Uncovering antagonisms in recovery planning for species at risk: A diagnostic approach

Amid Earth's ongoing sixth mass extinction event, numerous measures have been proposed to recover the populations of species at risk of extinction. However, the methods and objectives of different species' recovery plans sometimes conflict with each other, causing a conundrum we refer to as recovery–action antagonism. Recovery–action antagonism reduces the cost‐effectiveness of conservation programs and can increase the extinction risk of nontarget species. We describe a method to identify interactions between recovery actions, including antagonisms proposed for different at‐risk species in a given location. The method includes a process to evaluate potential drivers of recovery‐action antagonism and other interaction types using principal coordinates analysis and distance‐based redundancy analysis. We illustrate various applications of the method through case studies performed in Pelee Island and Rouge National Urban Park, two biodiverse areas in Ontario, Canada. Potential antagonism was identified between 1.5% (Pelee) and 5% (Rouge) of the evaluated recovery actions. Although the rate of antagonism was low in our case studies, the method allows the identification of a variety of interactions, which can help to prioritize similar and complementary actions that will benefit a large number of species while minimizing actions that may have competing outcomes.

Recovery planning and subsequent implementation is the leading approach for the conservation of at-risk species (Schwartz, 2008).Recovery plans identify threats to a species and prescribe recovery actions based on quantitative data and expert opinion (Bottrill et al., 2011).Recovery plans may be geared towards individual species ("single-species approach") or target several species ("multispecies approach") or entire ecosystems ("ecosystem approach"; Clark & Harvey, 2002).There is ongoing debate regarding the comparative efficacy of single-species versus multispecies and ecosystem approaches (Boersma et al., 2001;Clark & Harvey, 2002;Rejis Baptista et al., 2019).Generally, single-species recovery planning promotes the creation of better targeted conservation actions (Sheppard et al., 2005), whereas multispecies and ecosystem plans may better address the complexity of ecological dynamics (Franklin, 1993;Jewell, 2000;Simberloff, 1998).Recovery is probably best achieved using both strategy types together to adequately address functional, phylogenetic, and evolutionary diversity at both species and community levels and to accommodate for spatial discrepancies between species (Campbell et al., 2002;Lamothe, Dextrase, & Drake, 2019;Ortega-Argueta et al., 2017).
Yet, in practice, many countries have historically emphasized a single-species approach to recovery planning at the federal level (Clark & Harvey, 2002;Rejis Baptista et al., 2019).Thus, local conservation practitioners who oversee lands containing multiple at-risk species often must navigate several single-species recovery plans.As a result, a particular conundrum, which we refer to as recovery-action antagonism, may arise when the actions identified in different recovery plans contradict or conflict with one another (Chadès et al., 2012;Gumm et al., 2011;Jewell, 2000;Sigaud et al., 2020;Simberloff, 1998).A well-documented example of this scenario occurred in the United States, when a proposal to reduce water levels for the threatened Wood Stork (Mycteria americana) in Everglades National Park contradicted proposed measures aiming to increase water levels to benefit the endangered Everglades Snail Kite (Rostrhamus sociabilis plumbeu).This eventually led the US Fish and Wildlife Service to oppose Everglades National Park's proposal on the grounds that it would harm the kite (Simberloff, 1998).
While conservation conflicts are frequently overlooked in practice (Sigaud et al., 2020), other examples of antagonism have been described.Prescribed burns in Kansas, USA, were found to both benefit and harm different imperiled grassland and shrub-dependent bird species (Powell, 2006).The introduction of Red Squirrel (Tamiasciurus hudsonicus) in Newfoundland, Canada, to benefit the near-extirpated Pine Marten (Martes americana atrata) devastated the native Red Crossbill (Loxia curvirostra) population (Benkman, 2009).The removal of an invasive weed (Oxalis pescaprae) in central Portugal had unanticipated negative effects for native pollinators (Ferrero et al., 2013), while the removal of invasive feral Cat (Felis catus) on World Heritage Macquarie Island (sub-Antarctic region) led to a trophic cascade that caused several detrimental ecosystem changes (Bergstrom et al., 2009).Particularly challenging cases of antagonism have been described wherein one at-risk species predates upon, or competes for resources with, another at-risk species, implying that certain recovery actions designed to benefit one species could negatively impact another (e.g., Chadès et al., 2012;Gumm et al., 2011;Oro et al., 2009;Thirgood et al., 2000;Williams et al., 2011).
