Trait‐based vulnerability reveals hotspots of potential impact for a global marine invader

Abstract

Predation from the invasive Indo‐Pacific lionfish is likely to amplify declines in marine fishes observed in multiple ocean basins. As the invasion intensifies and expands, there is an urgent need to identify species that are most at risk for extirpation—and possible extinction—from this added threat. To address this gap and inform conservation plans, we develop and apply a quantitative framework for classifying the relative vulnerability of fishes based on morphological and behavioural traits known to influence susceptibility to lionfish predation (e.g. body shape, water column position and aggregation behaviour), habitat overlap with lionfish, and degree of geographic range restriction. Applying the framework to fishes across the invaded Caribbean Sea and ahead of the invasion front in the southwestern Atlantic revealed the identity of at least 77 fishes with relatively small ranges that are likely to be most affected by lionfish predation. Trait‐based vulnerability scores significantly predict the probability of fishes appearing within the diets of lionfish across the invaded region. Spatial richness analyses reveal hotspots of vulnerable species in the Bahamas, Belize and Curaçao. Crucially, our framework identifies 29 vulnerable fishes endemic to Brazil, which has not yet been colonized by lionfish. Of these, we suggest reefs around offshore island groups occupied by a dozen highly vulnerable and range‐restricted species as priorities for intervention should lionfish spread to the region. Observations of the rate of lionfish spread across the invaded range suggest that an average of 5 years (with a median of nearly 2 years) elapses from first sighting to maximum observed densities. This lag may allow managers to mobilize plans to suppress lionfish ahead of an invasion front in priority locations. Our framework also provides a method for assessing the relative vulnerability of cryptobenthic and/or deep‐reef fishes, for which population‐monitoring data are limited.

We classified Western Atlantic fish species' vulnerabilities based on range restriction and seven traits that influence susceptibility to lionfish predation to identify species most at risk of extirpation due to invasive Indo‐Pacific lionfish. We identified 77 potential at‐risk fishes with relatively small ranges and highly vulnerable traits. Across the Caribbean, an average of 5 years (with a median of nearly 2 years) elapsed between first lionfish reports and maximum lionfish densities, indicating a possible lag during which lionfish suppression could be prioritized ahead of an invasion front, such as islands offshore of Brazil where a dozen highly vulnerable endemics occur.

1. INTRODUCTION

As shifts in range and abundance of species advance at an unprecedented pace in response to global change processes, there is a need to develop systematic frameworks that predict changes and inform targeted conservation action (Côté et al., 2016). Invasions are a major driver of global change. Substantial, large‐scale ecological and economic impacts have been documented following rapid range expansion and abundance increase by exotic species in a range of ecosystems (Pimentel et al., 2005; Simberloff et al., 2013). Invasive species are also a leading driver of extinctions in terrestrial ecosystems (Duenas et al., 2021). A review of 133 local and global marine extinctions by Dulvy et al. (2003) reveals interaction with an invasive species was listed as a primary cause for five of these events. Despite pervasive threats to species in both marine and terrestrial environments, the intrinsic difficulties associated with sampling marine habitats is thought to contribute to a lag in documenting marine extinctions compared to the well‐documented losses on land (McGeoch et al., 2010; Roberts & Hawkins, 1999; Seebens et al., 2017; Webb & Mindel, 2015), suggesting the number of events could be higher than currently known. While relatively fewer marine extinctions have been attributed to invasive species, extinction events may increase as the numbers of invasions (and their associated impacts) continue to climb (Grosholz, 2002). For example, interactions between the invasive Northern Pacific seastar (Asterias amurensis) and the Critically Endangered spotted handfish (Brachionichthys hirsutus), which is facing several other serious threats and shares traits and overlapping distribution with the Extinct smooth handfish (Sympterichthys unipennis; Stuart‐Smith et al., 2020), may push the species below minimum viable population sizes.

Natural resource managers and conservation practitioners are often playing catch up when it comes to invasive species, addressing invasion effects only after they have been detected in recipient ecosystems. Truly mitigating anticipated ecological and economic impacts of an invasion before they occur requires knowledge products and frameworks that identify the types of effects most likely to manifest and facilitate predicting where these effects might be most intense. Direct observations of impacts to rare and/or cryptic taxa are generally lacking due to sampling challenges, which causes lesser‐known species to be overlooked during conservation decision‐making. Predictive frameworks that examine the vulnerability of understudied species based on comprehensive species‐level traits data from similar and well‐studied fauna can act as a proxy option to estimate threat impact (e.g. Dulvy et al., 2003). For example, trait‐based approaches have been used to predict the vulnerability of marine species caught as bycatch in fisheries (Stobutzki et al., 2001), to climate change (Hare et al., 2016) and to petrochemical exposure (Polidoro et al., 2021). Threats from invasive species are expected to continue to increase in both terrestrial and aquatic systems, but studies applying trait‐based approaches to predict the outcome of interactions between invaders and native biodiversity are limited (e.g. Soto‐Shoender et al., 2020). Given that predation is a main mechanism of invasion‐mediated extinction (Sax & Gaines, 2008), understanding the traits of species most vulnerable to the effects of invading predators may help to guide interventions that substantially reduce extinction risk.

Here we describe a predictive framework for forecasting invasion effects based on morphological and behavioural traits known to mediate the strength of trophic interactions between invasive and native species, and geographic range information. We apply this approach to forecast hotspots of ecological impact, in terms of increasing extinction risk for native species, within the invaded range and ahead of the invasion front for the Indo‐Pacific lionfish (Pterois volitans and P. miles), a globally invasive marine predator.

The invasion of predatory Indo‐Pacific lionfish into highly diverse and threatened Caribbean fish communities has garnered much attention by natural resource managers, who seek to mitigate predation effects on coastal biodiversity through local culling programs that suppress lionfish abundance (Morris, 2012). The invasion in the western Atlantic covers the entirety of the Gulf of Mexico and Caribbean Sea as well as northward to Bermuda and the U.S. east coast (Schofield et al., 2014). As for other broadly distributed non‐native species, the current geographic extent of the invasion precludes complete eradication, and is predicted to spread into additional regions of the Atlantic, Pacific and within the Mediterranean (Bariche et al., 2017; Evangelista et al., 2016; Ferreira et al., 2015). An extinction event driven by the invasive lionfish threat remains a possibility (Côté & Smith, 2018), and compiling available data on invasion history can be used to predict future impacts and prevent worst case scenarios (Kulhanek et al., 2011). Managers within the invaded range and ahead of the invasion front must therefore identify priority locations at which to expend limited resources for monitoring and mitigation (Green & Grosholz, 2021). But where might impacts from the invasion, most likely through predation on naive species, be most intense?

