A practical tool for assessing ecosystem services enhancement and degradation associated with invasive alien species

Abstract Current approaches for assessing the effects of invasive alien species (IAS) are biased toward the negative effects of these species, resulting in an incomplete picture of their real effects. This can result in an inefficient IAS management. We address this issue by describing the INvasive Species Effects Assessment Tool (INSEAT) that enables expert elicitation for rapidly assessing the ecological consequences of IAS using the ecosystem services (ES) framework. INSEAT scores the ecosystem service “gains and losses” using a scale that accounted for the magnitude and the reversibility of its effects. We tested INSEAT on 18 IAS in Great Britain. Here, we highlighted four case studies: Harmonia axyridis (Harlequin ladybird), Astacus leptodactylus (Turkish crayfish), Pacifastacus leniusculus (Signal crayfish) and Impatiens glandulifera (Himalayan balsam). The results demonstrated that a collation of different experts’ opinions using INSEAT could yield valuable information on the invasive aliens’ ecological and social effects. The users can identify certain IAS as ES providers and the trade‐offs between the ES provision and loss associated with them. This practical tool can be useful for evidence‐based policy and management decisions that consider the potential role of invasive species in delivering human well‐being.


| 3919
MARTINEZ-CILLERO ET AL. Olden, 2011;Schlaepfer et al., 2012). In fact, there are relatively few empirical studies that present information about the benefits provided by IAS, although the focus on this literature has been increasing in the last years (Kull et al., 2011;Shackleton et al., 2007;Tassin & Kull, 2015). The so-called "conflict species" can be highly regarded for the benefits they provide. But they can also be considered as a serious environmental threat from a management perspective.
Many risk and impact assessments have been developed to prioritize IAS control and management, with a focus on the negative environmental impacts and economic damages (Roy et al., 2014).
Prevention has been increasingly recognized as the most cost-effective strategy to ensure pristine ecosystems remaining free of IAS (Genovesi & Monaco, 2013;Meyerson & Mooney, 2007), even though it is not foolproof (Chornesky et al., 2005). IAS control and eradication are often advocated as consequent management operations and require huge financial resources (Boonman-Berson, Turnhout, & van Tatenhove, 2014;Ewel & Putz, 2004). Yet, high rates of species invasions are projected to increase in the future. Suggestions have been proposed toward building or maintaining ecosystem resilience and services, rather than restoring IAS-free ecosystems that may be futile (Lin & Petersen, 2013;Pyšek & Richardson, 2010). Although this approach is controversial due to the importance of the evolutionary context in species interactions (Richardson & Ricciardi, 2003) and the unpredictability of some negative consequences of invasions, there is nevertheless a pragmatic need for management alternatives to IAS removal (Hulme, Pyšek, Nentwig, & Vilà, 2009;McMahon et al., 2006).
Ecosystem services (ES) are the processes, functions or ecological characteristics through which ecosystems sustain and fulfill human life, either directly (e.g., provision of food) or indirectly (e.g., pollination) (Costanza et al., 2017;Daily, 1997). IAS may cause changes in these services by altering the ecosystems (Peh et al., 2015;Vilà et al., 2010;Vilà & Hulme, 2017). Therefore tools, such as risk-assessment schemes, that help to evaluate such impacts and aid for the prioritization and management of IAS are essential. Roy et al., (2018) identified 14 minimum attributes a risk-assessment scheme should include, of which two are related to human well-being: "Assessment of impact on ecosystem services" and "Assessment of socio-economic impacts." These attributes were also two of the most notable gaps in our knowledge required for completing risk assessments.
However, IAS ES impact assessments are always challenging and require substantial resources for three reasons: first, ES are governed by complex interactions that make them difficult to measure over space and time; second, long-term, large-scale data often do not exist (Eviner, Garbach, Baty, & Hoskinson, 2012;Kremen, 2005); and last, current measures of many ES are still crude (Bennett, Peterson, & Gordon, 2009;Naidoo et al., 2008).
Yet, new standards to evaluate IAS effects on human well-being have been developed (Çinar, Arianoutsou, Zenetos, & Golani, 2014;Dickie et al., 2014;McLaughlan, Gallardo, & Aldridge, 2014;Pejchar & Mooney, 2009). An important example is the Socio-Economic Impact Classification of Alien Taxa (SEICAT; Bacher et al., 2018) that evaluates the impacts on human welfares using changes in human activities as metric; a sister-scheme of the Environmental Impact Classification of Alien Taxa (EICAT) which is officially ad- This practical tool, however, would not yet address complexities such as discerning effects that are temporally or spatially scale-dependent, or accounting for biological factors such as lag-times, dispersal, interactive effects, and environmental context. Nevertheless, INSEAT can yield valuable information for IAS managers by enabling them to (a) evaluate rapidly experts' opinions on how IAS affects ES delivery, including positive IAS effects; (b) gather knowledge and information to enable exploration of alternative management options; (c) produce simple, graphical representation of synergies and trade-offs among the effects of IAS; and (d) assess the management effort required to eradicate an alien species. This would make IAS management more efficient and diverse, in terms of exploring management potential that is overlooked under current methodologies.
Information obtained by using INSEAT can then be fed into an integrated approach which, among other activities, involves seeking stakeholder opinions on the way forward (Cook & Proctor, 2007;Liu, Proctor, & Cook, 2010).
In this study, we piloted INSEAT to assess the effects of 18 wellknown IAS in Great Britain (GB) on ecosystem service provision.
However, due to space constraint, we described only four case studies here: Harmonia axyridis (Harlequin ladybird), Astacus leptodactylus (Turkish crayfish), Pacifastacus leniusculus (Signal crayfish) and Impatiens glandulifera (Himalayan balsam). The feedback from the experts then led to a further refinement of the tool which includes an improved impact scale definition; an assessment of uncertainty on the experts' responses; and a request of supporting information from the experts.