At present, recovery planning is often undertaken with limited understanding of interspecific relations and ecological processes (Campbell et al., 2002;Dunwiddie et al., 2016;Sigaud et al., 2020).While it is unrealistic to expect a comprehensive understanding of how all spatially co-occurring species interact (Sigaud et al., 2020) and their role in various ecological processes (Lamothe et al., 2019b), failure to consider potential interaction types, including antagonisms between recovery actions, could result in the implementation of measures that unintentionally harm nontarget at-risk species, thus potentially increasing their risk of extinction (Chadès et al., 2012).Furthermore, recovery planning can waste the limited resources available for species conservation if recovery actions are redundant, inefficient, or undermine each other (Barrows et al., 2005;Burgess et al., 2019;Campbell et al., 2002;Sigaud et al., 2020).Most current conservation tools, such as population viability analyses, continue to draw predominantly from single-species models and lack "biological realism" (Sabo, 2008).There is a lack of conservation tools that can be utilized to evaluate interspecific interactions in multispecies systems (but see Soulè et al., 2003Soulè et al., , 2005)).As the number of at-risk species and their associated recovery strategies continue to grow, the potential for antagonism between recovery actions and objectives increases (Campbell et al., 2002;Dunwiddie et al., 2016;Sabo, 2008;Sigaud et al., 2020).Thus, approaches to identify and address antagonisms are vital for conservation actions directed at species recovery.
Given limited resources for conservation, several methods have been developed to prioritize recovery actions for a given space.Many of these procedures promote cost-effectiveness and feasibility (Carwardine et al., 2018;Joseph et al., 2008;Martin et al., 2018), including prioritizing recovery actions that have benefits for multiple at-risk species (Chadés et al., 2014;Tulloch et al., 2016).However, to our knowledge, none of these methods provide specific guidance on how to identify and address potential recovery-action antagonism.
There may be opportunities to predict recovery action antagonisms and address them proactively.Thus, this article presents a novel methodology for assessing and prioritizing conservation measures from multiple recovery plans.Specifically, we describe a procedure for characterizing the likely interactions between recovery actions proposed for different at-risk species in a specific area, enabling the identification of both antagonistic recovery actions (measures to be avoided) and recovery actions that are likely to be neutral or have multispecies benefits (measures to be prioritized).Using two case studies, we focus on the application of the method for identifying potential antagonism between recovery actions and describe a process for exploring the potential drivers of recovery-action antagonism.Researchers and environmental managers can apply this methodology at different spatial scales to inform the prioritization of recovery actions, support the development of multispecies and ecosystem recovery plans, and evaluate the cost-effectiveness of existing conservation programs.

| MATERIALS AND METHODS
Here, we detail the application of a methodology to assess how recovery actions from different recovery strategies interact in multispecies systems.We apply the method to two case studies for at-risk species inhabiting two protected areas in southern Ontario, Canada: Pelee Island (PI) and Rouge National Urban Park (RNUP).Figure 1 provides a schematic representation of the methodology.

| Define study area
Discrete study areas are first identified to spatially bound the analysis.PI and RNUP were selected as case locations because they are priority conservation areas with high numbers of at-risk species (Chu et al., 2015;Deguise & Kerr, 2006;Kraus & Hebb, 2020;Parks Canada, 2021), suggesting that potential recovery-action antagonisms within their boundaries could be highly consequential for biodiversity conservation in Canada.

| Identify recovery plans
Next, recovery actions are extracted from recoveryplanning documents.In Canada, the federal government oversees recovery planning for species listed as threatened, endangered, or extirpated under the federal Species at Risk Act (SARA, 2002;Mooers et al., 2010).Thus, recovery strategies developed under SARA were used as the source of recovery actions for the case studies.

