Get 20M+ Full-Text Papers For Less Than $1.50/day. Start a 14-Day Trial for You or Your Team.

Learn More →

Predicting IUCN Extinction Risk Categories for the World's Data Deficient Groupers (Teleostei: Epinephelidae)

Predicting IUCN Extinction Risk Categories for the World's Data Deficient Groupers (Teleostei:... Body size; coral reef fish; depth refuge; Groupers are highly susceptible to human-induced impacts, making them one Epinephelidae; fisheries; geographic range; Red of the most threatened fish families globally. Extinction risk assessments are List. important in endangered threatened species management, however the most Correspondence comprehensive—the International Union for Conservation of Nature (IUCN) Osmar J. Luiz, Department of Biological Red List—cannot classify approximately one-third of grouper species due to Sciences, Macquarie University, Balaclava Road, data deficiency. We used an ordinal analytical approach to model relationships Sydney, NSW 2109, Australia. between species-level traits and extinction risk categories. We found that larger Tel: +61(02) 9850 6271; species and those with shallower maximum depths and smaller geographic fax: +61(02) 02 9850 8667. ranges had higher extinction risk. Using our best fitting model, we classified E-mail: osmarjluiz@gmail.com data deficient grouper species into IUCN’s extinction risk categories based on Received traits. Most of these species were predicted to be of least concern. However, 13 August 2015 12% were predicted to be endangered or vulnerable, suggesting that they may Accepted be of conservation interest. Importantly, we provide a quantitative method for 5 January 2016 overcoming data gaps that can be applied to conservation of other species. Editor Zabel, Richard doi: 10.1111/conl.12230 species before extinctions may occur (Howard & Bickford Introduction 2014; Bland et al. 2015b) has resulted in a growing num- The International Union for Conservation of Nature ber of studies aimed at overcoming uncertainty regard- (IUCN) Red List is regarded as the most objective and ing the conservation status of data deficient species. The authoritative system available for classifying species in approaches within these studies vary widely. Some au- terms of their risk of extinction (Rodrigues et al. 2006). thors have used a subset of the criteria used in Red List Assessments must be backed up by data and species for assessments to evaluate extinction risk of data deficient which insufficient data are available to make an assess- species (Morais et al. 2013), while others have used al- ment of extinction risk are termed data deficient (Mace ternative data sources to reevaluate previous assessments et al. 2008). Since species in the data deficient category (Good et al. 2006; Sousa-Baena et al. 2014). Others still could fall into any of the other Red List categories, gen- have modeled correlations between existing species’ trait uinely threatened species may be neglected by conserva- data and extinction risk among data sufficient species to tion programs due to their uncertain conservation status predict the conservation status of data deficient species (Bland et al. 2015a). Therefore, there is an urgent need (Bland et al. 2015a). Sophisticated models for predicting to prioritize research on data deficient species and to gen- conservation status of data deficient species also incorpo- erate data that will accurately assign them into a threat rate phylogenetic and environmental information along category. Unfortunately, due to both time and financial with traits (Bland et al. 2015b; Jetz & Freckleton 2015). limitations, funds are rarely directed to filling these gaps However, there is a trade-off between model complexity (Hoffman et al. 2008). and data availability. Data deficient species are, by defini- The realization that there is not enough time nor tion, poorly studied. Hence, they often lack phylogenetic, enough resources to collect basic data on all data deficient spatial, or trait information, which may hamper the use 342 Conservation Letters, September/October 2016, 9(5), 342–350 Copyright and Photocopying: 2016 The Authors. Conservation Letters published by Wiley Periodicals, Inc. This is an open access article under the terms of the Creative Commons Attribution License, which permits use, distribution and reproduction in any medium, provided the original work is properly cited. O. J. Luiz et al. Threatened status of data deficient groupers of data-hungry modeling approaches. This is especially Materials and methods the case with marine taxa, which are considerably less Data collection well studied on average than nonmarine taxa. As a re- sult, marine species have double the proportion of data Data on the Red List categories of grouper species were deficient species as their nonmarine counterparts (Webb taken from Craig et al. (2011), which are based on the & Mindel 2015). most recent assessment made by the GWSG at the time A commercially important group of marine fishes for of writing. For this study, we focused on five key eco- which approximately one-third of species are data de- logical traits that are most likely to influence the extinc- ficient are the groupers (Epinephelidae). Consisting of tion risk of groupers. First, large body size has often been 163 species, groupers are an iconic family of marine linked to elevated extinction risk in reef fishes (Jennings predators that occur on coral reefs worldwide. Groupers et al. 1999; Bender et al. 2013). Second, shallow water have high market prices, and are consequently heavily marine environments have being increasingly exposed targeted by commercial, recreational, and artisanal fish- to human impacts that threaten fish populations (Bridge eries (Sadovy de Mitcheson et al. 2013). Groupers are et al. 2013). Fishes capable of utilizing deep-reef habi- also popular among recreational divers, providing nonex- tats are considered to be at lower risk of local extinction tractive economic value to the tourism industry and in- than fishes confined to shallow habitats (Lindfield et al. fluencing the economic viability of marine-protected ar- 2014). Third, species with large geographic range sizes are eas (Rudd & Tupper 2002). Their economic importance generally at less risk than those that occur in restricted aside, groupers play a critical ecological role in moderat- ranges, as a broad distribution permits a large population ing the abundance (Stallings 2008) and behavior (Madin size and/or a buffer against habitat loss, therefore, able et al. 2010) of prey species, with numerous indirect effects to withstand local extirpation without risk of global ex- on ecosystems. Therefore, the loss of grouper species has tinction (Hawkins et al. 2000). Fourth, species that form substantial socioeconomic and ecological implications. spawning aggregations are likely to be vulnerable to over- Groupers possess a series of life history traits that fishing, as evidenced by the loss of aggregations of com- make them the teleost family with the highest number mercially important species in many locations due to un- of threatened species on coral reefs (McClenachan 2015). sustainable fishing (Sadovy de Mitcheson & Colin 2012). Despite the wide recognition that groupers are vulnera- Fifth, species that occupy a large range of habitats may be ble to high fishing pressure, catches have been increasing expected to be more resilient to disturbance than habitat- unabatedly for more than 50 years (Sadovy de Mitche- specific species. Here, habitat generalists included species son et al. 2013). This fact has raised concerns about their that use other habitats in addition to structural reefs (soft potential risk of extinction. As a result, since 1998, the bottoms, seagrass/macroalgae beds, mangroves, and es- IUCN Grouper and Wrasses Specialist Group (GWSG) has tuaries). Our analysis also includes species distributions been assessing the threat status of all taxonomically valid among three distinct regions: the Indo-Central Pacific, grouper species using the Red List criteria (Sadovy de the Atlantic, and the Tropical Eastern Pacific, to allow for Mitcheson et al. 2013). Among them, a relatively high tests of the effect of regional variation in the predictive proportion of species (30%) are considered to be data de- factors. ficient (Sadovy de Mitcheson et al. 2013). Data on body length, maximum depth, and mul- Data deficient species contribute to considerable uncer- tihabitat use were obtained from Craig et al. (2011), tainty in global patterns of extinction risk and conserva- Robertson & Allen (2012), Robertson & Van Tassel tion prioritization. The use of preexisting biological data (2012). Geographical range sizes were calculated as the to model the conservation status of data sufficient species total area of all constituent polygons in each species’ dis- to predict the status of data deficient species has been tribution map available from the IUCN Red List spatial proven a very cost-effective methodology relative to com- database (IUCN 2015). prehensive risk assessments (Bland et al. 2015b). Here, Species that form spawning aggregations were ob- we modeled correlates of extinction risk among species tained from Sadovy de Mitcheson & Colin (2012). They of one the most threatened families of marine fishes us- provide lists of species that form spawning aggregations ing an ordinal regression approach that accounts for bio- delineated by quality of evidence. We tested the effect of logical traits that influence extinction risk in reef fishes, forming spawning aggregations on groupers’ extinction that is, body size, maximum depth of occurrence, breadth risk at two levels of data quality. One model considered of habitat use, geographic range size, and aggregative aggregators only the species in which spawning aggrega- spawning behavior. We then used the best fitting model tions were confirmed by direct evidence (Table S1). A sec- to estimate the probabilities of data deficient species being ond, less conservative model, broadened the definition of assigned into each of the Red List categories. aggregators by including species in which information on Conservation Letters, September/October 2016, 9(5), 342–350 Copyright and Photocopying: 2016 The Authors. Conservation Letters published by Wiley Periodicals, Inc. 343 Threatened status of data deficient groupers O. J. Luiz et al. 0.8 0.6 0.4 0.2 0.0 DD LC NT VU EN CR DD LC NT VU EN CR IUCN assessed Data Deficient predicted Figure 1 Distribution of assessed and predicted extinction risk. Extinction risk distribution for International Union for the Conservation of Nature (IUCN) assessed (left, n = 163) and CLMM predictions for data deficient (right, n = 50) groupers as percentage of the species pool in each of the IUCN extinction risk categories. DD, data deficient; LC, least concern; NT, near threatened; VU, vulnerable; EN, endangered; CR, critically endangered. Table 1 Parameters of the final CLMM with the IUCN Red List category as gorical response variable. This approach prevents infor- an ordinal categorical response variable, including significant interactions mation loss, such as happens with aggregated binomial between biogeographic region and range size and depth classifications (e.g., pooling categories in threatened vs. nonthreatened species), as well as avoids elevated type I Variable Estimate SE z Value P value error rates caused by assuming that differences between Body size 3.441 0.636 5.410 <0.001 adjacent risk levels are equivalent, such as when trans- Region forming categorical data to a numeric index for multiple IWP 6.607 3.398 1.944 0.051 regression (e.g., from 1 to 5). TEP −1.176 6.858 −0.172 0.863 Range size −0.0006 0.0002 −2.558 0.010 The response variable in our model is the species Red Max. depth: ATL 0.092 0.552 0.168 0.866 List category as an ordered categorical factor (least con- Max. depth: IWPICP −1.433 0.673 −2.128 0.033 cern [LC] < near threatened [NT] < vulnerable [VU] < Max. depth: TEP −0.021 1.570 −0.014 0.988 endangered [EN] < critically endangered [CR]). Fixed variables were body size, maximum depth, geographic Coefficient estimates of fixed variables, standard error (SE), test statistic (z range, formation of spawning aggregations, and broad- value), and probability (P value) are included. Coefficient in bold indicates that p-value is significant (P < 0.05). habitat use. Biogeographical region was also included as a fixed factor to test regional variability in the predictor variables. Finally, we included taxonomic genus as a ran- dom effect in our models to account for potential effects spawning aggregations is derived from indirect evidence of shared ancestry. (Table S2). For model selection, we did a backward stepwise re- moval of nonsignificant fixed-effect terms from the full Data analysis model, based on log-likelihood ratio tests. The models In order to investigate the relationship between species were fitted using the function “clmm” from the package traits and extinction risk, and subsequently estimate the “ordinal” (Christensen 2013) in R (R Core Team 2015). probability for data deficient species to be assigned to We used the coefficients of our final model to estimate each Red List category, we used cumulative link mixed- the Red List category of the data deficient species. effects modeling (CLMM). CLMM’s are ideal for analyz- Overall model fit was quantified using the percent- ing ranked categorical response variables, such as the age of categorized species that were correctly classified IUCN Red List categories, because they preserve the vari- by our final model (percentage correct classified, PCC). ance structure of the original ordinal ranks of the cate- Because of the small sample size of our training set 344 Conservation Letters, September/October 2016, 9(5), 342–350 Copyright and Photocopying: 2016 The Authors. Conservation Letters published by Wiley Periodicals, Inc. Proportion of species pool O. J. Luiz et al. Threatened status of data deficient groupers 1,000 samples for the testing set from the list of all data (a) 5.5 sufficient species and calculated the mean PCC and SE. Model estimates were recalculated for each sample. We defined specificity and sensitivity as the combined per- 5.0 centage of correct classifications among nonthreatened (LC, NT) and threatened (VU, EN, CR) categories, respec- tively. Cohen’s kappa statistic (function “kappa2” in R 4.5 package “irr”; Gamer et al. 2012) was used to measure the agreement between predicted and actual categorizations while correcting for agreement caused by chance (Hand 4.0 2012). (b) ATL Results ICP 6.0 TEP Among all 163 grouper species evaluated by the GWSG, 71 were classedasLC, 22 as NT,12asVU,5asEN, and 5.5 3 as CR. Fifty species were data deficient (Figure 1). Our analysis showed that body size and geographic range 5.0 size were important predictors of extinction risk among groupers, with threatened status generally increasing 4.5 with increasing body size and decreasing range sizes across all regions globally (Table 1; Figure 2). We also 4.0 found a significant interaction between biogeographic region and maximum depth (Table S1). A negative relationship between maximum depth of occurrence and (c) threatened status was observed only in the Indo-Central Pacific (Table 1). Including only species with direct evidence for spawn- ing aggregation in the aggregation formation category re- sulted in a similar result as including species with indirect evidence for spawning aggregation as well (Table S2). In neither case was spawning aggregation a significant pre- dictor of extinction risk. The range of trait values and the modeled rela- tionship between traits and extinction risk were the same among data deficient and data sufficient species (Figure S1). The overall accuracy of our model is satis- IUCN Cat. LC NT VU EN CR factory (Table S3; PCC 74%; Cohen’s kappa = 0.52, P < 0.001) given that it predicted species classification into Threat level five categories, although the accuracy varied among cat- egories (LC = 87%; NT = 54%; VU = 41%; EN = 60%; Figure 2 Average maximum body size (a), maximum depth (b), and geo- CR = 66%). We caution that model performance tests graphic range size (c) of groupers in each Red List category. In (b), the data based on the full training set may overestimate accuracy were grouped by biogeographical region to illustrate the interaction de- estimates, however our choice is justified based on the tected in the model. ATL, Atlantic; ICP, Indo-Central Pacific; TEP, Tropical Eastern Pacific; DD, data deficient; LC, least concern; NT, near threatened; low sample size and high imbalance among response cat- VU, vulnerable; EN, endangered; CR, critically endangered. egories (Table S4). The uncertainty of predictions for data deficient species increased with extinction risk (Table S5), a likely re- (113 spp.), we tested the model fit both using the whole sult of the progressively lower number of species in the training set Additionally, we tested the model fit and with higher risk categories. Only three data deficient species the data set split into two subsets (75% for training; 25% for which neither maximum depth of occurrence or ge- for testing). For the divided data set, we randomly drew ographic range size was available were not assigned a Conservation Letters, September/October 2016, 9(5), 342–350 Copyright and Photocopying: 2016 The Authors. Conservation Letters published by Wiley Periodicals, Inc. 345 sqrt geographic range (km ) log maximum depth (m) log body size (cm) Threatened status of data deficient groupers O. J. Luiz et al. LC NT VU EN CR Epinephelus quinquefasciatus Dermatolepis striolata Probability Epinephelus caninus Epinephelus posteli Epinephelus goreensis Mycteroperca cidi Epinephelus costae Plectropomus punctatus Hyporthodus haifensis Epinephelus latifasciatus Epinephelus indistinctus Epinephelus chabaudi Epinephelus magniscuttis Hyporthodus exsul Epinephelus undulosus Epinephelus summana Epinephelus undulatostriatus Epinephelus awoara Epinephelus erythrurus Hyporthodus darwinensis Epinephelus suborbitalis Hyporthodus octofasciatus 0.5 Epinephelus tauvina Hyporthodus perplexus Cephalopholis aitha Epinephelus corallicola Epinephelus fasciatomaculosus Epinephelus stoliczkae Epinephelus epistictus Aethaloperca rogaa Epinephelus bilobatus Epinephelus bontoides Cephalopholis taeniops Epinephelus melanostigma Hyporthodus niphobles Epinephelus retouti