Wide‐scale reef decline has already impacted the same fishes that are now also consumed by invasive lionfish in the Atlantic (Alvarez‐Filip et al., 2015; Paddack et al., 2009), and the combination of these threats could reduce species' ability to recover, leading to localized extirpation or even extinction. To date, over 200 Atlantic fish species have been identified within the diets of invasive lionfish (Acero et al., 2019; Peake et al., 2018), which use a stalking hunting strategy that is unique among predators in the region (Côté & Maljković, 2010). Although few diet studies have focused on deep‐sea habitats (50–300 m depth), lionfish are known to consume many poorly studied mesophotic fishes that occur there, suggesting that negative impacts in these environments are likely underestimated (Tornabene & Baldwin, 2017). Due to regional differences in fish fauna across the Atlantic basin, the list of prey species is likely to grow as more areas are sampled. However, trait‐based analyses of lionfish prey selection reveal that within this broad species list, lionfish preferentially feed on small, solitary, narrow‐bodied fishes that occur on or just above reef habitats (Green & Côté, 2014; Green et al., 2019). Systematically examining the ecological and morphological traits of fish species in the region to gauge their susceptibility to lionfish could therefore inform conservation action.

Incorporating information on geographic range is also key for examining the relative vulnerability of species to invasion impacts. In particular, range‐restricted species are already at an elevated extinction risk and to prevent biodiversity loss in the face of major threats such as biological invasion, these species could be considered conservation priorities (Mace et al., 2008). Small‐bodied reef fishes like those consumed by lionfish are often range‐restricted due to limited dispersal and thus relatively low connectivity with neighbouring populations, which can reduce the probability of population replenishment following localized extirpation events (Beltrán et al., 2017; Taylor & Hellberg, 2003). Small‐bodied fishes are also key to reef ecosystem function (Bellwood et al., 2012; Brandl et al., 2019; Hawkins et al., 2000), highlighting the need to understand how fish species’ vulnerability to predation by lionfish and geographic range size intersect to affect risk of extirpation and ultimately ecological services on reefs.

Here we classify the vulnerability of all known Greater Caribbean bony fish fauna and endemic Brazilian bony fishes to predation by invasive lionfish based on traits that influence the predation process. We combine this trait classification with information on species' geographic ranges and habitat associations to identify: (1) native fish species that are most likely to be at risk of extirpation or extinction as a result of the invasion, and (2) regions containing the greatest concentration of vulnerable species within the invaded Caribbean Basin and beyond the invasion front along the coast of Brazil in the Southwestern Atlantic. We then ask how well vulnerability assessed via our trait‐based framework predicts observed lionfish diet composition from the invaded range. We also examine the rate at which increases in lionfish abundance have occurred across the region to illustrate how quickly the invasion proceeds and estimate how quickly a response must be mounted to implement control programs in highly vulnerable areas. Finally, we identify knowledge gaps regarding impacts and management of the invasion, focused particularly on vulnerable and understudied cryptic and deep‐water fish fauna across the region.

2. METHODS

2.1. Caribbean fish fauna inventory

To predict the vulnerability of fish species to lionfish predation, we compiled seven traits (Table 1) related to predation vulnerability for all marine bony fishes that occur in the Greater Caribbean between 0‐ and 300‐m depth, excluding mesopelagic species such as lanternfishes (Myctophiformes). The Greater Caribbean extends from Cape Hatteras, North Carolina in the U.S. to the border of French Guiana and Brazil, including Bermuda, the Gulf of Mexico and Caribbean Sea (Robertson & Cramer, 2014). We sourced the species from the Caribbean IUCN Red List initiative (Linardich et al., 2018; N = 1360) and the Smithsonian Tropical Research Institute's (STRI) database of Greater Caribbean shorefishes (Robertson & Van Tassell, 2015, N = 148), manually adding newly described species since 2014 (N = 19). The taxonomy of this inventory follows Eschmeyer's Catalog of Fishes (Fricke et al., 2020). We also sourced a list of 81 marine bony fishes endemic to Brazil (Pinheiro et al., 2018) from which to classify species' traits in a related region ahead of the invasion front.

TABLE 1.

Description of the seven traits used to score lionfish predation vulnerability

Biological trait	Description	High score (received value of ‘1’)	Low score (received value of ‘0’)
Habitat	Invasive lionfish have been documented in a wide variety of habitat types; however, the highest densities are associated with structurally complex hard habitats such as reefs a . We therefore considered reef‐associated species to be more vulnerable to lionfish predation vulnerability compared with species not associated with coral or rocky reefs at any time in their life history (i.e. inhabits soft bottoms, estuaries, seagrass beds and/or mangroves). Three species in Lucifuga are restricted to inland, anchialine caves and were considered to have low vulnerability	inhabits either rocky and/or coral reef	does not inhabit reefs or is restricted to inland, anchialine caves
Nocturnal behaviour	Lionfish hunt primarily during low‐light crepuscular periods b . Species that are most active at night (e.g. many cardinalfishes), and thus active during crepuscular times, are likely to be more vulnerable to predation than species with daytime activity patterns	known to be active at night	not known to be active at night
Obligate cleaning behaviour	Obligate cleaners in the Caribbean likely contain a chemical defence that causes lionfish to avoid consuming them c . All 19 goby species in the genus Elacatinus, which are mostly obligate cleaners, are considered to have a lower vulnerability	not in the genus Elacatinus	in the genus Elacatinus
Water column position	Lionfish primarily swim and hunt just above the substrate d . Demersal species, defined as living <2 m off the bottom, are considered to have a higher vulnerability compared with benthic species, defined as bottom burrowers (e.g. most eels, jawfishes and wormfishes), and pelagic species, defined as living >2 m off the bottom. In cases where species underwent ontogenetic shifts in water column position (e.g. juvenile barracuda are ‘demersal’, but adult barracuda are ‘pelagic’), we recorded it as ‘demersal’ because the species could be eaten by lionfish during the juvenile stage	demersal	benthic or pelagic
Aggregation behaviour	Schooling behaviour is an effective anti‐predator strategy against lionfish d , e . We therefore considered species that are solitary or exhibit shoaling behaviour to have a higher vulnerability than those that exhibit schooling behaviour	Solitary or shoaling	Schooling
Maximum length	Small fishes are more vulnerable to predation by lionfish d and diet studies reveal that lionfish typically do not consume prey items that are longer than 15 cm (~40% of an adult lionfish's total length) f . We therefore considered species with maximum length <15 cm TL to have a high vulnerability because individuals can be consumed over their entire life history (i.e. as both juveniles and adults) compared with species that grow larger than 15 cm TL	maximum body length less than 15 cm	maximum body length greater than 15 cm
Body shape	Deep‐bodied species are less vulnerable to predation by gape‐limited lionfish compared with slender species d . We considered species with a body shape ratio (length:height) >3 to have higher vulnerability than species with a ratio <3. Body height was measured as the linear distance from the dorsal ridge to the pelvic girdle except for dorsoventrally flattened species (e.g. flounders and batfishes), which was measured similarly, but as a width instead of height. Body length was measured as the distance from the tip of the mouth to end of the caudal fin, or total length	body shape ratio greater than 3 (shallow‐bodied)	body shape ratio less than 3 (deep‐bodied)

^a Hunt et al. (2019).

