1 Threat to the point: improving the value of comparative extinction risk analysis for 2 conservation action 3 4 Running head: Addressing threats in extinction risk studies 5 6 Authors: Kris A. Murray1*, Luis D. Verde Arregoitia 2, Ana Davidson3, Moreno Di Marco4, 7 and Martina M. I. Di Fonzo2,5 8 9 Author addresses: 10 1EcoHealth Alliance, 460 West 34th Street, New York, NY, USA 11 2School of Biological Sciences, University of Queensland, QLD, 4072, Australia 12 3Department of Ecology and Evolution, Stony Brook University, Stony Brook, NY, USA 13 4Global Mammal Assessment program, Department of Biology and Biotechnology, Sapienza 14 University of Rome, Viale dell' Università 32, I-00185 Rome, Italy. 15 5ARC Centre of Excellence for Environmental Decisions, the NERP Environmental 16 Decisions Hub, Centre for Biodiversity & Conservation Science, University of Queensland, 17 Brisbane, QLD, 4072, Australia 18 19 *Corresponding author: [email protected] 20 Mailing address as above; ph +1 347 475 9367; fax +1 212 380 4465. 21 22 Keywords: threats, intrinsic traits, extrinsic traits, IUCN Red-List, management, 23 prioritisation, declines, prediction, biodiversity 24 25 Type of paper: Primary Research Article 1 26 27 Words: Abstract (298), Main text (5805) 28 References: General references (71); Data source references (98) 29 Figures = 6, Tables = 2 30 31 2 32 Abstract 33 Comparative extinction risk analysis is a common approach for assessing the relative plight 34 of biodiversity and making conservation recommendations. However, the usefulness of such 35 analyses for conservation practice has been questioned. One reason for underperformance 36 may be that threats arising from global environmental changes (e.g., habitat loss, invasive 37 species, climate change) are often overlooked, despite being widely regarded as proximal 38 drivers of species’ endangerment. We explore this problem by 1) reviewing the use of threats 39 in this field, and 2) quantitatively investigating the effects of threat exclusion on the 40 interpretation and potential application of extinction risk model results. We show that threat 41 variables are routinely (59%) identified as significant predictors of extinction risk, yet while 42 most studies (78%) include extrinsic factors of some kind (e.g., geographic or bioclimatic 43 information), the majority (63%) do not include threats. Despite low overall usage, studies 44 are increasingly employing threats to explain patterns of extinction risk. However, most 45 continue to employ methods developed for the analysis of heritable traits (e.g., body size, 46 fecundity), which may be poorly suited to the treatment of non-heritable predictors including 47 threats. In our global mammal and continental amphibian extinction risk case studies, 48 omitting threats reduced model predictive performance, but more importantly 1) reduced 49 mechanistic information relevant to management, 2) resulted in considerable disagreement in 50 species classifications (12% and 5% for amphibians and mammals, respectively, translating 51 to dozens and hundreds of species), and 3) caused even greater disagreement (20-60%) in a 52 downstream conservation application (species ranking). We conclude that the use of threats 53 in comparative extinction risk analysis is important and increasing but currently in the early 54 stages of development. Priorities for future studies include improving uptake, availability, 55 quality and quantification of threat data, and developing analytical methods that yield more 56 robust, relevant and tangible products for conservation applications. 3 57 Introduction 58 Present rates of global species extinction are two to three orders of magnitude greater than 59 background levels recorded in geologic history (Barnosky et al., Pimm et al., 1988, Pimm et 60 al., 1995). Around 20% of assessed vertebrate species are currently listed as threatened under 61 the International Union for the Conservation of Nature (IUCN) Red List (Hoffmann et al., 62 2010). Human impacts, such as agricultural expansion, logging, overexploitation, climate 63 change and invasive species, are widely recognised as the major drivers of the current global 64 extinction crisis (e.g., Davies et al., 2006). Inadequate alleviation of such threats is the key 65 factor limiting the effectiveness of current conservation strategies aiming to reduce or reverse 66 biodiversity loss (Hoffmann et al., 2010, Tranquilli et al., 2012). 67 68 The risk of a species’ decline in abundance or extinction (often measured as a relative level of 69 endangerment, e.g., as per the IUCN Red List) is nevertheless also strongly influenced by its 70 intrinsic life-history and ecological traits, such as body size, fecundity or ecological 71 versatility (Cardillo et al., 2004, Fisher et al., 2003, Lee & Jetz, 2011, Purvis et al., 2005). 