Epinephelus amblycephalus Epinephelus faveatus Cephalopholis nigripinnis Cephalopholis igarashiensis Saloptia powelli Epinephelus timorensis Epinephelus sexfasciatus Gracila albomarginata Epinephelus heniochus Epinephelus trophis Cephalopholis aurantia Figure 3 Probability distributions for data deficient species to be assigned into each of the IUCN’s Red List categories. Three species (Epinephelus chlorocephalus, E. lebretonianus,and E. polystigma) were not predicted to an extinction risk category because they lack data on the maximum depth of occurrence or geographic range size. LC, least concern; NT, near threatened; VU, vulnerable; EN, endangered; CR, critically endangered. predicted category. Among those categorized, two are that same category. Twelve percent (n = 6) of the data predictedtobeEN, four to be VU,nineNT, and32of deficient species are predicted to be EN or VU and may LC (Figure 3). therefore be of particular conservation interest. Our analysis also highlights traits of groupers that are associated with extinction risk. In general, large body size Discussion was correlated with grouper endangerment. Large ma- In this study, we used a novel method to estimate rine animals tend to have limited intrinsic rebound po- the threat status of data deficient grouper species. The tential (Jennings et al. 1999; Bender et al. 2013), and large method, which can be broadly applied to other taxo- fish species are more likely to be targets for fishing (Levin nomic groups in any system type, models relationships & Grimes 2002). Even moderate artisanal fishing effort between species’ traits and extinction risk categories for has been shown to deplete local stocks of some of the IUCN-assessed species. In general, data deficient groupers ´ largest groupers in the Tropical Eastern Pacific (Saenz- were predicted to be slightly less threatened than data Arroyo et al. 2005) and in the Western Atlantic (Giglio sufficient species. Sixty-four percent of the data deficient et al. 2015). species are predicted to be of LC, which is roughly one- The negative relationship between extinction risk third more of the percentage of data sufficient species in and maximum depth of occurrence, which is predicted 346 Conservation Letters, September/October 2016, 9(5), 342–350 Copyright and Photocopying: 2016 The Authors. Conservation Letters published by Wiley Periodicals, Inc. O. J. Luiz et al. Threatened status of data deficient groupers Figure 4 Maps of the distribution of all data deficient, data deficient predicted to be least concern, data deficient predicted to be vulnerable and data deficient predicted to be endangered species. (a) Total number of data deficient species. (b) Total number of data deficient species predicted to be least concern. (c) Distribution of data deficient species predicted to be vulnerable or endangered; a: Epinephelus quinquefasciatus,b: Dermatolepis striolata, c: Mycteroperca cidi,d: E. costae,e: E. posteli,f: E. goreensis,and g: E. caninus. under the “depth refuge” hypothesis, was significant only Pacific, and its extensive network of archipelagos, offers in the Indo-Central Pacific. This may be explained by many refuges where fishing is limited to artisanal sub- the contrasting geographical settings and differing fishing sistence, often performed with depth-restrictive equip- practices among regions. The vastness of the Indo-Central ment (e.g., spearfishing; Lindfield et al. 2014). In contrast, Conservation Letters, September/October 2016, 9(5), 342–350 Copyright and Photocopying: 2016 The Authors. Conservation Letters published by Wiley Periodicals, Inc. 347 Threatened status of data deficient groupers O. J. Luiz et al. large-scale commercial harvest of groupers is common ited by low sample sizes, especially when small data sets in the Atlantic, where large fleets with advanced fishing are associated with highly imbalanced response categories equipment, such as bottom longlines (Russell et al. 1988), (Table S6), because MLM split the data during the anal- allow commercial fishers to fish at greater depths than yses, progressively diminishing observations available for artisanal fishers. the next node construction (Elith et al. 2008). The formation of spawning aggregations, which is In addition, inconsistent conclusions about the traits largely assumed to be a primary driver of extinction explaining variation in extinction risk among species are risk among groupers, was not a significant predictor for derived from the use of comparative models of broad tax- threatened status. Apparently, reproductive aggregations onomic and geographic scope (Cardillo & Meijaard 2012). are linked to extinction risk only when associated with For these reasons, it has been suggested that the most large-bodied species. powerful and informative comparative models of extinc- Although conservation prioritization must con- tion risk will be those of narrow scope, restricted to rel- sider important socioeconomical factors in addition to atively lower taxonomic groups (Fisher & Owens 2004). observed or predicted extinction risk (Possingham et al. Our approach provides an alternative method to MLM 2002; Miller et al. 2006), the latter is a key input variable when predicting the conservation status of narrow taxo- for priority setting. Therefore, identifying species of nomic groups or groups in which only a few subsets of conservation concern among data deficient groupers is species have been assessed by IUCN. We must caution, considered high priority research for their management however, that our model provides lower sensitivity rel- (Sadovy de Mitcheson et al. 2013). Along with continued ative to specificity. This means that it is more prone to protection of species already recognized as being at risk, incorrectly predict species to be nonthreatened than the there needs to be a selective mechanism for pinpointing converse. However, the quantification of this uncertainty the data deficient species that are likely to be at most risk for each species Red List category combination (Figure 3; before their population declines go unnoticed. An impor- Table S4) is an important advantage over MLM in this tant outcome of the analytical approach used here is the case. probability estimates for a species being within each of Data deficiency is a major hindrance to conserva- the Red List categories (Figure 3, Table S5), which allow tion planning in taxonomic groups with a large propor- ranking of data deficient species both among and within tion of data deficient species. Despite the precaution- categories. This fine-scale categorization of extinction risk ary recommendation that data deficient species should among data deficient species has important implications be afforded the same degree of protection as threat- for management, policy and conservation planning. For ened taxa (Mace et al. 2008), this is often not the case example, it has been suggested that the areas with higher in practice (Hoffmann et al. 2008). This can result in numbers of data deficient species should be prioritized genuinely threatened species receiving little conserva- in order to tackle data deficiency (Brito 2010). However, tion attention until their populations decline substantially doing so may lead to misplaced effort if the goal is to (Howard & Bickford 2014). Modeling risk status of data maximize protection of threatened species. For example, sufficient species is a very cost-effective way to iden- while the Coral Triangle in SE Asia is the global hotspot tify high-risk data deficient species for preferential re- for data deficient groupers, none of the data deficient assessment within a reasonable time frame (Bland et al. species predicted to be threatened by our model occur in 2015b). The method presented here provides one simple, the Coral Triangle. Instead, they are found in the Western defensible way to overcome this pressing conservation Indian Ocean, the Tropical Eastern Pacific, in the Eastern problem. Atlantic, and in the Mediterranean Sea (Figure 4). To date, most models aimed at predicting the conser- vation status of data deficient species have been based on Acknowledgments machine-learning methods (MLM). This approach is well suited to data sets with many hundreds or thousands of O.J.L. thanks the Quantitative Ecology and Evolution species, as is the case of broader taxonomic groups with Group at Macquarie University for financial support. good coverage by IUCN assessments such as mammals O.J.L. was supported by the New South Wales Envi- and amphibians (Howard & Bickford 2014; Bland et al. ronmental Trust. E.M.P.M. was supported by the World 2015a; Jetz & Freckleton 2015). However, predicting the Wildlife Fund’s Kathryn S. Fuller Science for Nature Fund status of data deficient species among narrow taxonomic and an Australian Research Council DECRA Fellowship groups (single families or genera) or groups for which (project number DE120102614). J.S.M. was supported data are available for just a small subset of the species by an Australian Research Council Future Fellowship. remains challenging. The performance of MLM is lim- We thank A. Grech for help with mapping, A. Allen for 348 Conservation Letters, September/October 2016, 9(5), 342–350 Copyright and