Bonn zoological Bulletin Volume 60 Issue 1 pp. 3-8 Bonn, May 2011 Estimating the herpetofaunal species richness of Pangkor Island, Peninsular Malaysia Johan van Rooijen1 -, Chan Kin Onn\ L. Lee Grismer4 & Norhayati Ahmad5 Netherlands Centre for Biodiversity Naturalis, section Zoological Museum Amsterdam, AD Mauritskade 61, NL-1092 Amsterdam, The Netherlands 2Tulpentuin 313, NL-2272 EH, Voorburg, The Netherlands; E-mail: [email protected] -'Institute for Environment and Development (LESTARI), Universiti Kebangsaan Malaysia, 43600 Bangi, Selangor Darul Ehsan, Malaysia; E-mail: [email protected] 4Department ofBiology, La Sierra University, 4500 Riverwalk Parkway, Riverside, California, 92515-8247 USA; E-mail: [email protected] 5School ofEnvironment and Natural Resource Sciences, Faculty ofScience and Technology, Universiti Kebangsaan Malaysia, 43600 Bangi, Selangor Darul Ehsan, Malaysia; E-mail: [email protected]. Abstract. Herpetological surveys ofSoutheastAsian tropical ecosystems rarely, ifever, result in complete inventories. This is due to the fact that surveying to completion requires huge investments in terms ofsearch effort. As a result, the presentedspecies listsusuallyrepresentsubsets ofthetotal herpetofaunal assemblageandconsequently donotshed light on the total species richness ofthe investigated area. This is regrettable as species richness is an elementary measure of biodiversitythatunderliesmany ecological models andconservation strategies. By recording notjustspeciesbut species perunit ofsearch effort, an extended datasetresults which can be usedto generate estimates oftotal species richness. In this study, the herpetofauna ofPangkor Island, Peninsular Malaysia is used as an example. In 2009 and 2010, the first herpetological surveys were carried out on this small, 18 km2 island. Those surveys recorded 43 species ofreptiles and , 13 species ofamphibians. Inthis study, totalreptile species richness was estimated by fitting several modelsto the sam- ple-based rarefaction curve as well as by application ofthe nonparametric Chao-I estimator. Ofthe applied models, the 4-parameter Weibull function was shown to be superior, a finding that is in line with several other studies. Consequent- ly, the use ofthis model is recommended. On the basis ofthe fitted 4-parameterWeibull-function, 69 reptile species are expected to occur on Pangkor Island. As for amphibians, total species richness was estimated to be 17. As such, a re- markably extensive herpetofaunal assemblage inhabits this small island. Key words. Pangkor Island, Malaysia; amphibians, reptiles, species-richness, sample-based rarefaction curve, negative exponentialfunction, 3-parameterWeibull function,4-parameterWeibull function,Chapman-Richardsmodel,Chao-Ies- timator INTRODUCTION Herpetological surveys ofspecies-richtropical ecosystems exhaustiveness ofthe survey (e.g. Murphy etal. 1994; Zug rarely result in complete inventories (e.g. Lloyd et al. et al. 1998; Ziegler2002). However, extensive and sophis- 1968; Murphy et al. 1994; Hofer & Bersier 2001; Van ticated literature exists pertaining to methods and models Rooijen 2009). This is due to the asymptotical nature of usedto quantitatively estimate species richness, onthe ba- the species accumulationprocess incombination with lim- sis ofeitherrarefied species accumulationcurves orabun- ited survey-investments.As a consequence, theuse ofes- dance patterns (e.g. Colwell & Coddington 1994; Flather timationtechniques is unavoidable when the intendedob- 1996; Gotelli & Colwell2001; Longino etal. 2002; Brose jective is to assess species richness, an elementary meas- et al. 2003). ure ofbiodiversity that underlies many ecological mod- els andconservation strategies. One such estimationtech- The offshore archipelagos of Peninsular Malaysia have nique consists of fitting an appropriate function to the been the subject of increasing interest in recent years. species accumulation curve. The asymptote ofthe fitted Many ofthese islands have never been surveyed and re- functions can then be regarded as an estimate of total cent explorations are only beginning to uncover the hid- species richness. Species accumulation curves are regu- den diversity and endemism that they shelter (Chan et al. & larly applied in herpetofaunal surveys but they are most- 2009a; Grismer 2008; Grismer Norhayati 2008; Gris- ly used to arrive at qualitative judgements about the mer& Pan 2008; Grismeretal. 