Using structured eradication feasibility assessment to prioritize the management of new and emerging invasive alien species in Europe

Prioritizing the management of invasive alien species (IAS) is of global importance and within Europe integral to the EU IAS regulation. To prioritize management effectively, the risks posed by IAS need to be assessed, but so too does the feasibility of their management. While the risk of IAS to the EU has been assessed, the feasibility of management has not. We assessed the feasibility of eradicating 60 new (not yet established) and 35 emerging (established with limited distribution) species that pose a threat to the EU, as identified by horizon scanning. The assessment was carried out by 34 experts in invasion management from across Europe, applying the Non‐Native Risk Management scheme to defined invasion scenarios and eradication strategies for each species, assessing the feasibility of eradication using seven key risk management criteria. Management priorities were identified by combining scores for risk (derived from horizon scanning) and feasibility of eradication. The results show eradication feasibility score and risk score were not correlated, indicating that risk management criteria evaluate different information than risk assessment. In all, 17 new species were identified as particularly high priorities for eradication should they establish in the future, whereas 14 emerging species were identified as priorities for eradication now. A number of species considered highest priority for eradication were terrestrial vertebrates, a group that has been the focus of a number of eradication attempts in Europe. However, eradication priorities also included a diverse range of other taxa (plants, invertebrates and fish) suggesting there is scope to broaden the taxonomic range of attempted eradication in Europe. We demonstrate that broad scale structured assessments of management feasibility can help prioritize IAS for management. Such frameworks are needed to support evidence‐based decision‐making.


| INTRODUC TI ON
Managing the increasing risks and impacts of invasive alien species (IAS, cf. invasive non-native, invasive non-indigenous species) is one of the great societal challenges of the 21 st century (Seebens et al., 2018;Simberloff et al., 2013;Vilà et al., 2011).
Ambitious international goals aim to reduce or halt these rising impacts, including Aichi Target 9 of the Convention on Biological Diversity (CBD, 2014), which commits signatories to control or eradicate priority species. This commitment is reflected in European Union (EU) regulation 1143/2014 on IAS (EU, 2014). However, the control or eradication of IAS can be expensive.
With numerous species and limited resources, decision-makers must carefully prioritize which species to manage and how (McGeoch et al., 2016).
Risk assessment, the process by which the likelihood and magnitude of impact is assessed, is commonly used to support the prioritization of IAS and has been well used in Europe and elsewhere (Roy et al., 2018). However, simply assessing the risks and impacts of IAS is of limited use for prioritizing their management, as it fails to take into account the feasibility of delivering an effective response (Booy et al., 2017). Failure to account for management feasibility can result in species being prioritized that may be unmanageable or for which management is unlikely to be economically viable (Branquart et al., 2016;Cassey, García-Díaz, Lockwood, & Blackburn, 2018;Courtois, Figuieres, Mulier, & Weill, 2018). As a result, resources could be wasted or used inefficiently and confidence in decision-making could be reduced.
A number of approaches are available to support the assessment of IAS management feasibility, its costs and benefits.
Economic cost-benefit analysis (CBA) and cost-effectiveness analysis (CEA) have been used to assess aspects of management for particular species and in some cases to approve management plans prior to implementation (Blackwood, Hastings, & Costello, 2010;Born, Rauschmayer, & Bräuer, 2005;Courtois et al., 2018). However, purely economic CBA and CEA approaches generally require large quantities of empirical information, are costly and time-consuming to produce (Reyns et al., 2018). There are also complexities in how to effectively monetize the full range of social, environmental, animal welfare and biodiversity consequences of IAS management (Hoagland & Jin, 2006). As a result, CBA and CEA are generally applied to individual IAS and particular situations (Panzacchi, Cocchi, Genovesi, & Bertolino, 2007;Rajmis, Thiele, & Marggraf, 2016), but are difficult to apply across large numbers of different species to identify broad management priorities.
Multi-criteria approaches (Born et al., 2005), including Multi Criteria Decision Analysis (MCDA), provide a means of assessing and comparing between larger numbers of species using available data against a wide range of different criteria, without the need for monetization. As such, they are commonly used to support risk assessment, as well as risk management evaluations in some cases (EPPO, 2011;Mehta, Haight, & Homans, 2010;OiE, 2017).
One such approach is the Non-Native Risk Management (NNRM) scheme (Booy et al., 2017), which uses multiple criteria relevant to decision-makers (beyond solely monetary considerations) to score different aspects of IAS management, based on predefined invasion scenarios and strategies. Within this scheme, species are assessed using expert judgement and elicitation methods, incorporating empirical information where available and including a framework for assessing confidence (Roy, Peyton, & Booy, 2020). This approach is similar to methods used for IAS risk assessment (Baker et al., 2008;Brunel et al., 2010;Copp et al., 2016;Essl et al., 2011;Mumford et al., 2010;Vanderhoeven et al., 2017) and increasingly throughout the field of ecological conservation (Adem & Geneletti, 2018;Burgman et al., 2011).
To date, the NNRM has been applied at regional (Osunkoya, Froese, & Nicol, 2019) and national scales (Adriaens, Branquart, Gosse, Reniers, & Vanderhoeven, 2019;Booy et al., 2017); however, there are advantages of applying it at larger scales. IAS pose threats to multiple countries and do not respect national boundaries, meaning that management responses will often require cooperation and resource sharing between states to be effective (Robertson et al., 2015). Large-scale prioritization is currently of particular relevance in the EU to support the implementation of the Regulation 1143/2014 on the prevention and management of the introduction and spread of IAS.
Here we apply the NNRM at a large scale to evaluate an existing multi-taxa list of new and emerging IAS that threaten the EU as identified by horizon scanning (Roy et al., 2015. We use this evaluation of species along with existing risk assessment scores (derived from horizon scanning) to consider potential priorities for management within Europe. In particular, we consider priorities for (a) early detection and rapid eradication of new species should they start to establish in Europe; and (b) eradication of species that are currently established in Europe, but with limited distributions. In addition, we provide an insight into potential priorities for (c) prevention and (d) long-term management. We explore the suitability of using this approach for large-scale prioritization and consider patterns in the feasibility of eradication in different environments and at different scales.