| ME THODS
A concise, yet informative, ES classification scheme is essential for IAS managers to understand the different types of ES. We built an integrated ES classification scheme (Appendix 1) based on three widely accepted ES classifications from the Millennium Ecosystem Assessment (Millennium Ecosystem Assessment, 2005), the UK National Ecosystem Assessment (UK NEA; Mace et al., 2011) and The Economics of Ecosystem and Biodiversity (TEEB, 2016). We excluded supporting services in our ES classification scheme to avoid double-counting since all the other services are underpinned by them (Haines-Young & Potschin, 2012).
Assessing IAS effects on ES requires a qualitative and broad evaluation (Roy et al., 2014). INSEAT is designed to be completed by experts on a particular IAS by scoring its effect on a range of ES from our ES classification scheme (although other ES classifications could also be used). We created an integrated assessment proforma (Figure 1) that included questions designed to assess (a) the strength and direction of IAS effects on ES provision; (b) IAS potential to provide ES; and (c) the management effort required to eradicate the alien species.

| Using experts' opinions
The INSEAT protocol relies on expert judgment, which is often sought when there is scientific uncertainty or when data are absent

Level of confidence Description
High Score is based on (a) scientific evidence/existing data derived within your country; and/or (b) personal observation within your country; and/or (c) professional opinion/view/judgement based on knowledge, fact or work/research experience that is specific to the country.

Medium
Score is based on (a) scientific evidence/existing data derived from a region that may be a good surrogate for your country; and/or (b) personal observation in a region which is similar to your country; and/or (c) professional opinion/view/judgement based on knowledge, fact or work/research experience that is specific to a region similar to your country.

Low
Score is based on (a) scientific evidence/existing data derived from an unknown locality or area that may not be a good surrogate for your country; and/or (b) personal observation that is specific to a region not similar to your country; and/or (c) anecdotal evidence. Any suppositions, assumptions or hypotheses shall also fall into this category.

Indicate your level of confidence: High/Medium/Low (use the guidance table in Question 1)
Score Please describe the spreading capacity of this species 1 Low potential -the species spreads slowly 2 Medium/moderate potential -the species spreads rapidly but does not double its range in less than 10 years 3 High potential -the species spreads rapidly, doubling its range in less than 10 years -I do not know

Indicate your level of confidence: High/Medium/Low (use the guidance table in Question 1)
Score Please select the management effort necessary to eradicate this species 1 Low efforts -the species can be eradicated locally with low resources input (easy access, no need for machinery, skilled staff or materials such as pesticides) 2 Medium efforts -the species can be eradicated with medium resources input (access requires certain effort, might need machinery, skilled staff or materials such as pesticides) 3 High efforts-can be eradicated but is resource intensive (due to complicated access, need of machinery, skilled staff and materials such as pesticides).