2008, 2009a, b). One such ) 4 Johan van Rooijen et al. Fig. 1. A: View into the primary forest ofPangkor Island; B: Boiga drapiezii (Boie, 1827); C: Cnemaspis shahruli Grismer et al., 2010; D: Diyophiops rubescens (Gray, 1835). island is Pulau Pangkor, approximately 18 km2 and situ- tiles were ignored altogether. The predominantly applied ated 3.5 km from the west coast ofPeninsular Malaysia. survey method corresponds to visual encounter survey The island's interior remains heavily forested and main- (VES), a simple method which has been shown to be ef- tains a river system, Sungai Pinang, which supplies a sig- fective for surveying rainforest herpetofauna (Doan, nificant source ofpermanent fresh water in the form of 2003). VES was earned out both duringday andnight (e.g. multiple streams. Chan et al. (2010) provided the first re- Coddington et al., 1996; Doan, 2003) in a way similar to port on the (non-marine) herpetofauna ofthis island and that applied by Minh (2007). Existing trails as well as sev- documented 43 reptiles and 13 amphibians. A few illus- eral trails made by the Department ofForestry ofPerak trations are provided in figures 1 and 2. and PeninsularMalaysia were used. These trails traversed dipterocarp forest, mangrove forest and cultivated areas The study described in this paper had a dual objective. and provided ample access to forest streams. The second First, data collected by Chan et al. (2010) were used to collection method consisted of searching for road-kills. evaluate the performance of several estimators. Second, The third entailed turning logs, fallen tree bark and sim- the most appropriate estimator was used to estimate the ilarobjects in orderto uncoveranimals hidingunderneath. total herpetofaunal species richness harboured by Pangkor Sampling effort was expressed in terms of search-days Island, West Malaysia. where one search-day was defined as roughly 4.5 search- hours. As two ofthe surveys underlying the data collect- ed by Chan et al. (2010) focused solely on reptiles, sub- MATERIALAND METHODS stantially more information was available regarding rep- tiles allowing formore thorough analyses. Therefore, this The data underlying this study were based on the surveys study focuses predominantly on reptiles, amphibians be- carried out by Chan et al. (2010). These were conducted ing treated separately and in less detail. in the periods March 15-17, May 4-8, June 13 to July 8, 2009, and February 22 to March 8, 2010. During the lat- Sample-basedrarefaction curves (Gotelli & Colwell 2001 tertwo periods, only reptiles were surveyed. Marine rep- were generated with the program Estimates (Colwell Bonn zoological Bulletin 60 (1): 3-8 ©ZFMK Estimating herpetofaunal species richness 5 Fig. 2. A: Heosemys spinosa (Gray, 1831); B: Kaloulapulchra Gray, 1831; C: Naja sumatrana (Mtiller, 1890); D: Tropidolae- mus wagleri (Boie, 1827). 2005). Four models were fitted to the rarefaction curve. The four models were fitted to the sample-based rarefac- The first corresponds with a negative exponential model tion curveusing nonlinearregression analysis (e.g. Noru- (Colwell & Coddington 1994; Flather 1996; Van Rooijen sis and SPSS 1994) with SPSS (release 14 February 1996; 2009). It is based on the assumption that the number of SPSS Inc.). new species found per search day is proportional to the number ofas yet undiscovered species, in mathematical Extrapolation using different models for the species ac- terms: dYldt = c {A-Y) where A is the total number of cumulation process can provide different asymptotes and species present in the area under investigation, Yis the