Does size matter? A comment on "The Perils of Picky Eating: Dietary Breadth is related to Extinction Risk in Insectivorous Bats" by Justin G. Boyles and Jonathan J. Storm published 2007 in Plos One. Kamran Safi University of Zurich, Winterthurerstrasse 190, CH-8057 Zurich August 2007

Introduction Identifying factors that determine the extinction risk of species is an important task for conservation biologists [1]. Recognizing the correlates of extinction proneness and understanding the mechanisms influencing population decline and species extinction allow to accurately assess extinction risk. This can ultimately be used to formulate effective action plans for the conservation of threatened species [1,2,3,4,5,6,7,8,9]. Boyles and Storm have shown in their comparative study that in insectivorous vespertilionid bats (family Vespertilionidae, Chiroptera) dietary niche breadth correlates negatively with increasing conservation relevance measured with IUCN thread categories. It may be intuitively obvious that species with narrow dietary niches, i.e. specialists, are at a larger risk of extinction than generalists with broad dietary niches [10]. However, this is not supported by empirical evidence, as some studies report a relationship between dietary specialization [6,9,11,12] while others do not [6,8,13,14,15,16]. The findings of Boyles and Storm are important in this ontext as they allow to identify mechanistic pathways of population decline caused by humans in a series of bat species. They also show a relationship with the degree of dietary specialisation. Clearly the data that Boyles and Storm have used indicate that with increasing dietary specialisation bat species tend to become more prone to extinction. This is in line with theoretical predictions [17] and the evidence from some empirical studies [6,9,11,12]. However, it also and more importantly refutes the recent results of Safi and Kerth [18]. A study where no relationship between dietary specialisation and extinction risk was found. Interestingly Safi and Kerth also used IUCN ranking and the same dietary specialisation index for an overlapping dataset of species. The question arises why did Boyles and Storm find what Safi and Kerth could not. Below I discuss several possible explanations for the diverging results, not only due to the personal and scientific importance, but mainly because of the conservation relevance of the topic. The questions are whether there is a relationship between dietary specialisation and extinction risk. Finding the cause(s) for the different results of the two studies and their consequences for the conservation of bats should emphasize the importance of Boyles and Storm's finding in the light of the previous studies. Identifying the cause of discrepancy is also relevant as it may help to gain new insight, which result from contrasting the results between the studies. Thus by investigating the factors that have lead to the differences between the two studies, we may find evidence for confounding effects influencing extinction risk beyond what each of the two studies could have found alone. Four potential reasons for the differences between the studies can be identified (and in most cases rejected). Different methods of phylogenetic correction Theoretically the use of different methods for phylogenetic correction coud have caused differences in the results. Safi and Kerth used independent contrasts generated with the software CAIC [19,20] Boyles and Storm used PDAP which performs phylogenetically controlled analyses of variance [21]. The major advantage of PDAP is the power in correcting for phylogenetic relationship without loosing datapoints and degrees of freedom, perhaps rendering the test more powerful than independent contrasts [21,22].