^b Green et al. (2011).

^c Tuttle (2017).

^d Green and Côté (2014).

^e Green et al. (2019).

^f Green et al. (2012).

2.2. Trait‐based predation vulnerability framework

The seven traits we used to predict vulnerability to lionfish predation include preferred habitat type, nocturnal behaviour, cleaning behaviour, position in the water column, aggregation behaviour, maximum body length and body shape (Table 1); characteristics that have been demonstrated to influence lionfish prey selection (Green & Côté, 2014; Green et al., 2019). For each species, we assigned each trait either a ‘1’ or ‘0’, with ‘1’ indicating that the species’ trait form is associated with increased vulnerability to predation, and ‘0’ indicating the form of the trait is associated with decreased vulnerability. Three species restricted to inland, anchialine caves (genus Lucifuga) and 85 pelagic species (e.g. tunas, billfishes, molas and flyingfishes) were assigned a vulnerability score of ‘0’ as they occupy habitats that are unlikely to be utilized by adult lionfish, and thus unlikely to be consumed.

For all other species, we summed vulnerability scores for the seven traits to obtain a cumulative vulnerability score for each species ranging from 0 to 7. Species with scores of 6 or 7 are predicted to have high vulnerability to predation by lionfish (i.e. all the known characteristics of preferred prey), scores of 4 or 5 have moderate vulnerability and scores of 1, 2 or 3 have low vulnerability. A straight sum was used because empirical data are not available to accurately measure the relative weight or importance of each trait among the particular suite of traits used here (which differs slightly from those considered in Green & Côté, 2014 and Green et al., 2019), and therefore, anything other than equal would be arbitrary.

Species‐specific data on habitat type, the presence of nocturnal behaviour, cleaning behaviour and aggregation behaviour, position in the water column, and maximum body length were sourced from IUCN Red List assessments, which are freely available for use in scientific analyses and/or conservation planning on iucnredlist.org. The body shape ratio, which is the body length divided by height, was estimated according to methodology by Green and Côté (2014). We measured these dimensions in a minimum of three photos of each species (primarily from the STRI database; Robertson & Van Tassell, 2015) using the free software ImageJ by tracing body length and height.

We estimated the geographic range size of each species using the distribution maps that accompany the Red List assessments. Due to the generalized style of these maps, which are drawn by a standardized methodology developed by the IUCN Marine Biodiversity Unit, the area within the distribution polygon is often an overestimate of the actual area occupied by the species. Most common Caribbean fishes are widely distributed in the region, and many also range southward to Brazil (Robertson & Cramer, 2014). A histogram of the range sizes of all Greater Caribbean endemic bony shorefishes (N = 674) with a bin size of 20,000 km² revealed that range size is strongly right skewed, with a natural break at 80,000 km² (Linardich, 2016). We added a 20,000 km² buffer to this natural break in range size, and considered a species to be ‘range‐restricted’ if the area of the distribution polygon was less than 100,000 km². Across all 815 endemics (53% of the fauna) in this study, the mean range size is 904,800 km²; 158 species (19%) have a range size of <80,000 km², 167 species' (20%) have ranges <100,000 km².

To evaluate the extent to which the identity of species observed in lionfish diets can be predicted by our trait‐based vulnerability framework, we first constructed a non‐metric multidimensional scaling (NMDS) ordination to visualize the difference in multivariate trait values between species with low, medium and high cumulative vulnerability scores. Specifically, we calculated Jaccard dissimilarity distance measures between each fish species in two dimensions, with seven traits in our dissimilarity matrix. We then grouped the distances for each taxa by low, medium and high cumulative vulnerability to visualize the difference between the three groups. Finally, we plotted the trait‐space location of the top five prey taxa reported in the diets of lionfish, in terms of numerical frequency (Peake et al., 2018), to illustrate their position within the three trait‐based vulnerability groups (low, medium or high).

Additionally, we tested the extent to which the presence of fish species consumed by lionfish could be explained by their cumulative trait‐based vulnerability score and abundance in the environment (a factor likely to influence probability of being consumed by chance alone; Manly et al., 2007). To do this, we combined data on the identities of fish species reported within diets of lionfish from six invaded regions of the Caribbean synthesized in Peake et al., 2018 with the average abundance of fish species documented on visual surveys of reef fishes conducted by the Reef Environmental Education Foundation (REEF) Fish Survey Project in the same locations and the same time periods as the lionfish diet assessments were conducted. During REEF surveys, trained scuba divers record the abundance of fish species, including lionfish, on a log scale, using codes of 1–4 for each species during a roving search of the dive site (1 = a single individual sighted, 2 = 2–10 individuals, 3 = 11–100 individuals and 4 = >100 individuals; REEF, 2018). Each fish species reported on REEF surveys within a focal region and time period was scored as a 0 (i.e. not identified in the corresponding diet study of lionfish) or 1 (i.e. positively identified in the corresponding diet study).

We then evaluated the probability of species being consumed (0 or 1) as a function of the species' vulnerability score and their relative abundance in that region using generalized mixed‐effects models. The models employed a binomial distribution, and involved two explanatory variables (i.e. fixed effects), cumulative trait‐based vulnerability score and average abundance, with region as the random effect on the intercept. We also assessed whether there was an interaction between the two explanatory variables (vulnerability score and abundance) to address the hypothesis that rarer species are more likely to be consumed when they have higher vulnerability scores. We fit all combinations of these variables (i.e. five models total; Table 2) and used Akaike information criterion (AIC; Akaike, 1974) to perform model selection. Since two models had similar AIC values (i.e. performed equally well), we used both models, and calculated model‐averaged predictions (Burnham & Anderson, 2004; Cade, 2015) with 95% confidence intervals. We performed our analysis in R using the lme4 package (Bates et al., 2007), in R version 4.0.1 (R Core Team, 2020). The R code for all statistical analyses presented in this paper is available in an open‐access repository (see Acknowledgements section).

TABLE 2.