| Compile list of at-risk species
Species of conservation concern are then identified for the study area.For both PI and RNUP, species lists were obtained from park authorities.These lists were then refined using the following criteria: (1) species are terrestrial during at least one life stage (e.g., fishes excluded, amphibians included); (2) species occur regularly within the boundary of the study area; and, (3) species are listed as threatened, endangered, or extirpated under SARA and have an available recovery strategy, either in draft or final form.

| Acquire recovery plans
Next, recovery plans are acquired for the refined list of species, as available.For the case studies, where federal recovery strategies under SARA had not yet been developed or made available, corresponding provincial recovery strategies developed under Ontario's Endangered Species Act (2007) were utilized, as they were considered a reasonable approximation for federal strategies given the degree of coordination between provincial and federal recovery planning (Canada-Ontario Agreement on Species at Risk, 2010).Species that did not have a recovery strategy available under either the provincial or federal law were excluded from the study.

| Compile recovery actions
Recovery actions are then compiled from the identified recovery plans.For the case studies, a focus was placed on actions with substantive, on-the-ground effects for habitat in the study areas.To facilitate the identification of such actions, several inclusion criteria were defined in Appendix S1, which describes, in general terms, the types of "methods" (physical intervention) and "objectives" (intended outcome) of recovery actions included in the analysis.While more than one method and/or objective might apply to a recovery action, any action that could not be classified as at least one of the methods and objectives in Appendix S1 was excluded.For example, recovery actions that pertained to fostering partnerships, research/monitoring, and outreach were excluded from the study, as these actions were means to formulate more inclusive teams and better-informed decisions, but did not themselves result in a substantive, on-the-ground effect in the study area.
To ensure that the compilation of recovery actions was standardized among assessors involved in data collection, a Delphi-type approach was employed using a set of sample data (McCune et al., 2013;Montgomery et al., 2021;Ortega-Argueta et al., 2017).Disagreements between assessors regarding recovery actions to be included were addressed through iterative debrief sessions.Trials were repeated until 80% agreement between assessors was attained.
The completion of this step yielded a list of recovery actions associated with the at-risk species inhabiting each study area.To facilitate future steps of the methodology, recovery actions were recorded verbatim from the recovery strategy (Appendix S2) and classified as outlined in Appendix S1.Additional information on the method and objective, such as habitat types addressed by the recovery action, was recorded in square brackets (see Appendix S2).

| Pairwise evaluation of recovery actions
Next, all recovery actions are evaluated against each other in a pairwise fashion to identify the nature of the relationship between each action.Every possible pair of recovery actions was evaluated for potential antagonism.
Five types of possible interactions between recovery action pairs were defined: F I G U R E 1 Schematic representation of the methodological steps, with arrows indicating the workflow, used to examine interactions of recovery actions for species-at-risk within a specific case location.
• Similar-recovery actions that propose physical interventions ("methods") that are essentially the same and are working towards a similar outcome ("objective"); • Complementary-recovery actions that propose different methods, but that are trying to achieve similar outcomes; F I G U R E 2 Decision-tree framework to guide the pair-wise comparison of recovery actions between two species.
• Neutral-recovery actions with different methods and objectives; the methods and objectives are unlikely to conflict with each other; • Antagonistic-recovery actions where the method or objective (or both) of one action conflicts with the method and/or objective of the other action; • Unknown-recovery actions that cannot be classified as similar, complementary, neutral, or antagonistic due to inadequate information being provided in the recovery plan.
Examples of each of these recovery-action interaction types from the case studies are provided in Appendix S3.
Each unique pair of recovery actions is classified as one of these interaction types.A decision-tree framework was developed to guide the qualitative assessment of recovery-action pairs (Figure 2).The first node of the decision tree prompts users to evaluate if the two actions could possibly be antagonistic, before any of the other interaction types are contemplated.The rationale for classifying recovery actions as antagonistic was often subtle.Namely, we designated actions as antagonistic if there was reasonable cause to suspect that the two actions would meet the definition for antagonism in a certain plausible case, even if, for the majority of cases, the antagonistic dynamic would likely be avoided through well-informed environmental management.When evaluating recovery-action pairs, we considered the possible outcomes of implementing the two actions in the same place at the same time, assuming that virtually no consideration was granted for any nontarget species.An example of the use of the decision-tree with two actual recovery actions is provided in Appendix S4.
Before evaluating pairs of recovery actions, a Delphi-type approach was employed to ensure that the qualitative assessment of a sample dataset was being conducted consistently between assessors; specifically, until 80% agreement between assessors was attained.In scoring the sample dataset, questions arose regarding the degree to which potential indirect effects should be considered, which external sources of information could be used, and on the appropriate threshold for classifying a pair of recovery actions as antagonistic.Consensus building on these topics was facilitated through iterative debrief sessions, focusing on areas for which there was disagreement between assessors.Key decisions made collaboratively during these discussions were recorded so they could be referred to during the actual assessment and accessed by future researchers (see Appendix S5).