However, the discrepancies found between the studies are present at the species level and consequently the methods correcting for shared ancestry cannot explain them. In addition, the results at species and phylogenetically corrected level do not differ between the studies, indicating that there is no or only negligible phylogenetic signal in the data [22,23]. Thus, as Boyles and Storm have also recognized, the methodology for correcting for common ancestry can not be responsible for the differences. I will therefore also focus only on species level differences. Sample size and species composition Boyles and Storm used a larger data set than Safi and Kerth (N=44 vs. N=35), which may have resulted in improved statistical resolution. In addition, Boyles and Storm used only vespertilionid bats. Safi and Kerth also included rhinolophid species, which use a fundamentally different way of echolocation [24,25,26]. These differences in species composition may have masked an effect which is mainly found in the family Vespertilionidae but not in other bats. Finally Boyles and Storm also included species from Australia which Safi and Kerth did not do. I therefore tested whether the species composition and sample size explain the differences between the studies by repeating their analysis with the dataset of Safi and Kerth. Using IUCN as continuous vs. categorical variable The differences between the two studies may be a consequence of the different ways IUCN extinction risk categories were treated in the statistical analyses. Safi and Kerth assumed IUCN categories of extinction risk to be a continuous measure of extinction risk, which is an assumption other authors have also used [1]. However, this may be problematic, since it cannot be assumed that the distances between categories are linear and equal. To avoid this Boyles and Storm used two different approaches: a dichotomous classification dividing the species into conservation relevant and non-relevant species and using the categories in classical ANOVAs (analysis of variance). If treating IUCN categories as continuous variables has lead to the loss of a relationship between dietary niche breadth and extinction risk, one would expect that the results reported by Boyles and Storm would also become non-significant if IUCN ranks were treated as a continuous measure. Methods in determining dietary composition Boyles and Storm included only studies quantifying insect components in bat faeces by their percentage of volume (%V). Studies using %V determine the volumes of each recognizable insect order in each faecal pellet, dividing it by the total volume of all insect remains and multiplying it by 100 (total n = 100%) [27]. In contrast, Safi & Kerth included also studies determining dietary composition by percentage of occurrence (%O) and percentage of frequency (%F). Percentage of occurrence (%O) is defined as the number of times a prey type was encountered, divided by the number of analysed droppings, multiplied by 100 (total n > 100%) [27]. Percentage of frequency (%F) is defined as the number of occurrences of a particular prey type, divided by the total number of occurrences of all prey types, multiplied by 100 (total n = 100%) [27]. As %O and %F can be converted into each other by a simple equation, I will henceforth refer only to %F. In contrast, %V cannot be easily converted into corresponding values of %F (or %O) and vice versa. Safi and Kerth used a best fit line to convert %F into %V and used only the former for their analyses. Although Safi and Kerth give detailed information about the accuracy of their conversion, the estimation using a best fit line represents only an approximation and it is possible that by doing so the converted values contained too much statistical noise. This would imply that prior to the conversion the data contained a relationship with extinction risk. Another possible explanation, could be that Boyles and Storm used %V only. The method used in determination of insect remnants in faeces produces biases in the relative importance

of prey. Whereas studies determining %V overestimate the contribution of large prey in the diet, studies using %F overestimate the relative contribution of small insects in the diet [27,28]. In addition, small and soft bodied insects may be partly or wholly digested and rendered unidentified and legs and wings of larger insects may be culled before consumption [27,28]. Thus calculating the dietary niche breadth using %V or %F may lead to different values. The differences in niche breadth from such methodological biases have potential implications for conservation . A relationship between dietary niche breadth and conservation relevance from studies using %V only would mean that species relying on large prey and/or prey that is not very well digested are actually at a a higher risk of extinction. Therefore I reevaluated the unconverted data of Safi and Kerth and tested for a correlation between conservation relevance and niche breadth once for studies determining %F and once for studies using %V. In addition, I compared which insect taxa show the largest shift in relative contribution in studies where both %F and %V data were given. An insect taxon, which is consistently overestimated in %V studies compared to %F datasets, could cause a narrower dietary niche breadth estimation and therefore be an indicator of conservation relevance.

Methods Due to limited sample size I used only the dichotomous classificatiof conservation relevant (near threatened or above) or non-relevant (least concern). I used the additional online material for the analysis of Boyles and Storm‘s data. For the re-evaluation of Safi and Kerth’s I used unpublished data which was used as the basis for the study (and references therein) and which I expanded by including a few recent publications on diet in bats (cite?). For the nonparametric statistics I calculated exact p-values namely Wilcoxon two sample tests using SAS-lab statistical package version 9.1. Tests were considered significant at p≤0.05. The contrasts between the %F and %V were calculated from studies giving both values. For each paired value in these studies (referenced in 18) I calculated the difference in the percentages (%F-%V). In a generalized linear model, I tested whether independent of the study, these differences deviated from zero depending on the insect taxon.

