Scanning the horizon for invasive plant threats to Florida, USA

Early detection and eradication of invasive plants are more cost-effective than managing well-established invasive plant populations and their impacts. However, there is high uncertainty around which taxa are likely to become invasive in a given area. Horizon scanning, which pairs rapid risk assessment with consensus building among experts, can help identify invasion threats. We performed a horizon scan of potential invasive plant threats to Florida, USA—a state with a high influx of introduced species, conditions that are favorable for plant establishment, and a history of negative impacts from invasive plants. We began with a list of 2128 non-native plant species and subspecies that are crop pests or invasive somewhere in the world and used publicly available data to prioritize 100 taxa for rapid risk assessment. We derived overall invasion risk scores by evaluating the likelihood and certainty of each of the 100 taxa arriving, establishing, and having an impact in Florida. Through the rapid risk assessments and a consensus-building discussion, we identified six plant taxa with high overall risk scores ranging from 75 to 100 out of a possible 125. The six taxa are globally distributed, easily transported to new areas, found in regions with climates similar to Florida’s, and can impact native plant communities, human health, or agriculture. We recommend more thorough risk assessments for each of these six species and, if appropriate, policy and management actions to limit invasive plant introduction and establishment in Florida. 


Introduction
by the Centre for Agriculture and Biosciences International (CABI), and used publicly 105 available data to identify the 100 plant taxa most likely to arrive and have impacts in Florida. 106 Then, during the risk assessment phase, experts assigned scores for likelihood of arrival, 107 establishment, and impacts in Florida to the 100 taxa. Risk assessments were peer-reviewed, 108 and taxa were ranked based on their scores. Finally, in the consensus building phase, experts 109 reviewed the ranked list and suggested modifications. Here, we present the ranked list of 110 potential invasive plant threats to Florida, which can be used to inform research, 111 management, and policy aimed at reducing invasive plant impacts in Florida. 112 113 Methods 114 We adapted the horizon scanning method outlined by Sutherland et al. (2011) and Roy et al. 115 (2014Roy et al. 115 ( , 2015 to develop a ranked list of invasive plant threats and their potential pathways for 116 arrival to the state of Florida over the next ten years (2020-2030). We chose a ten year time 117 frame to capture threats in the near future and to establish a minimum frequency (once every 118 ten years) for updating the horizon scan with new information. 119 120 Expert panel and workshop 121 We (the authors) formed the expert panel, providing knowledge of Florida's natural systems 122 and existing invasive plants, including experience in invasion ecology, botany, policy, and 123 data analysis. We are employed by governmental, academic, and non-profit organizations 124 (Environmental Defense Fund, Florida Department of Agriculture and Consumer Services, 125 Florida Fish and Wildlife Commission, University of Florida, and United States Geological 126 Survey), which supported our participation. We organized and attended a workshop in 127 December 2019, during which we designed criteria for selecting taxa to assess (see 128 Assembling a list), discussed the rapid risk assessment tool (see Assessing and scoring the 129 taxa), and identified online resources for completing the rapid risk assessments. 130 131 Assembling a list 132 Using the horizon scan tool developed by CABI (an inter-governmental not-for-profit 133 organization that provides information and expertise on agriculture and the environment), we 134 generated a preliminary list of invasive taxa and crop pests that are not known to be present 135 in Florida based on CABI's databases (CABI 2018). The tool consolidates information from 136 the CABI Invasive Species Compendium and Crop Protection Compendium, which are 137 science-based encyclopedic databases. The tool generated a list of 9629 taxa, 2128 of which 138 were in the kingdom Plantae. 139 140 We first corrected the preliminary list for synonyms by compiling accepted names and 141 synonyms from the Atlas of Florida Plants (Wunderlin et al. 2019), the Taxonomic Name 142 Resolution Service (TNRS; Boyle et al. 2015), and the Integrated Taxonomic Information  143 System (ITIS; ITIS n.d.; see Suppl. material 1: Methods S1 for more details), which increased 144 our list to 2360 taxa. The modified CABI list was then trimmed based on several criteria ( Fig.  145 1), including: climatically matched to Florida, not naturalized in Florida (i.e., a self-146 sustaining population), not on a Florida or federal noxious weed list, naturalized outside of its 147 native range, historically weedy, and commonness (Suppl. material 2: Table S1). If taxa met a 148 criterion, they were retained for further assessment in the next stage. We assessed 149 commonness by the number of global occurrences (GBIF.org 2021a) and selected the top 100 150 (12%) for further assessment, which was the largest number of taxa that the experts felt they 151 could feasibly evaluate. 152 153 154 Figure 1. Criteria  habitats where they did not co-evolve with other species, were selected (912 taxa); (7) taxa 164 with a record of "weediness", suggesting the ability to produce a self-sustaining population 165 and have at least mild impacts on the surrounding environment (Randall 2017), were selected 166 (808 taxa); (8) the top 100 taxa, ranked by number of global occurrences (GBIF.org 2021a), 167 were selected. More details on the datasets used to inform these criteria can be found in 168 Suppl. material 1: Methods S1. 169 170 We used expert opinion to remove and add taxa from the list before and after finalizing the 171 top 100. Pastinaca sativa was previously assessed by one of the experts and determined to be 172 not a threat to Florida (removed before commonness was assessed). Two taxa were removed 173 from the list of 100 taxa: Rosmarinus officinalis and Galeopsis tetrahit. Rosmarinus 174 officinalis was known by experts to be naturalized in Florida and found to have 13 recent 175 herbarium records. There was not enough information on Galeopsis tetrahit available to 176 complete the rapid risk assessment and very few of its GBIF occurrences were inside 177 Florida's Köppen-Geiger climatic zones (Galeopsis tetrahit L. in GBIF Secretariat 2021). 178 Because GloNAF, the database used for criterion 6 ( Fig. 1 score was the product of arrival, establishment, and impact likelihood scores. We provided 215 brief justifications for our scores and assigned certainty scores that ranged from very low 216 (i.e., all scores were equally likely) to high (i.e., could confidently eliminate all other scores). 217 The overall certainty score was the score most consistent with three component certainty 218 scores. 219 220 Closest observation to Florida is within the contiguous U.S., but not in the southeastern | U.S., and the closest online seller to Florida is outside of the contiguous U.S.