Selection statistics for full model set

Abundance	Vulnerability score	Abundance × Vulnerability	Negative log likelihood	ΔAIC	AIC weights	Marginal R 2 value
+	+		−34.674	0.00	0.724	0.378
+	+	+	−34.548	1.94	0.275	0.392
+			−42.817	14.14	0.001	–
	+		−43.437	15.38	0.000	–
			−47.274	20.94	0.000	–

2.3. Identifying vulnerability hotspots

We combined data on species' cumulative vulnerability scores and geographic ranges to identify hotspots where the greatest number of fish taxa could be most impacted by predation from high densities of invasive lionfish, estimated as the number of taxa (i.e. species richness) in a given area with cumulative vulnerability scores of 6 or 7, and geographic ranges <100,000 km². We conducted richness analyses in ArcGIS 10.7 on two different scales; by 1 × 1 km² grid cell (fine‐scale), and also by Exclusive Economic Zone (EEZ) in order to relate hotspots to countries’ geographic boundaries for national decision‐making authority for coastal resource and conservation issues across the region. For fine‐scale richness, we converted each distribution polygon to a raster with 1 × 1 km² grid cells and summed all rasters for a given cell. For EEZ richness, we summed the number of species that occur in each EEZ. In addition to visualizing geographic richness for all range‐restricted and vulnerable Caribbean fishes together, we also compared the locations of hotspots for deep‐living (upper depth ranges deeper than 60 m) and shallow‐living (<60 m) range‐restricted and vulnerable species separately. To highlight hotspots of species’ vulnerability ahead of an invasion front, we conducted a fine‐scale richness analysis of endemic fishes that met these same criteria along coastal Brazil in the Southwestern Atlantic (i.e. cumulative vulnerability scores of 6 or 7, and geographic ranges <100,000 km²).

2.4. Estimating management intervention timelines

In designing management plans for invasive species, initial sightings of the invader within a jurisdiction is often a trigger for enhanced monitoring and control activities (Green & Grosholz, 2021). However, the rate at which abundance increases following initial reports limits the time management authorities have to mount adequate responses to mitigate ecological effects. To provide jurisdictions ahead of the invasion front with an estimate of potential rates of lionfish population increase, and thus the timeline over which proactive intervention plans might be implemented, we sourced data on when initial lionfish sighting and peak lionfish abundance occurred across the invaded region from underwater surveys conducted in the Caribbean through the Reef Environmental Education Foundation (REEF) Volunteer Fish Survey Project from 2000 to 2017 (REEF, 2018). Geographically referenced survey data for eight major regions, containing 37 subregions, are publicly available for use in research (REEF, 2018; Figure S1). We used the REEF dataset, and noted both (1) the date of the first sighted occurrence of lionfish in each subregion; (2) the maximum abundance sighted in each subregion on any single survey; and (3) the earliest date the maximum abundance was sighted. We excluded subregions (N = 27) if (1) lionfish had not yet been recorded on a REEF survey by the end of 2015 (yielding a minimum of 3 years in each time series); (2) lionfish abundance never reached a level of at least 3 (11–100 individuals) on any survey or (3) lionfish abundance never reached a level above the initial recorded level. From these data, we calculated the time between the dates of the survey reporting the first lionfish and the survey first reporting maximum abundance observed in each subregion (i.e. time to maximum abundance). We also estimated a proxy for population growth rate by calculating mean lionfish abundance on surveys in the temporal midpoint (from all surveys during that month) in the time series between first sighting and first maximum abundance. We then divided the temporal midpoint abundance by the length of the time interval between first sighting and the first maximum sighting. This value was taken as a proxy for population growth rate for both regions and subregions.

3. RESULTS

3.1. Vulnerability of Caribbean fishes

Our trait‐based analysis predicts that 26% (400 of 1527 species) of Caribbean bony fishes are highly vulnerable to predation by invasive lionfish, in that most are reef‐associated, small, narrow‐bodied, solitary and do not burrow within the substrate (Figure 1). Species with these trait forms represent 154 genera from 45 families, with eighty‐percent of taxa representing blennioids (122 species), gobioids (94), serranids (43), cusk eels/brotulas (25), clingfishes (24), cardinalfishes (23), basslets/grammas (17) and scorpionfishes (11). Of these fishes, 160 have restricted geographic ranges (<100,000 km²). Further filtering those with restricted geographic ranges by high vulnerability, allowed us to identify 77 taxa likely to face the greatest risk of extirpation in the event of dense lionfish populations invading their ranges, comprising 35 genera from 14 families (Table S1). Of these, over half are gobioids (26 species) or blennioids (20), with the remainder classified as basslets/grammas (8), brotulas (8), clingfishes (7), serranids (4) and single species each of wrasse, scorpionfish, anchovy and toadfish.

FIGURE 1.

Vulnerability of Caribbean fishes to invasive lionfish predation; (a) all species and (b) only species endemic to the Caribbean. Low vulnerability (scores of 0, 2, 3 or 4) is characterized by species with habitat association traits that make them inaccessible for predation and/or body plans that reduce encounter and capture success (e.g. the skipjack tuna [E] is epipelagic, the broadnose worm eel [F] burrows in soft sediment, the Orinoco sea catfish [K] occurs on soft bottoms in estuaries and the tropical flounder [L] is wide‐bodied). Species with moderate vulnerability (score of 5) are typically demersal and narrow‐bodied, but sometimes have a maximum length larger than 15 cm and/or are not reef‐associated (e.g. the western comb grouper [C] has a maximum length of 80 cm, the sand diver [D] has a maximum length of 45 cm, the butter hamlet [I] is deep‐bodied and the rainbow parrotfish [J] has a maximum length of 120 cm). High vulnerability (scores of 6 or 7) is characterized by reef‐association, small maximum length and narrow body (e.g. the striped cardinalfish [A]), the colon goby [B], the reef squirrelfish [G] and the candy basslet [H]). No species had a score of 1. Photos are not to scale and all were sourced from https://biogeodb.stri.si.edu. Photo credits: (A) Skipjack Tuna (Katsuwonus pelamis) by R. Frietas; (B) Broadnose Worm Eel (Myrophis platyrhynchus) by J. Van Tassell and D. Robertson; (C) Western Comb Grouper (Mycteroperca acutirostris) by J. Van Tassell and D. Robertson; (D) Sand Diver (Synodus intermedius) by J. Van Tassell and D. Robertson; (E) Striped Cardinalfish (Apogon robbyi) by J. Van Tassell and D. Robertson; (F) Colon Goby (Coryphopterus dicrus) by D. Robertson; (G) Orinoco Sea Catfish (Cathorops nuchalis) by J. Van Tassell and D. Robertson; (H) Tropical Flounder (Paralichthys tropicus) by J. Van Tassell and D. Robertson; I) Butter Hamlet (Hypoplectrus unicolor) by K. Bryant; (J) Rainbow Parrotfish (Scarus guacamaia) by K. Bryant; (K) Reef Squirrelfish (Sargocentron coruscum) by J. Williams; (L) Candy Basslet (Liopropoma carmabi) by J. Van Tassell and D. Robertson [Colour figure can be viewed at wileyonlinelibrary.com]

At least 222 Caribbean fishes (~15% of fish fauna) have been reported from one or more of 24 lionfish diet studies conducted in ~16 different localities across the invaded region (Table S2). Importantly, the top prey species ingested by lionfish in diet analyses from five subregions of the Caribbean (Peake et al., 2018) are classified as high (N = 4) or high/medium (N = 1) vulnerability levels through our framework, the trait space which occupies the least amount of multivariate trait space (Figure 2a). However, of the 400 species we identify as having high predation vulnerability, only 21% (84 species) have been reported within these in diet studies. Moreover, only two of the 77 species with both restricted geographic ranges and high predation vulnerability have been directly observed in the diet: the social wrasse (Halichoeres socialis) and the Arawak blenny (Emblemariopsis arawak).