Results

Niche breadth: Bowyles & Storm

In general in both studies niche breadth was comparable (figure 1). Therefore the bias introduced by a conversion of the data did not result in completely different dietary niche breadths. 5

y = 0.9482x + 0.3408 R2 = 0.6719

4 3 2 1 0 0

1

2

3

4

5

Niche breadth: Safi & Kerth

Figure 1: Correlation between dietary niche breadth in the studies by Boyles and Storm and Safi and Kerth (r=0.82, p<0.0001). Dark dots are species of conservation relevance whereas the grey dots depict species which are non-relevant.

Sample size and species composition Sample size and species composition seem not to influence the qualitative outcome of the results. After reducing the data presented in Boyles and Storm to those species which were used in both studies, the correlation between niche breadth and conservation relevance remains significant for the Boyles and Storm data (non-parametric: Nrelevant species=6 Nnon-relevant species=15, chi-square=5.8, d.f.=1, p=0.02; parametric: F1,19=5.4, p=0.03). The same analysis using species present in both studies remains non-significant for the data of Safi & Kerth (p>0.1). Using IUCN as continuous vs. categorical variable A simple correlation on the data of Boyles and Storm reveals that that IUCN rank treated as a continuous variable and dietary niche breadth correlate negatively (r=-0.44 p=0.003) in Boyles and Storm data set. Therefore the fact that Safi & Kerth treated IUCN classification as continuous characters cannot explain the discrepancy either, since it is not possible to reproduce the same outcome as previously found by Safi & Kerth by treating IUCN ranks as continuous. Methods in determining dietary composition The re-evaluation of the data of Safi & Kerth reveals that mean dietary niche breadth calculated from %F studies does not correlate with conservation relevance (table 1). However the same analysis for niche breadth calculated from %V, results in a significant negative relationship also in the data presented in Safi & Kerth (table 1). Therefore the lack of a relationship between dietary niche breadth and extinction risk in the study of Safi and Kerth seems to be a result of including %F and %O to determine dietary niche breadth in the analyses. Table 1: Non-parametric tests of the relationship between dietary niche breadth and conservation relevance once for niche breadth calculated from %F studies and once for niche breadth calculated from %V. Conservation non- S Mean score Mean score non- Std Dev P Type Conservation relevant species relevant species relevant species relevant species under H0 (exact) %F 5 23 56 11.2 15.2 16.7 0.35 %V 8 19 71 8.9 16.2 18.8 0.03

Determining invertebrate taxa using %F and %V did not result in a consistent overestimation of a particular taxon (figure 2). In those studies which determined both %F and %V only the contribution of lepidopterans was significantly underestimated irrespective of the study (GLM estimate for Lepidoptera=-9.1±2.8, t=-3.2, p=0.002). In all other taxa the difference between the %F and %V values did not differ from zero, however, %F values seem to be on average larger than the corresponding %V values, since most contrasts have positive differences (figure 2).

Trichoptera

Orthoptera

Neuroptera

Lepidoptera

Hymenoptera

Hemiptera/Heteroptera

Ephemeroptera

Diptera

Coleoptera

Arachnida 10 5 0 -5 -10 -15

Figure 2: GLM estimates of the differences between %F and %V for the different taxa for which both methods of dietary composition were applied. Only Lepidoptera deviate negatively from zero after correction for the influence of the study (p=0.002).