| MATERIAL S AND ME THODS
A list of 95 species were identified as high or very high risk through the horizon scanning of Roy et al. (2015). This comprised terrestrial, freshwater and marine taxa that were categorized as either new to the EU (i.e. not yet established) or emerging (i.e. established with limited distributions; Table 1). For each species, a risk management assessment was completed using a modified version of the NNRM scheme (Booy et al., 2017). A key modification was to standardize invasion scenarios using pre-defined categories for the number of discrete populations (1-3, 4-10, 10-50, +50) and total combined area of all populations (<1ha, 1-10 ha, 10 ha-1 km 2 , 1-10 km 2 , 10-100 km 2 , >100 km 2 ; for more guidance refer to Methods S1). This helped take into account the greater complexity of assessment at the European scale and also allowed for patterns in feasibility of eradication at increasing area and number of populations to be analysed. Species were included that had a range of areas and populations ( To aid in this, each group leader presented the initial scores of their group, discussed any areas of potential ambiguity and agreed on clarifications. This was then repeated in plenary so that participants could go through the scoring guidance with the organizers and ensure consistency in application. The main workshop proceeded with a simplified, facilitated Delphi approach (Mukherjee et al., 2015) including two rounds of consensus within and across expert groups: 1. Group leaders presented an overview of the initial scores from their groups to all participants, who were encouraged to discuss and challenge the scores.
2. Expert groups reviewed and refined the scores of their group, taking into account the discussions from session 1. Each group was provided with the median response and confidence scores for each of their species and asked to discuss disagreement on scores and refine them where necessary.

| Analysis
All analyses were undertaken in R (R Core Team, 2020).