FIGURE 2.

(a) A non‐metric multidimensional scaling (NMDS) ordination showing the multivariate trait space occupied by Caribbean fish species, grouped by their vulnerability to predation by invasive lionfish. The trait space was constructed from seven morphological and behavioural traits that confer vulnerability to predation by invasive lionfish (Table 1). Labelled species are the top fish species identified within lionfish diet studies from five subregions in the invaded Caribbean. (b) Model‐averaged predictions for whether or not a species will be consumed as a probability, at each vulnerability score and abundance. Light coloured ribbons represent the 95% confidence intervals for each estimated curve [Colour figure can be viewed at wileyonlinelibrary.com]

Our quantitative analyses of lionfish diet composition reveal that both trait‐based vulnerability and abundance in the environment significantly influence the probability of species being consumed by lionfish, and that there exists a strong interaction between vulnerability and abundance such that low‐abundance species are significantly more likely to be to be present in the diet when they have high vulnerability scores (Table 2; Figure 2b). Two top models emerged (Table 2); model‐averaged predictions of the probability of consumption for species at various abundances and vulnerabilities (Figure 2b). Our analysis indicates that vulnerable species were more likely to be consumed by lionfish at all abundances than those with lower vulnerability scores (Figure 2b). While the lower vulnerability scores (0–3) show an exponential‐like relationship in predicted likelihood of being consumed, even species with the highest level of abundance still are not predicted to be consumed at a probability higher than 0.5. Higher vulnerability species however (4+) have a distinctively different pattern, where species with high vulnerability increase in probability of being consumed at much lower abundances. Indeed, even at very low abundances, predicted consumption of species with the highest vulnerability score (7) are almost one‐fourth.

3.2. Hotspots of potential invasion impact

Our analysis reveals two major hotspots of range‐restricted fishes that are highly vulnerable to lionfish predation in the Caribbean basin; areas of coastal habitat near Belize (12 species) and Curaçao (16 species; Figure 3b; Table S3). While these two locations are hotspots for range‐restricted fishes in general (Figure 3a), our analyses highlight that at least 49% and 85%, respectively, of their endemic fish fauna are likely to be at heightened risk of extirpation or extinction if subject to high levels of predation by invasive lionfish. Interestingly, most of the Curaçao species (13) occur only at deep depths (63–290 m; see Figure 4 for examples), while the majority of Belizean species (10) inhabit shallow depths (0–40 m), with only two vulnerable, range‐restricted species occupying reefs at depths between 180 and 290 m. In fact, separating our hotspot analysis by depth reveals that over half (13) of the 25 vulnerable and range‐restricted species with upper depth ranges deeper than 60 m (i.e. deep‐living species) occur in Curaçao, followed by a secondary hot spot of seven species occurring in San Salvador Island in the Bahamas, presumably due to deep ROV sampling effort in those areas (Figure S2). Visualizing the distributions of the 52 shallow‐living (<60 m) vulnerable and range‐restricted species revealed a primary hotspot in Belize (10 species) followed by five species each in Grand Cayman Island, several nearshore localities off Panamá and Isla de Margarita in Venezuela (Figure S2).

FIGURE 3.

Density of Caribbean fishes with (a) range restricted (i.e. <100,000 km²) and (b) range restricted and high vulnerability to predation by invasive lionfish (i.e. vulnerability score = 6 or 7) per 1 km² grid cell. These same species sets are grouped by country Exclusive Economic Zone (EEZ) in (c) and (d). Dark red represents the areas with the highest numbers of species [Colour figure can be viewed at wileyonlinelibrary.com]

FIGURE 4.

Six of the 25 lionfish vulnerable and range restricted Caribbean deep‐water (i.e. >60 m depth) reef species. All photos were sourced from https://biogeodb.stri.si.edu. Baseline population data are limited for most of these species, meaning that detecting effects of stressors such as predation from invasive lionfish is challenging. Photo credits: (A) Decorated Split‐fin Goby (Varicus decorum) by C. Baldwin and D. Robertson; (B) Yellow‐spotted Basslet (Liopropoma olneyi) by C. Baldwin and D. Robertson; (C) Godzilla Goby (Varicus lacerta) by C. Baldwin; (D) Yellow‐spotted Sand Goby (Coryphopterus curasub) by C. Baldwin and D. Robertson; (E) Tail‐spot Basslet (Liopropoma santi) by L. Tornabene and D. Robertson; (F) Blue‐spotted Basslet (Lipogramma barrettorum) by C. Baldwin and D. Robertson [Colour figure can be viewed at wileyonlinelibrary.com]

Grouping occurrences by EEZ (rather than by fine‐scale [i.e. 1 × 1 km²] geographic regions within countries) reveals that Mexico, Curaçao, Belize, the Bahamas and Honduras, harbour the highest numbers of range‐restricted species overall (i.e. 20–27 species per jurisdiction; Figure 3c). Venezuela, the United States and Colombia also have high numbers (i.e. >15 species per jurisdiction) compared to elsewhere in the Caribbean (Figure 3c). The highest numbers of range‐restricted and vulnerable species occur in similar jurisdictions: Curaçao, Belize and the Bahamas (i.e. 11–16 species; Figure 3d).

Looking ahead of the invasion front in the Southwestern Atlantic along South America, 36% of endemic fishes (i.e. 29 of 81 species) found only along the coast of Brazil are likely to be highly vulnerable to predation and occupy restricted geographic ranges. Coastal regions that harbour the greatest number of vulnerable, range‐restricted species are in the states of São Paulo and Santa Catarina (nine species each) followed by the Fernando de Noronha Archipelago and Rocas Atoll (eight species; Figure 5). Twelve of these 29 species are restricted to coastal reefs that surround only one of the three small, offshore, oceanic island groups: six species in the Fernando de Noronha Archipelago and Rocas Atoll, four species in the Trindade–Martin Vaz insular complex and two species in St. Peter and St. Paul's Rocks.

FIGURE 5.