2
Closest observation to Florida and closest online seller to Florida are within the contiguous U.S., but not in the southeastern U.S. or closest observation to Florida is in the southeastern U.S., but not in Florida, and the closest online seller to Florida is outside the contiguous U.S.

3
Closest observation to Florida is within the southeastern U.S., but not in Florida, and the closest online seller is within the contiguous U.S. or the southeastern U.S., but not in Florida.

4
The taxon has been observed or sold within Florida. 5

Establish ment †
No observations in areas with matching Köppen-Geiger (KG) zones to Florida ¶ . 1 Few observations in one area with matching KG zones to Florida. 2 Many observations in one area or few observations in multiple areas with matching KG zones to Florida.

3
Many observations in multiple areas with matching KG zones to Florida. 4 Criteria for score 4 plus evidence of a biological strategy that aids establishment or evidence of establishment in Florida.

Impact
Unlikely to cause negative impacts on the native biota or abiotic environment, human well-being, or economic systems.

1
Likely to cause (a) declines in the performance (e.g., biomass, body size) of native biota, but no decline in native population sizes or (b) income loss, minor health problems, higher effort or expense to participate in activities, increased difficulty in accessing goods, or minor disruption of social activities, but no significant impact on participation in normal activities. The taxon has no other impacts that would cause it to be classified in a higher impact category.

2
Likely to cause (a) declines in the population size(s) of native species, but no changes to the structure of communities or to the abiotic or biotic composition of ecosystems or (b) changes in the size of social activities, with fewer people participating, but the activity is still carried out. These changes to social activities could be linked to accessibility to the activity area or mild effects to human health (e.g., allergies). The taxon has no impacts that would cause it to be classified in a higher impact category.

3
Likely to cause (a) the local or population extinction of at least one native species, leading to reversible changes in the structure of communities, the abiotic or biotic composition of ecosystems or (b) the local disappearance of a social or economic activity from all or part of the area invaded by the alien taxon, collapse of the specific activity, switch to other activities, abandonment of activity without replacement, emigration from region, or moderate effects to human health. The taxon has no impacts that would cause it to be classified in a higher impact category.