| Descriptive analysis of recoveryaction interactions
The results of the pairwise recovery-action evaluations are compiled in a recovery-action-to-recovery-action matrix (see Table 1).From this matrix, the overall percentage of each interaction type, considering all evaluated pairs of actions, can be calculated to provide a cumulative perspective on the compatibility of recovery actions within a study area.Counts or percentages of each interaction type can also be determined for individual recovery actions, providing focused insight on the relative alignment of a specific recovery action with the conservation objectives of other at-risk species in the study area.
To evaluate the data at the species level, five speciesto-species matrices are developed that, respectively, contain the proportion of comparisons classified for each recovery-action interaction type, relative to the total number of comparisons between the two species (see Table 2).If, in the original recovery-action-to-recovery- action matrix, each classification type is considered independently (such that a given classification is either present or absent for a pair), then these five matrices represent the simple-matching coefficient for each interaction type (Gower, 1971;Sokal & Michener, 1958).In each of these five tables, average values for each species can be calculated, providing information about the relative alignment of a given species' recovery plan with the recovery objectives of other at-risk species in the study area.These tables also enable the identification of species possessing the highest and lowest proportion of each interaction type.

| Part C: Exploring drivers of recovery-action interactions
The previous steps of the methodology can stand alone as an independent analysis of the degree of recoveryaction alignment and antagonism in a study area.However, some users of the methodology may wish to explore the potential drivers of recovery-action antagonism (or other interaction types) within a study area.
Steps 8 and 9 of the methodology describe a process for exploring the potential influence of explanatory variables on the interaction between species' recovery actions.

| Compile explanatory variables
The first step of this process is to identify explanatory variables believed to potentially influence the interaction of recovery actions.For the RNUP and PI case studies, the following explanatory variables were identified and summarized in a table (Table 3): (1) listing status under SARA (threatened, endangered, or extirpated); (2) year that the final recovery plan was released; (3) whether a species was part of a multispecies or single-species recovery plan; (4) whether the recovery plan assessed was federal or provincial; (5) goal of the recovery plan (e.g., population recovery, slowing the rate of decline, etc.); (6) whether critical habitat was designated; (7) species taxonomic group; (8) whether species is a habitat specialist; (9) habitat types; and, (10) threats to species as represented by the International Union for Conservation of Nature (IUCN) threat categories.These variables can generally be classified as describing either species information or a characteristic of a recovery strategy (McCune et al., 2013).Detailed descriptions and datacollection approaches for these variables are described in Appendix S6.