to- thus predict different values ofspecies richness (e.g. Col- tal numberofspecies found andcis a constant. This equa- well & Coddington 1994; Flather 1996). Therefore, care tion can be represented as anegative exponential function has to be taken to selectthe most appropriate model in or- (e.g. Van Rooijen 2009): Y=A (\-erct). The basic assump- derto minimize bias. In this study, the appropriateness of tion underlying the negative exponential model (hence- each model was evaluated on the basis ofthree criteria. forth NE) may be overly simplistic given that abundance The first criterion was goodness-of-fit. The second crite- patterns are usually strongly skewed (e.g. Lloyd et al. rion entailed the behaviour ofthe richness-estimate with 1968; Coddington et al. 1996; Limpert et al. 2001; Longi- increasing cumulative search effort. The final criterion was no et al. 2002; Thompson et al. 2003). In order to model the difference with the nonparametric Chao-I estimator more complex species accumulationprocesses, theNE can (Chao 1984; Coddington et al. 1996; Hofer & Bersier be refined in various ways by adding one (d) or two pa- 2001; Veith et al. 2004). This estimator is anoften applied rameters (d and p), resulting in the Chapman-Richards representative ofa class ofestimators that uses a differ- model (henceforth CR), 3- and 4-parameter Weibull cu- ent approach as they are based on abundance patterns in- mulative distribution functions (henceforth 3pW and stead ofthe accumulation curve. The Chao-I estimator is 4pW): based on the observed number of rare species, Y A{\-e-«) (CR), A = Y+ (a2/2b), where A is the total number ofspecies, Y=A(l-e<c')d) (3pW), Yis the observed number ofspecies, a is the number of Y= A (\-e-M'-p))d (4pW) observed species represented by a single specimen and b ) Bonn zoological Bulletin 60 (1): 3-8 ©ZFMK 6 Johan van Rooijen et al. is the number ofobserved species represented by exact- ly two specimens. 46 RESULTS Figure 3 depicts the deviations ofthe fitted functions from the sample-basedrarefaction curve (residuals) forreptiles. As expected, the NE exhibits a very poorgoodness-of-fit. At first, the fitted curve is situatedbeneath the rarefaction curve, then above and finally drops beneath the rarefac- tion curve again. The suboptimal fit ofthe NE as well as the pattern ofresiduals are in agreement with results ob- 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 tainedwith regardto avian diversity (Flather 1996) aswell cumulativenuniheiofsearchdays as snake diversity (Van Rooijen 2009). Although the RC and 3pW fit substantially better than the NE, they exhib- Fig. 4. Reptile species sample-based rarefaction curve with it a similar pattern ofresiduals. The 4pW function obvi- fitted 4-parameter Weibull function. ously exhibits a superior, near-perfect, fit. 1.5 has reached a more or less stable level. A similar result (unpublished) was obtained on the basis ofa dataset that underlies a study ofBornean snakes (Van Rooijen 2009): the 4pW-based estimate reached an approximately stable ~^ji~> "1 9 11 13 15 17. ^^T^*— plateau ratherearlywiththe CR- and 3pW-basedestimates approaching the 4pW-based estimate with increasing search days. On the basis ofthese results, the 4pW-based estimate, 69, is assumed to be the least biased. Finally, the estimate negativeexponential - - - 43--ppaarraammeetteerrWWeeiibbuullll based on the nonparametric Chao I estimator, 62, agrees Chapman-Richards reasonably well with the 4pW-based estimate. Taking the average, 65 reptile species are expected to inhabit searchdays Pangkor Island. As for amphibians species richness, the Fig. 3. Deviationsofthe fittedmodels from the sample-based fitted 4pW and Chao I estimator both resulted in an esti- rarefaction curve (residuals). mate of 17 species. Figure4 depicts the sample-based rarefaction curve aswell as the fitted4pW. The only noticeable deviation ofthe fit- ted 4pW from the rarefaction curve lies in the fact that it passes through (p,0) instead ofthe origin (0,0). The close fit of the 4pW is in agreement with results obtained in studies ofavian diversity (Flather 1996), snake diversity (Van Rooijen 2009) and diversity of small reptiles (Thompson et al. 2003). The progression ofthe rarefac- tion curve clearly indicates that an asymptote has not yet been reached, thus the surveys have not been exhaustive. 