Herbarium records do not predict rediscovery of presumed nationally extinct species

Transcription

1 Biodivers Conserv (2012) 21: DOI /s x ORIGINAL PAPER Herbarium records do not predict rediscovery of presumed nationally extinct species Kwek Yan Chong Serena M. L. Lee Aik Teck Gwee Paul K. F. Leong Samsuri Ahmad Wee Foong Ang Alvin F. S. L. Lok Chow Khoon Yeo Richard T. Corlett Hugh T. W. Tan Received: 2 January 2012 / Accepted: 14 June 2012 / Published online: 28 June 2012 Ó Springer Science+Business Media B.V Abstract Rediscoveries of species previously thought to be extinct present a dilemma to conservation biology. On one hand, such instances offer the chance to change the course of events away from one that would have led to extinctions. On the other hand, public support for conservation may wane if scientists are frequently seen to overstate and prematurely declare extinctions. Recent studies have adopted a probabilistic approach to infer extinction, using sightings or collections and statistical models to calculate the chance that a species may still be extant. We conduct the first broad-scale test of such models using a recently compiled national red list and national herbarium collection records, including collections of presumed nationally extinct species made after the red list publication, which constitute rediscoveries. There was little evidence that the probabilities calculated by these models were associated with rediscoveries over a 3.5-year period. Current probabilistic models of extinction using sighting records could hence be inadequate for use with most natural history collection data. Keywords Extinct species Rediscovery Lazarus effect Herbarium collections Sighting rate equations Preliminary results were presented at the 15th Biological Sciences Graduate Congress at the University of Malaya. Electronic supplementary material The online version of this article (doi: /s x) contains supplementary material, which is available to authorized users. K. Y. Chong (&) W. F. Ang A. F. S. L. Lok C. K. Yeo R. T. Corlett H. T. W. Tan Department of Biological Sciences, National University of Singapore, 14 Science Drive 4, Singapore , Singapore kwek@nus.edu.sg S. M. L. Lee A. T. Gwee P. K. F. Leong S. Ahmad Singapore Botanic Gardens, National Parks Board of Singapore, 1 Cluny Road, Singapore , Singapore

2 2590 Biodivers Conserv (2012) 21: Introduction Species extinction and their prevention are central interests in conservation biology (Pimm and Jenkins 2010). Extinctions, where global, are irreversible, and species missing from an ecosystem have impacts on ecological processes (Sodhi et al. 2009). Yet, extinctions are notoriously difficult to confirm with any degree of confidence. The most famous recent rediscovery would be the ivory-billed woodpecker (Campephilus principalis) thought to be extinct from the southeastern United States with hopes of its persistence kept alive by the occasional unconfirmed sighting, until a team of ornithologists video-recorded what they believed to be the missing species (Fitzpatrick et al. 2005). Evidence of the rediscovery, however, was strongly challenged and debated (Jackson 2006; Sibley et al. 2006a, b), although the authors maintained that their report was scientifically rigorous (Fitzpatrick et al. 2006a, b, c). The controversy surrounding the supposed rediscovery helped to generate media attention and renewed support for conservation of the potential habitat for ivory-billed woodpecker (Mikusinski et al. 2010). In other circumstances, rediscoveries may not have such positive consequences. The IUCN Red List guidelines for classifying species as extinct do not state a time period for carrying out exhaustive surveys, taking into consideration that different taxa have different detection probabilities. For example, cryptic nocturnal animals and small-sized crown epiphytes or lianas would require much more search effort than large, colorful, grown-dwelling animal or plant species. However, exhaustive surveys without a definite timeline to confirm single species extinction is often impractical, and many national and regional red lists have often adopted a window period, varying from 10 years (de Iongh and Bal 2007) to 20 years (Keller et al. 2005) to 50 years (Keith and Burgman 2004) without sightings or collections of a species to categorize them as extinct. Many species may be initially classified as extinct and later rediscovered, resulting in a phenomenon of shrinking lists of extinct species, termed by Keith and Burgman (2004) as the Lazarus effect. As the threat of looming extinctions is often used to urge conservation action (e.g., No. 12 of the targets of the Convention on Biological Diversity for the year 2020 ( mass rediscoveries of species presumed to be extinct may instead desensitize the public to warnings by conservationists and undermine support for biodiversity (Ladle et al. 2004). Premature declarations of extinction may write a species off when it may still have a chance of survival. This has been termed a Romeo error, the most often cited example of which is the Cebu flowerpecker (Diaceum quadricolor) and several other sub-species of