| Assessing strength and direction of IAS effects
Semiquantitative Likert scales are used to rank environmental and socio-economic impacts, following other assessments such as the Generic Impact Scoring System (Nentwig, Bacher, Pyšek, Vilà, & Kumschick, 2016). Each scale level is well-defined to avoid ambiguities and also to make categories and taxa comparable. The scale ranges from −4 to 4, each level combining the strength ("no effect," "too small," "noticeable," "substantial," and "intense") and the reversibility of the impact if the species is removed ("reversible" or "irreversible"). We consider that only "intense" effects can be irreversible, as for less extreme impacts the ecosystems would naturally recover to their original state.
We used the variability of agreement among the respondents as a measure of robustness in the knowledge of a species in terms of its impact on a particular ES. Low agreement, inferred by a high variability in the scoring, helps to identify knowledge gaps about the effect of that species.
We assumed that the effect of a widely distributed species to be greater than if it were more narrowly distributed. Therefore, the "Impact Index," was determined by weighing the species impact (from −4 to 4) score with its spatial occupation score (from 1 to 3) (i.e., Impact index = impact*occupation). The spatial occupation score of the invasive species in their non-native range-ranging from 1 (localized occupation) to 3 (nationwide occupation)-was obtained from the respondents. Hence, Impact index scores range from −12 to 12: scores from −12 to −4 indicate strong negative impacts, scores from −4 to 4 indicate mild or null effects and scores from 4 to 12 indicate strong positive effects. The color code on the "Index graphs" Finally, we wanted to know the similarities and contrasts in the effects among species. This might be useful to answer ecological questions-such as "Do IAS from same taxonomic groups have similar effects, and do those effects differ between taxonomic

Question 4. Ecosystem services impact assessment
We have designed a semi-quantitative scale that assesses both positive and negative effects of this species. Each score is defined as: Score Impact score definition 4 The species leads to an increase in the provision of the ecosystem service, which is both intense and irreversible 3 Substantial increase in the provision of the ecosystem service; the effect is reversible if the species is managed or removed 2 Noticeable increase in the provision of the ecosystem service but reversible if the species is managed or removed 1 The increase in the provision of the ecosystem service is too small to be significant 0 No impacts detectable/ecosystem services not applicable to this species The reduction in the provision of the ecosystem service is too small to be significant -2 Noticeable reduction in the provision of the ecosystem service; it is reversible if the species is managed or removed -3 Substantial reduction in the provision of the ecosystem service; the effect is reversible if the species is managed or removed -4 The species leads to a reduction in the provision of the ecosystem service, which is both intense and irreversible groups?"-that may ultimately help to design management plans.
Then, we used k-means clustering algorithm (Hartigan & Wong, 1979) to determine the naturally occurring groups within the dataset, and the Silhouette Plot method (Appendix 3) to measure the fitness of the clustering (Kaufman & Rousseeuw, 1990).

| Assessing species potential to provide ES
We assumed that IAS has a potential to provide ecological or cultural benefits under appropriate management (defined as any management scenario that would lead to the improvement of a

| Assessing species manageability
Prioritization of cost-effective IAS management is often essential for site managers, due to limited resources. Risk management is a tool

Question 5. Species potential
Let's suppose that the negative impacts of this species on the environment can be mitigated with an appropriate management of its wild populations.
Do you think this species would then have the potential to provide any of the following benefits? Please, choose yes for the ecosystem services that can be improved.

Indicate your level of confidence: High/Medium/Low (use the guidance table in Question 1)
Provisioning services:  Booy et al. (2017), uses seven key criteria: Effectiveness, Practicality, Cost, Impact, Acceptability, Opportunity window and Likelihood of re-invasion.
As part of the quick IAS assessment proposed here, we developed a basic manageability assessment for assessing the feasibility of eradicating an IAS. This complements the results of the IAS effects assessment by providing a more comprehensive information about the ecology of the species in question. We based the manageability of the species on their spreading capacity (i.e., invasiveness), and the management effort (i.e., practicality-e.g., physical access and resources such as overall costs, dependent on machinery, staff and materials such as pesticides) that would be required for its eradication locally (see Booy et al., 2017).

| Piloting INSEAT: Case studies
Approximately 3,864 alien species are currently established in Great Britain (Zieritz, Armas, & Aldridge, 2014 All the graphical outputs and statistical analysis were performed using RStudio 3.3.1 (R Core Team, 2016), R packages "gg-plot2" (Wickham, 2009), "ggrepel" (Slowikowski, 2016) and "Flexible Procedures for Clustering" (Hennig, 2015). The pilot assessment form can be found in Appendix 2; this assessment form improved after the pilot thanks to the feedback provided by the respondents and reviewers. The final assessment form is shown in Figure 1.