3-parameterWeibull • " " 4-parameterWeibull Chacman-Rir.hard;. Figure 5 shows how the richness estimates (as opposed to observedrichness orfittedmodel-values) develop with aniBit.twemutterofseoichdays increasing search days during the second halfofthe sur- vey. The NE is neglected dueto its poorfit. Evidently, the 3pW- as well as CR-based estimates are still increasing Fig. 5. Richnessestimatesinrelationto cumulativesearchef- at the end ofthe survey whereas the 4pW-based estimate fort. Bonn zoological Bulletin 60 (1): 3-8 ©ZFMK Estimating herpetofaunal species richness 7 DISCUSSION may also be expected to be rather uniform. Second, the higherthe numberofparameters ofamodel, the betterthe Estimation by extrapolation, as opposed to interpolation, fit. Thus, the 4pW may simply be expected to fit better into the unknown entails a high degree of uncertainty. than similar functions with fewerparameters, irrespective However, the alternative, surveying until completion, is ofthe field ofapplication. simplyunfeasible when the survey is aimedat the herpeto- fauna of a Southeast Asian rainforest ecosystem (Van 65 reptilian species are estimated to inhabit Pangkor Is- Rooijen 2009). As such, there simply is no alternative. land, 43 ofwhich have been recorded (Chan et al. 2010). A Thus, further refinement ofestimation methods is ofim- major part ofthe as yet unrecorded species will con- portance. That said, approximately complete species lists cern snakes as members ofthis taxon are notoriously hard are obviously still crucial for many zoogeographic stud- to find due to their elusive habits and low densities (e.g. ies. Such species lists can be (and are) compiled by com- Lloyd et al. 1968; Inger & Colwell 1977; Hofer & Bersi- bining results ofvarious surveys. However, this usually er 2001; Orlov et al. 2003; Van Rooijen 2009). This no- entails the combination ofspecies lists pertaining to dif- tion is strengthened by several indirect observations. Lo- ferent locations. As such, completeness comes at a price cals provided accounts regarding observations ofBoiga as one has to accept a huge decrease in spatial resolution. dendrophila (Boie, 1827), Cryptelytropspurpureomacu- In mostzoogeographic studies however, a high resolution latus (Gray, 1832), Ophiophagus hannah (Cantor, 1836) is not essential as species compositions of major land andMaticora btvirgata (Boie, 1827). In addition, therehas masses are compared (e.g. In Den Bosch 1985; How and been a visual record of a Coelognathus radiants (Boie, Kitchener 1997; IngerandVoris 2001).Atthe spatial scale 1827) (Schultz, pers. comm.). As amphibian species rich- which is relevant for conservation though, complete ness is estimated to be at least 17, herpetofaunal species species lists are a Utopia in the majority ofcases and es- richness is expected to equal at least 82 species, which is timation will therefore gain in importance (Van Rooijen quite impressive given the factthatPangkorIslandencom- 2009). This study underscores the notion that a combina- passes merely 18 km2 On the other hand, whether . tion ofcriteria needs to be applied to select the most ap- Pangkor Island harbours a comparatively extensive her- propriate estimator. The4-parameterWeibull function ex- petofauna is impossible to determine at the moment as hibitedthe highest goodness-of-fit. Moreover, it produced species richness on other Malaysian islands has not yet the most stable estimate near the end ofthe survey (fig- been estimated. ure 5). Finally, itresulted in arichness-estimatethatagreed well with the estimate based on the nonparametric Chao- Estimating species richness ofspecific sites suchas islands I estimator. Even then, the estimate may be