Contributed Paper Effects of Oil-Palm Plantations on Diversity of Tropical Anurans

AISYAH FARUK,∗† †† DAICUS BELABUT,‡§ NORHAYATI AHMAD,§∗∗ ROBERT J. KNELL,∗ AND TRENTON W. J. GARNER†

∗School of Biological and Chemical Sciences, Queen Mary University of London, Mile End Road, E1 4NS, London, United Kingdom †Institute of Zoology, Zoological Society of London, Regents Park, NW1 4RY London, United Kingdom ‡Institute of Biological Sciences, Faculty of Science, University of Malaya, 50603 Kuala Lumpur, §Institute for Environment and Development (LESTARI), Universiti Kebangsaan Malaysia, 43600 Bangi Selangor Darul Ehsan, Malaysia ∗∗Faculty of Science and Technology, Universiti Kebangsaan Malaysia, 43600 Bangi Selangor Darul Ehsan, Malaysia

Abstract: Agriculturally altered vegetation, especially oil-palm plantations, is rapidly increasing in South- east Asia. Low species diversity is associated with this commodity, but data on anuran diversity in oil- palm plantations are lacking. We investigated how anuran biological diversity differs between and oil-palm plantation, and whether observed differences in biological diversity of these areas is linked to specific environmental factors. We hypothesized that biological diversity is lower in plantations and that plantations support a larger proportion of disturbance-tolerant species than forest. We compared species richness, abundance, and community composition between plantation and forest areas and between site types within plantation and forest (forest stream vs. plantation stream, forest riparian vs. plantation ri- parian, forest terrestrial vs. plantation terrestrial). Not all measures of biological diversity differed between oil-palm plantations and secondary forest sites. Anuran community composition, however, differed greatly between forest and plantation, and communities of anurans in plantations contained species that prosper in disturbed areas. Although plantations supported large numbers of breeding anurans, we concluded the community consisted of common species that were of little conservation concern (commonly found species include , Microhyla heymonsi,andHylarana erythrea). We believe that with a number of management interventions, oil-palm plantations can provide habitat for species that dwell in secondary .

Keywords: agriculture, Asia, indicators, inventory and monitoring, protected areas

Efectos de las Plantaciones de Palma de Aceite sobre la Diversidad de Anuros Tropicales Faruk et al.

Resumen: La vegetacion´ alterada por la agricultura, especialmente las plantaciones de palma de aceite, esta´ incrementando rapidamente´ en el sureste de Asia. Una baja diversidad de especies estaasociadacon´ este bien tangible, pero existen pocos datos sobre la diversidad de anuros en estas plantaciones. Investigamos como difiere la biodiversidad de anuros entre bosques y plantaciones de palma de aceites y si las diferen- cias observadas en la biodiversidad de estas areas´ estan´ relacionadas con factores ambientales espec´ıficos. Partimos de la hipotesis´ de que la biodiversidad es mas´ baja en las plantaciones y que ´estas soportan una mayor proporcion´ de especies tolerantes a los disturbios que los bosques. Comparamos riqueza de especies, abundancia y composicion´ de la comunidad entre areas´ de bosque y de plantaciones y entre sitios tipo dentro de plantaciones y bisques (franja de bosque contra franja de plantacion,´ riparia de bosque contra riparia de plantacion,´ terrestre de bosque contra terrestre de plantacion).´ No todas las medidas de biodiversidad difirieron entre las plantaciones de palma de aceite y sitios de bosque secundario. Sin embargo, la com- posicion´ de la comunidad de anuros vario´ de gran manera entre bosques y plantaciones; y las comunidades de anuros en las plantaciones tuvieron especies que prosperan en areas´ perturbadas. Aunque las plantaciones soportaron un gran numero´ de anuros reproductivos, concluimos que la comunidad consist´ıa de especies que son de poca preocupacion´ para la conservacion´ (especies comunmente´ encontradas que incluyen a

††Address for correspondence: Institute of Zoology, email [email protected] Paper submitted June 3, 2012; revised manuscript accepted October 9, 2012. 615 Conservation Biology,Volume27,No.3,615–624 C 2013 Society for Conservation Biology ⃝ DOI: 10.1111/cobi.12062 616 Effects of Oil-Palm Plantations on Anurans

Fejervary limnocharis, Microhyla heymonsi y erythrea). Creemos que con un numero´ de intervenciones de manejo las plantaciones de palma de aceite pueden brindar un habitat´ para las especies que radican en bosques secundarios.

Palabras Clave: agricultura, ´areas protegidas, Asia, indicadores, inventarios y monitoreo