| Categorizing level of confidence
We acknowledge the feedback from the testing of INSEAT that the pilot proforma lacks the capacity for the experts to validate their responses. The fact that respondents did not need to justify their answers or indicate their degree of uncertainty may strongly reduce the reliability of the assessment. Although the strength of INSEAT lies on its ability to rapidly obtain responses from a large number of experts, scores derived from this tool will inevitably have varying degree of uncertainty associated with them. In order to keep a balance between practicality and reliability, we added a section in the revised proforma asking the respondents to report the confidence level of their assessment for each ES (as High, Medium or Low; for definitions, see Figure 1). We also added a request to the respondents for information (e.g., scientific evidence, personal observations, professional opinions) that support their scores in general.
Understanding the uncertainty of the responses and its implications can help to further inform IAS management decisions.

| RE SULTS
Our pilot survey, covering 18 IAS, was completed by 78 IAS experts in total (i.e., response rate of 17%). The average number of species to Scotland (Roy, 2015). This invasive species was assessed by 12 experts in this study. The experts agreed that Harmonia axyridis has a positive impact through its effect on pest regulation. This also has a synergistic association with other benefits such as the production of cultivated goods (Figure 2a). Furthermore, 30% of the experts considered that this ladybird is potentially beneficial for provision of fuels (i.e., beneficial for standing vegetation) and harvested wild goods (Figure 2b). However, the experts had also identified some negative effects associated with this IAS; primarily this species could adversely affect wild species diversity, or genetic diversity (with a median score of −2). Therefore, this case study demonstrates how the tool could be employed to detect important trade-offs between the provision and loss of services associated with an invasive species (Figure 2c).
• Astacus leptodactylus (Figure 3)-Turkish crayfish occupies lakes, ponds, and rivers, but it has also been recorded in brackish water (Aldridge, 2016). This species was first recorded in 1975.
Currently, it is well established in England with isolated populations in Wales as well. This invasive species was assessed by 12 experts. The overall effect of this species in the country is considered as "mild" as none of the effect index is higher than 3 or lower than −3 (Figure 3c). This case study, however, highlighted a discrepancy among the experts in terms of their views on the usefulness of this species used as a food source ( Figure   3a). Nevertheless, 50% of the respondents indicated that there is a potential of this species to be used as a harvested wild good ( Figure 3b).
• Pacifastacus leniusculus ( Figure 4) (Day, 2015). We had 26 experts assessing this species. The majority of the effects of this invasive species were considered negative (Figure 5a). The level of congruence for two particular ES is low (i.e., high uncertainty): erosion regulation (median score of −3; quartiles ranging from 0 to −4), and pollination, (median score of 1; quartiles ranging from 3 to −3). Nevertheless, the impact index scores clearly indicated that this species as highly damaging to the environment (Figure 5c).

| Manageability and clustering analysis
Overall, the manageability of all 18 IAS in this study is low, with a man- The clustering analysis indicated that the best number of clusters for our species sample is three, with an average silhouette width of 0.27. This silhouette width is substantially low, indicating a weak clustering structure (Appendix 3). Hence, no statistically significant cluster was found among the 18 IAS in the study (Kaufman & Rousseeuw, 1990).