Density of range restricted Brazil endemic fishes predicted to be highly vulnerable to predation by invasive lionfish, by 1 km² grid cell. (A)–(C) 39% of these species, such as those in (A)–(C), are known only from reefs around offshore oceanic islands. Photo credits: (A) O. Luiz, Jr.; (B, C) R. Macieira [Colour figure can be viewed at wileyonlinelibrary.com]

3.3. Estimating management intervention timelines

According to the United States Geological Survey Nonindigenous Aquatic Species database, the first lionfish record occurred in 1985 off southeastern Florida, but the invasion did not start until 2000 (15 years later), when it spread north along the U.S. coast and Bermuda. From 2004 to 2007, the invasion expanded to the Bahamas and then throughout the Caribbean basin before reaching the southwestern Gulf of Mexico in 2011/2012. Across the 29 subregions of the Greater Caribbean covered by REEF surveys, lionfish were first recorded by divers in Bermuda in 2001 and the most recent first sighting occurred in the Gulf of Mexico off the northern Yucatán Peninsula in 2017 (Figure S4). Overall, most first sightings occurred in the years between 2009 and 2014 (Figure 6; Figure S4). While generally correlated in time, subregions in close proximity to already invaded subregions were not necessarily quickly invaded themselves. In the northwest Caribbean region, which includes Cuba, the Cayman Islands and Jamaica and the Mexican Caribbean to Nicaragua, almost 5 years elapsed between the initial invasion and the invasion of all subregions within it. Similarly, in Florida, 7 years separates the initial invasion of the region from the complete invasion of all subregions.

FIGURE 6.

Time lapse video of the average abundance of lionfish by subregion for the years 2001–2017 in the tropical western Atlantic according to Reef Environmental Education Foundation (REEF) data. The animation begins when the first lionfish was recorded in the REEF database, which occurred in Bermuda in the year 2001. Geographic text labels are for each primary region (1–8). Subregions with hatching indicate no REEF surveys were recorded during that year. Subregions with no color indicate no lionfish were recorded in REEF surveys during that year. Abundance increases with increasing darkness of the red color [Colour figure can be viewed at wileyonlinelibrary.com]

Seven subregions never recorded lionfish abundance greater than ‘2’ (2–10 sighted per survey), 21 reached a maximum of ‘3’ (11–100 sighted) and five had ‘4’ (>100 sighted) as the maximum binned abundance. The duration of time to maximum population (TMP) varied from 7.3 months to 11.8 years with an average of 3 years. For subregions where a level of ‘4’ was the maximum abundance, TMP spanned a mean of 5.1 years, a minimum of 1.9 years and a maximum of 9.7 years. For the 10 subregions where a midpoint was available, only three out of 10 subregions showed a relatively linear increase in abundance through time. The majority showed some manner of lag time where, after initial abundance sighting, the abundance of lionfish remained low (single individuals), for anywhere between 5 months to almost 3 years (Figure 7; Table S4). The minimum (3 days), maximum (11 years, 10 months) and median (1 year, 10 months) elapsed times are to be interpreted with the caveat that survey efforts were not uniform across the subregions in frequency or intensity, so it is expected that, in some cases, lionfish were likely in a region before being recorded in the REEF surveys. Generally, TMP was highly variable, but rarely was less than a full year.

FIGURE 7.

Invasion timeline and population growth rate estimates for Indo‐Pacific lionfish across subregions of the Caribbean Basin as recorded by the Reef Environmental Education Foundation's Volunteer Fish Survey Project. (a) Time between the date on which lionfish were first reported on surveys in the subregions (n = 26) and the first date the highest abundance was recorded ranges between 0 and 12 years. (b) Example time series of yearly population estimates for subregions representing different invasion timelines in panel (A). Mean abundance per year is plotted as a dashed line while 95% confidence interval forms the shaded region bounded by the solid lines [Colour figure can be viewed at wileyonlinelibrary.com]

4. DISCUSSION

4.1. Trait‐based vulnerability to invasion impacts

This study brings awareness to potential biodiversity loss by predicting vulnerable, range‐restricted prey species of invasive lionfish in the Western Atlantic Ocean. Using a trait‐based approach, we identified 400 Caribbean fishes that are likely to be vulnerable to predation by invasive lionfish, approximately twice the number of species recorded across 24 diet studies from the region thus far (222 species). Diet studies are useful for identifying common, widespread prey items and understanding general patterns of predator trophic ecology. However, given that most fishes engage in brief periods of foraging marked by low capture success and rapidly digest whole prey within highly acidic guts, the frequency and spatial scale of sampling possible for most diet studies means they are unlikely to detect rare or range‐restricted taxa. The largest synthesis of lionfish diet studies to date identified 128 fish species from over 8000 stomachs sampled at 10 locations scattered across the Caribbean (Peake et al., 2018). Out of the nearly 3000 prey items examined from stomachs taken in Belize, the social wrasse (H. socialis), a Belizean endemic that represents a high proportion of the local lionfish diet (Rocha et al., 2015), was not detected by Peake et al. (2018). Since the stomachs from Belize were only collected from areas outside the range of the social wrasse (Peake et al., 2018), its exclusion was presumably due to undersampling. Field studies are needed in areas where these species occur to better understand the impact that lionfish may cause.

Importantly, our trait‐based approach to predicting predation vulnerability identified 77 range‐restricted fishes, only two of which have been recorded in diet studies, but exhibit characteristics of preferred lionfish prey. This includes four clingfishes, six brotulas, one anchovy and one toadfish, which are members of families that have no species represented in diet studies (Gobiesocidae, Bythitidae, Dinematichthyidae, Engraulidae and Batrachoididae). Except for the anchovy, these 12 species are highly cryptic, which may cause predation events to be rare. Alternatively, detection of these small‐bodied fishes in diet studies may be lower due to high digestion rate and/or difficulty in identification (Beukers‐Stewart & Jones, 2004). In general, cryptic and diminutive species are infrequently studied. There are 13 cryptobenthic Caribbean reef fish families (Brandl et al., 2018), and 69%, or 276 of the species with high vulnerability to predation are members of these families and 19% of these (52 species) have been recorded in diet studies. If cryptic behaviour does reduce the vulnerability of a species to lionfish predation, then it should be considered in future trait‐based approaches. The impact of lionfish on cryptobenthic fishes, many of which have traits that predispose them to predation, is also relatively unknown. The critical role that cryptobenthic species fill in reef ecosystem functioning has only recently been measured (Brandl et al., 2019), and in light of myriad pervasive threats, there is an urgent need for increasing studies of these species (Baldwin et al., 2018; Smith‐Vaniz et al., 2006; Tornabene & Baldwin, 2017). In addition to interactions with invasive lionfish, Greater Caribbean shorefishes assessed as threatened on the IUCN Red List (indicating an elevated estimated extinction risk) face pervasive overexploitation and habitat degradation (especially of coral reefs and estuaries; Linardich et al., 2018). Unless threat mitigation is implemented, the intensification of multiple stressors in marine systems is expected to amplify loss of species across the Caribbean (Alvarez‐Filip et al., 2015), in Brazilian waters (Pinheiro et al., 2015) and on a global level (McCauley et al., 2015). Further examination of the intersection between biological and ecological traits of species that impart vulnerability to predation by lionfish with these other stressors may reveal additional species‐ and location‐specific priorities for intensive management intervention.