birds endemic to the island of Cebu in the Philippines (Collar 1998). Rapid deforestation of the island led Rabor (1959) to consider that these taxa could be extinct. This was subsequently accepted uncritically as a declaration of extinction. Dutson et al. (1993) later discovered that the Cebu flowerpecker, along with four other endemic subspecies, was surviving in the island s last strip of forest. This fragment was precariously at risk of being cleared throughout this time, which would likely have resulted in genuine extinctions. Some researchers have hence suggested that any statement of extinction should be phrased to convey the probability of making a misjudgment (Butchart et al. 2006; Roberts 2006). Several statistical models have been proposed to estimate the probability that a species remains extant, and the associated confidence interval for the estimated time of extinction (Solow 2005; Rivadeneira et al. 2009), from a chronological record of encounters. These models based on sighting records have been demonstrated for the ivory-billed woodpecker (Roberts 2006; Roberts et al. 2010), as well as other well-known extinctions such as the

3 Biodivers Conserv (2012) 21: dodo (Roberts and Solow 2003) and the Carribean monk seal (Monachus tropicalis; Solow 1993a, 2005), and have recently been further extended to collection records from herbaria and natural history museums (McInerny et al. 2006; Roberts et al. 2010). If the probabilistic models can accurately estimate the probability of an extinction event from the valuable information contained in natural history collections, the status of local and global biodiversity can be more efficiently assessed to inform conservation planning. However, so far the probabilistic models have only been demonstrated for their potential but not yet assessed for their actual utility when applied to a full database of collections. Studies have applied the models to the sets of extinct taxa ranging from dung beetles (Carpaneto et al. (2007), to birds (Collen et al. 2010; Roberts et al. 2010), and mammals (Collen et al. 2010), but without tests of accuracy. Duffy et al. (2009) only qualitatively discussed the utility of the models based on field surveys of four orchid species. McInerny et al. (2006) found significant correlation of the p values from their modified version of the equation of Solow (1993a) with the IUCN status of 23 orchid species, but this was a test of ability to infer threat and not extinction as none of these species was extinct. Studies of how Type I and II error rates vary with different trajectories of decline to extinction have been conducted using simulated data (Rivadeneira et al. 2009; Collen et al. 2010), but these may not reflect the limitations of real sighting data or natural history records. A direct test (of whether species predicted to be extinct are really extinct) would be difficult: one would need to know that a set of species is truly extinct to be able to test predictions against it; instead, we tested if species predicted to be extinct are less likely to be subsequently rediscovered. We used the recently digitized database of collections from the Singapore Botanic Gardens Herbarium (SING) and lists of vascular plant species presumed to be extinct in the latest edition of the Singapore Red Data Book (Tan BC et al. 2008; Tan HTW et al. 2008). Extinct vascular plant species in this Red Data Book list was defined as species that have not been seen or collected in the 30 years before the book s compilation in After the book went to press, a number of species presumed nationally extinct were rediscovered. We hypothesized that species assigned with a lower probability of still being extant (i.e., higher probability of being extinct) by the probabilistic models would be less likely to be subsequently rediscovered. Methods The three probabilistic models that we tested all use the n sightings of each species in periods arranged in chronological order i.e., t 1 \t 2 \...\t i 1 \t i \t iþ1 \...\t n 1 \t n. Each unit of the sighting period in our case was the year of collection, t i. The first year of collection for a species is t 1, and the last year that a species was collected is t n. Even if a species was collected multiple times in a given year, it was considered only as a positive sighting for the year. The model of Solow (1993a) assumes a constant rate of sighting a species under the null hypothesis that the species is extant. The probability p that the species is still extant at year T after the end of the collection period is given by: p ¼ ½ðt n t 1 Þ= ðt t 1 Þ The model of Solow and Roberts (2003) is non-parametric and only uses the last two positive sightings of the species: Š n 1