Photocopying: 2016 The Authors. Conservation Letters published by Wiley Periodicals, Inc. O. J. Luiz et al. Threatened status of data deficient groupers statistical advice, and V. Giglio and two anonymous re- Brito, D. (2010). Overcoming the Linnean shortfall: data viewers for comments on the manuscript. deficiency and biological survey priorities. Basic Appl. Ecol., 11, 709-713. Cardillo, M. & Meijaard, E. (2012). Are comparative studies of Supporting Information extinction risk useful for conservation? Trends Ecol. Evol., 27, 167-171. Additional Supporting Information may be found in the Christensen, R.H.B. (2013). Package “Ordinal”: regression models online version of this article at the publisher’s web site: for ordinal data. Copenhagen, Denmark. (Version Figure S1. (A) Boxplots of the significant explanatory 2012.09-11). variables in data sufficient (n = 113) and data deficient http://www.cran.r-project.org/package=ordinal/. species (n = 50). (B) Modeled relationships among sig- Craig, M.T., Mitcheson, Y.J.S. & Heemstra, P.C. (2011). Groupers of the world. NISC, Grahamstown, South Africa. nificant explanatory variables and data deficient species. Elith, J., Leathwick, J.R. & Hastie, T. (2008). A working guide Table S1. Model selection for a) the effect of including to boosted regression trees. J. Anim. Ecol., 77, 802-813. the interaction of region with each variable and b) the ef- Fisher, D.O. & Owens, I.P. (2004). The comparative method fect of dropping variables in a backward stepwise manner in conservation biology. Trends Ecol. Evol., 19, 391-398. Table S2. Model selection for a) the effect of including Gamer, M., Lemon, J. & Fellows, I. (2012). Package irr: the interaction of region with each variable and b) the ef- various coefficients of interrater reliability and agreement. fect of dropping variables in a backward stepwise manner (Version 0.84). http://CRAN.R-project.org/package=irr Table S3. Confusion matrix of the training data pre- Giglio, V.J., Luiz, O.J. & Gerhardinger, L.C. (2015). Depletion dictions and accuracy measures for the selected CLMM of marine megafauna and shifting baselines among predicting the IUCN Red List conservation status category artisanal fishers in eastern Brazil. Anim. Conserv., 18, in groupers (n = 113) 348-358. Table S4. Accuracy measures (PCC) for the training Good, T.C., Zjhra, M.L. & Kremen, C. (2006). Addressing data data predictions of the selected CLM on the full training deficiency in classifying extinction risk: a case study of a set and on the model calibrated on a 75% training set radiation of Bignoniaceae from Madagascar. Conserv. Biol., and 25% validation set. The split set was randomly drawn 20, 1099-1110. 1,000 times to calculate the mean and standard error (SE) Hand, D.J. (2012). Assessing the performance of classification Table S5. Estimates of probabilities of data deficient methods. Int. Stat. Rev., 80, 400-414. species being classed in each Red List category Hawkins, J.P., Roberts, C.M. & Clark, V. (2000). The Table S6. Confusion matrix for the training data threatened status of restricted-range coral reef fish species. predictions and accuracy measures for the machine- Anim. Conserv., 3, 81-88. learning∗ model predicting the IUCN Red List conserva- Hoffmann, M., Brooks, T.M., da Fonseca, G.A. et al. (2008). tion status category in groupers (n = 113) Conservation planning and the IUCN Red List. Endang. This material is available as part of the online ar- Species Res., 6, 113-125. Howard, S.D. & Bickford, D.P. (2014). Amphibians over the ticle from: http://www.blackwell-synergy.com/doi/full/ edge: silent extinction risk of data deficient species. Divers. 10.1111/j.1755–263X.2008.00002.x Distrib., 20, 837-846. (This link will take you to the article abstract). IUCN. (2015). The IUCN red list of threatened species. Version 2015-4. http://www.iucnredlist.org. Downloaded Accessed References on 19 October 2015. Jennings, S., Reynolds, J.D. & Polunin, N.V. (1999). Predicting Bender, M.G., Floeter, S.R., Mayer, F.P. et al. (2013). the vulnerability of tropical reef fishes to exploitation with Biological attributes and major threats as predictors of the phylogenies and life histories. Conserv. Biol., 13, 1466-1475. vulnerability of species: a case study with Brazilian reef Jetz, W. & Freckleton, R.P. (2015). Towards a general fishes. Orix, 47, 259-265. framework for predicting threat status of data-deficient Bland, L.M., Collen, B., Orme, C.D.L. & Bielby, J. (2015a). species from phylogenetic, spatial and environmental Predicting the conservation status of data-deficient species. information. Philos. Trans. R. Soc. B, 370, 20140016. Conserv. Biol., 29, 250-259. Levin, P.S. & Grimes, C.B. (2002). Reef fish ecology and Bland, L.M., Orme, C.D.L., Bielby, J., Collen, B., Nicholson, E. grouper conservation and management. Pages 377-390 in & McCarthy, M.A. (2015b). Cost-effective assessment of P.F. Sale, editor. Coral reef fishes: dynamics and diversity in a extinction risk with limited information. J. Appl. Ecol., 52, complex ecosystem. Academic Press, San Diego. 861-870. Lindfield, S.J., McIlwain, J.L. & Harvey, E.S. (2014). Depth Bridge, T.C., Hughes, T.P., Guinotte, J.M. & Bongaerts, P. refuge and the impacts of SCUBA spearfishing on coral reef (2013). Call to protect all coral reefs. Nat. Clim. Change, 3, fishes. PloS One, 9, e92628. 528-530. Conservation Letters, September/October 2016, 9(5), 342–350 Copyright and Photocopying: 2016 The Authors. Conservation Letters published by Wiley Periodicals, Inc. 349 Threatened status of data deficient groupers O. J. Luiz et al. Mace, G.M., Collar, N.J., Gaston, K.J. et al. (2008). us/app/fishes-greater-caribbean/id570048678?mt=8. Quantification of extinction risk: IUCN’s system for Accessed September 2015. classifying threatened species. Conserv. Biol., 22, Rodrigues, A.S., Pilgrim, J.D., Lamoreux, J.F., Hoffmann, M. 1424-1442. & Brooks, T.M. (2006). The value of the IUCN Red List for Madin, E.M.P., Gaines, S.D. & Warner, R.R. (2010). Field conservation. Trends Ecol. Evol., 21, 71-76. evidence for pervasive indirect effects of fishing on prey Rudd, M.A. & Tupper, M.H. (2002). The impact of Nassau foraging behavior. Ecology, 91, 3563-3571. grouper size and abundance on scuba diver site selection McClenachan, L. (2015). Extinction risk in reef fishes. Pages and MPA economics. Coast. Manage., 30, 133-151. 199-207 in C. Mora, editor. Ecology of fishes on coral reefs. Russell, G.M., Gutherz, E.J. & Barans, C.A. (1988). Evaluation Cambridge University Press, Cambridge. of demersal longline gear off South Carolina and Puerto Miller, R.M., Rodr´ıguez, J.P., Aniskowicz-Fowler, T. et al. Rico with emphasis on deep-water reef fish stocks. Mar. (2006). Extinction risk and conservation priorities. Science, Fish. Rev., 50, 26-31. 313, 441-441. Sadovy de Mitcheson, Y. & Colin, P., editors. (2012). Reef fish Morais, A.R., Siqueira, M.N., Lemes, P., Maciel, N.M., De spawning aggregations: biology, research and management. Marco, P. & Brito, D. (2013). Unraveling the conservation Springer, New York. status of data deficient species. Biol. Conserv., 166, 98- Sadovy de Mitcheson, Y., Craig, M.T., Bertoncini, A.A. et al. 102. (2013). Fishing groupers towards extinction: a global Possingham, H.P., Andelman, S.J., Burgman, M.A., Medellı́n, assessment of threats and extinction risks in a billion dollar R.A., Master, L.L. & Keith, D.A. (2002). Limits to the use of fishery. Fish Fish., 14, 119-136. threatened species lists. Trends Ecol. Evol., 17, 503-507. Saenz-Arroyo, ´ A., Roberts, C.M., Torre, J. & Carino-Olvera, ˜ R Core Team. (2015). R: a language and environment for M. (2005). Using fishers’ anecdotes, naturalists’ statistical computing. R Foundation for Statistical Computing, observations and grey literature to reassess marine species Vienna, Austria. at risk: the case of the Gulf grouper in the Gulf of Robertson, D.R. & Allen, G. (2012). Fishes: east pacific. An California, Mexico. Fish Fish., 6, 121-133. identification guide to the shore-fish fauna of the tropical eastern Sousa-Baena, M.S., Garcia, L.C. & Peterson, A.T. (2014). pacific. (iOS App. Copyright Smithsonian Institution, Left Knowledge behind conservation status decisions: data basis Coast R&C, Santa Cruz, California). Available at: https:// for “Data Deficient” Brazilian plant species. Biol. Conserv., itunes.apple.com/us/app/fishes-east-pacific/ 173, 80-89. id494644648?mt=8. Accessed September 2015. Stallings, C.D. (2008). Indirect effects of an exploited Robertson, D.R. & Van Tassel, J. (2012). Fishes: Greater predator on recruitment of coral-reef fishes. Ecology, 89, Caribbean. An identification guide to the shore-fish fauna of the 2090-2095. Caribbean and adjacent areas. (iOS App. Copyright Webb, T.J. & Mindel, B.L. (2015). Global patterns of Smithsonian Institution, Left Coast R&C, Santa extinction risk in marine and non-marine systems. Curr. Cruz, California). Available at: https://itunes.apple.com/ Biol., 25, 506-511. 350 Conservation Letters, September/October 2016, 9(5), 342–350 Copyright and Photocopying: 2016 The Authors. Conservation Letters published by Wiley Periodicals, Inc. http://www.deepdyve.com/assets/images/DeepDyve-Logo-lg.png Conservation Letters Wiley