Optimal foraging theory predicts that the proportion of any given prey species in the diet is a function of encounter rate (i.e. prey abundance) and rate of successful attack and capture, with growing empirical evidence that traits of potential prey items will strongly influence the rates of these two final steps in the predation process. Our quantitative analyses of lionfish diet composition reveal that both trait‐based vulnerability scores and environmental abundance significantly influence the probability of species being consumed by lionfish. We also found a strong interaction between vulnerability and environmental abundance of potential prey species such that low‐abundance species are significantly more likely to be present in the diet when they have high vulnerability scores. Taken together, these patterns provide strong independent support for the utility of traits for identifying locations with high numbers of species that are vulnerable to this novel predator. Importantly, the trait‐based framework can be flexibly applied to risk assessments for newly discovered species, and more practically, for species that lionfish have yet to encounter (i.e. those on the south‐eastern coast of South America and in the western Mediterranean Basin).

While our trait‐based approach can estimate vulnerability to predation, further work is required to link relative probabilities of consumption with mortality rates and ultimately population dynamics. Intermediate analytical steps between predicting relative vulnerability to interactions with invasive species and constructing population dynamics models that incorporate trait‐based mortality from those interactions include network (e.g. food web) modelling, which aims to elucidate the strength of interaction with the novel species in the context of existing interactions (e.g. Eklöf et al., 2013). Trait‐based food web models have been used to predict predator–prey relationships in pelagic fishes facing impacts from species range shifts under global change scenarios (Gravel et al., 2013); we suggest that our traits framework could be extended to this approach as well. However, such an approach requires that abundance estimates and feeding relationships between species within a particular food web are well‐resolved.

4.2. Hotspots of potential invasion impact in the invaded Caribbean

The greatest concentration of range‐restricted species occur in Belize and Curaçao followed by the Bahamas and Honduras. Belize and the Honduran Bay Islands are known hotspots of microendemism for Caribbean marine species that may have risen from isolating oceanographic conditions in that region (Moran et al., 2019). Curaçao, however, is part of an offshore island chain of the Lesser Antilles within which there is a relatively high level of connectivity (Robertson & Cramer, 2014) and is not known to have a unique richness of marine species. As most Curaçao species captured within this richness analysis inhabit mesophotic depths only, Curaçao as a hotspot is a reflection of the higher sampling of deep reef biodiversity in that country (Figure S2). In fact, at least 13 papers have been published since 2013 describing many new mesophotic species collected during deep reef sampling in Curaçao, Bonaire, Honduras, St. Eustatius and Dominica (Baldwin et al., 2018; Tornabene & Baldwin, 2019). Most Caribbean fishes, including the endemics, are widely distributed (Robertson & Cramer, 2014) and much of the mesophotic Caribbean fauna may be similarly widespread (Baldwin et al., 2018). Until further sampling of mesophotic fishes occurs, the extent of their distributions will remain poorly understood.

The conservation status of marine species known from few specimens is often confounded by a lack of distribution and population information (Linardich et al., 2018). Over two dozen of the range‐restricted species have been collected from multiple localities separated by substantial distances, but whether they occur continuously between these localities is not yet confirmed due to insufficient sampling of cryptic and/or deep‐living species. For example, the locally abundant, mesophotic ember goby (Palatogobius incendius) was discovered during opportunistic deep‐submersible dives in Curaçao, Dominica and Honduras (Roatán), and was even filmed being consumed by lionfish (Tornabene & Baldwin, 2017). These three localities are separated by 900–2000 km and many areas between them contain mesophotic reef habitats that have not been explored, suggesting it likely has a larger range size than is currently understood. Conversely, species known from only a single locality or few neighbouring areas may be range‐restricted, but are not classified as such due to insufficient sampling. Considering that some diminutive reef fishes have short‐distance dispersal abilities in the Caribbean (D'Aloia et al., 2015), additional studies are needed before excluding them as range‐restricted. The social wrasse (H. socialis) and Cuban gramma (Gramma dejongi) are examples of confirmed range‐restricted species. An inhabitant of the deeper limits of recreational scuba diving (20–30 m), the Cuban gramma is restricted to a small area off southeastern Cuba and is lesser known than its sister species, the widespread fairy basslet (Gramma loreto). Localized extirpations of the fairy basslet in the Bahamas due to lionfish predation (Ingeman, 2016) justifies high concern for the range‐restricted Cuban gramma.

Gaps in sampling effort, especially for small or deep‐living species, can influence richness patterns. We hypothesized there is a correlation between the sighting of shallow (i.e. <40 m), range‐restricted species and the number of surveys in regional databases of fish fauna conducted via traditional scuba diving survey methods (i.e. sampling effort). This pattern is unlikely for range‐restricted species in mesophotic (i.e. >40 m) habitats, which are typically excluded from surveys due to depth restrictions. To test these hypotheses, we used survey data for the years 2000–2017 from the REEF Volunteer Survey Project Database where most of the surveys are conducted by divers at depths shallower than 40 m (REEF, 2018). We separated the 160 range‐restricted species into shallow (<40 m) and deep (>40 m) and identified the number of species occurring in each country. To identify any differences in correlation, with respect to deep or shallow species, between the number of surveys performed and the number of species observed in a given country, we performed and plotted a linear regression (Figure S3). For both shallow and deep species, the number of range‐restricted species observed increased with an increasing number of surveys. There was no statistical difference between shallow and deep species with respect to the number of range‐restricted species observed at different numbers of surveys.

4.3. Hotspots of potential impact ahead of the invasion front

Concern remains high for the likely future establishment of lionfish in Brazilian waters and subsequent native fish declines (Bumbeer et al., 2018). Since 2014, five lionfish have been confirmed from three widely spaced localities in Brazil, including one collected from the Fernando de Noronha Archipelago in 2020 (Ferreira et al., 2015; Luiz et al. 2021). We found that half of Brazilian endemic fishes have high vulnerability to lionfish predation. Of particular concern are 12 highly vulnerable endemics occupying habitats around offshore islands, each with an average range size of less than 12,000 km². The colonization of invasive lionfish into Brazil's three offshore island groups would potentially represent a major threat, and likely qualify these species for the highest extinction risk categories of Critically Endangered or Endangered according to IUCN Red List methodology. For example, the St. Paul's blenny (Enneanectes smithi) and salmon‐spotted jewelfish (Choranthias salmopunctatus), which are endemic to St. Peter and St. Paul's Rocks, have a range size of less than 1 km². This is amongst the smallest of ranges across the world's marine fishes (Pinheiro et al., 2020).

