Integrating local knowledge to prioritise invasive species management

1Centre for Biodiversity and Conservation Science, School of Biological Sciences, University of Queensland, St. Lucia, Queensland, Australia; 2ARC Centre of Excellence for Environmental Decisions, University of Queensland, St. Lucia, Queensland, Australia; 3NESP Threatened Species Recovery Hub, School of Biological Sciences, University of Queensland, St. Lucia, Queensland, Australia; 4United Nations Development Programme (UNDP), New York, New York; 5 The Nature Conservancy, Arlington, Virginia; 6Land, Environment, Economics and Policy Institute, University of Exeter Business School, Exeter, UK and 7Centre for Environmental Economics and Policy, UWA School of Agriculture and Environment, University of Western Australia, Crawley, Western Australia, Australia

Incorporating local values and preferences into early planning stages can be challenging (Ford-Thompson et al., 2012;Oppel, Beaven, Bolton, Vickery, et al., 2011). However, through engagement it is possible to clarify and diminish any safety or social concerns, such as fear of water supply contamination, risks to pets and children from poisoned baits, visually unappealing signs, animal welfare concerns or closure of tourist areas (Glen et al., 2013). Engaging with local stakeholders and the general community can mitigate possible opposition to the implementation of eradication projects, and ensure the local community is informed of the socio-economic, health and ecological benefits (and costs) (Vane & Runhaar, 2016) that could arise through the implementation of eradication plans. This is particularly important in invasive species management, given that the survival of few-invasive-individuals can undermine the whole project (Glen et al., 2013).
Existing approaches to incorporate the preferences and values of local communities and practitioners have often targeted a single-stage of the eradication planning process (Ford-Thompson et al., 2012;Novoa et al., 2018) for example: engagement (Luyet, Schlaepfer, Parlange, & Buttler, 2012), eliciting information (Larson et al., 2011) or informing perceptions (Bardsley & Edwards-Jones, 2006). In this work, we present a novel systematic approach based on adaptive management (Holling, 1978) to address the multidimensional challenges posed by invasive species management on inhabited islands (Berkes, 2007;Glen et al., 2013;Kerr, Baxter, Salguero-Gómez, Wardle, & Buckley, 2016). On inhabited islands, the social preferences of local communities must be addressed, in addition to ecological components and economic constraints. Here, we incorporate local knowledge and management preferences throughout the eradication planning process. We engaged with multiple stakeholders, elicited local knowledge, included natural resource managers' perceptions and budgetary constraints to compare contrasting management scenarios using a cost-benefit analysis tool, the Investment Framework for Environmental Resources (INFFER) .
Adaptive management is a special instance of structured decision making, which is a step-wise decision support framework that promotes the integration of scientific information and stakeholder's values to assist decision makers identify optimal management strategies (Bower et al., 2018;Gregory et al., 2012;Murphy & Weiland, 2014). The main principle of adaptive management is 'learn while doing', which makes it particularly useful for iterative approaches (Holling, 1978;USGS, 2019;Williams, 2011). We use INFFER  as it complements an adaptive management approach and can be easily implemented by decision makers to: (a) assess the perceptions and preferences of stakeholders regarding invasive species management (social dimension); (b) assess the feasibility, cost and impacts of alternative projects (economic dimension); and (c) incorporate stakeholders' expertise and perceptions to better inform invasive species management plans and their possible conservation benefits (ecological dimension).
We applied the proposed approach on Minjerribah-North Stradbroke Island, located in Queensland, Australia (hereafter Minjerribah), where we co-developed and evaluated six management scenarios, with different investment levels, each designed to control the impacts of European red foxes (Vulpes vulpes, Linnaeus, 1758) and feral cats (Felis catus, Linnaeus, 1758). We elicited stakeholder data through a semi-structured online survey (eSurvey) (Appendix S1) to assess six different scenarios and define which scenario would deliver the most cost-effective benefits to threatened and culturally relevant species (Appendix S2), and to the local community on Minjerribah.

| ME THODOLOG IC AL ANALYS IS AND CONTE X T
The objective of this study was to evaluate the cost-effectiveness of management strategies to control the impacts of invasive species on F I G U R E 1 Stages of the proposed framework to develop, assess and select invasive species management strategies. Boxes with broken outlines represent complementary actions that need to be undertaken to complete the main goal in every stage, which is represented by boxes with solid outlines the threatened and culturally relevant species of Minjerribah. In this section, we provide details about our case study, Minjerribah, the stakeholder-engagement process, application of INFFER , data collection and development and analysis to select the best strategies to control invasive species impacts on native and culturally relevant species. This wider process is described in Figure 1.