| Statistical analysis of drivers of recovery-action interactions
Potential drivers of recovery interactions can be explored using multivariate analyses.Specifically, principal coordinates analysis (PCoA) and distance-based redundancy analysis (db-RDA) can be used to evaluate whether the explanatory variables are correlated with variation in each of the five species matrices.First, the simple matching values in the five species matrices are converted to distances by subtracting each value from one (Legendre & Legendre, 1998, p. 275).The table of explanatory variables is the "predictor" data, while the interaction between species' recovery plans, as represented by the five species distance matrices, is the "response" data.Distances between species in the response data are visualized using PCoA (Gower, 1966;Legendre & Legendre, 1998;McCune et al., 2013), which can be executed in R using the "pco" function from the ecodist package (Goslee & Urban, 2007; R Core Team, 2020).The db-RDA can then be used to assess whether the explanatory variables are correlated with the distance matrices (Legendre & Anderson, 1999).The db-RDA extends multiple linear regression by evaluating the effect of multiple explanatory terms on multiple principal coordinates (Buttigieg & Ramette, 2014).The db-RDA can be executed using the "capscale" function from the vegan package in R (Oksanen et al., 2020;R Core Team, 2020).Additional detail on our application of the db-RDA, including key assumptions and limitations, as well as our approach to developing and evaluating db-RDA models, is provided in Appendix S7.

| RESULTS
Results are provided for two applications of our methodology for assessing and prioritizing conservation measures in multiple recovery plans.

| Descriptive analysis of recoveryaction interactions
Between the RNUP and PI case studies, a total of 37 species (with 4 species occurring in both study areas) met all inclusion criteria for the study (see Appendices S8 and S9).Together, these species had 301 recovery actions and, consequently, over 16,000 unique action-to-action comparisons were evaluated.Pairwise (species-to-species) simple-matching values derived from the evaluations conducted for RNUP and PI were calculated, subdivided by interaction type.In both locations, the majority of action-to-action comparisons were designated as neutral, followed by unknown, then complementary (Table 4).Only 1% of comparisons were designated as similar for both locations (Table 4).
The percentage of action-to-action comparisons designated as antagonistic was greater for RNUP (5%) than for PI (1.5%), representing the most distinct difference between the results of the two case studies (Table 4).However, the number of species-to-species relationships with at least one pair of antagonistic recovery actions was relatively similar between RNUP (40%) and PI (32.5%).Pairs of species with some antagonism tended to have a greater number of antagonistic recovery actions in RNUP compared with PI.As 13.5% and 16% of the interactions in PI and RNUP, respectively, were designated as "unknown," the full extent of antagonism may have been underestimated in the case studies.

| Drivers of recovery-action interactions
The influence of some explanatory variables-notably, habitat types occupied by species (see Figure 3)-was apparent in the PCoA plot for RNUP.In the RNUP case study, the variables "Habitat: Coniferous" (p = .021;R 2 = .063)and "Habitat Specialist" (p = .028;R 2 = .058)were correlated with the antagonistic distance matrix.The "Habitat: Coniferous" and "Habitat Specialist" variables had similar Akaike's Information Criteria (AIC) and adjusted R 2 scores when assessed independently."Habitat: Coniferous" and "Habitat Specialist" were largely redundant; including both terms in a db-RDA increased the AIC score compared with models that included each term individually.In PI, "Habitat: Mixed Forest" was significant for the antagonistic distance matrix (p = .032;R 2 = .015).Db-RDA models were developed with conceptually meaningful combinations of explanatory variables to explore which factors, or combinations of factors, may contribute to each recovery action interaction type (see Table 5).