4 2592 Biodivers Conserv (2012) 21: Fig. 1 Total number of vascular plant collections from Singapore deposited in the Singapore Botanic Gardens Herbarium (SING) each year p ¼ ½ðt n t n 1 Þ= ðt t n 1 ÞŠ McInerny et al. (2006) modified the model of Solow (1993a) to take into account different first sightings of observations (and hence different lengths of the sighting periods), which would affect comparability between species. The resulting equation was: p ¼ f1 ½n= ðt n t 1 ÞŠ g T tn For calculations using these equations (henceforth referred to as the extinction equations ), we used SING records only up to 21 Sep 2006 because these were part of the dataset that was used by compilers of the Singapore Red Data Book to determine if a species is considered nationally extinct. Ideally, we should include only wild collections and exclude all collections from cultivated individuals, but this was difficult to determine. Instead, we excluded from our calculations all records in which the collection locality was listed as a nursery or the Singapore Botanic Gardens, with the exception of the rainforest fragment within the Botanic Gardens grounds. Use of synonyms in the database was another major problem that was overcome through thorough checks on alternative names and spellings. Because a key assumption under the null hypothesis underlying all of the probabilistic models is that sighting effort should be non-zero for any time unit of the sighting period, we investigated the trends in numbers of herbarium collections per year as a proxy of overall effort in making botanical collections in each year (e i ; Fig. 1). Botanical collections per year were consistently non-zero starting from 1,874, hence only records from this year onwards were used in calculations. Although the numbers of collections varied across each year, botanical effort only dipped noticeably but briefly during the 43-month long Japanese occupation of Singapore during the Second World War. To account for possible unevenness in collection effort, we employed the modified equation by McCarthy (1998): p ¼ Xtn i¼1 e i = XT i¼1 e i! n 1

5 Biodivers Conserv (2012) 21: We checked the herbarium database for errors where species had been sighted and collected in the 30 years preceding the compilation of the Red Data Book (i.e., inclusive) but which were listed as nationally extinct. We excluded these errors for the analysis below, but their implications are discussed. After removing these species, the last year of any collection would be limited to To determine if an extinct species was rediscovered, we used all collection records from 22 Sep 2006 to 13 April 2010 (approximately 3.5 years duration). If a nationally extinct species was collected within this period, we considered it as rediscovered. Specimens of such species were determined by authors who are members of the herbarium s species identification team (S.M.L.L., A.T.G., P.K.F.L., and A.S.) using published floras and by matching with other specimens whose identities were verified by experts. We also checked local published literature for reliable, expert records of nationally extinct species that were sighted recently but whose specimens were not deposited in the herbarium. We modeled the rediscovery of the nationally extinct species as a binary response in a simple logistic regression with the probability that the species is still extant calculated from the three equations as the predictors in separate models. Some criticism may be leveled at our approach of using what are essentially p values from null hypothesis testing (see Solow 2005) as predictors in a regression. In this study, however, we were testing if these calculated probabilities were reliable in their inference of extinction by being associated with chances of rediscovery. Furthermore, instead of utilizing an arbitrary cut-off value (e.g., 0.05), we used the raw probability value for the regression an approach similar to McInerny et al. (2006) who tested the correlation of these probabilities against the IUCN threat statuses. We calculated Akaike s Information Criterion (AIC) corrected for small sample sizes (AICc) for each of these regression models: AICc ¼ AIC þ 2 k ðk þ 1Þ= ðn k 1Þ where k is the number of parameters in the model (i.e., 0 for the null model without any predictors, 1 for all other models) and N is the sample size of species that were considered extinct without error, and had at least two separate years of collections for calculating the probability values. To compare between the three probabilistic models and a null model with no predictors, we calculated model weights w, from the ratio of the relative likelihood (rl) of each j th model relative to the sum of all models combined, i.e., where w j ¼ rl j =RrL rl ¼ expð 2 DAICcÞ and DAICc is the difference between the AICc of each model and the lowest AICc among the models being compared. Model weights ranged from 0 to 1 and represented the probability that the regression model is the best model among those being compared (Burnham and Anderson 2002). We also compared the simple use of the last sighting to infer the probability of extinction, by regressing the rediscoveries as a binary response against the last year of collection in the herbarium database. Because only at least one collection was needed in this case, the sample size was larger than that of the probabilistic models that require at least two collections, we therefore calculated AICc, relative likelihood, and model weights against a separate null model. All calculations and analyses were implemented using R version (R Development Core Team 2006).