Island (Queensland, Australia)
Minjerribah has unique ecological, economic and culturally relevant values for the local and national Australian population. These values are currently being impacted, directly or indirectly, by invasive species. Minjerribah's ecological uniqueness and internationally important cultural heritage make it one of the top 50 offshore islands prioritised for protection in Australia (Ecosure, 2009). The Island is located approximately 40 km east of Brisbane (Queensland, Australia) ( Figure 2). It is the second largest sand island in the world (approximately 285 km 2 ) (Laycock, 1978), and the largest of the Moreton Bay Islands (Queensland, Australia) (27°30′S, 153°28′E). Minjerribah hosts a wide variety of habitats (Queensland Herbarium, 2009) that support many native sedentary and migratory species. The island is a stepping stone along the East Asian-Australian Flyway and is a 'Wetland of International Importance' (Ramsar Convention, 1971) making it an important site for Australian bird resident species as well as for intercontinental migrants (Wilson, Kendall, Fuller, Milton, & Possingham, 2011).
The Island has been inhabited by the Quandamooka people for at least 21,000 years (Barram, Carew, Hill, & Phillips, 2016). The Quandamooka people are the historical custodians of the Moreton Bay. In 2011, this was recognised by the Federal Court of Australia (National Native Title Tribunal, 2011), highlighting the cultural significance of the area. Since the 1940s, the Island has also been the source of extensive sand mining operations. The mining activities are scheduled to end by late 2019; a period which marks the end of an industrial era on Minjerribah, and the prospect of major change and potential economic growth for local businesses, tourism and the local community.
Fifteen vertebrate invasive species have been recorded on the island, including red foxes and cats (Appendix S2) (Threatened Island Biodiversity Database Partners, 2014). Red foxes and feral cats are two of the most damaging invasive species in the world (Courchamp et al., 2003;Doherty et al., 2016;Lowe, Browne, Boudjelas, & Poorter, 2000), and on Australian islands they are a main driver of native species decline (Doherty et al., 2015;Glen & Dickman, 2005;Legge et al., 2016;Saunders, Gentle, & Dickman, 2010). Red foxes and feral cats species not only have direct and indirect impacts on the threatened and culturally relevant species of the Island, but also affect its cultural heritage, and economically valuable local industries, such as tourism (Gong, Sinden, Braysher, & Jones, 2009;Jones, Saunders, & Balogh, 2006). In response to this threat, the local pest management authorities formed the Straddie Pest Management Group (SPMG). The aim of this group was to manage the impacts of invasive species on the Island. The diversity of local stakeholders, including indigenous and non-indigenous residents, and economic activities, as well as its biological uniqueness make Minjerribah the perfect location to assess optimal invasive species management approaches.

| Stakeholder engagement
The goal of this process was to engage with a representative sample of local stakeholders that actively participate or could be affected by the implementation of conservation actions that target invasive species on Minjerribah-North Stradbroke Island. The first step of this study was to contact (through email and phone) senior managers from a wide range of federal and local government authorities and non-government organisations. These managers were all involved in the development, assessment and implementation of invasive F I G U R E 2 Location of Minjerribah (North Stradbroke Island) in Queensland, Australia. Light grey areas indicate urban development species management strategies on Minjerribah. Initial meetings with these groups and individuals were unstructured and, after a process known as Snowballing Sampling (Atkinson & Flint, 2004), we were able to identify and engage with a broader group of local stakeholders involved in SPMG (Stage I in Figure 1), with whom we collaborated throughout all stages of our study. We communicated with the members of the SPMG through a gatekeeper who oversaw the distribution of information regarding our research project (according to the requirements of the Human Research Ethics Committee (HREC) of The University of Queensland), but the members of the SPMG could directly contact us for general queries.
The SPMG, at the time of this study, was formed by 13 representatives from six government agencies, five non-government organisations and community groups and two private organisations. SPMG members have been working on invasive species management for almost 10 years. Collectively, members have extensive experience managing invasive species on the Island and are familiar with the views of the local community regarding invasive species management. In consultation with members of the SPMG, we jointly defined a Species of Interest list by reviewing the existing literature (i.e. grey and scientific literature) and discussed not only invasive and threatened species that are found on the Island, but also species that have some cultural or local significance (Appendix S2). We used the Species of Interest list to design and assess a set of management scenarios to control the impacts of red foxes and feral cats on Minjerribah's Species of Interest (Table 1).