Introduction Effects of habitat loss and alteration on anuran bio- logical diversity have been documented. However, re- Biological diversity loss is most commonly attributed to sults of studies of anuran responses to such changes are habitat loss and alteration (Stuart et al. 2004; Todd & inconsistent among studies (Hecnar & M’Closkey 1997; Rothermel 2006). The rise in global demand for agri- Gibbs 1998; Parris 2004). In some cases, land-use changes cultural commodities is a major driver of habitat mod- can make an area unsuitable for anuran reproduction ification, and the amount of land under cultivation is and survival. In Australia conversion of Eucalyptus forest predicted to expand substantially in the next 4 decades to pine monoculture reduced the number of detected (Wanger et al. 2009). A large component of this con- species 4-fold (Parris 2004). However, in other sit- version to cropland is for oil-palm plantations, much of uations disturbed or converted areas may still support which has occurred in (Koh 2007). Oil- a substantial subset of the original anuran community palm plantations dominate the landscape of Malaysia and (Gibbs 1998). Anuran species that tolerate disturbed en- and have steadily replaced other crops in these vironments often have high fecundity (Williams & Hero countries due to low maintenance requirements and high 1998), so the negative effects of conversion on abun- yield per unit area (Donald 2004; Koh & Wilcove 2008; dance, but not on species richness, may decrease over Danielsen et al. 2009). There has been a call to discourage time. the planting of oil palm on primary and secondary forests When an area is rendered unsuitable for the original and restrict it to cleared grasslands and other agricultural community, it may be colonized by previously rare or fields (Koh & Wilcove 2008). undetected species. A decline in species richness and Although large areas of old-growth forest and of forests abundance evident soon after conversion may eventu- with low levels of disturbance are protected in Malaysia ally be masked by species that can exploit the dis- and Indonesia from large-scale conversion to agriculture, turbed area. However, species diversity should exhibit smaller areas of secondary forest remain unprotected. strong correlations with habitat variables in disturbed Koh and Wilcove (2008) reported that from 1990 to 2005, and undisturbed landscapes. In North America the in- 55–59% of oil-palm expansion was at the expense of sec- troduction of non-native predatory fish has led, in some ondary forest. Secondary forests in Malaysia, although not cases, to the loss of small palatable na- as diverse as old-growth forests, retain substantial biologi- tive to the site and colonization by larger, unpalatable cal diversity and may contain unique assemblages relative amphibians (Hecnar & M’Closkey 1997). Changes to to monoculture plantations (Chazdon et al. 2009). the landscape may also result in a subset of the exist- Oil-palm plantations are associated with reductions in ing anuran community persisting in the disturbed area species richness, and species diversity and shifts in com- and new species colonizing the newly created niche munity composition (beetles: Chung et al. 2000; reptiles: space. Given time, little or no reduction in species Glor et al. 2001; birds: Koh & Wilcove 2008; ants: Bruhl¨ richness or abundance may be detectable, again due & Eltz 2010), but to date there has been only one study to colonization by disturbance-tolerant species. How- of anuran communities in oil-palm plantations (Gillespie ever, species correlations with habitat variables would be et al. 2012). The International Union for Conservation of inconsistent. Nature (IUCN) Red List reports 41% of 6300 described, We sought to describe abundance, species richness, extant species are at risk of extinction (IUCN and community composition of anurans in oil-palm plan- 2010); thus, Amphibia are the most threatened vertebrate tations and secondary forest on Peninsular Malaysia. By class assessed to date. Pollution, overharvesting, and in- using multiple methods, we avoided dependence on sin- fectious diseases have contributed to a global amphibian gle indexes, which cannot distinguish between the range decline, but habitat alteration consistently ranks as the of potential responses an anuran community may have to largest threat. Southeast Asia is an amphibian biological forest conversion. On the basis of the results of previ- diversity hotspot, yet amphibian studies in this region ous studies of the effects of agricultural landscapes on are rare (Sodhi et al. 2007, 2008). Eighty-two percent of anuran biological diversity (Gardner et al. 2007; Wanger global amphibian species are forest dependent, a pattern et al. 2009), we hypothesized that oil-palm plantations reflected in Southeast Asia, so even slight alterations of have less diverse anuran communities than forest and are forested landscapes may affect a large proportion of trop- dominated by an anuran community suited to thrive in ical amphibians (Ernst & Rodel¨ 2005). the disturbed monoculture oil-palm plantations.

Conservation Biology Volume 27, No. 3, 2013 Faruk et al. 617

Figure 1. Location of study sites in the Temerloh district, Malaysia (shaded area, forested; white, plantation mosaic [oil palm, rubber, fruit, and Acacia spp.] [Malaysian Ministry of Agriculture and Agro-Based Industry 2008]; squares, survey sites within forest; circles, survey sites within plantation). Insert of Peninsular Malaysia shows location of Pahang (square).

Methods We sampled a total of 57 plots (30 2m):20plots in streams, 20 plots in riparian, and 17 plots× in terrestrial Study Sites and Field Methods areas. A total of 22 plots were in forest and 35 plots were among 4 oil-palm plantation sites (Fig. 2). We marked one The district of Temerloh is in the state of Pahang, Penin- stream at each forest and plantation site with flagging sular Malaysia (340 N, 10210 E). This state has the largest ′ ′ tape every 10 m up to 400 m or until the stream was dry. remaining forest cover, forest reserves, and protected Start points for 30-m stream plots were selected randomly area within Peninsular Malaysia (Perhilitan 2001), but it within the 10 m length. Stream plots included 1 m of has deforested areas due to urbanization and the expan- stream bank and the width of the wetted channel within sion of palm plantations. Oil palm and other types of that 30-m plot. Start points for the riparian and terrestrial plantations cover 43% of the state (Struebig et al. 2009) survey plots were also randomly selected from among the (Fig. 1). Our study areas were in mature oil-palm plan- 10-m lengths marked on the stream. Riparian plots were tations (established >30 years ago) and secondary forest 2 m wide and started 1 m out from the stream bank. sites (previously logged >50 years ago). Two periods of Terrestrial plots were of the same width but located a maximum precipitation occur between September and minimum of 10 m out from the stream bank. Riparian December and between March and May, and tempera- and terrestrial plots were surveyed in the direction of ture is also seasonal (highest temperature occurring from stream flow but did not overlap each other. All plot types April to September) (Struebig et al. 2009). Because tem- (stream, riparian, and terrestrial) were surveyed weekly perature and rainfall patterns affect anuran activity and in a random order to standardize temporal variability of detectability (Pellet & Schmidt 2005), we surveyed at the anuran presence and density. same time each year, after the spring wet season. We We searched each plot for anurans 7 times between also collected data on rainfall and temperature to con- May and July in 2009 (total time 3568 min) and 4 times firm their consistency. We used rain gauges to measure between May and June in 2010 (total time 2195 min). We rainfall and hand-held thermometers to measure night conducted direct visual searches to find stream-dwelling time temperature each night before we surveyed plots and arboreal anurans. To find anurans in riparian and for anurans.