| Risk management scores
We assessed the interrelation between the seven risk management components scores and the overall feasibility of eradication score in ordinal space using a factor plot and non-metric multi-dimensional scaling. A distance matrix of species by component was analysed using the isoMDS function in the MASS (Venables & Ripley, 2002) package and then visualized using FactoMineR package (Le, Josse, & Husson, 2008), colouring each species by the independent overall score. Underlying patterns of correlation between components (variables) were visualized in a factor plot.
Polychoric correlations (R package 'Polychor'; Fox, 2019) were used to compare the ordinal scores for overall risk (derived from horizon scanning) and the overall feasibility of eradication scores (derived from this exercise). Correlation between the two assessments implies they measure similar underlying information; we did not expect to find strong correlation.
TA B L E 2 Count of species by scenario code for extent. Letters A-D represent the number of discrete populations (respectively 1-3, 4-10, 10-50, +50) and numbers 1-6 represent total combined area (respectively <1 ha, 1-10 ha, 10 ha-1 km 2 , 1-10 km 2 , 10-100 km 2 , >100 km 2 ). For example, the code B2 indicate a species with 4-10 populations covering a total area 1-10 ha Note that these analyses were used to investigate the relationship between the assessed variables, but are not a requirement for those applying the risk management scheme in the future.

| Effect of extent and environment on overall feasibility
To assess the relationship between the score for overall feasibility of eradication (ordinal response) and environment (terrestrial, freshwater, marine), total area and number of populations, a cumulative link model (CLM) was fitted using the R package 'Ordinal' (Christensen, 2018). It was hypothesized that the overall feasibility of eradication score for each species would decline with increasing spatial extent (total area and number of populations) and be dependent on the environment in which the species occurred. Population categories 'C' and 'D' were pooled into one category (10+ populations) as were areas >10 ha (greater than category 3) owing to sparse data at these ranges. Ordinal regression assumes proportional odds (i.e. the relationship between each pair of outcome groups is the same). Statistical tests for proportional odds have been criticized as they tend to falsely reject the null hypothesis, so proportionality was assessed using a graphical method following Bender and Grouven (1997) and Gould (2000). This method uses plots of predicted values derived from a series of binary logistic regressions to check the assumption that coefficients are equally separated across cut-points.
The final model was used to predict the feasibility of eradication for every combination of environment, total area and number of populations. Model predictions were expressed as the probability of the overall feasibility of eradication score being each of the five response levels (very high to very low) and visualized using the R package 'Ggplot2' (Wickham, 2009).

| Prioritization
To indicate priorities for eradication, we combined the overall risk assessment scores (derived from horizon scanning) with the overall feasibility of eradication scores (from this risk management exercise) in a prioritization matrix (following Booy et al., 2017). As both the overall risk and overall feasibility of eradication scores used a five-point scale (very low to very high), the result was a 5 × 5 prioritization matrix, with priorities ranging from lowest (1:1) to highest (5:5; Table 3). However, as only species with risk assessment scores of high and very high were included in this exercise, only positions in the top two rows of the matrix could be achieved, resulting in priorities ranging from medium-low (4:1) to highest (5:5).
The matrix was also used to investigate other priorities, including prevention and long-term management. For new species, prevention was likely to be a particular priority if the species posed a high risk and the feasibility of eradication after arrival was low. For emerging species, long-term management (e.g. containment, slowing spread, control) was likely to be a particular priority if the species posed a high risk and the feasibility of eradication was low. These priorities corresponded to the top left corner of the matrix and are marked: ++ highest, and +high priority for prevention/long-term management (Table 3).

| Data
The data underpinning the analysis reported in this paper are deposited in the Dryad Data Repository .