6 2594 Biodivers Conserv (2012) 21: Results Of the 629 vascular plant species previously considered nationally extinct (Table 1), 35 were found to be erroneously listed, either based on herbarium specimens collected from 1976 to 2006 or were published as rediscovered in an issue in the Botanic Gardens newsletter in 2005 (Lee et al. 2005; Leong 2005). A total of 61 were rediscovered in the period following the compilation, including sightings of three species published in an online, peer-reviewed natural history journal (Ang et al. 2010a, b; Lok et al. 2010). Species names for these errors and rediscoveries are given in Table S1 (see Online Resource). The 225 species of all species previously considered extinct had at least two specimens with collection dates from 1874 to 1975 and could be assessed using the extinction equations, and of these, 27 were rediscovered after the Red Data Book was compiled (Table 1). Of 363 species with at least one specimen with a collection date and which could be assessed with its last year of collection, 40 were rediscovered (Table 1). The remaining = 231 species did not have specimens deposited in the herbarium, did not have wild collections, or only had undated collections (Table S1) and hence could not be used for further analyses. There was little support that the probability values from the extinction equations or the last year of collection were associated with rediscoveries. The models of Solow (1993a) and Solow and Roberts (2003) performed only slightly better than the null model, while the other two models performed worse (Table 2). The last year of collection, as a parameter, also only performed slightly better than the null model (Table 3), although its level of support against the null model given by the ratio of model relative likelihoods, Table 1 No. of vascular plant species listed as nationally extinct in the Singapore Red Data Book (Tan BC et al. 2008; Tan HTW et al. 2008), and subsequently found to be errors or were rediscovered. Only a subset of these species had at least two specimens with collection dates required to calculate probability values using the sighting rate equations or at least one specimen with a collection date required to use last date of collection as a predictor variable No. of species Total C1 dated collection C2 dated collections Listed as extinct 629 Erroneous Rediscovered Table 2 Summary of relative model performance when probabilities calculated from the extinction equations were used as predictors of extinct species rediscovery in a binary logistic regression, and comparing these against a null model with no predictors (N = 225). rl and w, respectively refer to the relative likelihood and weight of the model; formulas are given in the text Model AIC c DAIC c rl w Solow (1993a) Solow and Roberts (2003) Null McInerny et al. (2006) McCarthy (1998)

7 Biodivers Conserv (2012) 21: Table 3 Summary of model performance when the last year of collection was used as a predictor of extinct species rediscovery in a binary logistic regression compared against a null model (N = 363). rl and w, respectively refer to the relative likelihood and weight of the model; formulas are given in the text Model AIC c DAIC c rl w Last year of collection Null =0:241 ¼ 4:14, is slightly higher than the level of support for the model of Solow (1993a) against the null model, 1=0:392 ¼ 2:55. Discussion We found that probability values calculated from extinction equations were not useful in predicting which of the plant species presumed nationally extinct would be rediscovered. The last year of collection also performed poorly when used as a predictor. These results are sobering, considering the potential that these probabilistic models would have had for global plant conservation assessment if they could be applied with high reliability to whole databases of herbarium records. However, we emphasize that our results do not put in doubt the accuracy of the probabilistic models when applied to sighting records that satisfy the assumptions of the model, but