Conservation Biology Volume 27, No. 3, 2013 618 Effects of Oil-Palm Plantations on Anurans

Figure 2. Schematic representation of sampling of amphibians in forest and oil-palm plantations. terrestrial plots, we used visual and disturbance searches Within each stream plot, we estimated the mean width and displaced leaf litter, logs, and rocks. When possible, (average of maximum and minimum length) and maxi- we captured individual and photographed them mum depth of the wet channel. We estimated density of for later confirmation of species identification. We re- vegetation up to 1 m from the wet channel on the basis of leased individuals at point of capture at the end of each Keller et al.’s (2009) method, in which the apex of indi- plot survey. Individuals that were not captured but were vidual plants is categorized into 3 height categories (0–10 clearly identifiable to species were included in the data m, 11–100 m, and 101–200 m). We measured water pH, set. dissolved oxygen, conductivity, and total dissolved solids Prior to each plot survey we measured macrohabitat with a multivariable-YSI kit (YSI, Yellow Springs, Ohio, variables. We were careful not to disturb anurans in or USA) at the beginning, middle, and end of each stream out of survey plots when we took these measurements plot. We estimated the percentage of each substrate type and did not observe any short-term movement associated in plots of 2 2 m at the beginning, middle, and end with this effort. We estimated percent canopy cover, soil of each stream× plot. Apart from detritus (dead, floating pH, woody debris (percent leaf-litter cover, depth and vegetation) and sand, we assumed substrate material was weight of leaf litter and felled logs [<50 cm and 50 cm rounded and based its classification on estimated diam- maximum diameter]), percent cover of undergrowth≥ veg- eters: gravel ( 10 mm, <30 mm), pebbles ( 30 mm, etation, and number of trees (<50 cm and 50 cm diam- <60 mm), cobble≥ ( 60 mm, <200 mm), and≥ boulders eter at breast height) in all riparian and terrestrial≥ plots. ( 200 mm). ≥ We estimated canopy cover from digital photographs of ≥ the canopy taken at the beginning, middle, and end of Data Analyses the plot (D40 SLR camera, 10.5mm f/2.8 DX fisheye lens, Nikon, Tokyo). We used the ImageJ software (Rasband We compared average nightly temperature and weekly 2011) to set a minimum threshold value of 100 pixels, and rainfall between forest and plantation and among years pixels that exceeded this value were considered canopy with a Pairwise Wilcoxon test. We identified differences cover (black pixels). We averaged the percentage of black in frog abundance between forest and plantation and pixels across the 3 points to estimate canopy cover for among plot types with generalized linear models (GLMs) each plot. We measured soil pH, estimated leaf-litter char- with Poisson errors in R statistical software (R Core Devel- acteristics, and percent undergrowth vegetation at the opment Team 2009). We detected overdispersion while beginning, middle, and end of each plot and averaged fitting the Poisson GLM of the abundance data and subse- measurements per plot. We visually estimated percent quently corrected the standard errors with a quasi-GLM leaf-litter cover in 1 1 m quadrats and percent cover model in which variance is a product of the mean and × of undergrowth vegetation in 2 2 m quadrats. We esti- dispersion parameter. We simplified and tested models × mated depth of leaf litter 3 times at the beginning, middle, by deleting terms and comparing changes in deviance and end of each riparian and terrestrial plot. We collected with F tests (Zuur et al. 2009). leaf litter from every quadrat and measured its wet weight Missed or undetected species are universal problems in the field. Macrohabitat measurements were repeated in biological diversity studies due to imperfect detec- in each sampling year. tion probabilities (Pollock et al. 2002; Pellet & Schmidt