| D ISCUSS I ON
Preventing IAS spread is the most cost-effective strategy to build IAS-free ecosystems (Richardon & Ricciardi, 2013). However, such management approach is unlikely to be 100% effective (Chornesky et al., 2005); and the ongoing rapid rates of species invasion suggest that eradication of IAS may not be economically feasible in the future. In such scenario, goals of coexistence would be more viable and realistic (Hobbs et al., 2006;Hobbs, Higgs, & Harris, 2009;Walther et al., 2009).  (Humair, Edwards, Siegrist, & Kueffer, 2014). This is because the notion of IAS as concepts have similar but not identical meaning to different group of experts and stakeholders; this interpretative flexibility bears the risk of introducing misunderstandings. Humair et al. (2014) urged IAS experts to acknowledge uncertainties, to engage transparently in deliberation about conflicting issues and to take the role of impartial mediators of policy alternatives rather than of issue advocates. INSEAT supports this observation, with an aspiration that our results will aid in this deliberation.
In some IAS, the direction of their effects on certain ES remained equivocal. For instance, the impact score of Himalayan balsam on pollination ranged from 3 to −3. Furthermore, the socio-cultural attitudes of the respondents toward a particular species could also vary.
This was prominently reflected by the significant variations (ranging from positive to negative) in the impact scores for cultural ES-such as "aesthetics"-in many cases. As the assessments on cultural services are dependent on personal views, it could therefore inevitably be opened to more ambiguous outcomes.
Having incorporated the positive effects provided by IAS, INSEAT provides a more comprehensive assessment of the IAS consequences across different types of ES, as opposed to focusing on the negative aspects exclusively. This will provide users new insights into the species, allowing diversification of management actions. Once the prevention measures have failed, goals of coexistence are more feasible than eradication in terms of economic resources, time and management effort (Davis, 2009;Wittenberg & Cock, 2005). Hence, these management strategies should be preferred whenever it is possible. Successful management strategies often acknowledge "that the primary and inevitable constant of the natural world is change" (Davis, 2009). Therefore, we suggest an adaptive management approach to deal with IAS (Murray & Marmorek, 2003)  There are accounts of how the removal of an alien species could compromise the provision of cultural ES in a local context and lead to strong public opposition (Bennett, 2016;Bonanno, 2016;Dickie et al., 2014). Information gathered about the effects of an invasive species can be used, in combination with local knowledge, to work with stakeholders to identify the most appropriate management plan. For example, Sciurus carolinensis (gray squirrel)-one of the pilot species in this study-had received positive impact scores on multiple cultural ES and comments such as "for some people in the most urbanized areas, gray squirrels are their only experience of wildlife." The removal of gray squirrel had led to strong public opposition in the past (Bremner & Park, 2007); INSEAT would have allowed wildlife managers to circumvent public outrage by identifying alternative, socially acceptable squirrel management plans.
One useful feature of INSEAT is that it could highlight the potential benefits that an invasive species could provide under appropriate management (Figures 1b, 2b, 3b and 4b). Under certain climate change scenarios, some non-native species have even been considered necessary to assure local ecosystem function continuity (Lin & Petersen, 2013;Walther et al., 2009). Cases of IAS providing refuge for native species have also been reported (Chiba, 2010 should complement other risk assessments (e.g., Booy et al., 2017) and be used to build awareness, detect knowledge gaps and aid in the design of alternative management strategies. In fact, a bridge between INSEAT and EICAT, which evaluates, compares, and predicts the magnitudes of the environmental impacts of different IAS taxa (Hawkins et al., 2015), would be beneficial for both IAS management and policy.
Decision-makers could then evaluate all the knowledge available, while exploring management alternatives, by focusing on the functional role rather than on the origin of the species (Bonanno, 2016).

APPENDIX 1
Correspondence To determine uncertainty, the revised proforma asks respondents to report the level of confidence in their assessments, and to provide information that support their scores. (**) Management effort. The scale used to define the management effort was improved, inspired by the work of Booy et al. (2017). (***) In the revised version, the impact categories are better defined, and use terminologies that avoid making judgements on the effect of IAS on the ES. The new version also includes the option "Data deficient" for acknowledging that the lack of knowledge is because of no existing data or evidence, rather than the unawareness of an individual. (****) Question 6 was added to gather information that support the expert's responses.

K-M E A N S CLUS TE R I N G
Best number of clusters and interpretation of the results.
Clustering algorithms find naturally occurring groups in a dataset.
The K-means method divides the data points into "k" number of groups in which the sum of squares from these points to the clusters center is minimized (Hartigan & Wong, 1979). To select the best "k" (number of clusters), the Silhouette Plot method is widely used. A Silhouette Plot is a representation of the cohesion among the points of a cluster and the separation between the points of different clusters ( Figure A3b) (Rousseeuw, 1987). Using the notation in the Figure, the silhouette width (s i ) is a representation of the suitability of the object to belong to a cluster (j), and the silhouette of a cluster is a plot of the of the s i of all the objects (n j ) of the cluster ranked in decreasing order. The average silhouette plot width (S i ) is the average of the s i of all the clusters and can be used to measure of fitness of the clustering. Values <0.25 indicate that no substantial structure has been found; 0.26-0.50 indicate that the structure is weak and could be artificial; 0.51-0.70 indicate that a reasonable structure has been found, and from 0.71 to 1.0 indicate a strong structure. From all the average silhouette widths of all "k," the optimum (closest to 1) indicate the "best" number of "k," that is, the most "naturally" occurring number of clusters.
When interpreting the graph, one should attend both to the average silhouette plot width (S i ) to find the "best" k and the silhouette width (s i ) of each cluster for the validation of consistency: to identify outliers and objects that lie well within their cluster of the ones are F I G U R E A 3 K-mean clustering plot and Silhouette plot obtained for the species of the case studies. (a) Clustering plot. The 16 case studies are represented in two dimensions and grouped into three clusters. Components 1 and 2 explain 55.03% of the point variability.
(b) Silhouette plot used to select the best number of clusters. From all the possible number of clusters, k = 3 has the highest Average Silhouette Width (S (i) = 0.27). This low value indicates weak cohesion among the data points. j: cluster; n j : number objects (species) in each cluster; ave i∈Cj s i : average width of each cluster