only their applicability to large herbarium databases such as the one utilized here. A possible reason for the models poor performance could be their sensitivity to inconstant survey effort over time, but the model of McCarthy (1998), designed specifically to attempt to solve this problem, performed the worst among all the models tested (Table 2). We have also attempted to check the total number of herbarium collections for the years without collections (Fig. 1), but this may be too crude a measure as probability values are calculated for individual species. Sampling effort for a plant group may increase with the expertise of a botanist who specializes in its taxonomy, and may decrease drastically or completely stop when the specialist leaves the region or retires. Overall sampling effort may also fluctuate significantly with political events such as wars, and changes in the administration of institutions can result in decreased interest for botanical exploration. The brief period of wartime occupation in Singapore s recent history experienced significantly decreased botanical activity as British botanists were imprisoned or their activities were restricted by the Imperial Japanese Army. Regions in most urgent need of conservation assessment are generally those that have experienced greater geo-political instability and lack of funding for basic taxonomic research. We expect the SING records to be at least typical of herbarium data from other developing, tropical countries, if not of higher quality. SING has been the center of research for regional botanists since it was established, and the Singapore flora has been well-studied and well-collected owing to its compact size and the ease of accessibility to most habitats. Hence if Singapore s flora is too under-collected for the probabilistic models to work reliably to predict the species that are locally extinct, it would almost certainly be the same case elsewhere as well, although tests using data from other natural history collections are still needed. There are other probabilistic models that were not tested here, but these models typically require a larger number of collections per species. For example, Solow (2005) recommended that the number of sightings when using the Weilbull model (Roberts and

8 2596 Biodivers Conserv (2012) 21: Solow 2003) should be neither too small nor too large. Collen et al. (2010) found that at least five sightings are needed, which would result in only a small proportion of the species available for testing in this study. Complicated calculations in the model of Solow (1993b) resulted in nonsense probability values (results not shown), and Solow (2005) noted that it was difficult to justify the assumption for an exponential decline in species populations for this model. Our study only considered sighting information that was supported by actual herbarium specimens. Roberts et al. (2010) compared the addition of expert opinion sightings in addition to using only physical records, and found that the expected probability of extinction can increase if gaps are filled in between earlier sightings or decrease if such sightings occur after the last physical collection. For charismatic species such as birds and mammals, detailed sighting records by experts and citizen scientists can complement data from museums, but systematic notes of encounters with plants are very rarely available. Once again, the paucity of sighting information outside of herbarium records is likely to be the same for botanical records elsewhere and not just restricted to our study. The error rate of categorizing species as nationally extinct for Singapore vascular plants was high. With 35 species as errors and 61 species rediscovered out of 629 species, the extinct species list shrank by 15.3 % in the 3.5-year period after the compilation of the Red Data Book, and this Lazarus effect is likely to increase over time. Fisher and Blomberg (2011) found that for mammals, species presumed extinct owing to habitat loss were most likely to be rediscovered, and