| Risk management scores
The workshop resulted in consensus risk management scores for all species.
Scores for overall risk (derived from horizon scanning) and overall feasibility of eradication (derived from this exercise) were not cor- The scores for overall feasibility of eradication aligned in sequence with the individual component scores (i.e. effectiveness, practicality, cost, impact, acceptability, window of opportunity and likelihood of reinvasion) with some overlap ( Figure S1). This suggests that while component scores were in general agreement with the overall score TA B L E 3 Priority matrix based on risk assessment scores (derived from horizon scanning) and scores for overall feasibility of eradication (derived from this risk management exercise). Only high and very high-risk species were included in this study (hence, it was not possible for species to be placed in greyed out parts of the matrix). The matrix indicates priorities for eradication (background colour and cell text). Potential priorities for prevention and long-term management are marked + (high) and ++ (highest priority) Overall risk assessment score (derived from horizon scanning)

Very low (1) Low (2) Medium (3) High (4) Very high (5)
Very high (5) Medium ++ Medium-high + High Very high Highest it was not possible to consistently determine the overall score based on individual components. Five of the risk management components (effectiveness, practicality, cost, impact and acceptability) were correlated with overall feasibility of eradication, while window of opportunity and likelihood of reinvasion were not ( Figure S2).

| Effect of extent and environment on the overall feasibility of eradication
The assumptions of proportionality were met for the CLM as the thresholds (intercepts) for each covariate were broadly similar distances apart ( Figure S3). All variables (environment, total area and number of populations) were significant predictors of the scores for overall feasibility of eradication ( Figure S4).
In general, the scores for overall feasibility of eradication were lowest for marine species and highest for terrestrial species, with freshwater species in between. In each environment, overall feasibility of eradication decreased as total area occupied or number of populations of the IAS increased ( Figure S4).
Increasing total area and number of populations reduced the probability of very high and high scores for overall feasibility of eradication in all environments (Figure 1). For terrestrial species, high overall scores for feasibility of eradication were more probable F I G U R E 1 Cumulative link model predictions for the overall feasibility of eradication in different environments at different spatial scales. The probability of the overall feasibility of eradication being each of the five response levels very high (VH) to very low (VL) is given (on the y-axis) for each combination of variables, with 95% confidence intervals. Note that colours indicate feasibility of eradication (green = higher feasibility, red = lower feasibility), these are different to those used (e.g. in Table 3) to indicate priority (where red = higher priority and green = lower priority) Overall feasibilty of eradication Predicted probability Populations than low scores at every combination of total area and number of population. In the freshwater environment, high scores were probable when either the total area was small (<1 ha) or there were few populations (<1 to 3), but beyond this low scores were more probable. For marine species, low scores were more probable than high scores at all combinations.

| Prioritization
Combining scores for overall risk (derived from horizon scanning) and overall feasibility of eradication resulted in six levels of eradication priority: highest (1 species), very high (20), high (36), med-high (20), medium (14) and med-low (4) (Figure 2). These were further F I G U R E 2 Counts of species within the priority matrix for (a) new and (b) emerging species. The colour of the matrix reflects priority (derived from   divided into priorities for future rapid eradication of new species should they establish ( Figure 2a) and eradication priorities for emerging species that are already established (Figure 2b). In addition, new (i.e. not yet established) species for which overall feasibility of eradication on detection was low were considered priorities for prevention (Table S1); while emerging (i.e. already established) species with low feasibility of eradication were considered priorities for long-term management (e.g. control, slowing spread, containment) (Table S2). Detail on key eradication priorities is provided below and in Tables 4 and 5 (scores for all species are available in   Tables S1 and S2).

| Priorities for future rapid eradication of new species
Of the 60 new species, Faxonius rusticus (rusty crayfish) scored the highest priority for eradication, with both the overall risk and overall feasibility of eradication scoring very high (  Figure 2a). The invasion scenarios for these species suggested that the majority were likely to be in one to three populations covering <1 ha or 1-10 ha at the point of detection.