| Scenario development
Over a 2-year period (2015-2017), we met biannually with the SPMG and attended the group's annual general meeting. The meetings started with an open forum, where members of the community expressed their queries regarding conservation management projects that would take place on the island. Following the open forum, we met with members of the SPMG to discuss management strategies to control the impacts of invasive species (i.e. red foxes and feral cats) on the native and culturally relevant species of Minjerribah. During the meetings, we initially discussed broad topics, such as the latest management approaches, then focused on specific details, such as which actions were considered acceptable by the community, the effectiveness of past actions and the effectiveness of similar plans implemented in other locations. During this period, we also received direct email-based feedback from individuals and organisations, although most of the information was collated and then transmitted by the gatekeeper. All information was incorporated in the co-development of the six proposed management strategies (Stage III in Figure 1).
During this 2-year period, we co-developed six scenarios to manage invasive species by reviewing relevant literature and drawing on the previous experience of the SPMG members (Stage III in Figure 1).
In conjunction with members from the SPMG, we defined relevant attributes for the proposed management scenarios, such as cost, timeframe, feasibility according to local regulations, and accounted for management concerns from the public. All the information we gathered was treated equally (not weighted) and all members of the SPMG had an equal opportunity to contribute to the co-development process during our study (Stage III in Figure 1). The final six proposed scenarios were based on different investment levels (i.e. the total cost of the management strategy over a period of 25 years) defining the management intensity (i.e. low, medium and high). The management intensity varied according to the frequency of baiting campaigns (i.e. low = 1 campaign p.a.; medium = 2 campaigns p.a.; and high = 3 campaigns p.a.) and bait density that would be used (i.e. low = 2 baits/km 2 ; medium = 5 baits/km 2 ; and high = 10 baits/ km 2 ) throughout the year over a 3-year implementation window (a summary of the scenarios can be found in Table 1). The goal of the different scenarios was to diminish the impacts caused by red foxes and feral cats, by eradicating these species from the Island, hence increasing the probability of survival of culturally relevant and threatened species. Three scenarios were designed to target only red foxes (Scenarios 1 to Scenario 3) and three more scenarios that target red foxes and feral cats (Scenarios 4 to Scenarios 6). We did not develop management scenarios aimed at 'only cat' control as this action was deemed 'infeasible' in the absence of long-term commitments to control resident pet cat populations.
Each scenario varied in its management intensity (i.e. number of traps deployed, number of baits/km 2 , number of stations/km 2 and number of baiting campaigns per annum), and length of implementation of the different control methods throughout the year (i.e. baiting, trapping, hunting and den fumigation; see Appendix S3 for a detailed description, including cost information). The cost of each scenario was constructed using a combination of data provided by members of the SPMG, scientific and grey literature and quotes by private distributors of the consumables, goods and capital assets (Auerbach, Tulloch, & Possingham, 2014;Holmes et al., 2015Holmes et al., , 2016Mcleod & Saunders, 2010). We report in detail the costs in

Target species
Scenario # Investment level

Management intensity
Only red foxes (Vulpes vulpes)

| Data collection
To identify which of the scenarios would offer the greatest return on investment, we used INFFER (described below). We elicited the input parameters for INFFER (Stage IV in Figure 1) by sending out an online, semi-structured questionnaire (eSurvey, found in Appendix S1

| Analysis framework: INFFER analysis
We then used the PAF from the Investment Framework for organised approach based on a benefit-cost ratio (BCR) to identify management actions that will achieve the most cost-effective environmental outcome , the steps of the INFFER approach are shown in Table 2.