habitat loss was also the major reason attributed to species extinctions in Singapore (Brook et al. 2003). Many of the rediscoveries of the Singapore flora were made in a remnant patch of primary freshwater swamp forest in the nature reserves (e.g., Ang et al. 2010a, b; Lok et al. 2010), which has been largely unexplored in recent years and may harbor many more of the species that have been presumed nationally extinct, especially epiphytes and climbers that grow high in the tree crowns. While this indicates that there is still some hope to conserve remnants of the original local populations of these species, it also implies that current lists of extinct vascular plant species are not stable enough to be used as indicators of environmental change, at least until botanical exploration of species-rich sites are satisfactorily updated. Rediscovery hotspots such as the freshwater swamp forest should be given high priority for future exploration. Monitoring indices such as the Cities Biodiversity Index ( authorities/gettinginvolved/cbi.shtml) should not misinterpret rediscovered extinct species when making serial comparisons of the status of biodiversity. Studies that make use of the list of nationally extinct species for statistical analysis, such as Sodhi et al. (2008) which looked for ecological correlates of extinction proneness, and Brook et al. (2003) which projected extinction rates for Southeast Asia from Singapore data, will also have to be revised in light of the significant Lazarus effect. At first glance, a window period that was too short (30 years) for declaring species extinct may be the easiest factor to pinpoint for the high incidence of rediscoveries. However, this brings the discussion back to the question of whether there an optimal window period to adopt. The evidence ratio for a model with last date of collection as a predictor versus the null model was not sufficiently high for us to be confident of its utility to be used as an indicator of extinction, even if it is expected to perform better than all the sighting rate equations for our dataset. In addition, an optimum window period will have to take into account the sampling effort in each separate scenario. This is precisely the motivation behind the formulation of the extinction equations. While the definition of extinction in the IUCN states exhaustive surveys as a requirement, exhaustive itself can be arbitrary without some form of quantification. While the sighting rate equations are

9 Biodivers Conserv (2012) 21: considerable developments towards quantifying the probability of an extinction event, our study shows that these may be too data-hungry for broad-scale application, even for wellcollected areas such as Singapore. There seems, therefore, to be no escape from the need to escalate collection efforts. In addition, collection efforts need to be consistent or at least be without bias, in terms of temporal and spatial patterns, and kinds of taxa. This would be a tall order given the realities: institutions have limited space and expertise, and must prioritise their current research efforts on these resources available, hence leading to temporal fluctuations and taxon-group biases in collection effort. To minimize future errors in lists of extinct species, an assessment of deficiencies and undersampling of hotspot areas and difficult taxa should be made first, followed by targeted, intensive expeditions. We used the data from Singapore to test rediscoveries following local extinctions, not global extinctions. Given the proximity of Singapore to Peninsular Malaysia, with which it shares almost all of its flora, some recolonisation by species extinct in Singapore but still extant in Peninsular Malaysia may occur, but we think the context of most of the rediscoveries make this explanation only a minor consideration. Additionally, with the loss of large frugivores from Singapore s landscape, dispersal from the nearest source in mainland Peninsular Malaysia would be too far and highly unlikely to occur (Corlett 2009). We