Conservation Biology Volume 27, No. 3, 2013 Faruk et al. 619

2005).To account for a lack of detection of species, we correlation coefficients express the rank (tau) and lin- used Species Prediction and Diversity Estimation (SPADE) ear (r) relations between the variables and ordination software (Chao & Shen 2010) to estimate species rich- scores. Strong correlations (positive or negative) are rep- ness. We used abundance-based estimates (ACE) to com- resented as vectors on the ordination plots. Variables pare species richness between forest and plantation and and species with R2 values or tau values >0.4 or < 0.4 among plot types (Chao & Lee 1992; Chao 2005). We di- were considered, respectively, positively or negatively− vided abundance data into 2 groups, rare and abundant. correlated. The abundant group represented anuran species with We used a variant of the traditional multivariate 10 individuals. We did this because SPADE can be used analysis of variance (MANOVA), known as permutation- only≥ with frequency statistics of rare species to estimate based multivariate analysis of variance (PerMANOVA) the number of undetected species (i.e., abundant species (Anderson 2001) to test the hypothesis of no difference are detected too frequently to provide information on between community composition between forest and missing species) (Chao & Shen 2003). The ACE calcula- plantation sites. In contrast to traditional MANOVA, tion takes into account undetected species by incorpo- PerMANOVA does not require distribution assumptions rating a rule, originally proposed by Turing (Good 1953), such as normality and homogeneity of variance, which is where the proportion of known singletons (species only rarely met in data on ecological communities. The same found once) is the same as the fraction of undetected distance matrix we used for the ordination was used to species in the sample. This can be intuitively assumed compare community composition between forest and because the detection of a previously undetected species plantation and among plots in forest and plantation. We must be a singleton in the sample that includes an addi- used the adonis function in the vegan program of R to per- tional individual. form the PerMANOVA (R Core Development Team 2009). Chao and Shen (2010) recommend the use of the de- This method partitions sums of squares of multivariate fault cut-off point of 10 individuals between rare and data sets and allows for the use of non-Euclidean distance abundant groups if the ACE value is larger than the Chao1 measures. The pseudo-F ratios from PerMANOVA cal- estimate. Results from initial analyses indicated the ACE culations were obtained without calculating the central value was higher than the Chao1 estimate for all our sam- location of groups (centroid) because the centroid does ples (Results); therefore, 10 individuals were retained as not necessarily represent the central tendency in a non- the cut-off point in subsequent analyses. The higher the Euclidean space (McCune et al. 2002). A total of 1000 coefficient of variation (CV) for rare species the more permutations of raw data were used to obtain p values. heterogeneous the sample. Thus, models that are based on homogenous distributions (ACE) may substantially un- derestimate the species richness of a site. We classified Results samples with CVs >0.8 as highly heterogenous and used ACE-1 as an alternative measure of species richness for Mean nightly temperature did not differ significantly be- those sites (Chao & Shen 2010). tween plantation and forest in either year (2009: W 121, Community-composition data were transformed into p 0.383; 2010: W 162, p 0.098), except for= forest adistancematrixwithBray-Curtisasadistancemea- sites,= which were moderately= = warmer in 2010 compared sure to illustrate habitat-specific anuran community as- with 2009 (W 67.5, p 0.026). Mean nightly rainfall semblages and associated macrohabitat characteristics also did not differ= significantly= between plantation and (Keller et al. 2009). We pooled data from both years forest in either year (2009: W 1176, p 0.6; 2010: W and used a nonmetric multidimensional scaling (NMDS) 1486, p 0.835). = = analysis in PC-ORD (version 5; MjM Software, Glene- = Anuran abundance= did not differ significantly between den Beach, Oregon, USA) to construct ordinations of plantation and forest (F[1,46] 0.752, p 0.391); mean anuran communities between forest and plantation for abundance between the 2 site= types was= similar over- each plot type (stream, riparian, and terrestrial). Out- all (forest:10.4, 95% CI 3.16, n 20; plantation: 18.8, puts grouped species into assemblages representing co- 95% CI 8.19, n 26). When plot= type was added to occurring species associated with given axes. We set the our explanatory= variables, plot type and site interacted program to generate 250 iterations and 250 runs of real significantly (Table 1). Mean abundance of anurans in and randomized data to produce the final ordination of plantation streams were lower than mean abundance minimum stress and the best fit for all 3 plot types. We in forest streams, whereas mean abundance was greater did not use p values to interpret environmental differ- in plantation terrestrial plots than forest terrestrial plots ences between forest and plantation sites because ordi- (Fig. 3a). nation scores violated the assumptions of independence. Species richness did not vary significantly between Instead, we used correlation coefficients between each forest and plantation (forest: 22.2, 95% CI 0.613, n site and their corresponding axes to evaluate relations 20; plantation: 22.4, 95% CI 0.829, n 26), and CV did= between individual variables and ordination axes. These not indicate forest and plantation were= heterogeneous.

Conservation Biology Volume 27, No. 3, 2013 620 Effects of Oil-Palm Plantations on Anurans