Discussion Converting the insect proportions using the formula used by Safi & Kerth does not seem to change the relative importance of the insect orders in comparison to a truly determined %V. What obviously does matter is how the diet was quantified using %F or %V, a finding, which could be of conservation relevance. The analyses above suggest that dietary niche breadth calculated from %V does not only estimate the dietary specialization in general but contains a not unimportant component related with the specialisation on prey size. Determining the proportions of insect components by measuring the volume of remnants (or their weight) will result in a over representation of large (and possibly hard bodied) insects. The present evidence suggests that different accuracy of the methods may depend more on prey size, because no systematic error in specific prey taxa was found. Large species of bees and beetles suffer from higher population losses that small bodied species [29]. Also for other taxa a relationship between body-size and extinction rate has been demonstrated [30,31]. Therefore bat species specializing on large prey could face a larger problem than those which opportunistically take all prey sizes. Therefore, although both methods cannot depict the reality truly, a systematic attribute of dietary determination using %V seems to indicate that there might be indeed a peril of picky eating, in this case specializing on a few insect orders from which large prey are preferred. Ultimately we would need studies simultaneously determining %V and %F in a large range of conservation relevant and abundant species to reliably assess those variables with the power of predicting extinction proneness and dietary niche breadth. Determining the preferred prey size in a series of bat species would also help to disentangle potential influences of specialization on prey diversity and prey size. Finally, those species in the study of Boyles and Storm, which are not strong specialists but are categorized as conservation relevant may not specialize on particular insect orders but generally take large prey. Such bat species would have a relatively large dietary niche breadth, but due to the uniformly large prey resemble a

generalist much the same the way as a species capturing small but similarly diverse insects. And accordingly, there might be species which are specialists, but considered conservationirrelevant, simply because they specialize on small species from a few insect orders. A potential impact of specialisation on large prey could also go hand in hand with the detected relationship between habitat specialization and extinction risk in bats. Large hard bodied prey could be more strongly restricted to living on substrate than the rest of insects, because of large body mass. Many bat species which take large prey would consequently be gleaners. Gleaning bats are characterized by broad wings which render them highly manoeuvrable but inefficient flyers. Although Boyles and Storm do not find a relationship between dietary niche breadth and wing morphology, such species may face a higher extinction risk by two mutually acting processes. In conclusion, Boyles and Storm have demonstrated an important relationship between diet and extinction risk emphasizing the merits of dietary analyses in particular the determination of percent of the volume which prey categories make up in the diet, for conservation. Interestingly the discrepancy with previous findings could in fact result in a fruitful investigation. First in a thorough determination of the reason why extinction risk correlates with %V and not with %F and second determining the importance of prey size for extinction risk in bats. Whether extinction risk is directly determined by prey size, hardness, or is ultimately a consequence of morphological adaptation to particular foraging strategies and prey size, which all additionally limit habitat availability and accessibility, remains to be investigated.

Acknowledgement I would like to thank D.K.N. Dechmann for help on this comment.

References 1. Purvis A, Gittleman JL, Cowlishaw G, Mace GM (2000) Predicting extinction risk in declining species. Proceedings of the Royal Society of London Series B-Biological Sciences 267: 1947-1952. 2. Brashares JS (2003) Ecological, behavioral, and life-history correlates of mammal extinctions in West Africa. Conservation Biology 17: 733-743. 3. Cardillo M, Bromham L (2001) Body size and risk of extinction in Australian mammals. Conservation Biology 15: 1435-1440. 4. Crandall KA, Bininda-Emonds ORP, Mace GM, Wayne RK (2000) Considering evolutionary processes in conservation biology. Trends in Ecology & Evolution 15: 290-295. 5. Crooks KR (2002) Relative sensitivities of mammalian carnivores to habitat fragmentation. Conservation Biology 16: 488-502. 6. Harcourt AH, Coppeto SA, Parks SA (2002) Rarity, specialization and extinction in primates. Journal of Biogeography 29: 445-456. 7. Koh LP, Sodhi NS, Brook BW (2004) Ecological correlates of extinction proneness in tropical butterflies. Conservation Biology 18: 1571-1578. 8. Laurance WF (1991) Ecological correlates of extinction proneness in Australian tropical rain-forest mammals. Conservation Biology 5: 79-89. 9. Owens IPF, Bennett PM (2000) Ecological basis of extinction risk in birds: Habitat loss versus human persecution and introduced predators. Proceedings of the National Academy of Sciences of the United States of America 97: 12144-12148. 10. Davies KF, Margules CR, Lawrence JF (2004) A synergistic effect puts rare, specialized species at greater risk of extinction. Ecology 85: 265-271. 11. Foufopoulos J, Ives AR (1999) Reptile extinctions on land-bridge islands: Life-history attributes and vulnerability to extinction. American Naturalist 153: 1-25. 12. Hughes JB, Daily GC, Ehrlich PR (2000) Conservation of insect diversity: a habitat approach. Conservation Biology 14: 1788-1797. 13. Hopkins GW, Thacker JI, Dixon AFG, Waring P, Telfer MG (2002) Identifying rarity in insects: the importance of host plant range. Biological Conservation 105: 293-307.