Relevant and Time-Bound) projects, INFFER helps clarify what
is required to achieve proposed outcomes (Bottrill et al., 2008).
This assessment process is the core of INFFER, and provides the basis to assess whether a project is cost-effective (i.e. if BCR > 1, then a project is deemed cost-effective), as calculated by the BCR F is the multiplier for technical feasibility risk. P is the probability that socio-political factors will not derail the project, and that the required changes take place; this value was obtained from the eSurvey ('What do you think is the chance that one of the social or political situations described below will prevent eradication from being achieved?'). Socio-political risks include: noncooperation by the different organisations and the impacts of social, administrative or political constraints. G is the probability of obtaining long-term funding. DFb is the discount factor. C is the short-term project cost ($ million in total, over the life-span of the project). M is the total cost of maintaining the outcomes ($ million per year, beyond the immediate project). PV(M) is the present value to convert a stream of future annual maintenance costs (assumed constant in real terms) to their present-day value (in $ millions) . Further information about the rationale for the BCR algorithm and the underpinning theoretical background can be found in Pannell et al. (2012Pannell et al. ( , 2013. Subsequently, the results from the INFFER analysis were sent out to the SPMG members for review, and to assess whether the scenarios were appropriate (Stage V in Figure 1). The INFFER parameters were collected from the eSurvey that was sent to the participants as described in Section 2.42.3 (Data Collection). It is  (Bateman, Mace, Fezzi, Atkinson, & Turner, 2011), and in the case where primary values are sought, these can be highly influenced by individual preferences, and are often overestimated by local stakeholders (Jakobsson & Dragun, 2001;Portney, 1994).
Heterogeneous responses can also confound the proper interpretation of this parameter.
We used a ranking-based assessment for the six proposed scenarios. We obtained an Overall ranking for the six scenarios and two, more detailed, Internal rankings: one for red fox-only control, and a second for joint-management. By using a structured decision-making approach based on INFFER, we were able to account for intrinsic biases, information-gaps and respondents' valuation heterogeneity, thereby facilitating the overall analysis and increasing the robustness of policy recommendations.

| Respondents summary
All sectors involved in invasive species management on the Island were represented in the surveyed respondents: 46% were representatives of government agencies (six responses); 39% were from community or non-government organisations (five responses); and 15% were from private organisations (two responses), which matches the total number of representatives we initially identified (i.e. 13). A key aspect of the INFFER assessment is to define the significance of the environmental asset that a project will affect. Respondents scored the probability of eradication of red foxes under Scenario 1 as low-77% (medium-23%), Scenario 2 as medium-46% (low-23% and high-31%) and Scenario 3 as high-85% (medium-15%). The probability of joint-eradication (red foxes and feral cats) under Scenario 4 was scored as low-77% (medium-23%), Scenario 5 as medium-54% (low-23% and high-23%) and Scenario 6 was scored as high-77% (low-15% and medium-8%).

| INFFER analysis
We present the results of the INFFER parameters in Table 3. We found that respondents' asset valuation-V was highly heterogeneous. Hence, we assessed the BCR of each scenario under three different assumptions regarding the value of this parameter, the When V is equal to 1, the BCRs are less than one for all scenarios, except for Scenario 3. When the BCR value is higher than 1, it represents the 'break even' point of the project, meaning that project benefits exceed project costs. When V = 15 and V = 50, all scenarios have BCRs higher than 1. Despite the changes in the BCR according to changes of the asset value, the rankings do not change.
By comparing the scenarios under different perspectives of the asset value (V), we were able to assess the robustness of our results to different stakeholders' values. Across all six scenarios, the highest-ranking strategy was Scenario 3 (BCR = 1.15), as shown in Overall and Internal ranking for fox-only management in Table 3, which was the fox-only 'High' management intensity Scenario. For fox-only management scenarios, Scenario 3 was also the most expensive approach (AU$m 5.33). Scenario 1 (AU$m 3.48) was approximately 35% cheaper than Scenario 3, whereas Scenario 2 (AU$m 4.08) was 24% cheaper than Scenario 3. Across all three scenarios targeting only red foxes there was little variability in the socio-political risk (P), however, the impact of works-W varies con-