need to continue to discover ways to better utilize data from natural history collections for conservation, especially for threat and extinction assessment. Expeditions continue to be mounted for species that are presumed extinct owing to a long absence of sightings (Hance 2011), and such surveys can be expensive to fund. A possibility could be to develop probabilistic models that identify rediscovery hotspots for targeted expeditions, and the new survey data can be integrated to inform yet further expeditions in an adaptive learning cycle. Acknowledgments K.Y.C. is grateful to the Department of Biological Sciences, National University of Singapore, for sponsoring the conference expenses. We thank Wong Khoon Meng and the reviewers for their thoughtful comments and suggestions. We would also like to thank the Singapore Botanic Gardens Herbarium and the National Parks Board (Singapore) for access to the collections database, and Jeremy Woon for facilitating the Research Permits (RP977 and RP943). References Ang WF, Lok AFSL, Tan HTW (2010a) Rediscovery in Singapore of Pinanga simplicifrons (Miq.) Becc. (Arecaceae). Nat Singap 3: Available online: nis pdf Ang WF, Lok AFSL, Yeo CK, Tan SY, Tan HTW (2010b) Rediscovery of Dendrobium aloifolium (Blume) Rchb.f. (Orchidaceae) in Singapore. Nat Singap 3: Available online: nis/bulletin2010/2010nis pdf Brook BW, Sodhi NS, Ng PKL (2003) Catastrophic extinctions follow deforestation in Singapore. Nature 424: Burnham KP, Anderson DR (2002) Model selection and multi-model inference: a practical informationtheoretic approach. Springer, New York Butchart SHM, Stattersfield AJ, Brooks TM (2006) Going or gone: defining possibly extinct species to give a truer picture of recent extinctions. Bull Br Ornithol Club 126A:7 24 Carpaneto GM, Mazziotta A, Valerio L (2007) Inferring species decline from collection records: roller dung beetles in Italy (Coleoptera, Scarabaeidae). Divers Distrib 13: Collar NJ (1998) Extinction by assumption; or, the Romeo error on Cebu. Oryx 32: Collen B, Purvis A, Mace GM (2010) When is a species really extinct? Testing extinction inference from a sighting record to inform conservation assessment. Divers Distrib 16: Corlett RT (2009) Seed dispersal distances and plant migration potential in Tropical East Asia. Biotropica 41:

10 2598 Biodivers Conserv (2012) 21: de Iongh HH, Bal D (2007) Harmonization of Red Lists in Europe: some lessons learned in the Netherlands when applying the new IUCN Red List Categories and Critieria version 3.1. Endanger Species Res 3:53 60 Duffy KJ, Kingston NE, Sayers BA, Roberts DL, Stout JC (2009) Inferring national and regional declines of rare orchid species with probabilistic models. Conserv Biol 23: Dutson GCL, Magsalay PM, Timmins RJ (1993) The rediscovery of the Cebu Flowerpecker Diaceum quadricolor, with notes on other forest birds on Cebu, Philippines. Bird Conserv Int 3: Fisher DO, Blomberg SP (2011) Correlates of rediscovery and the detectability of extinction in mammals. Proc R Soc B 278: Fitzpatrick JW, Lammertink M, Luneau DM Jr, Gallagher TW, Harrison BR, Sparling GM, Rosenberg KV, Rohrbaugh RW, Swarthout ECH, Wrege PH, Swarthout SB, Dantzker MS, Charif RA, Barksdale TR, Remsen JV Jr, Simon SD, Douglas Z (2005) Ivory-billed Woodpecker (Campephilus principalis) persists in continental North America. Science 308: Fitzpatrick JW, Lammertink M, Luneau DM Jr, Gallagher TW, Rosenberg KV (2006a) Response to comment on Ivory-billed Woodpecker (Campephilus principalis) persists in continental North America. Science 311:1555b Fitzpatrick JW, Lammertink M, Luneau DM Jr, Rosenberg KV, Gallagher TW, Rohrbaugh RW (2006b) Response to Ivory-billed or Pileated Woodpecker?. Science 315:1496 Fitzpatrick JW, Lammertink M, Luneau DM Jr, Gallagher TW, Harrison BR, Sparling GM, Rosenberg KV, Rohrbaugh RW, Swarthout ECH, Wrege PH, Swarthout SB, Dantzker MS, Charif RA, Barksdale TR, Remsen JV Jr, Simon SD, Douglas Z (2006c) Clarifications about current research on the status of Ivory-billed Woodpecker (Campephilus principalis) in