A biogeographical analysis by Pinheiro et al. (2018) located the highest richness of endemic Brazilian reef fishes in the continental shelf area extending from the states of Bahia to Santa Catarina. Our study identified a part of this same area—coastal São Paulo and Santa Catarina—as having the second‐highest richness of range‐restricted species with high vulnerability to lionfish predation. This part of Brazil is influenced by the Brazil Current and characterized by a wide continental shelf and coastal islands with rocky reefs where tropical and subtropical/temperate species overlap. The Fernando de Noronha Archipelago and Rocas Atoll oceanic islands, which we identified as having the highest richness of range‐restricted species with high vulnerability, is not particularly high in overall reef fish biodiversity as compared to other parts of Brazil (Pinheiro et al., 2018). The six Fernando de Noronha endemics that drive the hotspot result in our study are shallow reef fishes with an average maximum length of 6 cm.

Though not the focus of our empirical analyses, this same trait‐based framework could also be used to identify hotspots of potential impact from lionfish at its other invasion front: the Mediterranean Sea. First reported off the coast of Lebanon in 2012 (Bariche et al., 2013), subsequent increases in lionfish abundance and distribution in the Mediterranean are generating serious concern among stakeholders about potential impacts to native fauna and the economies they support (Kleitou et al., 2019; Kletou et al., 2016). We suggest that this exercise is particularly important and timely given the high‐diversity of range‐restricted and endemic species in this region (Mouillot et al., 2011), and the myriad threats facing fish assemblages across this relatively small geographic area (Ben Rais Lasram et al., 2010; Nieto et al., 2015).

4.4. Estimating management intervention timelines

Due to the volunteer methodology that REEF uses to collect data, effort varies widely between the 37 subregions and this contributed to uncertainty in estimating the invasion spread in some areas, including the Gulf of Mexico coasts of Cuba and Mexico. The invasion timeline according to REEF surveys is not linear among subregions within the same region; a pattern mirrored in other sources used to track the invasion's spread such as the U.S. Geological Survey (2021). Explanations of time to complete invasion include influences from currents, barriers to dispersal, and extreme weather events. It is also possible that the population growth rate is slow upon initial invasion in many subregions for a period of time before some tipping point leads to substantial increase in growth. If only a small number of mature adults are present in the adjacent subregion, the chance of juveniles dispersing to a neighbouring subregion may be low. In our study, most newly invaded areas showed a period of low abundance and little growth followed by a period of quick sustained growth. It should be noted that our proxy for population growth is only relevant for regions that have reached high abundance (i.e. greater than 11–100 lionfish per REEF survey).

This study highlighted the speed at which lionfish populations can grow to large numbers. Our results strongly suggest that a traits‐based approach, rather than a time‐ and resource‐intensive series of field studies, may provide enough preliminary information to guide policymakers and managers to begin targeted control programs aimed at suppressing lionfish populations before they become too large (i.e. above density thresholds that cause impacts to native fauna; e.g. Green et al., 2014). Across 10 subregions of the Caribbean, the average time period from initial sighting to maximum population abundance was about 5 years with a median of nearly 2 years. In most subregions where the midpoint was higher than the initial observed abundance, the rate of increase after the midpoint was much quicker than beforehand, possible evidence of exponential growth in the population over time. In the Florida Keys, lionfish were first sighted in 2009 and remained uncommon during that year until 2010–2011, during which abundance increased by three‐ to sixfold (Ruttenberg et al., 2012). Similar rapid population growth was recorded in the Flower Garden Banks from 2012 to 2013 where the first sighting occurred in 2011 (Johnston et al. 2016). This initial period of low abundance, or 1‐ to 2‐year lag time before rapid population growth, could be targeted as a key time to initiate culling activities with the goal of suppressing lionfish abundance below harmful thresholds. Conservation managers in parts of the world where invasions are currently predicted or nascent, such as Brazil and parts of the Mediterranean and eastern Pacific, could develop a timeline for action based on this information combined with monitoring surveys. In the invaded range, early detection and rapid response programs, including training of divers for lionfish removal, were employed in some areas of the Caribbean, and could be sourced for lessons learned and good practice (Green et al., 2017; Morris, 2012). With culling activities now widespread across the invaded Atlantic region, there is much to be learned about the factors likely affecting the success of this strategy including what the goals for intervention are (suppression vs. eradication), resources in terms of personnel and equipment allocated to culling, the timing and frequency of removal, and environmental conditions (e.g. Andradi‐Brown et al., 2017; Côté et al., 2014; Dahl et al., 2016; Green et al., 2017; Smith et al., 2017). Our approach provides one way to identify where priority areas might exist to concentrate limited resources for intervention, and could be combined with spatial information on these other important considerations to enhance culling success.

5. CONCLUSIONS

This study provides a framework for conservation planners that reduces the chance of overlooking a high‐risk species during priority‐setting, increases awareness for the global impact of invasive lionfish, highlights a need for studies on poorly known species and deep lionfish ecology and informs future on‐the‐ground action to reduce lionfish populations ahead of an invasion front. The incorporation of range‐restricted, small‐bodied, reef fishes into country‐specific and/or regional‐level conservation plans will help avoid loss of unique marine biodiversity, especially in the face of multiple stressors. Additional sampling effort targeting cryptobenthic species and mesophotic reefs is needed to understand the distribution of prey species and the ecology of invasive lionfish in the deeper part of their range (50–300 m). In cases where a threat, such as the invasive lionfish, occurs over a substantial portion of the globe, freely available global and regional‐level data (e.g. in this study: the IUCN Red List, the STRI and the REEF) can facilitate large‐scale biodiversity meta‐analyses that inform conservation. Comprehensive marine Red List initiatives similar to the one that assessed Caribbean shorefishes (Linardich et al., 2018) have also been completed for the eastern tropical Pacific (Polidoro et al., 2012) and the Mediterranean Sea (Abdul Malak et al., 2011), as these are areas where the lionfish invasion may expand to, those Red List data could be used to apply the trait‐based approach developed in this study to predict at‐risk species ahead of the lionfish invasion front within those fish faunas. Studies on factors that influence invasive lionfish population growth would also inform planning. Research on innovations that improve the efficiency of lionfish culling continues and should be encouraged (Hunt et al., 2019).

Supporting information

Supplementary Material

Linardich, C. , Brookson, C. B. , & Green, S. J. (2021). Trait‐based vulnerability reveals hotspots of potential impact for a global marine invader. Global Change Biology, 27, 4322–4338. 10.1111/gcb.15732

DATA AVAILABILITY STATEMENT

The data that support the findings of this study are available from the corresponding author upon reasonable request. Some of the data that support the findings of this study are available in the public domain on the IUCN Red List of Threatened Species at iucnredlist.org. The data and code for the statistical analysis presented in this paper is available in the open‐access repository: https://github.com/colebrookson/lionfish‐invasion‐timeline.