| Sensitivity analysis (SA)
We conducted a sensitivity analysis to determine the sensitivity of management recommendations to changes in three of the INFFER parameters: (a) Impact of works-W, (b) Socio-political risk-P (c) and Lag time-L. We chose these parameters because they demonstrated the greatest heterogeneity or were identified in the literature (Glen et al., 2013) as having a large impact on the success of invasive species management. We assessed changes in the three parameters across the Best Performing Scenarios (Scenarios 3 and 5), and calculated a Sensitivity Index (SI) (Alexander, 1989) for each parameter, as well as a BCR Difference (%) (see Table 4). A high SI score indicates a high sensitivity of the BCR to changes in that parameter. Across the three parameters, the BCR was most sensitive to changes in Sociopolitical risk-P (SI = 0.88 and 0.87 in Scenarios 3 and 5 respectively).
After socio-political risk, Scenario 3 was more sensitive to changes in Impacts of the works-W (SI = 0.69), than to variation in Lag time-L (SI = 0.60), whereas Scenario 5 was more sensitive to changes in Lag time-L (SI = 0.77), than to changes in Impacts of the works-W.

| D ISCUSS I ON
We assessed the BCR of six invasive species management scenarios on Minjerribah by including the perspectives of local government and community members into a cost-benefit analysis, INFFER. The analysis showed that fox-only control with 'high' intensity (Scenario 3) was the best strategy, as well as the only strategy under a conservative estimate of asset value (V = 1) that had a BCR greater than 1 (1.15), implying that the benefits of implementing this action exceeded the costs.
Among the fox-only Scenarios, Scenario 3 had a shorter time lag (3 years vs. 7). This result suggests that higher investment levels will TA B L E 3 Results of benefit-cost ratios and correspondent parameters calculated in INFFER  lead to quicker outcomes, relative to lower investment levels. The dominance of this strategy can be explained by the perceived greater knowledge of red fox ecology among respondents, the current understanding of eradication measures and wider political and community support to control a species that is not considered a companion animal (like cats). Among the scenarios aimed at joint-management of feral cats and red foxes, Scenario 5 ('Medium' investment levels) had the highest BCR (BCR = 1.15). Invasive species managers on the island judged that Scenarios 5 and 6 (high investment levels) would have equivalent impact of works, socio-political risk and lag times.
However, the higher cost of Scenario 6 resulted in a lower BCR relative to Scenario 5. It is important to note that Scenario 5 corresponds to current, recommended feral cat management strategies Overall, the impacts of feral cats on native species are well documented Denny & Dickman, 2010;Dickman, 1996;Doherty et al., 2015;Medina et al., 2011). What is not well understood is how to operationalise invasive management activities, such as baiting and banning companion animals on islands, without incurring significant community resistance. Existing management actions (i.e. hunting, trapping and baiting), which target feral cats are unlikely to be effective on inhabited islands in the long term, as pet cats can be a source for re-establishment of feral cat populations (Denny & Dickman, 2010). This is captured by the Lag time (L) for joint-management scenarios, which was 30 years (Scenario 4) compared to 10 years for Scenarios 5 and 6. In this project, none of the scenarios required behavioural changes (A) by the communitywhich we know is needed-which is why the perceived Impacts of the Works-W value for joint-management scenarios might not have been higher. Notwithstanding the lack of a standard procedure to tackle these species ( (Baker, Gordon, & Bode, 2016), Optimal eradication schedules (Bode, Baker, & Plein, 2015) and Optimal surveillance (Holden, Nyrop, & Ellner, 2016;Rout, Hauser, McCarthy, & Moore, 2017) have proven to be valuable techniques to identify potential ecosystem impacts from single-species management, and to optimise invasive species eradication.
Sensitivity analyses (Section 3.33.1 and Table 4) are a valuable tool to assess the sensitivity of management recommendations to changes in parameter values (Pannell, 1997). A sensitivity analysis showed that Scenarios 3 and 5 were highly sensitive to variation in Socio-political risk (P). This result highlights the need to account for local socio-political conditions and its possible variations between early planning stages and later implementation of the proposed actions. Moreover, it also emphasises the importance of facilitating the participation of multiple stakeholders to capture their perspectives, and the need for methods to evaluate local stakeholders' preferences, as new conditions could derail the future success of the proposed strategies. Scenario 3 is also sensitive to changes in the effectiveness of the proposed management strategy (Impacts of the works-W); if effectiveness were to decline, Scenario 3 would no longer be a viable management strategy (BCR < 1). Scenario 5, which was the best performing scenario within the 'joint-management' strategies, was also sensitive to changes in Lag time (L), followed by variations in effectiveness of the works (W). Even with a 'best' approach for these parameters (i.e. where all parameters are set to their most optimistic value, relative to project success), Scenario 5 does not have a BCR > 1, implying that the benefits of implementing this scenario would not exceed the costs. As previously suggested, information gaps could be affecting a proper assessment of jointmanagement strategies, as the removal of a single species has shown to be a suboptimal strategy when red foxes and feral cats coexist (Ballari et al., 2016).