| Priorities for eradication of currently established emerging species
Of the 35 emerging species assessed, four were identified as very high priority for eradication and a further 10 were identified as high priority (Table 5; Figure 2b).
The top four priority species were terrestrial vertebrates with very high scores for overall risk and high scores for overall feasibility of eradication. The invasion scenario for these species The key eradication methods identified included netting, trapping, manual capture and shooting, which were not considered to cause significant adverse environmental, social or economic harm. Acceptability scores were high, except for N. nasua, which scored medium. The window of opportunity for all of these species was 1-3 years.
The 10 high priority established species comprised three terrestrial plants, one freshwater plant, two terrestrial vertebrates, two freshwater animals, one insect and one marine tunicate (

| Prevention and long-term management priorities
Where a species that has not yet established poses a high overall risk, but overall feasibility of eradication on detection is low, it is likely to be a priority for prevention. Three species were identified as particularly important for prevention based on very high overall risk and low or very low scores for overall feasibility of eradication: Plotosus lineatus (striped eel catfish), Homarus americanus (American lobster) and Codium parvulum (a green algae; Figure 2a; Table S1).
For already established species with low scores for overall feasibility of eradication, long-term management (e.g. containment, slowing spread, control) may be a high priority. In all, 11 species were identified as potentially high priorities for longterm management on this basis (Figure 2b; Table S2). Three