Predicting IUCN Extinction Risk Categories for the World's Data Deficient Groupers (Teleostei: Epinephelidae)

Download PDF
9 pages

Loading next page...
 
/lp/wiley/predicting-iucn-extinction-risk-categories-for-the-world-s-data-A9zJxiGV6q

References (44)

Publisher
Wiley
Copyright
Copyright © 2016 Wiley Periodicals, Inc.
ISSN
1755-263X
eISSN
1755-263X
DOI
10.1111/conl.12230
Publisher site
See Article on Publisher Site

Abstract

Body size; coral reef fish; depth refuge; Groupers are highly susceptible to human-induced impacts, making them one Epinephelidae; fisheries; geographic range; Red of the most threatened fish families globally. Extinction risk assessments are List. important in endangered threatened species management, however the most Correspondence comprehensive—the International Union for Conservation of Nature (IUCN) Osmar J. Luiz, Department of Biological Red List—cannot classify approximately one-third of grouper species due to Sciences, Macquarie University, Balaclava Road, data deficiency. We used an ordinal analytical approach to model relationships Sydney, NSW 2109, Australia. between species-level traits and extinction risk categories. We found that larger Tel: +61(02) 9850 6271; species and those with shallower maximum depths and smaller geographic fax: +61(02) 02 9850 8667. ranges had higher extinction risk. Using our best fitting model, we classified E-mail: osmarjluiz@gmail.com data deficient grouper species into IUCN’s extinction risk categories based on Received traits. Most of these species were predicted to be of least concern. However, 13 August 2015 12% were predicted to be endangered or vulnerable, suggesting that they may Accepted be of conservation interest. Importantly, we provide a quantitative method for 5 January 2016 overcoming data gaps that can be applied to conservation of other species. Editor Zabel, Richard doi: 10.1111/conl.12230 species before extinctions may occur (Howard & Bickford Introduction 2014; Bland et al. 2015b) has resulted in a growing num- The International Union for Conservation of Nature ber of studies aimed at overcoming uncertainty regard- (IUCN) Red List is regarded as the most objective and ing the conservation status of data deficient species. The authoritative system available for classifying species in approaches within these studies vary widely. Some au- terms of their risk of extinction (Rodrigues et al. 2006). thors have used a subset of the criteria used in Red List Assessments must be backed up by data and species for assessments to evaluate extinction risk of data deficient which insufficient data are available to make an assess- species (Morais et al. 2013), while others have used al- ment of extinction risk are termed data deficient (Mace ternative data sources to reevaluate previous assessments et al. 2008). Since species in the data deficient category (Good et al. 2006; Sousa-Baena et al. 2014). Others still could fall into any of the other Red List categories, gen- have modeled correlations between existing species’ trait uinely threatened species may be neglected by conserva- data and extinction risk among data sufficient species to tion programs due to their uncertain conservation status predict the conservation status of data deficient species (Bland et al. 2015a). Therefore, there is an urgent need (Bland et al. 2015a). Sophisticated models for predicting to prioritize research on data deficient species and to gen- conservation status of data deficient species also incorpo- erate data that will accurately assign them into a threat rate phylogenetic and environmental information along category. Unfortunately, due to both time and financial with traits (Bland et al. 2015b; Jetz & Freckleton 2015). limitations, funds are rarely directed to filling these gaps However, there is a trade-off between model complexity (Hoffman et al. 2008). and data availability. Data deficient species are, by defini- The realization that there is not enough time nor tion, poorly studied. Hence, they often lack phylogenetic, enough resources to collect basic data on all data deficient spatial, or trait information, which may hamper the use 342 Conservation Letters, September/October 2016, 9(5), 342–350 Copyright and Photocopying: 2016 The Authors. Conservation Letters published by Wiley Periodicals, Inc. This is an open access article under the terms of the Creative Commons Attribution License, which permits use, distribution and reproduction in any medium, provided the original work is properly cited. O. J. Luiz et al. Threatened status of data deficient groupers of data-hungry modeling approaches. This is especially Materials and methods the case with marine taxa, which are considerably less Data collection well studied on average than nonmarine taxa. As a re- sult, marine species have double the proportion of data Data on the Red List categories of grouper species were deficient species as their nonmarine counterparts (Webb taken from Craig et al. (2011), which are based on the & Mindel 2015). most recent assessment made by the GWSG at the time A commercially important group of marine fishes for of writing. For this study, we focused on five key eco- which approximately one-third of species are data de- logical traits that are most likely to influence the extinc- ficient are the groupers (Epinephelidae). Consisting of tion risk of groupers. First, large body size has often been 163 species, groupers are an iconic family of marine linked to elevated extinction risk in reef fishes (Jennings predators that occur on coral reefs worldwide. Groupers et al. 1999; Bender et al. 2013). Second, shallow water have high market prices, and are consequently heavily marine environments have being increasingly exposed targeted by commercial, recreational, and artisanal fish- to human impacts that threaten fish populations (Bridge eries (Sadovy de Mitcheson et al. 2013). Groupers are et al. 2013). Fishes capable of utilizing deep-reef habi- also popular among recreational divers, providing nonex- tats are considered to be at lower risk of local extinction tractive economic value to the tourism industry and in- than fishes confined to shallow habitats (Lindfield et al. fluencing the economic viability of marine-protected ar- 2014). Third, species with large geographic range sizes are eas (Rudd & Tupper 2002). Their economic importance generally at less risk than those that occur in restricted aside, groupers play a critical ecological role in moderat- ranges, as a broad distribution permits a large population ing the abundance (Stallings 2008) and behavior (Madin size and/or a buffer against habitat loss, therefore, able et al. 2010) of prey species, with numerous indirect effects to withstand local extirpation without risk of global ex- on ecosystems. Therefore, the loss of grouper species has tinction (Hawkins et al. 2000). Fourth, species that form substantial socioeconomic and ecological implications. spawning aggregations are likely to be vulnerable to over- Groupers possess a series of life history traits that fishing, as evidenced by the loss of aggregations of com- make them the teleost family with the highest number mercially important species in many locations due to un- of threatened species on coral reefs (McClenachan 2015). sustainable fishing (Sadovy de Mitcheson & Colin 2012). Despite the wide recognition that groupers are vulnera- Fifth, species that occupy a large range of habitats may be ble to high fishing pressure, catches have been increasing expected to be more resilient to disturbance than habitat- unabatedly for more than 50 years (Sadovy de Mitche- specific species. Here, habitat generalists included species son et al. 2013). This fact has raised concerns about their that use other habitats in addition to structural reefs (soft potential risk of extinction. As a result, since 1998, the bottoms, seagrass/macroalgae beds, mangroves, and es- IUCN Grouper and Wrasses Specialist Group (GWSG) has tuaries). Our analysis also includes species distributions been assessing the threat status of all taxonomically valid among three distinct regions: the Indo-Central Pacific, grouper species using the Red List criteria (Sadovy de the Atlantic, and the Tropical Eastern Pacific, to allow for Mitcheson et al. 2013). Among them, a relatively high tests of the effect of regional variation in the predictive proportion of species (30%) are considered to be data de- factors. ficient (Sadovy de Mitcheson et al. 2013). Data on body length, maximum depth, and mul- Data deficient species contribute to considerable uncer- tihabitat use were obtained from Craig et al. (2011), tainty in global patterns of extinction risk and conserva- Robertson & Allen (2012), Robertson & Van Tassel tion prioritization. The use of preexisting biological data (2012). Geographical range sizes were calculated as the to model the conservation status of data sufficient species total area of all constituent polygons in each species’ dis- to predict the status of data deficient species has been tribution map available from the IUCN Red List spatial proven a very cost-effective methodology relative to com- database (IUCN 2015). prehensive risk assessments (Bland et al. 2015b). Here, Species that form spawning aggregations were ob- we