14. Reed JM (1995) Relative vulnerability extirpation of montane breeding birds in the Great-Basin. Great Basin Naturalist 55: 342-351. 15. Reed RN, Shine R (2002) Lying in wait for extinction: Ecological correlates of conservation status among Australian elapid snakes. Conservation Biology 16: 451-461. 16. Wilson JD, Morris AJ, Arroyo BE, Clark SC, Bradbury RB (1999) A review of the abundance and diversity of invertebrate and plant foods of granivorous birds in northern Europe in relation to agricultural change. Agriculture Ecosystems & Environment 75: 13-30. 17. Begon ME, Harper JL, Townsend CR (1996) Ecology, Third Edition. Held A, translator. Oxford: Blackwell Science Limited. 18. Safi K, Kerth G (2004) A comparative analysis of specialization and extinction risk in temperate-zone bats. Conservation Biology 18: 1293-1303. 19. Felsenstein J (1985) Phylogenies and the comparative method. American Naturalist 125: 1-15. 20. Purvis A, Rambaut A (1995) Comparative-analysis by independent contrasts (CAIC) - an Apple-Macintosh application for analyzing comparative data. Computer Applications in the Biosciences 11: 247-251. 21. Garland T, Dickerman AW, Janis CM, Jones JA (1993) Phylogenetic analysis of covariance by computersimulation. Systematic Biology 42: 265-292. 22. Garland T, Harvey PH, Ives AR (1992) Procedures for the analysis of comparative data using phylogenetically independent contrasts. Systematic Biology 41: 18-32. 23. Diaz-Uriarte R, Garland T (1996) Testing hypotheses of correlated evolution using phylogenetically independent contrasts: Sensitivity to deviations from Brownian motion. Systematic Biology 45: 27-47. 24. Schnitzler HU, Moss CF, Denzinger A (2003) From spatial orientation to food acquisition in echolocating bats. Trends in Ecology & Evolution 18: 386-394. 25. Schnitzler HU, Kalko EKV (2001) Echolocation by insect-eating bats. Bioscience 51: 557-569. 26. Neuweiler G (1984) Foraging, echolocation and audition in bats. Naturwissenschaften 71: 446-455. 27. McAney CM, Shiel CB, Sullivan C, Fairley JS (1991) The Analysis of Bat Droppings. London: The Mammal Society. 28. Kunz TH, Whitaker JO (1983) An evaluation of fecal analysis for determining food habits of insectivorous bats. Canadian Journal of Zoology-Revue Canadienne De Zoologie 61: 1317-1321. 29. Larsen TH, Williams NM, Kremen C (2005) Extinction order and altered community structure rapidly disrupt ecosystem functioning. Ecology Letters 8: 538-547. 30. Cardillo M (2003) Biological determinants of extinction risk: why are smaller species less vulnerable? Animal Conservation 6: 63-69. 31. Gaston KJ, Blackburn TM (1995) Birds, body-size and the threat of extinction. Philosophical Transactions of the Royal Society of London Series B-Biological Sciences 347: 205-212.