4
Likely to cause (a) the replacement and local extinction of native species and will produce irreversible changes in the structure of communities and the abiotic or biotic 5 "escape from confinement" (intentional commodity that escapes, such as a horticultural 237 taxon), "transport contaminant" (associated with the transport of a specific commodity, such 238 as a seed contaminant), "transport stowaway" (other forms of unintentional transport, such as 239 through soil on equipment), "corridor" (through human infrastructure linking previously 240 unconnected areas, such as a waterway), and unaided (natural dispersal). 241 242 Reviewing and modifying scores 243 In the first round of reviews, ten of us reviewed ten assessments each. We searched for each 244 taxon in references that we found helpful when completing risk assessments and filled in 245 information where the assessor indicated that the certainty was "low" or "very low". We 246 changed scores, edited justifications, and wrote comments, differentiating our text from the 247 original assessor by using red font (Suppl. material 3: Table S2). We aimed to complete the 248 reviews efficiently, spending 30 minutes or fewer on each. We then ranked the taxa by their 249 overall risk score and all group members read the reviewed risk assessments. 250 251 Consensus building 252 During the two hour virtual meeting, we discussed taxa in descending order of scores. We 253 spent extra time discussing taxa with more controversial scores, such as taxa with large 254 discrepancies between scores assigned by the assessor and reviewer, which were reviewed 255 again following consensus building. We determined overall risk score thresholds to categorize taxa as high, medium, or low risk: 262 a taxon scoring ≥ 64 (i.e., an average score of 4 for each variable of arrival, establishment, 263 and impact) was categorized as high risk, a taxon with a score between 27 (i.e., an average 264 score of 3 for each variable) and 63 as medium risk, and a taxon with a score less than 27 as 265 low risk. 266 267 Analysis of risk scores 268 We evaluated whether peer-review and consensus building significantly affected overall risk 269 scores with a paired two-sample t-test, comparing scores from the first assessments to those 270 of the final list. We also evaluated how variation among experts and characteristics of the 271 taxa affected the overall risk scores. We fit a generalized linear regression with a negative 272 binomial error structure to the overall risk scores with the expert who completed the 273 assessment (N = 9), expert certainty about the overall score (very low, low, medium, or high), 274 whether the typical habitat of the plant taxon is terrestrial or aquatic, the number of records of 275 the taxon in the U.S., and the year of the earliest record of the taxon in the U.S. as 276 independent variables. To determine the number of records and the earliest record of each 277 taxon in the U.S., we used the package 'rgbif' (Chamberlain et al. 2021) to extract all GBIF 278 records in the U.S. for each taxon, selecting records that had coordinates and no geospatial 279 issues (GBIF.org 2021b). One taxon had no records in the U.S., so we used the current year 280 for its earliest record value. Number of records and earliest record were centered and scaled 281 and were not significantly correlated with each other (r = 0.04, P = 0.68). We fit the model 282 using the 'MASS' package (Venables and Ripley 2002), evaluated the fit using the 283 'DHARMa' package (Hartig and Lohse 2020), tested the significance of each independent 284 variable using likelihood ratio tests, and compared factor levels using the 'emmeans' package 285 (Lenth et al. 2021). All analyses were conducted in R version 4.0.2 (R Core Team 2020). 286 287 Taxa characteristics 288 We evaluated whether plant taxonomic families were under-or overrepresented in the CABI 289 plant list and in the final list using a resampling procedure (Daehler 1998 High risk taxa 313 Six plant taxa received risk scores greater than or equal to 64 (Fig. 2) without naturalization within the last 20 years (fewer than three records in wild areas); and 321 records of seeds sold within the U.S. at the time of the assessment (Suppl. material 3: Table  322 S2). All six taxa were considered very likely to establish in Florida (establishment score = 5 323 out of 5) because they occur in other regions of the world with climates similar to Florida and 324 in some cases, the taxon is known to have high reproductive capacity (Suppl. material 3: 325 Table S2). While none of the taxa were considered very likely to cause economic or 326 environmental impacts in Florida, four taxa received the next highest impact score (impact 327 score = 4 out of 5; L. vulgare, C. scoparius, P. arundinacea, and A. capillaris). These four 328 taxa have impacted native vegetation through competition, produce pollen that can be a 329 human allergen, and some are agricultural weeds (Suppl. material 3: Table S2). Information 330 about the six species from a handful of sources can help inform potential future policy actions 331 ( including two for S. officinale that suggested escape (it is planted as an ornamental, Table  357 S2). Symphytum officinale was considered likely to establish in Florida (establishment score 358 = 4), but unlikely to have impacts (impact score = 2). Four taxa were considered very likely 359 to establish in Florida (establishment score = 5)-Hypericum perforatum, Malva sylvestris, 360 Matricaria chamomilla, and Mentha aquatica-because they occur in areas with climate 361 similar to Florida and M. chamomilla readily self-seeds (Table S2). Hypericum perforatum, 362 M. sylvestris, and M. aquatica were considered likely to arrive in Florida (arrival score = 4) 363 and potentially likely to have negative impacts (impact score = 3), but M. chamomilla was 364 considered likely to have only minimal negative impacts (impact score = 2). These four taxa 365 are sold as ornamental plants within the U.S., have been reported in the southeastern U.S. in 366 the past 20 years, and can naturally disperse (Table S2). None of the plant taxa in the medium 367 risk group were considered very likely to have negative impacts. We had high certainty about 368 the scores of two taxa, medium certainty about the scores for 18 taxa, and low certainty about 369 the scores for three taxa. The three taxa for which we had low certainty about their scores 370 received relatively low risk scores: Symphytum officinale (overall risk score = 40), Jacobaea 371 vulgaris (overall risk score = 27), and Calystegia sepium spp. sepium (overall risk score = 372 27). 373 374 Figure 3. Horizon scan scores. A The overall risk scores for 99 taxa, divided into groups of 376 high risk (score ≥ 64), medium risk (27 ≤ score < 64), and low risk (score < 27) and shaded 377 by overall certainty score. B The number of taxa associated with each of the pathways of 378 arrival. Multiple pathways could be assigned to a single taxon. C The relationship between 379 certainty and the overall risk score, averaged across all taxa. Letters above bars indicate 380 significant differences in overall risk score among certainty scores with P < 0.05. 381 382 Low risk taxa 383 Seventy taxa received low risk scores (less than 27; Fig. 3). One taxon, Poa trivalis (overall 384 risk score = 20, overall certainty = high), was considered very likely to arrive in Florida 385 (arrival score = 5) because it is in the southeastern U.S., has been used in at least one research 386 experiment in Florida, and it is planted in golf courses in the southeast both intentionally and 387 unintentionally (seed contaminant). Poa trivalis, however, is unlikely to establish in Florida 388 (establishment score = 2) and have impacts (impact score = 2). Sambucus nigra ssp. nigra 389 (overall risk score = 10, overall certainty = very low), was considered very likely to establish 390 in Florida (establishment score = 5) because the species Sambucus nigra occurs in multiple 391 locations with climate similar to Florida (Table S2). However, the subspecies has few 392 recorded occurrences globally, which led to very low certainty about the establishment score. 393 In addition, Sambucus nigra ssp. nigra is very unlikely to arrive in Florida (arrival score = 1) 394 and unlikely to have impacts (impact score = 2). None of the plant taxa in the low risk group 395 were considered very likely to have negative impacts. We had high certainty about the scores 396 of eight taxa, medium certainty about the scores of 43 taxa, low certainty about the scores of 397 16 taxa, and very low certainty about the scores of three taxa. The three taxa for which we 398 had very low certainty about their scores were Filipendula vulgaris (overall risk score = 12), 399 S. nigra ssp. nigra (overall risk score = 10), and Gnaphalium uliginosum (overall risk score = 400 4). 401 402 Pathways of arrival 403 The most likely pathway of arrival for the taxa on the final list was escape from confinement 404 (Fig. 3B). Taxa are also likely to arrive in Florida as transport contaminants, transport 405 stowaways, or with unaided dispersal. It is less likely that plants will arrive through 406 intentional release into nature or through a constructed corridor. 407 408 Analysis of risk scores 409 There was no significant difference in the overall risk scores before and after peer-review and 410 consensus building (t = -1.41, 95% CI = -4.43-1.61, df = 97, P = 0.357) with an average 411 score (± SE) of 21.3 ± 2.1 before and 22.7 ± 2.1 after. The assessor ( 2 = 27.0, df = 8, P < 412 0.001), certainty level ( 2 = 21.4, df = 3, P < 0.001), and earliest record in the U.S. ( 2 = 3.9, 413 df = 1, P = 0.050) significantly affected the overall risk score, while the habitat (terrestrial vs. 414 aquatic; 2 = 0.07, df = 1, P = 0.787) and number of records in the U.S. ( 2 = 1.7, df = 1, P = 415 0.196) did not. Four out of 36 pairwise comparisons of assessors were significantly different 416 with P < 0.05. Taxa with higher overall certainty scores also had higher overall risk scores 417 (Fig. 3C). Taxa with earlier first records in the U.S. received higher overall risk scores than 418 taxa with later first records (Fig. 4A), but taxa with more records in the U.S. did not receive 419 significantly higher overall risk scores than taxa with fewer records, although there was a 420 positive trend (Fig. 4B). Our analysis of the 99 plant taxa most likely to be introduced to Florida identified six that 460 have a high risk of becoming invasive in the state in the next ten years (2020-2030). The 461 horizon scanning process helped us identify taxa that should undergo more thorough risk 462 assessments and potentially receive policy restrictions or research priority. The process we 463 used is a reproducible methodology that can be applied to future horizon scans. 