modeled correlates of extinction risk among species tained from Sadovy de Mitcheson & Colin (2012). They of one the most threatened families of marine fishes us- provide lists of species that form spawning aggregations ing an ordinal regression approach that accounts for bio- delineated by quality of evidence. We tested the effect of logical traits that influence extinction risk in reef fishes, forming spawning aggregations on groupers’ extinction that is, body size, maximum depth of occurrence, breadth risk at two levels of data quality. One model considered of habitat use, geographic range size, and aggregative aggregators only the species in which spawning aggrega- spawning behavior. We then used the best fitting model tions were confirmed by direct evidence (Table S1). A sec- to estimate the probabilities of data deficient species being ond, less conservative model, broadened the definition of assigned into each of the Red List categories. aggregators by including species in which information on Conservation Letters, September/October 2016, 9(5), 342–350 Copyright and Photocopying: 2016 The Authors. Conservation Letters published by Wiley Periodicals, Inc. 343 Threatened status of data deficient groupers O. J. Luiz et al. 0.8 0.6 0.4 0.2 0.0 DD LC NT VU EN CR DD LC NT VU EN CR IUCN assessed Data Deficient predicted Figure 1 Distribution of assessed and predicted extinction risk. Extinction risk distribution for International Union for the Conservation of Nature (IUCN) assessed (left, n = 163) and CLMM predictions for data deficient (right, n = 50) groupers as percentage of the species pool in each of the IUCN extinction risk categories. DD, data deficient; LC, least concern; NT, near threatened; VU, vulnerable; EN, endangered; CR, critically endangered. Table 1 Parameters of the final CLMM with the IUCN Red List category as gorical response variable. This approach prevents infor- an ordinal categorical response variable, including significant interactions mation loss, such as happens with aggregated binomial between biogeographic region and range size and depth classifications (e.g., pooling categories in threatened vs. nonthreatened species), as well as avoids elevated type I Variable Estimate SE z Value P value error rates caused by assuming that differences between Body size 3.441 0.636 5.410 <0.001 adjacent risk levels are equivalent, such as when trans- Region forming categorical data to a numeric index for multiple IWP 6.607 3.398 1.944 0.051 regression (e.g., from 1 to 5). TEP −1.176 6.858 −0.172 0.863 Range size −0.0006 0.0002 −2.558 0.010 The response variable in our model is the species Red Max. depth: ATL 0.092 0.552 0.168 0.866 List category as an ordered categorical factor (least con- Max. depth: IWPICP −1.433 0.673 −2.128 0.033 cern [LC] < near threatened [NT] < vulnerable [VU] < Max. depth: TEP −0.021 1.570 −0.014 0.988 endangered [EN] < critically endangered [CR]). Fixed variables were body size, maximum depth, geographic Coefficient estimates of fixed variables, standard error (SE), test statistic (z range, formation of spawning aggregations, and broad- value), and probability (P value) are included. Coefficient in bold indicates that p-value is significant (P < 0.05). habitat use. Biogeographical region was also included as a fixed factor to test regional variability in the predictor variables. Finally, we included taxonomic genus as a ran- dom effect in our models to account for potential effects spawning aggregations is derived from indirect evidence of shared ancestry. (Table S2). For model selection, we did a backward stepwise re- moval of nonsignificant fixed-effect terms from the full Data analysis model, based on log-likelihood ratio tests. The models In order to investigate the relationship between species were fitted using the function “clmm” from the package traits and extinction risk, and subsequently estimate the “ordinal” (Christensen 2013) in R (R Core Team 2015). probability for data deficient species to be assigned to We used the coefficients of our final model to estimate each Red List category, we used cumulative link mixed- the Red List category of the data deficient species. effects modeling (CLMM). CLMM’s are ideal for analyz- Overall model fit was quantified using the percent- ing ranked categorical response variables, such as the age of categorized species that were correctly classified IUCN Red List categories, because they preserve the vari- by our final model (percentage correct classified, PCC). ance structure of the original ordinal ranks of the cate- Because of the small sample size of our training set 344 Conservation Letters, September/October 2016, 9(5), 342–350 Copyright and Photocopying: 2016 The Authors. Conservation Letters published by Wiley Periodicals, Inc. Proportion of species pool O. J. Luiz et al. Threatened status of data deficient groupers 1,000 samples for the testing set from the list of all data (a) 5.5 sufficient species and calculated the mean PCC and SE. Model estimates were recalculated for each sample. We defined specificity and sensitivity as the combined per- 5.0 centage of correct classifications among nonthreatened (LC, NT) and threatened (VU, EN, CR) categories, respec- tively. Cohen’s kappa statistic (function “kappa2” in R 4.5 package “irr”; Gamer et al. 2012) was used to measure the agreement between predicted and actual categorizations while correcting for agreement caused by chance (Hand 4.0 2012). (b) ATL Results ICP 6.0 TEP Among all 163 grouper species evaluated by the GWSG, 71 were classedasLC, 22 as NT,12asVU,5asEN, and 5.5 3 as CR. Fifty species were data deficient (Figure 1). Our analysis showed that body size and geographic range 5.0 size were important predictors of extinction risk among groupers, with threatened status generally increasing 4.5 with increasing body size and decreasing range sizes across all regions globally (Table 1; Figure 2). We also 4.0 found a significant interaction between biogeographic region and maximum depth (Table S1). A negative relationship between maximum depth of occurrence and (c) threatened status was observed only in the Indo-Central Pacific (Table 1). Including only species with direct evidence for spawn- ing aggregation in the aggregation formation category re- sulted in a similar result as including species with indirect evidence for spawning aggregation as well (Table S2). In neither case was spawning aggregation a significant pre- dictor of extinction risk. The range of trait values and the modeled rela- tionship between traits and extinction risk were the same among data deficient and data sufficient species (Figure S1). The overall accuracy of our model is satis- IUCN Cat. LC NT VU EN CR factory (Table S3; PCC 74%; Cohen’s kappa = 0.52, P < 0.001) given that it predicted species classification into Threat level five categories, although the accuracy varied among cat- egories (LC = 87%; NT = 54%; VU = 41%; EN = 60%; Figure 2 Average maximum body size (a), maximum depth (b), and geo- CR = 66%). We caution that model performance tests graphic range size (c) of groupers in each Red List category. In (b), the data based on the full training set may overestimate accuracy were grouped by biogeographical region to illustrate the interaction de- estimates, however our choice is justified based on the tected in the model. ATL, Atlantic; ICP, Indo-Central Pacific; TEP, Tropical Eastern Pacific; DD, data deficient; LC, least concern; NT, near threatened; low sample size and high imbalance among response cat- VU, vulnerable; EN, endangered; CR, critically endangered. egories (Table S4). The uncertainty of predictions for data deficient species increased with extinction risk (Table S5), a likely re- (113 spp.), we tested the model fit both using the whole sult of the progressively lower number of species in the training set Additionally, we tested the model fit and with higher risk categories. Only three data deficient species the data set split into two subsets (75% for training; 25% for which neither maximum depth of occurrence or ge- for testing). For the divided data set, we randomly drew ographic range size was available were not assigned a Conservation Letters, September/October 2016, 9(5), 342–350 Copyright and Photocopying: 2016 The Authors. Conservation Letters published by Wiley Periodicals, Inc. 345 sqrt geographic range (km ) log maximum depth (m) log body size (cm) Threatened status of data deficient groupers O. J. Luiz et al. LC NT VU EN CR Epinephelus quinquefasciatus Dermatolepis striolata Probability Epinephelus caninus Epinephelus posteli Epinephelus goreensis Mycteroperca cidi Epinephelus costae Plectropomus punctatus Hyporthodus haifensis Epinephelus latifasciatus Epinephelus indistinctus Epinephelus chabaudi Epinephelus magniscuttis Hyporthodus exsul Epinephelus undulosus Epinephelus summana Epinephelus undulatostriatus Epinephelus awoara Epinephelus erythrurus Hyporthodus darwinensis Epinephelus suborbitalis Hyporthodus octofasciatus 0.5 Epinephelus tauvina Hyporthodus perplexus Cephalopholis aitha Epinephelus corallicola Epinephelus fasciatomaculosus Epinephelus stoliczkae Epinephelus epistictus Aethaloperca rogaa Epinephelus bilobatus Epinephelus bontoides Cephalopholis taeniops Epinephelus melanostigma Hyporthodus