| D ISCUSS I ON
We identified priorities for the eradication of new and emerging IAS in Europe using a structured risk management tool combined with risk assessment scores derived from horizon scanning. This exercise not only indicated priorities for the eradication of emerging species and contingency planning for new species, but potential priorities for prevention and long-term management as well. While the NNRM has previously been applied at regional and national scales Booy et al., 2017;Osunkoya et al., 2019), this is the first application across multiple countries. Despite increased complexity at this scale and a lack of information on the status of some species in Europe, we found that the scheme could be applied successfully at a continental scale.
Although the species-specific eradication feasibility scores resulting from this exercise provide support for those taking decisions about how and which IAS to manage, they are not straightforward management recommendations. The feasibility scores are linked to specific invasion scenarios and eradication strategies, which are subject to knowledge gaps and change, for example as a result of changes in species distributions and new eradication methods becoming technically or legally available.
As with other screening methods (including horizon scanning, rapid risk assessment and hazard identification), the results should be considered preliminary and subject to further in-depth assessment.
For example, detailed management plans would need to be drafted to implement the management priorities identified here and these should include further assessment in the field to confirm population sizes and distribution as well as the applicability of management methods. These need to accommodate for alternative strategies if eradication actions do not obtain the expected result (Gregory et al., 2012;Richardson, Mill, Davis, Jam, & Ward, 2020). Careful planning is necessary to evaluate the effort needed for eradication, which can be supported by modelling (e.g. Tattoni et al., 2006).
Further tools for in-depth assessment of the initial priorities identified here could include the use of CBA, CEA and eradication probability modelling (Drolet, Locke, Lewis, & Davidson, 2015).
We assessed high and very high-risk IAS identified by horizon scanning as these are likely candidates for prevention, early detection and rapid eradication given their absence or limited status in the EU (Roy et al., 2015). They are also of particular concern currently in the EU which has recently adopted regulation 1143/2014 on IAS that emphasizes the importance of prevention and rapid eradication (EU, 2014). While horizon scanning provides a useful method for reducing long lists of potentially thousands of species to a shorter list of those most likely to be threats (Peyton et al., 2019;Roy et al., 2015), it is of limited use for prioritizing specific actions as it does not take into account the feasibility of management (Booy et al., 2017;Vanderhoeven et al., 2017). By applying risk management criteria, our study refined this list into specific management priorities, aligning with the guiding three step hierarchical approach of IAS management set out in the Convention on Biological Diversity (UNEP, 2011).
The results of this study demonstrate the value of incorporating both risk assessment (here derived from horizon scanning) and risk management criteria when prioritizing IAS. There was no correlation between eradicating feasibility and risk assessment scores, indicating that risk management criteria evaluate information that is different to risk assessment. This additional information is an essential part of risk analysis, and fundamental to decision-makers, who must take into account a wide range of criteria that go beyond risk (Dana, Jeschke, & García-De-Lomas, 2014;Kerr, Baxter, Salguero-Gomez, Wardle, & Buckley, 2016;Simberloff, 2003). While risk management is traditionally included along with risk assessment as part of an overall approach to risk analysis in other disciplines, such as plant health, animal health and food safety (Ahl et al., 1993;EFSA, 2010;FAO, 2013;OiE, 2017), it has rarely been applied so systematically to IAS. This is particularly true in Europe, where risk assessment alone has been the dominant method used to support prioritization (Essl et al., 2011;Heikkilä, 2011;Kerr et al., 2016;Roy et al., 2018;Turbé et al., 2017;Vanderhoeven et al., 2017). Our results highlight the importance of incorporating this step and, by doing so, identifying refined priorities more specifically linked to management outcomes.
Modifying the NNRM scheme by standardizing invasion scenarios, based on the number of discrete populations and total combined area of all populations, allowed us to explore the feasibility of eradication at different spatial scales. Across all environments, the overall feasibility of eradication decreased as extent increased, which reflects the fact that elements of feasibility, such as cost and resource effort, are known to scale with extent (Brockerhoff, Liebhold, Richardson, & Suckling, 2010;Howald et al., 2007;Rejmánek & Pitcairn, 2002;Robertson et al., 2017).
Terrestrial species received highest scores for overall feasibility of eradication, followed by freshwater species and then marine species, which reflects the different challenges of eradication in these different environments (Booy et al., 2017). While the feasibility of eradicating terrestrial species was highest at smaller scales, it remained high even at larger scales, albeit with reduced confidence. Indeed, successful eradications on large land masses have been reported in Europe of invasive mammals and birds (Robertson et al., 2015. In contrast, the feasibility of eradicating freshwater species was likely to be feasible at small scales (i.e. few populations <1-3, or small area <1 ha), but unlikely to be feasible at larger scales (i.e. >1-3 populations and >1 ha). In the marine environment, feasibility was likely to be low, even at small extents. These results indicate that extent alone is not a good predictor of feasibility when comparing species from different environments. They also suggest that early detection and rapid eradication is particularly important for freshwater species, for which action at an early stage of invasion considerably increases the likelihood that eradication will be feasible. This appears to be less important for terrestrial species, for which eradication remains feasible across considerably larger scales, and for marine species, for which eradication even at small scales is unlikely to be feasible in most circumstances. Of course, eradication is not the only rapid response measure that could be deployed, and these results do not preclude the possibility that early detection and rapid action to contain or slow the spread of a marine species may be useful. in comparison to EU funding for other IAS projects (Scalera, 2009).