72 For instance, relatively larger-bodied and longer-lived bird species are more likely to go 73 extinct as a result of overhunting, whereas small-bodied species with high habitat 74 specialisation are more likely to decline in response to habitat loss (Owens & Bennett, 2000). 75 Extinction risk is thus shaped by interactions that arise between species’ intrinsic traits and 76 the extrinsic factors including threats to which it may be exposed (Fig 1). 77 78 Comparative extinction risk analysis has been one tool used to explore these complex 79 interactions. By synthesizing global or regional-scale information across large numbers of 80 species, these macroecological analyses can yield important conservation recommendations 81 by elucidating why some species are more vulnerable to extinction processes than others 4 82 (e.g., Cardillo et al., 2006, Cooper et al., 2008, Davidson et al., 2009, Di Marco et al., 2012, 83 González-Suárez & Revilla, 2012, Isaac et al., 2009, Jones et al., 2003, Koh et al., 2004, 84 Purvis et al., 2000). 85 86 In a recent review, however, Cardillo and Meijaard (2012) highlight a general lack of 87 influence that comparative extinction risk analyses have had on conservation practice and 88 policy (i.e., "primarily academic exercises"). One reason for this underperformance may be 89 that these studies seem to focus more heavily on species’ intrinsic traits, reflecting general 90 patterns of ‘vulnerability’ or ‘extinction proneness’ (Purvis et al., 2005, Reynolds, 2003), 91 than on their more immediate and potentially manageable causes of endangerment: extrinsic 92 threats (Fisher et al., 2003). 93 94 Although trait-based responses can sometimes reflect or help identify the nature of 95 underlying or unknown threats (e.g., Williams & Hero, 1998), failing to explicitly consider 96 threats in comparative extinction risk analyses could be limiting for several reasons. In 97 models that employ intrinsic traits only as predictor variables, there is an implicit assumption 98 that susceptibility to decline or extinction is the result of trait-based vulnerability to some 99 uniform, average or overarching threat (Fritz et al., 2009). However, we cannot assume that 100 all threats are equal, or that a species’ response to one threat will be correlated with its 101 response to others. 102 103 For example, in addition to interactions that might arise between extrinsic threats and 104 intrinsic traits (Fig. 1), threats often vary spatially and/or temporally in their presence and 105 magnitude (Davies et al., 2006, Harcourt & Parks, 2003, Myers, 1988, Wilcove et al., 1998), 106 and their prevalence can vary significantly between taxonomic groups (Mace & Balmford, 5 107 2000). Threats could also influence the decline of a population at multiple spatial or temporal 108 scales (Wilcove et al., 1998). Different threats may also cause different types of population 109 responses, depending on distinctions in the rate at which they cause mortality over time (Di 110 Fonzo et al., 2013, Mace et al., 2008), and multiple threats may impact a species interactively 111 or additively (e.g., Hof et al., 2011). If adequately quantified, these factors should explain 112 additional variation in extinction risk patterns among species, beyond the variation that might 113 be attributable to the mediating effects of intrinsic traits alone. 114 115 Due to the correlative nature of comparative extinction risk analyses, intrinsic trait-only 116 models must implicitly include the above characteristics and other nuances related to threats 117 (analogous to e.g., species interactions in correlative species distribution models; Pearson & 118 Dawson, 2003). As such, while performance of trait-based extinction risk analyses might be 119 reasonable for a given case study (e.g., Davidson et al. (2009) was able to classify threat 120 listings in mammals globally with >80% accuracy based solely on intrinsic trait data), models 121 failing to explicitly include threats may be poorly suited to extrapolation or have limited 122 transferability into different regions, taxa or timeframes (e.g., future projections). For 123 instance, trait-based models of extinction risk in farmland birds showed poor predictive 124 ability among different regions (Pocock, 2011), perhaps due to varying patterns of key 125 threats. Similarly, Fritz et al. (2009) showed that considerable geographic variation in 126 mammalian extinction risk derived from trait-based models could be explained by 127 considering varying patterns of past anthropogenic impacts. 128 129 Studies that do not consider threats might also suffer from a lack of mechanistic information 130 that may be of