Arkansas. Auk : Hance J (2011) Worldwide search for lost frogs ends with 4 % success, but some surprises. Accessed 31 Aug 2011 Jackson JA (2006) Ivoy-billed Woodpecker (Campephilus principalis): hope, and the interfaces of science, conservation, and politics. Auk :1 15 Keith DA, Burgman MA (2004) The Lazarus effect: can the dynamics of extinct species lists tell us anything about the status of biodiversity? Biol Conserv 117:41 48 Keller V, Zbinden N, Schmid H, Volet B (2005) A case study in applying the IUCN regional guidelines for national Red Lists and justifications for their modification. Conserv Biol 19: Ladle RJ, Jepson P, Araújo MB, Whittaker RJ (2004) Dangers of crying wolf over risk of extinctions. Nature 428:799 Lee SML, Samsuri A, Leong P, Gwee AT, Kiew R (2005) Extinct species. Gardenwise 24:12 Leong P (2005) Rediscovery of extinct native orchids. Gardenwise 24:12 13 Lok ASFL, Ang WF, Chong KY, Tan HTW (2010) Rediscovery of Liparis barbata Lindl. (Orchidaceae) in Singapore. Nature Singapore 3: Available online: nis pdf McCarthy MA (1998) Indentifying declining and threatened species with museum data. Biol Conserv 83:9 17 McInerny GJ, Roberts DL, Davy AJ, Cribb PJ (2006) Significance of sighting rate in inferring extinction and threat. Conserv Biol 20: Mikusinski G, Blicharska M, Baxter PWJ (2010) Costs and benefits of ivory-billed woodpecker rediscovery. Front Ecol Environ 8:460 Pimm SL, Jenkins CN (2010) Extinctions and the practice of preventing them. In: Sodhi NS, Ehrlich PR (eds) Conservation biology for all. Oxford University Press, UK, pp Rabor DS (1959) The impact of deforestation on birds of Cebu, Philippines, with new records for that island. Auk 76:37 43 Rivadeneira MM, Hunt G, Roy K (2009) The use of sighting records to infer species extinctions: an evaluation of different methods. Ecology 90: Roberts DL (2006) Extinct of possibly extinct? Science 312:997 Roberts DL, Solow AR (2003) When did the dodo become extinct? Nature 426:245 Roberts DL, Elphick CS, Reed JM (2010) Identifying anomalous reports of putatively extinct species and why it matters. Conserv Biol 24: R Development Core Team (2009) R: a language and environment for statistical computing. R Foundation for Statistical Computing, Vienna. Sibley DA, Bevier LR, Patten MA, Elphick CS (2006a) Comment on Ivory-billed Woodpecker (Campephilus principalis) persists in continental North America. Science 311:1555a Sibley DA, Bevier LR, Patten MA, Elphick CS (2006b) Ivory-billed or Pileated Woodpecker? Science 315:1495

11 Biodivers Conserv (2012) 21: Sodhi NS, Koh LP, Peh KS-H, Tan HTW, Chazdon RL, Corlett RT, Lee TM, Colwell RK, Brook BW, Sekercioglu CH, Bradshaw CJA (2008) Correlates of extinction proneness in tropical angiosperms. Divers Distrib 14:1 10 Sodhi NS, Brook BW, Bradshaw CJA (2009) Causes and consequences of species extinctions. In: Levin SA (ed) The Princeton guide to ecology. Princeton University Press, USA, pp Solow AR (1993a) Inferring extinction from sighting data. Ecology 74: Solow AR (1993b) Inferring extinction in a declining population. J Math Biol 32:79 82 Solow AR (2005) Inferring extinction from a sighting record. Math Biosci 195:47 55 Solow AR, Roberts DL (2003) A nonparametric test for extinction based on a sighting record. Ecology 84:1329 Tan BC, Tan HTW, Tan K-x, Ibrahim AB, Chew PT, Chua KS, Gwee AT, Kiew R, Lee SML, Leong P, Lok AFSL, Loo AHB, Lum SKY, Morgany T, Suran SB, Sim S, Ahmad HSBH, Wee YC, Yeo CK, Yong JWH (2008) Checklists of threatened species Ferns and fern allies. In: Davison GWH, Ng PKL, Ho HC (eds) The Singapore Red Data Book, 2nd edn. The Nature Society (Singapore), Singapore, pp Tan HTW, Tan K-x, Ibrahim AB, Chew PT, Chua KS, Duistermaat H, Ganesan SK, Goh MWK, Gwee AT, Kiew R, Lee SML, Leong P, Lim J, Lok AFSL, Loo AHB, Lum SKY, Morgany T, Suran SB, Sim S, Ahmad HSBH, Wee YC, Yap KF, Yeo CK, Yong JWH (2008) Checklists of Threatened species Seed plants In: Davison GWH, Ng PKL, Ho HC (eds) The Singapore Red Data Book, 2nd edn. The Nature Society (Singapore), Singapore, pp