| DISCUSSION
Several papers have previously examined antagonism between specific pairs of species (Chadès et al., 2012;Dunwiddie et al., 2016;Lamothe et al., 2019a;Simberloff, 1998).However, to date, conservation antagonisms have scarcely been examined on broader scales, such as within an entire protected area or ecosystem.Thus, the full extent of antagonism-and the resulting implications on the cost-effectiveness of conservation programs-is generally unknown.Our methodology addresses this gap by enabling the quantification of recovery-action antagonism, as well as other interaction types, among an entire community of at-risk species within specific locations, such as protected areas.
The overall percentage of antagonism identified (5% and 1% for RNUP and PI, respectively) suggests that recovery-action antagonism is a relatively minor phenomenon in the case-study locations, but present at sufficient scales to warrant further evaluation.Recovery actions were designated as antagonistic for a variety of reasons, meaning that some identified antagonisms may be more consequential than others in terms of practical risk to nontarget species and cost-effectiveness.Yet, under a precautionary approach, even negligible degrees of antagonism warrant further scrutiny, as antagonisms could increase the risk of species extinction (Chadès et al., 2012).
RNUP and PI are small, ecologically similar areas in southern Ontario, Canada, and it is plausible that antagonism could be more prevalent in other locations.Variables such as the geographic scale of the area assessed, type and diversity of habitats present (e.g., proportion of natural versus degraded sites), threats identified, and the local assemblage of at-risk species are all relevant factors that could lead to higher levels of antagonism in other spaces.Additional applications of the methodology in other case-study areas are needed to confirm whether findings from RNUP and PI are characteristic for other locations of conservation concern, which will help evaluate the implications of recovery-action antagonism in the context of global conservation efforts.
Given that fewer than half of species-to-species relationships had antagonistic recovery actions, antagonism is not an inevitable phenomenon, but rather the result of specific interactions between recovery actions and/or life T A B L E 5 Distance-based redundancy analysis models found to be explanatory of Similar, Complementary, and Antagonistic speciesto-species recovery-action interactions for species at risk in Rouge National Urban Park (RNUP).histories present between some, but not all, species.This is important because it suggests there may be opportunities to predict and, therefore, plan and manage recovery-action antagonism.Part C of the methodology describes a process for testing the correlation between predetermined "explanatory" variables and recoveryaction interactions, thus enabling the identification of factors that may be predictive of interaction types.Our application of the db-RDA builds upon a body of literature critiquing the SARA process (e.g., Bolliger et al., 2020;Buxton et al., 2022;Pawluk et al., 2019;Turcotte et al., 2021), including a paper that used similar statistical methods to test the relationship between threats to species and the ambition of recovery strategies (McCune et al., 2013).The use of db-RDA helped to corroborate ecological hypotheses regarding broad causes of recovery-action interactions.For example, the variable "Habitat Specialist" was significantly correlated with the antagonistic distance matrix (only for RNUP).This variable was a binomial that was developed by distinguishing species that occupied three or fewer habitats (specialist) and those that occupied greater than three habitats (generalist).Habitat specialists were more susceptible to antagonisms that occurred when habitat creation, restoration, and management measures risked shifting existing habitat away from a specialist's preferred state.An example of such an antagonism is between Yellow-breasted Chat (Icteria virens) and Bobolink (Dolichonyx oryzivorus) in RNUP, which are, respectively, obligate successional shrubland and grassland species.Several management measures proposed for Bobolink aim to maintain grassland habitat by suppressing the growth successional vegetation, thus precluding the growth of shrubs conducive to Yellow-breasted Chat.The same dynamic also worked in reverse, as measures encouraging the growth of shrubland for Yellow-breasted Chat risk the loss of grassland habitat suitable for Bobolink.

Model title
Existing literature has noted that habitat specialists are generally at greater risk of extinction compared with generalists because their habitat requirements are more easily lost (Bergamini et al., 2009;Segura et al., 2007).This result from the db-RDA, combined with our own interpretation of the data, led us to conclude that this same overarching factor may drive specialists' susceptibility to recoveryaction antagonism.Further research will be needed to corroborate whether such findings (e.g., concerning habit specialists) represent salient dynamics for antagonistic recovery actions, or whether they are phenomena unique to the RNUP and PI locations.More broadly, by applying this methodology to other case locations, the factors that result in conservation antagonisms may become clearer, enabling conservation practitioners to better predict and avoid conservation antagonism when developing policy or managing different environments.
The antagonisms identified between Yellow-breasted Chat and Bobolink also illustrate the subtlety inherent to the designation of recovery-action antagonisms.There are conceivable scenarios in which implementing one of these species' recovery actions would result in consequential habitat loss for the other, but also plausible cases in which both shrubland and grassland could be simultaneously managed to ensure adequate available habitat for both species.The risk of this antagonism leading to adverse effects is heightened when available habitat is lacking, which is generally the case in southern Ontario (Chu et al., 2015;Deguise & Kerr, 2006).In contrast, larger intact habitat would help to mitigate this antagonism's impact (Soulè et al., 2003).We stress that, when both antagonistic and nonantagonistic outcomes were conceivable, our approach erred towards designating actions as potentially antagonistic (e.g., Figure 2), which sometimes involved simplifying assumptions.