Table 1. Results of analysis of deviance for quasi-generalized linear The 95% confidence intervals of estimates of species model with Poisson errors of anuran abundance against habitat type richness overlapped among plot types across forest and (forest, plantation), plot type (stream, riparian, terrestrial), and a 2- plantation, indicating a lack of clear difference in species way interaction of habitat and plot.a richness (Fig. 3b). Variable Residual Residual Model Model The NMDS axes cumulatively accounted for 75% of Added deviance df deviance df Fp the overall variation in community composition in stream None 366 45 plots, 78% in riparian plots, and 65% in terrestrial plots. Habitat 360 44 6.29 1 0.866 0.358 Both axes for the stream-plot ordination contributed Plot 344 43 16.3 2 1.12 0.336 equivalently to variation, whereas NMDS axis 2 ac- Habitat:plot 293 40 50.2 2 3.45 0.041 counted for the majority of the variation in both ripar- aWe used F tests to compare models after each variable was added ian and terrestrial plots (Fig. 4). Axis 1 for stream plots sequentially. represented a gradient of decreasing canopy closure and increasing values of water quality measurements (pH, total dissolved solids, conductivity, and temperature). Streams in plantation had higher conductivity and water temperature, greater amounts of total dissolved solids, and higher pH than those in forest. There were no strong associations with any of the measured habitat variables among forest sites. However, the 2 forest sites were sep- arated along the second axis (percentage of cobble [R2 0.505, tau 0.483] and dissolved oxygen [R2 0.501,= tau 0.473])= − (Fig. 4a). = Although= − species richness was similar between forests and plantation, community composition was strikingly different between them (pseudo F[1,14] 5.94, p < 0.01). For example, the common green frog= (Hylarana erythrea) and Nicobar cricket-frog (Fejervarya nico- bariensis) were both strongly associated with plantation streams, whereas the Asian giant toad (Phrynoidis as- pera) and Hose’s frog (Odorana hosii)werepredomi- nantly found in forest streams. There was a weaker, but still significant difference in anuran communities among different forest streams (pseudo F[1,6] 11.5, p 0.028): white-lipped frogs (H. labialis)weremoreabundantin= = Kuala Lompat streams than in Bukit Rengit streams (Fig. 4a; for location of sites see Fig. 1). The NMDS 2 (y-axis) represented a gradient of increas- ing canopy closure and woody debris (small logs, percent- age of leaf litter, and weight of leaf litter) in riparian plots. Forest sites had significantly higher values for variables correlated to NMDS 2 than plantation sites (pseudo F[1,10] 6.11, p 0.002). Crab-eating frogs (F. cancrivora)and cricket= frogs= (F. limnocharis) were most abundant in plantation sites, whereas white-lipped frogs (H. labialis) and glandular frogs (H. glandulosa) were most abundant in forest sites (Fig. 4b). In contrast to stream plots, anu- ran communities in riparian plots did not differ among forest sites (pseudo F 4.51, p 0.092); however, [1,4] = = Figure 3. In forest and plantation, (a) mean frog communities in riparian plots differed among plantation abundance averaged across years for each plot type sites (pseudo F 2.32, p 0.022). Specifically, Sun- [2,6] = = (stream, riparian, and terrestrial) and (b) estimated gai Mai (for location of sites see Fig. 1) riparian sites amphibian species richness across plot types averaged were separated from the other 2 plantation riparian sites over the 2 years (lines, 95% CI). Data were pooled for (pseudo F 2.91, p 0.015), and abundance of the (1,7) = = all sampling plots. chubby frog (Kaloula pulchra)andtheAsiancommon toad (Duttaphrynus melanostictus) were greater at this plantation than the other 2.

Conservation Biology Volume 27, No. 3, 2013 Faruk et al. 621

Terrestrial plots in forest had higher canopy cover, more leaf litter, and higher densities of small and large trees than terrestrial plots in plantation. Cricket frogs (also abundant in riparian plots) and common tree frogs (Polypedates leucomystax) were more abundant in ter- restrial plots in plantation than in forest terrestrial plots (Fig. 4c). Again, these differences were significant be- tween forest and plantation sites (pseudo F[1,10] 3.62, p 0.001), but not within forest and plantation sites= (for- est:= pseudo F 2.15, p 0.099; plantation: pseudo [1,4] = = F[1,4] 1.11, p 0.494). Only one species (painted chorus= frog [Microhyla= butlerii]) was present in both plantation and forest and only in terrestrial plots.

Discussion

The importance of oil palm as a source of revenue in countries like Malaysia means plantations will be a persis- tent part of that landscape and must be included in local conservation planning. The intense agricultural practices associated with production of palm oil are linked to a reduction in diversity (Chung et al. 2000; Glor et al. 2001; Bruhl¨ & Eltz 2010). Other animal taxa may not respond so directly when forests are converted to oil-palm plantations. For those that do not exhibit such direct responses, oil-palm plantations may sustain their populations to some degree. For example, the inconsis- tent responses of North American and Australian anurans

! Figure 4. Nonmetric multidimensional scaling (NMDS), in which length and angle reflect the strength and direction of association, of amphibian assemblages in (a) stream, (b) riparian, and (c) terrestrial plots in forest (BR and KL) and plantation (SM, LAN, and KER) over 2 years (triangles, sites; distance on ordination, dissimilarities in amphibian species composition on the basis of Bray-Curtis coefficients; squares, species found only in forest sites; circles, species found only in plantations; star, species found in both forest and plantation). See Supporting Information for definitions of abbreviations. Vector loadings are macrohabitat variables with strong associations (see text) (canopy, percent canopy cover; conductivity units, siemens/m; temperature, degrees Celsius; cobble, percent cobble on river bed; oxygen, dissolved oxygen content in parts per million; TDS, total dissolved solids in parts per million; pH, water pH; log <50 m, number of logs with maximum diameter <50 m; litter, percentage of ground cover that was leaf litter; leaf wgt, weight difference between wet and dry leaf litter [g]; tree >50 m, number of trees with diameter at breast height >50 m; tree <50 m, number of trees with diameter at breast height <50 m).