niphobles Epinephelus retouti Epinephelus amblycephalus Epinephelus faveatus Cephalopholis nigripinnis Cephalopholis igarashiensis Saloptia powelli Epinephelus timorensis Epinephelus sexfasciatus Gracila albomarginata Epinephelus heniochus Epinephelus trophis Cephalopholis aurantia Figure 3 Probability distributions for data deficient species to be assigned into each of the IUCN’s Red List categories. Three species (Epinephelus chlorocephalus, E. lebretonianus,and E. polystigma) were not predicted to an extinction risk category because they lack data on the maximum depth of occurrence or geographic range size. LC, least concern; NT, near threatened; VU, vulnerable; EN, endangered; CR, critically endangered. predicted category. Among those categorized, two are that same category. Twelve percent (n = 6) of the data predictedtobeEN, four to be VU,nineNT, and32of deficient species are predicted to be EN or VU and may LC (Figure 3). therefore be of particular conservation interest. Our analysis also highlights traits of groupers that are associated with extinction risk. In general, large body size Discussion was correlated with grouper endangerment. Large ma- In this study, we used a novel method to estimate rine animals tend to have limited intrinsic rebound po- the threat status of data deficient grouper species. The tential (Jennings et al. 1999; Bender et al. 2013), and large method, which can be broadly applied to other taxo- fish species are more likely to be targets for fishing (Levin nomic groups in any system type, models relationships & Grimes 2002). Even moderate artisanal fishing effort between species’ traits and extinction risk categories for has been shown to deplete local stocks of some of the IUCN-assessed species. In general, data deficient groupers ´ largest groupers in the Tropical Eastern Pacific (Saenz- were predicted to be slightly less threatened than data Arroyo et al. 2005) and in the Western Atlantic (Giglio sufficient species. Sixty-four percent of the data deficient et al. 2015). species are predicted to be of LC, which is roughly one- The negative relationship between extinction risk third more of the percentage of data sufficient species in and maximum depth of occurrence, which is predicted 346 Conservation Letters, September/October 2016, 9(5), 342–350 Copyright and Photocopying: 2016 The Authors. Conservation Letters published by Wiley Periodicals, Inc. O. J. Luiz et al. Threatened status of data deficient groupers Figure 4 Maps of the distribution of all data deficient, data deficient predicted to be least concern, data deficient predicted to be vulnerable and data deficient predicted to be endangered species. (a) Total number of data deficient species. (b) Total number of data deficient species predicted to be least concern. (c) Distribution of data deficient species predicted to be vulnerable or endangered; a: Epinephelus quinquefasciatus,b: Dermatolepis striolata, c: Mycteroperca cidi,d: E. costae,e: E. posteli,f: E. goreensis,and g: E. caninus. under the “depth refuge” hypothesis, was significant only Pacific, and its extensive network of archipelagos, offers in the Indo-Central Pacific. This may be explained by many refuges where fishing is limited to artisanal sub- the contrasting geographical settings and differing fishing sistence, often performed with depth-restrictive equip- practices among regions. The vastness of the Indo-Central ment (e.g., spearfishing; Lindfield et al. 2014). In contrast, Conservation Letters, September/October 2016, 9(5), 342–350 Copyright and Photocopying: 2016 The Authors. Conservation Letters published by Wiley Periodicals, Inc. 347 Threatened status of data deficient groupers O. J. Luiz et al. large-scale commercial harvest of groupers is common ited by low sample sizes, especially when small data sets in the Atlantic, where large fleets with advanced fishing are associated with highly imbalanced response categories equipment, such as bottom longlines (Russell et al. 1988), (Table S6), because MLM split the data during the anal- allow commercial fishers to fish at greater depths than yses, progressively diminishing observations available for artisanal fishers. the next node construction (Elith et al. 2008). The formation of spawning aggregations, which is In addition, inconsistent conclusions about the traits largely assumed to be a primary driver of extinction explaining variation in extinction risk among species are risk among groupers, was not a significant predictor for derived from the use of comparative models of broad tax- threatened status. Apparently, reproductive aggregations onomic and geographic scope (Cardillo & Meijaard 2012). are linked to extinction risk only when associated with For these reasons, it has been suggested that the most large-bodied species. powerful and informative comparative models of extinc- Although conservation prioritization must con- tion risk will be those of narrow scope, restricted to rel- sider important socioeconomical factors in addition to atively lower taxonomic groups (Fisher & Owens 2004). observed or predicted extinction risk (Possingham et al. Our approach provides an alternative method to MLM 2002; Miller et al. 2006), the latter is a key input variable when predicting the conservation status of narrow taxo- for priority setting. Therefore, identifying species of nomic groups or groups in which only a few subsets of conservation concern among data deficient groupers is species have been assessed by IUCN. We must caution, considered high priority research for their management however, that our model provides lower sensitivity rel- (Sadovy de Mitcheson et al. 2013). Along with continued ative to specificity. This means that it is more prone to protection of species already recognized as being at risk, incorrectly predict species to be nonthreatened than the there needs to be a selective mechanism for pinpointing converse. However, the quantification of this uncertainty the data deficient species that are likely to be at most risk for each species Red List category combination (Figure 3; before their population declines go unnoticed. An impor- Table S4) is an important advantage over MLM in this tant outcome of the analytical approach used here is the case. probability estimates for a species being within each of Data deficiency is a major hindrance to conserva- the Red List categories (Figure 3, Table S5), which allow tion planning in taxonomic groups with a large propor- ranking of data deficient species both among and within tion of data deficient species. Despite the precaution- categories. This fine-scale categorization of extinction risk ary recommendation that data deficient species should among data deficient species has important implications be afforded the same degree of protection as threat- for management, policy and conservation planning. For ened taxa (Mace et al. 2008), this is often not the case example, it has been suggested that the areas with higher in practice (Hoffmann et al. 2008). This can result in numbers of data deficient species should be prioritized genuinely threatened species receiving little conserva- in order to tackle data deficiency (Brito 2010). However, tion attention until their populations decline substantially doing so may lead to misplaced effort if the goal is to (Howard & Bickford 2014). Modeling risk status of data maximize protection of threatened species. For example, sufficient species is a very cost-effective way to iden- while the Coral Triangle in SE Asia is the global hotspot tify high-risk data deficient species for preferential re- for data deficient groupers, none of the data deficient assessment within a reasonable time frame (Bland et al. species predicted to be threatened by our model occur in 2015b). The method presented here provides one simple, the Coral Triangle. Instead, they are found in the Western defensible way to overcome this pressing conservation Indian Ocean, the Tropical Eastern Pacific, in the Eastern problem. Atlantic, and in the Mediterranean Sea (Figure 4). To date, most models aimed at predicting the conser- vation status of data deficient species have been based on Acknowledgments machine-learning methods (MLM). This approach is well suited to data sets with many hundreds or thousands of O.J.L. thanks the Quantitative Ecology and Evolution species, as is the case of broader taxonomic groups with Group at Macquarie University for financial support. good coverage by IUCN assessments such as mammals O.J.L. was supported by the New South Wales Envi- and amphibians (Howard & Bickford 2014; Bland et al. ronmental Trust. E.M.P.M. was supported by the World 2015a; Jetz & Freckleton 2015). However, predicting the Wildlife Fund’s Kathryn S. Fuller Science for Nature Fund status of data deficient species among narrow taxonomic and an Australian Research Council DECRA Fellowship groups (single families or genera) or groups for which (project number DE120102614). J.S.M. was supported data are available for just a small subset of the species by an Australian Research Council Future Fellowship. remains challenging. The performance of MLM is lim- We thank A. Grech for help with mapping, A. Allen for 348 Conservation Letters, September/October 2016, 9(5), 342–350 Copyright and