downwardly obviously has added value as species richness is an ele- biased due to suboptimal sampling. First ofall, although mentary criterion a conservationist may use when select- searches were carried out during both day and night, no ing sites and is crucial formany ecological studies. How- searches were performed in the early morning (before ever, the function fitted to the rarefaction curve not only dawn). Secondly, sampling of canopy-microhabitat was provides an estimate oftotal species richness but also pro- obviously unfeasible. Finally, as true species richness vides insightin expectedreturn on furtherinvestment: how is underestimated by most estimators when sample size many previously unrecorded species may be expected to & is small (Colwell Coddington 1994; Canning-Clode et be found with additional search effort? For instance, on al. 2008), the estimate ofamphibian species richness may the basis ofthe fitted 4-parameter Weibull function, ten be downwardly biased since it was based on only five previously unrecorded reptiles are expected to be found search days. Nevertheless, the amphibian species richness on Pangkor Island when 20 additional search days are in- appears to be rather low which is in agreement with the vested. Such statistical expectations can be used as input impression ofherpetologists who carried out the surveys foracost-benefitevaluation whenchoiceshave to be made (Chan etal. 2010). between different sites for the investment of survey-ca- pacity. Alternatively, one might determine how much The fact that the 4-parameter Weibull function exhibits search effort would have to be invested to bring the sur- such a good fit when applied to diverse ecological com- vey to some specified level ofcompleteness. munities such as small reptiles in desert habitat (Thomp- son et al. 2003), birds (Flather 1996) and reptiles in rain- forest habitat (Van Rooijen 2009 and this report) is strik- REFERENCES ing. Two explanations can be put forward. First, relative species abundances follow very similar patterns over a BroseU, MartinezND,Williams RD (2003) Estimatingspecies wide range of ecological communities: relatively few richness: sensitivityto samplingcoverageandinsensitivityto species are abundant whereas most are rare (e.g. Hughes spatial patterns. Ecology 84: 2364-2377 M 1986). Thus, the shape ofthe species accumulationprocess Canning-Clode J, ValdiviaN, Molis M, Thomason JC, Wahl (2008)Estimation ofregional richness inmarinebenthiccom- Bonn zoological Bulletin 60 (1): 3-8 ©ZFMK : 8 Johan van Rooijen et al. munities: quantifying the error. Limnology and Oceanogra- How RA, Kitchener DJ (1997) Biogeography of Indonesian phy: Methods 6: 580-590 snakes. Journal ofBiogeography 24: 725-735 Chan KO, Grismer LL, Wood PL Jr, Grismer JL, Norhayati A Hughes RG (1986) Theories and models ofspecies abundance. (2009) PreliminaryChecklistoftheHerpetofaunaofPulau Be- American Naturalist 128: 879-899 sar, Melaka, Malaysia. Tropical Life Sciences Research,20(1): In Den Bosch HAJ (1985) Snakes of Sulawesi: checklist, key 81-87 andadditional biogeographical remarks. ZoologischeVerhan- Chan KO, Van Rooijen J, GrismerLL, Daicus B,Akil MAMM, delingen 217: 1-50 Jamaludin H, Gregory R, Norhayati A (2010) First report on Inger RF, Colwell RK (1977) Organization ofcontiguous com- the herpetofauna of Pulau Pangkor, Perak, Malaysia. The munities ofamphibians and reptiles in Thailand. Ecological Russian Journal ofHerpetology 17: 139-146 Monographs 47: 229-253 Chao A (1984) Non-parametric estimation of the number of IngerRF, Voris HK(2001) The biogeographical relations ofthe classes inapopulation. ScandinavianJournal ofStatistics 11 frogs and snakes ofSundaland. Journal ofBiogeography 28: 265-270 863-891 Coddington JA, Young LH, Coyle FA(1996) Estimating spider LimpertE, StahelWH,AbbtM (200 Log-normal distributions 1) speciesrichness inasouthernAppalachiancovehardwoodfor- across the sciences: keys and clues. Bioscience 51: 341-352 est. The Journal ofArachnology 24: 111-128 Lloyd