Conservation Biology Volume 27, No. 3, 2013 622 Effects of Oil-Palm Plantations on Anurans to intensive changes in land use suggest that anurans are both types of sites). This differs from the findings of good candidate taxa for investigating the potential of oil Gillespie et al. (2012). They did not detect any microhylid palm as anuran habitat. species in oil-palm plantation sites. This difference in To investigate this potential, the use of appropriate results may be due to study location; the common micro- methods is vitally important. Many authors have com- hylid species found in oil-palm plantations in Peninsular mented on the problems of decreased detection proba- Malaysia (dark-sided chorus frog [M. heymonsi]and bility and false negatives affecting the results of monitor- painted chorus frog) are not included in Bornean ing studies (Dodd & Dorazio 2004; MacKenzie & Royle amphibian checklists (Inger & Lian 1996; Sheridan et al. 2005; Royle 2006). Even with extensive repeated sam- 2012). The relative abundance of microhylid species in pling, individuals and even species can be missed, which Peninsular Malaysian oil-palm plantations contributed may partially explain the reduced abundance of frogs to the comparable species richness and abundance esti- we detected at nonstream sites in secondary forests. mates among forest and plantation. By contrast, Gillespie In contrast to anuran surveys conducted by Onn et al. et al. (2012) estimated species richness estimates is (2010) and Gillespie et al. (2012), we failed to detect any lower in oil-palm plantations than in secondary forest in Rhacophoridae species in forest, but we detected micro- Borneo. This difference between our results and their hylids in oil-palm plantations, which were not found by results suggests the effect of conversion of forest to Gillespie et al. (2012). This result does not imply that oil-palm plantation on anuran communities may not be rhacophorids were not present; species such as Nyctilus generalizable across Southeast Asian landscapes. pictus and Rhacophorus appendiculatis are known to Our results are consistent with the idea that habitat occur in secondary forests of Peninsular Malaysia (Onn conversion alters anuran communities and allows colo- et al. 2010). It also does not mean microhylids are absent nization of disturbed areas by species that can use the in the other study systems because tropical leaf-litter frogs newly created niche space. If the presence of endemic are notoriously difficult to detect, even with extensive forest species is a yardstick of conservation value of surveys (Veith et al. 2004). Nevertheless, labor-intensive an area, then the anuran community found in oil-palm surveys can be impractical for wildlife monitoring, abun- plantations we surveyed represents a community of lit- dance only provides one metric of biological diversity, tle conservation concern because the species that were and intensive surveys may actually distort estimates of present are ubiquitous across Malaysia and other parts abundance (Kery´ et al. 2009). The use of techniques that of Southeast Asia (Norhayati et al. 2009). If community provide an estimate of the proportion of missed species, unpredictability across sites is considered a measure of such as SPADE, may be suited to cases in which rapid biological diversity, then the riparian areas of oil-palm but robust comparisons of biological diversity metrics plantations, which exhibited variability similar to streams (e.g., species richness) are appropriate. We suggest that in secondary forest sites, appear to be of great conserva- future studies investigating how habitat alteration affects tion concern. Malaysian anurans avoid count statistics that fail to ac- In oil-palm plantations, some habitat characteristics count for missed species and be designed to measure appear homogenized (e.g., forest structure, simplified community composition of anurans. stream substrate, and decreased availability of leaf lit- Differences in anuran abundance, species richness, ter and woody debris) or altered in a way that may community composition, and species habitat associations have directly affected air and water temperature (opened cannot be compared with single estimators. Furthermore, canopy). The plantations we surveyed were created be- surveys designed to measure differences at a landscape fore the establishment of riparian-zone legislation, which level may not be suitable for detecting differences at now requires plantation owners to maintain a strip of ri- the patch level or be effective at linking differences to parian vegetation along streams. Plantation streams have specific habitat features. Our results are consistent with banks that are more straight sided, due to a lack of all of these points. Species richness and abundance in deep-rooted riparian vegetation, and are more prone to oil-palm plantations and secondary forest were similar, flooding, due to high rates of sedimentation, than streams yet abundance varied significantly among plots (riparian, in forest. Plantation owners use mechanical dredging to terrestrial, stream). For example, abundance was lower relieve water pressure during the rainy season (A.F., per- in plantation streams than in forest streams. Moreover, sonal observation), which removes large stream-substrate the congruence in abundance and species richness in material and vegetation matter in the stream. This dis- forest and plantation did not reflect the stark difference in turbance may explain the reduced frog abundance we anuran community structure between these 2 areas. Oil- found in plantation streams and why the species we palm plantations were dominated by species considered detected there were disturbance specialists. There are disturbance tolerant (Sheridan 2003; Wu et al. 2006; Gille- a number of land-management practices used to address spie et al. 2012), and the anuran community in plantation the effects of land-use changes on biological diversity, exhibited little, if any, overlap with the communities in including undisturbed strips of vegetation along edges secondary forests (only the painted chorus frog inhabited of agricultural fields, riparian buffers, and construction