Photocopying: 2016 The Authors. Conservation Letters published by Wiley Periodicals, Inc. O. J. Luiz et al. Threatened status of data deficient groupers statistical advice, and V. Giglio and two anonymous re- Brito, D. (2010). Overcoming the Linnean shortfall: data viewers for comments on the manuscript. deficiency and biological survey priorities. Basic Appl. Ecol., 11, 709-713. Cardillo, M. & Meijaard, E. (2012). Are comparative studies of Supporting Information extinction risk useful for conservation? Trends Ecol. Evol., 27, 167-171. Additional Supporting Information may be found in the Christensen, R.H.B. (2013). Package “Ordinal”: regression models online version of this article at the publisher’s web site: for ordinal data. Copenhagen, Denmark. (Version Figure S1. (A) Boxplots of the significant explanatory 2012.09-11). variables in data sufficient (n = 113) and data deficient http://www.cran.r-project.org/package=ordinal/. species (n = 50). (B) Modeled relationships among sig- Craig, M.T., Mitcheson, Y.J.S. & Heemstra, P.C. (2011). Groupers of the world. NISC, Grahamstown, South Africa. nificant explanatory variables and data deficient species. Elith, J., Leathwick, J.R. & Hastie, T. (2008). A working guide Table S1. Model selection for a) the effect of including to boosted regression trees. J. Anim. Ecol., 77, 802-813. the interaction of region with each variable and b) the ef- Fisher, D.O. & Owens, I.P. (2004). The comparative method fect of dropping variables in a backward stepwise manner in conservation biology. Trends Ecol. Evol., 19, 391-398. Table S2. Model selection for a) the effect of including Gamer, M., Lemon, J. & Fellows, I. (2012). Package irr: the interaction of region with each variable and b) the ef- various coefficients of interrater reliability and agreement. fect of dropping variables in a backward stepwise manner (Version 0.84). http://CRAN.R-project.org/package=irr Table S3. Confusion matrix of the training data pre- Giglio, V.J., Luiz, O.J. & Gerhardinger, L.C. (2015). Depletion dictions and accuracy measures for the selected CLMM of marine megafauna and shifting baselines among predicting the IUCN Red List conservation status category artisanal fishers in eastern Brazil. Anim. Conserv., 18, in groupers (n = 113) 348-358. Table S4. Accuracy measures (PCC) for the training Good, T.C., Zjhra, M.L. & Kremen, C. (2006). Addressing data data predictions of the selected CLM on the full training deficiency in classifying extinction risk: a case study of a set and on the model calibrated on a 75% training set radiation of Bignoniaceae from Madagascar. Conserv. Biol., and 25% validation set. The split set was randomly drawn 20, 1099-1110. 1,000 times to calculate the mean and standard error (SE) Hand, D.J. (2012). Assessing the performance of classification Table S5. Estimates of probabilities of data deficient methods. Int. Stat. Rev., 80, 400-414. species being classed in each Red List category Hawkins, J.P., Roberts, C.M. & Clark, V. (2000). The Table S6. Confusion matrix for the training data threatened status of restricted-range coral reef fish species. predictions and accuracy measures for the machine- Anim. Conserv., 3, 81-88. learning∗ model predicting the IUCN Red List conserva- Hoffmann, M., Brooks, T.M., da Fonseca, G.A. et al. (2008). tion status category in groupers (n = 113) Conservation planning and the IUCN Red List. Endang. This material is available as part of the online ar- Species Res., 6, 113-125. Howard, S.D. & Bickford, D.P. (2014). Amphibians over the ticle from: http://www.blackwell-synergy.com/doi/full/ edge: silent extinction risk of data deficient species. Divers. 10.1111/j.1755–263X.2008.00002.x Distrib., 20, 837-846. (This link will take you to the article abstract). IUCN. (2015). The IUCN red list of threatened species. Version 2015-4. http://www.iucnredlist.org. Downloaded Accessed References on 19 October 2015. Jennings, S., Reynolds, J.D. & Polunin, N.V. (1999). Predicting Bender, M.G., Floeter, S.R., Mayer, F.P. et al. (2013). the vulnerability of tropical reef fishes to exploitation with Biological attributes and major threats as predictors of the phylogenies and life histories. Conserv. Biol., 13, 1466-1475. vulnerability of species: a case study with Brazilian reef Jetz, W. & Freckleton, R.P. (2015). Towards a general fishes. Orix, 47, 259-265. framework for predicting threat status of data-deficient Bland, L.M., Collen, B., Orme, C.D.L. & Bielby, J. (2015a). species from phylogenetic, spatial and environmental Predicting the conservation status of data-deficient species. information. Philos. Trans. R. Soc. B, 370, 20140016. Conserv. Biol., 29, 250-259. Levin, P.S. & Grimes, C.B. (2002). Reef fish ecology and Bland, L.M., Orme, C.D.L., Bielby, J., Collen, B., Nicholson, E. grouper conservation and management. Pages 377-390 in & McCarthy, M.A. (2015b). Cost-effective assessment of P.F. Sale, editor. Coral reef fishes: dynamics and diversity in a extinction risk with limited information. J. Appl. Ecol., 52, complex ecosystem. Academic Press, San Diego. 861-870. Lindfield, S.J., McIlwain, J.L. & Harvey, E.S. (2014). Depth Bridge, T.C., Hughes, T.P., Guinotte, J.M. & Bongaerts, P. refuge and the impacts of SCUBA spearfishing on coral reef (2013). Call to protect all coral reefs. Nat. Clim. Change, 3, fishes. PloS One, 9, e92628. 528-530. Conservation Letters, September/October 2016, 9(5), 342–350 Copyright and Photocopying: 2016 The Authors. Conservation Letters published by Wiley Periodicals, Inc. 349 Threatened status of data deficient groupers O. J. Luiz et al. Mace, G.M., Collar, N.J., Gaston, K.J. et al. (2008). us/app/fishes-greater-caribbean/id570048678?mt=8. Quantification of extinction risk: IUCN’s system for Accessed September 2015. classifying threatened species. Conserv. Biol., 22, Rodrigues, A.S., Pilgrim, J.D., Lamoreux, J.F., Hoffmann, M. 1424-1442. & Brooks, T.M. (2006). The value of the IUCN Red List for Madin, E.M.P., Gaines, S.D. & Warner, R.R. (2010). Field conservation. Trends Ecol. Evol., 21, 71-76. evidence for pervasive indirect effects of fishing on prey Rudd, M.A. & Tupper, M.H. (2002). The impact of Nassau foraging behavior. Ecology, 91, 3563-3571. grouper size and abundance on scuba diver site selection McClenachan, L. (2015). Extinction risk in reef fishes. Pages and MPA economics. Coast. Manage., 30, 133-151. 199-207 in C. Mora, editor. Ecology of fishes on coral reefs. Russell, G.M., Gutherz, E.J. & Barans, C.A. (1988). Evaluation Cambridge University Press, Cambridge. of demersal longline gear off South Carolina and Puerto Miller, R.M., Rodr´ıguez, J.P., Aniskowicz-Fowler, T. et al. Rico with emphasis on deep-water reef fish stocks. Mar. (2006). Extinction risk and conservation priorities. Science, Fish. Rev., 50, 26-31. 313, 441-441. Sadovy de Mitcheson, Y. & Colin, P., editors. (2012). Reef fish Morais, A.R., Siqueira, M.N., Lemes, P., Maciel, N.M., De spawning aggregations: biology, research and management. Marco, P. & Brito, D. (2013). Unraveling the conservation Springer, New York. status of data deficient species. Biol. Conserv., 166, 98- Sadovy de Mitcheson, Y., Craig, M.T., Bertoncini, A.A. et al. 102. (2013). Fishing groupers towards extinction: a global Possingham, H.P., Andelman, S.J., Burgman, M.A., Medellı́n, assessment of threats and extinction risks in a billion dollar R.A., Master, L.L. & Keith, D.A. (2002). Limits to the use of fishery. Fish Fish., 14, 119-136. threatened species lists. Trends Ecol. Evol., 17, 503-507. Saenz-Arroyo, ´ A., Roberts, C.M., Torre, J. & Carino-Olvera, ˜ R Core Team. (2015). R: a language and environment for M. (2005). Using fishers’ anecdotes, naturalists’ statistical computing. R Foundation for Statistical Computing, observations and grey literature to reassess marine species Vienna, Austria. at risk: the case of the Gulf grouper in the Gulf of Robertson, D.R. & Allen, G. (2012). Fishes: east pacific. An California, Mexico. Fish Fish., 6, 121-133. identification guide to the shore-fish fauna of the tropical eastern Sousa-Baena, M.S., Garcia, L.C. & Peterson, A.T. (2014). pacific. (iOS App. Copyright Smithsonian Institution, Left Knowledge behind conservation status decisions: data basis Coast R&C, Santa Cruz, California). Available at: https:// for “Data Deficient” Brazilian plant species. Biol. Conserv., itunes.apple.com/us/app/fishes-east-pacific/ 173, 80-89. id494644648?mt=8. Accessed September 2015. Stallings, C.D. (2008). Indirect effects of an exploited Robertson, D.R. & Van Tassel, J. (2012). Fishes: Greater predator on recruitment of coral-reef fishes. Ecology, 89, Caribbean. An identification guide to the shore-fish fauna of the 2090-2095. Caribbean and adjacent areas. (iOS App. Copyright Webb, T.J. & Mindel, B.L. (2015). Global patterns of Smithsonian Institution, Left Coast R&C, Santa extinction risk in marine and non-marine systems. Curr. Cruz, California). Available at: https://itunes.apple.com/ Biol., 25, 506-511. 350 Conservation Letters, September/October 2016, 9(5), 342–350 Copyright and Photocopying: 2016 The Authors. Conservation Letters published by Wiley Periodicals, Inc.

Journal

Conservation Letters – Wiley

Published: Sep 1, 2016

Keywords: ; ; ; ; ; ;

There are no references for this article.