464 465 Four taxa (Ligustrum vulgare, Cytisus scoparius, Phalaris arundinacea, and Avena fatua) 466 had high risk scores with high certainty. We recommend that these taxa receive more 467 thorough risk assessments followed by consideration for monitoring or regulation by the 468 relevant state agencies. which interferes with crops and grazing in other regions (Suppl. material 3: Table S2). If the 477 high level of risk assigned to the top six taxa is supported following additional research and 478 more thorough risk assessments, we recommend regulators consider policy actions to limit 479 the introduction of these taxa to Florida. Specifically, the industries that use or 480 unintentionally disperse these taxa ( steps in the process of importing, processing, and storing goods that can be modified to 506 reduce plant survival (Hulme 2009). 507 508 The identity of the assessor, the assessor's certainty level, and the invasion history of the taxa 509 in the U.S. significantly affected the overall risk scores of the assessed taxa. Two experts, 510 who had extensive experience completing plant risk assessments, scored taxa consistently 511 higher than two other experts, who had less experience completing plant risk assessments. To 512 address this issue, future horizon scans could calibrate scores among experts with a set of test 513 taxa prior to beginning the rapid risk assessments. We hypothesize that overall risk scores are 514 correlated with overall certainty scores because more available data on a taxon contributes to 515 higher certainty and can provide more pieces of evidence that a taxon may arrive, establish, 516 or have impacts. Similarly, we hypothesize that taxa with earlier and more records of 517 occurrence in the U.S. are likely to be better represented in English-language texts than taxa 518 that are less common or more recently detected, leading to more evidence for arrival, 519 establishment, and impacts, which could explain their generally higher risk scores. Efforts to 520 synthesize and standardize information about invasive species ( concern. Future horizon scans could focus on taxa from geographic regions with strong trade 558 and tourism ties to the focal area, allowing for more targeted assessments. 559 560 GBIF is a powerful tool, connecting organizations and institutions that collect and store 561 biodiversity data and making that data publicly available (GBIF.org 2020). We used GBIF 562 data multiple times during our horizon scan: to select the top 100 taxa based on global 563 occurrences, to evaluate arrival risk based on how close occurrences were to Florida, and to 564 evaluate establishment risk based on whether taxa were found in areas with the same 565 Köppen-Geiger zones as Florida. Two potential sources of bias introduced by the GBIF 566 dataset are amateur identification of plants and records collected non-systematically across 567 geographies. Some plant species and subspecies are difficult to differentiate from one 568 another, leading to inaccurate records by amateur botanists (Scott and Hallam 2003 list of 2128 potential invasive taxa that was too large to perform rapid risk assessments in a 583 reasonable timeline. We therefore developed data-based criteria to filter the list to 100 taxa. 584 These methods could be applied to other horizon scans with similar resource constraints. 585 Second, the rapid risk assessment tool and associated rubrics led to enough consensus among 586 experts that our final rankings relied much more on scores than on discussion and consensus 587 building (e.g., in contrast to Roy et al. 2014, Lucy et al. 2020. A drawback of this approach 588 is the loss of nuanced expert opinion that falls outside of the rubrics, which is an important 589 component of horizon scans when information on a potential invasive species is limited in 590 peer-reviewed literature (Verbrugge et al. 2019). A major advantage, however, is that this 591 approach can be used with non-experts, which is relevant for efforts limited by available 592 expertise (Meyers et al. 2020). 593 594 Conclusion 595 Here we presented a horizon scan of 2128 plant taxa, identifying six with a high invasion risk 596 for Florida over the next ten years and 93 with medium or low invasion risk. The horizon 597 scan process therefore can reduce the potential number of taxa requiring thorough risk 598 assessments by three orders of magnitude. Although the process has room for improvement, 599 the results provide researchers, regulators, and private and public land managers with a clear 600 list of high risk species to focus on. Given the substantial impacts and costs of invaders in 601 Florida, the ability to differentiate and focus efforts on high probability bad actors is critical. 602 603 Funding 604 We would like to thank the Florida Fish and Wildlife Conservation Commission and 605 UF/IFAS Dean for Research for funding this project. 606 607 Competing interests 608 The authors have declared that no competing interests exist. 609 610