Conservation Biology Volume 27, No. 3, 2013 Faruk et al. 623 of ponds and tunnels under roads for migrating amphib- Danielsen, F., et al. 2009. Biofuel plantations on forested lands: double ians (Woltz et al. 2008; Garcia-Gonzalez & Garcia-Vazquez jeopardy for biodiversity and climate. Conservation Biology 23:348– 2010; Perry et al. 2011). On the basis of our results, we be- 358. Dodd, C., Jr., and R. Dorazio. 2004. Using counts to simultaneously lieve plantation streams should be a target for future man- estimate abundance and detection probabilities in a salamander agement, through the maintenance of stream complexity community. Herpetologica 60:468–478. and riparian buffers. At the very least, given the reduction Donald, P. F. 2004. Biodiversity impacts of some agricultural commodity in frog abundance in plantations streams, it would seem production systems. Conservation Biology 18:17–37. prudent to ascertain in the future whether plantations Ernst, R., and M. O. Rodel.¨ 2005. Anthropogenically induced changes of predictability in tropical anuran assemblages. Ecology 86:3111– with riparian buffer zones contain greater numbers of 3118. frogs or support more forest-associated species. Garcia-Gonzalez, C., and E. Garcia-Vazquez. 2010. The value of tradi- It is unclear how riparian and terrestrial areas in oil- tional troughs as freshwater shelters for amphibian diversity. Aquatic palm plantations may be managed to maintain anuran Conservation: Marine and freshwater ecosystems 21:74–81. biological diversity. Oil-palm harvesting practices are a Gardner, T. A., M. A. Ribeiro-Junior, J. Barlow, T. C. S. Avila-Pires, M. HoogMoed, and C. A. Peres. 2007. The value of primary, secondary, frequent disturbance and plantations are monocultures; and plantation forests for a Neotropical herpetofauna. Conservation thus, we see little chance of manipulating these practices Biology 21:775–787. in a way that would create habitat for forest-specialist Gibbs, J. P. 1998. Distribution of woodland amphibians along a forest anurans. Even if management of plantation streams were fragmentation gradient. Landscape Ecology 13:263–268. successful, to ensure anuran conservation in Malaysia Gillespie, G. R., E. Ahmad, B. Elahan, A. Evans, M. Ancrenaz, B. Goossens, and M. P. Scroggie. 2012. Conservation of amphibians such management would need to be accompanied by in borneo: relative value of secondary tropical forest and non-forest maintenance of forest reserves. habitats. Biological Conservation 152:136–144. Glor, R. E., A. S. Flecker, M. F. Bernard, and A. G. Power. 2001. Lizard di- versity and agricultural disturbance in a Caribbean forest landscape. Biodiversity Conservation 10:711–723. Supporting Information Good, I. J. 1953. The population frequencies of species and the estima- tion of population parameters. Biometrika 40:237–264. Hecnar, S. J., and R. T. M’Closkey. 1997. The effects of predatory fish Number of individuals found for each species per habitat on amphibian species richness and distribution. Biological Conser- sampled. Species abbreviations correspond to Figure 4a- vation 79:123–131. c. Threat status for each species are based on the IUCN Inger, R. F., and T. F. Lian. 1996. Checklist of the frogs of Borneo. The red list criteria (http://www.iucnredlist.org/); DD-data Raffles Bulletin of Zoology 44:551–574. deficient, Lc-least concern, nt- near threatened (Table IUCN (International Union of Conservation of Nature). 2010. Global amphibian assessment. IUCN, Gland, Switzerland. Available from S1 are available online). The authors are solely respon- http://www.amphibians.org/redlist/ (accessed May 2010). sible for the content and functionality of these materials. Keller, A., M. O. Rodel,¨ K. E. Linsenmair, and T. U. Grafe. 2009. The Queries (other than absence of the material) should be importance of environmental heterogeneity for species diversity directed to the corresponding author. and assemblage structure in Bornean stream frogs. Journal of Animal Ecology 78:305–314. Kery,´ M., R. M. Dorazio, L. Soldaat, A. van Strien, A. Zuiderwijk, and Literature Cited J. A. Royle. 2009. Trend estimation in populations with imperfect detection. Journal of Applied Ecology 46:1163–1172. Anderson, M. J. 2001. A new method for non-parametric multivariate Koh, L. 2007. Impending disaster or sliver of hope for Southeast Asian analysis of variance. Austral Ecology 26:32–46. forests? The devil may lie in the details. Biodiversity Conservation Bruhl,¨ C. A., and T. Eltz. 2010. Fuelling the biodiversity crisis: species 16:3935–3938. loss of ground-dwelling forest ants in oil palm plantations in Sabah, Koh, L. P., and D. Wilcove. 2008. Is oil palm agriculture really destroying Malaysia (Borneo). Biodiversity Conservation 19:519–529. tropical biodiversity? Conservation Letters 1:60–64. Chao, A. 2005. Species estimation and applications. Pages 7907–7916. MacKenzie, D., and J. Royle. 2005. Designing occupancy studies: gen- 2nd edition. Wiley, New York. eral advice and allocating survey effort. Ecology 42:1105–1114. Chao, A., and S. M. Lee. 1992. Estimating the number of classes via Malaysian Ministry of Agriculture and Agro-Based Industry. 2008. sample coverage. Journal of the American Statistical Association Malaysian Ministry of Agriculture and Agro-Based Industry, Putra- 87:210–217. jaya. Available from http://www.moa.gov.my/web/guest/home (ac- Chao, A., and T.-J. Shen. 2003. Nonparametric estimation of Shannon’s cessed November 2008). index of diversity when there are unseen species in sample. Envi- McCune, B., J. B. Grace, and D. L. Urban. 2002. Analysis of ecologi- ronmental and Ecological Statistics 10:429–443. cal communities. Pages 102–178. MjM Software Design, Gleneden Chao, A., and T. J. Shen. 2010. Program SPADE (species prediction Beach, Oregon. and diversity estimate). Available from http://chao.stat.nthu.edu.tw Norhayati, A., D. Belabut, and C. K. Onn. 2009. Amphibians. & (accessed January 2009). reptiles of Peninsular Malaysia. Available from http://amphibia. Chazdon, R. L., C. A. Peres, D. Dent, D. Sheil, A. E. Lugo, D. Lamb, N. E. my/index.php (accessed April 2009). Stork, and S. E. Miller. 2009. The potential of species conservation Onn, C. K., D. Belabut, and N. Ahmad. 2010. A revised checklist of the in tropical secondary forests. Conservation Biology 23:1406–1417. amphibians of Peninsular Malaysia. Russian Journal of Herpetology Chung, A. Y. C., P. Eggleton, M. R. Speight, P. M. Hammond, and 17:202–206. V. K. Chey. 2000. The diversity of beetle assemblages in different Parris, K. M. 2004. Environmental and spatial variables influence the habitat types in Sabah, Malaysia. Bulletin of Entomological Research composition of frog assemblages in sub-tropical eastern Australia. 90:475–496. Ecography 27:392–400.