| Environmental management
The methodology may have practical applications for environmental management.Specifically, we believe the method may have utility for: (1) narrowing a list of available recovery actions by prioritizing highly complementary/similar actions to facilitate ecosystem recovery, while proactively identifying potential antagonistic actions to ensure they are avoided; (2) informing the development of new recovery strategies, in particular multispecies or ecosystem-level plans; and (3) retroactively assessing the performance of conservation programs to identify areas for improvement or inform the planning/management of conservation resources.Some refinement to the methodology as described in this article may be required to address location-specific recovery-planning regimes, study areas, and species assemblages.
Environmental managers can use the methodology to select, out of a list of single-species recovery actions, measures that should move forward to implementation.The prioritization of recovery actions that benefit multiple species (i.e., actions designated as similar and complementary) and avoidance of potentially antagonistic actions will enhance the cost-effectiveness of conservation investments (Chadés et al., 2014).
In Canada, there are existing processes that consider the implications of recovery actions, including strategic environmental assessments (SEA) conducted during the development of recovery strategies, and analyses conducted as part of the permitting or funding of certain major conservation interventions (Cumming & Tavares, 2022).However, to our knowledge, these procedures do not offer a standardized analysis of all potential interactions between relevant species' recovery objectives and, therefore, may be insufficient to properly manage all antagonisms.Furthermore, our methodology could be used to inform analyses conducted as part of SEA or other assessment processes or could serve as a standalone risk analysis in jurisdictions and contexts in which existing controls are lacking.
There is an ongoing debate as to whether single, multispecies, or ecosystem-level recovery planning is best equipped to direct biodiversity conservation (Clark & Harvey, 2002;Rejis Baptista et al., 2019).The specific issue of recovery-action antagonism would be most effectively addressed through multispecies, and preferably ecosystem level, recovery planning.Our methodology can be used to guide the design of multispecies and ecosystem recovery plans that maximize cost-effectiveness and balance the needs of several at-risk species inhabiting a given space; specifically, by developing actions that avoid antagonisms to the extent possible and benefit the maximum number of species (i.e., prioritizing actions frequently classified as similar and/or complementary) (Chadés et al., 2014;Tulloch et al., 2016).
The notable difference in the extent of antagonism detected between RNUP and PI shows that the procedure is sensitive enough to yield distinct results between different case-study locations, even for relatively proximal areas with similar species assemblages, as is the case between PI and RNUP.This finding underscores the utility of the methodology for assessing the general degree of antagonism and other interaction types in an area.Such insights could have practical implications for evaluating the cost-effectiveness of existing conservation programs, or for informing regional budget allocation, given that areas with higher degrees of conservation antagonism are likely to require more specialized environmental management practices with higher funding requirements.