However, although cost is a very important factor in the overall feasibility of eradication (Booy et al., 2017), costing eradications is complex and comprehensive data on the cost of invasive species eradications are generally scarce Donlan & Wilcox, 2007) which warrants interpreting these crude ordinal cost estimates with caution. Also, the cost is very dependent on the specific invasion scenarios and management strategies drafted for this exercise. As the invasion extent of several species appeared poorly documented (e.g. A. tristis) or surrounded by considerable uncertainty (e.g. B. mauritanicus), costs could have been underestimated. Lastly, the extent of a species invasion can rapidly change. On the other hand, the cost for eradication could also be reduced by managing several co-occurring species with similar management approaches at once . Such concrete cost estimates are beyond the broad scale feasibility assessment performed in our study.
Lower scores for some risk management components suggest potential barriers to eradication that would need to be overcome.
These include the medium acceptability scores for eradicating the N.
nasua (coati), A. axis (Indian spotted deer) and R. americana (greater rhea), which indicates a potential lack of public or stakeholder acceptance for this work on perceived animal welfare grounds. While acceptance of the use of herbicides could be a barrier to eradicating invasive non-native plants, this was not considered a significant problem for the plants included in the high priority lists. However, acceptability was a potential barrier for the eradication of M. anguillicaudatus (oriental weatherfish) because of potential public concern over the use of piscicides. Furthermore, the use of piscicides in public waters is prone to meet legal barriers in most European countries which is reflected in medium scores for practicality. Gaining access is a potential barrier to the eradication of some plant species, especially where they grow in difficult terrain. This was the case for Euonymus fortunei, which received a low practicality score because the most likely invasion scenario included the potential for its establishment on cliff edges. While these barriers are challenging and would have to be addressed as part of an eradication strategy, they were not considered insurmountable by the assessors. While the main role of the NNRM is to identify priorities for eradication and contingency planning, it also identifies potential priorities for long-term management and prevention. Long-term management is likely to be a priority for established species where the overall feasibility of eradication is low and the overall risk is high.
For example, the feasibility of eradicating Arthurdendyus triangulatus (New Zealand flatworm) was considered very low, but it may be feasible to slow the spread of this species using phytosanitary measures (Boag & Yeates, 2001). Similarly, the NNRM can identify potential prevention priorities for species that are not yet established where the feasibility of eradication is low and the risk high. For example, should Homarus americanus (American lobster) establish in European waters it is unlikely that eradication would be feasible and so prevention, perhaps by tightening control of its release and escape pathways (Jørstad, Agnalt, & Farestveit, 2011;van der Meeren et al., 2016), should be considered a particularly high priority.
A limitation of the NNRM is that it does not currently evaluate the effectiveness of long-term management (e.g. containment, slowing spread, control) or prevention measures. This is important because long-term management may not always be feasible for species that cannot be eradicated. For example, long-term management may not have a lasting impact on the spreading population of Pterois miles (lion fish) in Europe, despite calls for its consideration (Kletou, Hall-Spencer, & Kleitou, 2016). Similarly, prevention may not always be feasible, as is likely to be the case for Plotosus lineatus (striped eel catfish) which seems set to establish in EU waters following its arrival through the Suez Canal (Edelist, Golani, Rilov, & Spanier, 2012).
Where considering future prevention and long-term management priorities, these factors need to be taken into account and this is a priority for further development of the NNRM.
The approach to prioritization presented here has application for IAS policy and management. Our results help focus more attention on the eradication of species with limited distributions and contingency planning for new arrivals where this is feasible. The availability of management methods, expected environmental non-target effects and the proportionality of the benefits and costs of eradication are important elements in the current decision-making on IAS management in Europe (EU, 2014). These elements of risk management are considered in our assessment and cannot be provided by risk assessment alone. Our approach thus helps to address these, including providing a method to assess the feasibility of eradication, supporting the development of management plans and evaluating the potential benefits of listing under the EU IAS regulation.
To date, there is no agreed method for determining whether eradication is feasible and so application is likely to be subjective and potentially inconsistent across Europe. Listing alone may not be sufficient to drive EU wide eradication and contingency planning for species identified as priorities. Other mechanisms may be needed to do this, for example specific eradication and contingency planning programmes under the EU LIFE funding stream. Such programmes would need to be coordinated across Europe and would benefit from sharing of expertise. While our results are focused on the European situation, the procedure here developed could be used in other part of the world to implement or improve strategies to limit the impact of IAS.
As numbers of IAS are predicted to increase and global management targets become more ambitious, transparent methods for prioritizing action are essential. We recommend that the structured assessment of risk management criteria, such as those included within the NNRM scheme, be applied routinely to IAS, as is commonplace in other biosecurity areas. While there are increasing calls for the application of risk assessment to more species (Carboneras et al., 2018), we suggest that there should be at least as great a focus on evaluating the feasibility of management in a future with increasingly limited resources for nature conservation.

ACK N OWLED G EM ENTS
This work was funded in part by Newcastle University (

DATA AVA I L A B I L I T Y S TAT E M E N T
The data that support the findings of this study are openly available in Dryad at https://doi.org/10.5061/dryad.8pk0p 2nk1 .