Conservation Biology Volume 27, No. 3, 2013 624 Effects of Oil-Palm Plantations on Anurans

Pellet, J., and B. R. Schmidt. 2005. Monitoring distributions using call Struebig, M. J., T. Kingston, A. Zubaid, S. C. Le Comber, A. Mohd- surveys: estimating site occupancy, detection probabilities and in- Adnan, A. Turner, J. Kelly, M. B., and S. J. Rossiter. 2009. Conser- ferring absence. Biological Conservation 123:27–35. vation importance of limestone karst outcrops for palaeotropical Perhilitan. 2001. Krau Wildlife Reserve management plan for 2002- bats in a fragmented landscape. Biological Conservation 10:2089– 2006. Perhilitan, Kuala Lumpur, Malaysia. 2096. Perry, R. W., T. B. Wigley, M. A. Melchiors, R. E. Thill, P. A. Tappe, Stuart, S. N., J. S. Chanson, N. A. Cox, B. E. Young, A. S. L. Rodrigues, and D. A. Miller. 2011. Width of riparian buffer and structure of D. L. Fischman, and R. Waller. 2004. Status and trends of amphibian adjacent plantations influence occupancy of conservation priority declines and extinctions worldwide. Science 309:1–8. birds. Biodiversity Conservation 20:625–642. Todd, B. D., and B. B. Rothermel. 2006. Assessing quality of clearcut Pollock, K. H., J. D. Nichols, T. R. Simons, G. L. Farnsworth, L. L. habitats for amphibians: effects on abundances versus vital rates Bailey, and J. R. Sauer. 2002. Large scale wildlife monitoring studies: in the southern toad (Bufo terrestris). Biological Conservation statistical methods for design and analysis. Environmetrics 13:105– 133:178–185. 119. Veith, M., S. Lotters,¨ F. Andreone, and M. O. Rodel.¨ 2004. Measuring R Core Development Team. 2009. R: a language and environment for and monitoring amphibian diversity in tropical forests. II. Estimat- statistical computing. R Foundation for Statistical Computing, Vi- ing species richness from standardized transect censing. Ecotropica enna. 10:85–99. Rasband, W. S. 2011. Imagej. U. S. National Institute of Health, Bethesda, Wanger, T. C., A. Saro, D. T. Iskandar, B. W. Brook, N. S. Sodhi, Y. Maryland. Available from http://imagej.nih.gov/ij/ (accessed January Clough, and T. Tscharntke. 2009. Conservation value of cacao agro- 2012). forestry for amphibians and reptiles in Southeast Asia: combining Royle, J. A. 2006. Site occupancy models with heterogeneous detection correlative models with follow-up field experiments. Journal of Ap- probabilities. Biometrics 62:97–102. plied Ecology 46:823–832. Sheridan, J. A. 2003. Reproductive variation corresponding to breeding Williams, S. E., and J.-M. Hero. 1998. Rainforest frogs of the Australian season length in three tropical frog species. Journal of Tropical wet tropics: guild classification and the ecological similarity of de- Ecology 25:583–592. clining species. Proceedings of the Royal Society Biological Sciences Sheridan, J. A., S. D. Howard, P. Yambun, J. L. Rice, R. Cadwallader- 265:597–602. Staub, A. Karoulus, and D. Bickford. 2012. Novel behaviors of Woltz, H. W., J. P. Gibbs, and P. K. Ducey. 2008. Road cross- Southeast Asian rhacophorid frogs (anura: Rhacophoridae) with ing structures for amphibians and reptiles: informing design an updated anuran species list for Danum Valley, Sabah, Malaysian through behavioural analysis. Biological Conservation 141:2745– Borneo. Tropical Natural History 12:1–8. 2750. Sodhi, N. S., D. Bickford, A. C. Diesmos, T. M. Lee, L. P. Koh, B. W. Wu, Z., Y. Li, and B. R. Murray. 2006. Insular shifts in body size of rice Brook, C. H. Sekercioglu, and C. J. A. Bradshaw. 2008. Measuring frogs in the Zhoushan archipelago, China. Journal of Animal Ecology the meltdown: drivers of global amphibian extinction and decline. 75:1071–1080. Biodiversity Conservation 19:317–328. Zuur, A. F., E. N. Ieno, N. J. Walker, A. A. Saveliev, and G. M. Smith. 2009. Sodhi, N. S., B. W. Brook, and C. J. A. Bradshaw. 2007. Tropical conser- Mixed effects models and extensions in ecology with R. Springer, vation biology. Blackwell Oxford, United Kingdom. New York.

Conservation Biology Volume 27, No. 3, 2013