Eliciting values for environmental goods is a difficult and complex process. Stakeholder valuation of local assets, like Minjerribah, can overestimate the intrinsic significance of the asset, and be sensitive to personal bias (Portney, 1994). The result is a high level of subjectivity and heterogeneity in provided answers (Marsh, Curatolo, Pannell, Park, & Roberts, 2010 (Jakobsson & Dragun, 2001). Using INFFER allowed us to incorporate these perspectives and preferences explicitly to support a transparent decision-making process (Marsh et al., 2010).
This study outlined an approach to incorporate social, ecolog- The framework presented in this study is based on benefit-cost analysis (BCA), which is a preferred method to compare alternative strategies or set priorities (Marshall, McNeill, & Reeve, 2011). BCA is used to determine if a given project improves outcomes relative to initial conditions. These conditions are measured by estimating the benefits and costs of a proposed project (Marshall et al., 2011).
One of the main concerns regarding BCA approaches is the difficulty assigning monetary value to environmental assets (Bateman et al., 2011;Marshall et al., 2011).
There are other approaches to making conservation investment decisions besides BCA, such as deliberative methods and multi-criteria analysis (MCA). Deliberative methods involve similar participatory approaches as BCA and MCA, however they do not use a formal organised model to calculate decision outcomes, which makes them a less reliable approach for decision making (Marshall et al., 2011).
MCA methods allow for the inclusion of multiple strategies, multiple objectives and different criteria simultaneously (Marshall et al., 2011); but have been criticised for their complexity, vulnerability to abuse by special-interest groups and large information requirements, which can make implementation and interpretation by decision makers difficult (Dobes & Bennett, 2009;Marshall et al., 2011).
Local conditions and the requirements of different studies will determine which approach, for example, BCA, deliberative methods or MCA, is most appropriate. In our case study, INFFER was the most suitable approach as it can be easily implemented by decision makers, allows for a more collaborative approach, explicitly integrates local perspectives and technical components (i.e. feasibility, cost and benefits), and provides accountability over the decision process and its outcomes.
The environmental uniqueness of Minjerribah is a key determinant of the island's environmental and cultural significance. However, native species on the island are threatened by European red foxes and feral cats. Involving stakeholders in invasive species management is a critical but difficult aspect of management (Ford-Thompson et al., 2012). We have overcome barriers to incorporate local stakeholder knowledge into invasive species management by following a multistakeholder engagement process based on adaptive management principles (Holling, 1978) and INFFER . Our approach allowed us to identify that a high level of investment targeting red foxes on Minjerribah would provide greater benefits relative to its costs. This result is a timely example of how invasive species management can be approached on inhabited islands but outlines the need for more research directed at feral cat management protocols.
We believe that provided the right pre-assessment, implementation and monitoring tools, Minjerribah is a suitable location to pursue eradication of feral cats and European red foxes. It is important to consider the existing socio-political environment, the technical experience of local natural resource managers, as well as community cohesiveness, engagement and overall support. Implementing these actions will ultimately protect the Island's unique biodiversity, future economic well-being and its unique cultural heritage.

R E S E A RCH E TH I C A L A PPROVA L
This research project complies with the requirements of the Human Research Ethics Committee (HREC) from the University of Queensland, approval number: 2016001001.

CO N FLI C T O F I NTE R E S T
Nothing to declare.

ACK N OWLED G EM ENTS
We acknowledge the Traditional Custodians of the Islands, waters and seas surrounding Minjerribah and pay our respects to Elders, past, present and future and hope this work is of benefit to the aspirations of the Quandamooka People. We thank the support of every-