaries. Additional options include use of an alternative platform such as iMapInvasives (Hinchey et al. 2013) in the United States or by organizing a bioblitz (Baker et al. 2014).

We used a national, digital list of nonnative species to identify recorded species for inclusion in watch lists. This approach facilitated the development of watch lists for NPS units across the United States. In some areas, a state/provincial (e.g., Keener et al. 2020) or regional (e.g., Weakley 2010) flora that distinguishes between locally native and nonnative species may be preferable to the national list, for example by including species that may be native to the country but nonnative to the region of interest.

Regardless of which reference list of nonnative species is used, there will be nomenclatural mismatches with the iNaturalist taxonomic backbone. Plants, with no widely accepted national list in the United States, are particularly prone to this problem. Use of synonyms when matching names can help lower the rate of missing nonnatives that have been recorded in a buffer but are not detected because the name used in iNaturalist is different from the name used in the reference list of nonnative species. In the absence of taxonomic reconciliation, managers should recognize that the resulting watch list may not include all nonnative species occurring in the buffer.

Although the method we present was tested on NPS units, the script could be used anywhere with both API access to georeferenced citizen science observations of the taxa of interest and a list of nonnative species. The method works on animal groups just as well as with plants. Observations from multiple citizen science schemes could be combined to increase the number of records available for detecting nonnative species. With no sign of abatement in the spread of nonnative species, tools such as the one we describe here can be important components of EDRR strategies to protect native ecosystems.