| Assumptions and limitations
Key assumptions and limitations should be carefully considered when applying or interpreting the result of the methodology.When designating antagonisms, we assumed that nontarget species were not thoroughly considered by practitioners seeking to implement recovery actions, unless explicitly specified in recovery documents.We recognize that this assumption often deviates from reality, as informed conservation management practices can mitigate or eliminate potential antagonisms (Dunwiddie et al., 2016).However, following the precautionary approach, we applied this assumption to all recovery actions to identify all potential antagonisms in the study areas.
A limitation of the methodology is the time required to assess all recovery-action interactions in a study area.The scale of analysis required to properly apply the method may be time consuming in certain contexts.In practice, users may opt to restrict the number and type of species and recovery actions evaluated, depending on the objectives of the analysis.In our study, we elected to exclude species that were not designated as Threatened or Endangered under SARA (i.e., those not legally requiring recovery strategies) or that were exclusively aquatic during all life stages.This approach aligned with our objective to focus on interactions between federal recovery actions for terrestrial at-risk species.However, it may have also led to certain antagonisms being overlooked.In future research, it may be particularly interesting to explore potential antagonisms between terrestrial and aquatic species.As the recovery of aquatic and terrestrial species is often managed by separate organizations, antagonisms between these groups may be more likely to be overlooked during the development and implementation of recovery actions.
Another key limitation of this study is its vulnerability to assessor bias.Data collection and the pairwise evaluation of recovery actions relied on qualitative assessments of recovery strategies.Such methods are susceptible to assessor biases, including differential interpretations of the language used in recovery strategies and personal knowledge of the species evaluated.We mitigated this limitation through standardized inclusion criteria (e.g., Appendix S1), a decision tree (Figure 2), specific definitions for each interaction type, and the use of a Delphi-type approach to test and enhance standardization between assessors.Future research can enhance the replicability of the process by employing robust Delphi-type approaches among a larger group of assessors.
Finally, it should be noted that this methodology focuses on antagonisms that are reasonably foreseeable due to the stated methods or objectives of recovery actions being in conflict with one another.This approach does not address several classes of potential conservation antagonism, including antagonisms resulting from unintended or indirect (e.g., second or third order) effects of recovery actions.As an example, it is conceivable that measures to reduce road mortality for an endangered reptile might also benefit local predators, such as feral cats, which could have negative ramifications for at-risk passerine species.Yet, this potential effect would not be recorded as an antagonism between the reptile and a passerine species because reducing mortality for local predators was not a stated or intended outcome of the road-mortality action.
Furthermore, antagonisms not directly encapsulated in an action-to-action conflict could also be overlooked.For example, an endangered amphibian might be harmed or killed by actions that propose the use of heavy machinery; however, if the amphibian's recovery strategy did not include actions directed at reducing mortality (e.g., if the recovery strategy instead focused on preventing habitat loss), there might be no actions deemed to directly conflict with the heavy-machinery measure.Overall, the method is a tool to screen recovery actions and identify predictable antagonisms; however, given the complexity of ecological dynamics, it cannot predict all potential negative consequences of recovery actions.

| Conclusions
Global assessments suggest that current conservation efforts are failing to address the biodiversity crisis (Ceballos et al., 2017;Wake & Vredenburg, 2008).Recovery-action antagonism can inadvertently harm nontarget species and poor recovery action prioritization can waste limited resources available for conservation.The results of two case studies performed in RNUP and PI showed that potential recovery-action antagonisms were present at non-negligible levels, suggesting that the issue may warrant further attention.The methodology presented in this article is a practical approach to identifying recovery-action interactions.It operates at a granular scale, enabling the identification of specific interactions between individual species; yet, it is also easily scalable to provide broader perspectives on the compatibility of multiple species' recovery needs within a given study area.Two case studies establish that this methodology can lead to interesting and useful insights for local conservation planning.Over time, wider application of the method may uncover broader trends that can inform and improve species recovery planning on the global stage.

F
Percentage of recovery action pairs designated as each interaction type for Rouge National Urban Park and Pelee Island.I G U R E 3 Plot of principal coordinates analysis (PCoA) of the Complementary (B) species-to-species multivariate distance matrix overlaid with habitat polygons for the Rouge National Urban Park case study.See Appendix S8 for key to species acronym.
In this example, we have populated the table based on the sample data from Table1for the Antagonistic ("D") interaction.Example of how data on explanatory variables was compiled for subsequent analysis (see step 9).
T A B L E 2 Example of a species-to-species comparison matrix containing the proportion of recovery action pairs for each recovery-action interaction type relative to the total number of comparisons between the two species.Note:Note: Appendix S6 provides the full list of explanatory variables that were assessed, as well as approaches for explanatory variable data collection.
Results for the Pelee Island case study are not shown.Results for Neutral and Unknown interactions are not shown.The presented models were developed based on Variance Inflation Factors (VIF), Akaike's Information Criteria (AIC), Adjusted R 2 , and expert knowledge of the dataset.The p-values are based on 999 permutations.Additional information on model development and assessment procedures is available in Appendix S7.Abbreviations: IUCN, International Union for Conservation of Nature; SARA, Species at Risk Act.