M, IngerRF, King FW (1968) On the diversity ofreptile Colwell RK, Coddington JA (1994) Estimating terrestrial bio- and amphibian species in a Bornean rain forest. The Ameri- diversitythrough extrapolation. PhilosophicalTransactions of can Naturalist 102: 497-515 the Royal Society ofLondon 345: 101-118 LonginoJT, CoddingtonJ, Colwell RK(2002)The ant faunaof RK Colwell (2005) Estimates Statistical estimation ofspecies a tropical rainforest: estimating species richness three differ- : richness andsharedspecies from samples. Version 7.5. User's ent ways. Ecology 83: 689-702 NQM Guide and application published at: http://purl.oclc.org/esti- Minn (2007) Estimates ofthe herpetofaunal diversity in mates the ShipsternNature Reserve, Belize. Biology master degree DoanTM (2003)Which methods are most effective forsurvey- thesis, Zoological Institute, UniversityofNeuchatel, Switser- ingrainforestherpetofauna?Journal ofHerpetology37: 72-81 land Flather CH (1996) Fitting species-accumulation functions and Murphy JC, Voris HK, Karns DL (1994) A field guide to the assessing regional land use impacts on avian diversity. Jour- snakes of the Danum Valley, a Bornean tropical rainforest nal ofBiogeography 23: 155-168 ecosystem. BulletinoftheChicago Herpetological Society29: Gotelli NJ, Colwell RK (2001) Quantifying biodiversity: pro- 133-151 cedures and pitfalls in the measurement and comparison of Norusis & SPSS Inc. (1993) SPSS Advanced Statistics, SPSS, species richness. Ecology letters 4: 379-391 Chicago GrismerLL(2008)Anewspecies ofinsularskink(GenusSphe- OrlovNL,Ryabov SA,NguyenVan Sang,NguenQuangTruong nomorphus Fitzinger 1843) from the LangkawiArchipelago, (2003) New records and data on the poorly known snakes of Kedah, West Malaysia with the first report ofthe herpetofau- Vietnam. Russian Journal ofHerpetology 10: 217-240 na ofPulau Singa Besar and an updated checklist ofthe her- Thompson GG, Withers PC, Pianka ER, Thompson SA (2003) petofauna ofPulau Langkawi. Zootaxa 1691: 53-66 Assessing biodiversity with species accumulation curves; in- Grismer LL, Chan KO, Grismer JL, Wood PL Jr., Daicus B ventoriesofsmallreptilesbypit-trappinginWesternAustralia. (2008) Three new species of Cyrtodactylus (Squamata: Austral Ecology 28: 361-383 Gekkonidae) from Peninsular Malaysia. Zootaxa 1921: 1-23 Van Rooijen J (2009) Estimating the snake species richness of Grismer LL, Pan KA (2008) Diversity, endemism and conser- the SantubongPeninsula(Borneo) intwodifferentways. Con- vation ofthe amphibians and reptiles ofSouthern peninsular tributions to Zoology 78: 141-147 Malaysia and its offshore islands. Herpetological Review Veith M, Lotters S, Andreone F, Rodel M-0 (2004) Measuring 39(3): 270-281 andmonitoringAmphibiandiversity intropical forests. II. Es- Grismer LL, Norhayati A(2008)Anew insular species ofCyr- timating species richness fromstandardizedtransectcensing. todactylus (Squamata: Gekkonidae) from the Langkawi Ar- Ecotropica 10: 85-99 chipelago, Kedah, PeninsularMalaysia. Zootaxa 1924: 53-68 Ziegler T (2002) Die Amphibien und Reptilien eines Tiefland- Grismer LL, Wood PL Jr., Grismer JL (2009a) A New Insular feuchtwald-Schutzgebietes in Vietnam. Minister: Natur und Species ofSkinkofthe Genus Sphenomorphus Strauch 1887 Tier-Verlag (Squamata: Scincidae) from Pulau Perhentian Besar, Tereng- Zug GR, Win H, Thin T, Min TZ, Lhon WZ, Kyaw K (1998) ganu, Peninsular Malaysia. Tropical Life Sciences Research HerpetofaunaoftheChatthinWildlife Sanctuary, North-Cen- 20(1): 51-69 tral Myanmar with preliminary observations oftheir natural Grismer LL, Norhayati A, Chan KO, Daicus B, Muin, MA, history. Hamadryad 23: 111-120 WoodPLJr., GrismerJL(2009b)Twonewdiminutive species of Cnemaspis Strauch 1887 (Squamata: Gekkonidae) from Peninsular Malaysia. Zootaxa 2019: 40-56 Received: 21.06.2010 HoferU, BersierL-F (2001) Herpetofaunal diversity and abun- Accepted: 14.03.2011 dance in tropical upland forests of Cameroon and Panama. Corresponding editor: F. Herder Biotropica 33: 142-152 Bonn zoological Bulletin 60 (1): 3-8 ©ZFMK