Importance of Reserves, Fragments, and Parks for Butterfly Conservation in a Tropical Urban Landscape Author(s): Lian Pin Koh and Navjot S. Sodhi Source: Ecological Applications, Vol. 14, No. 6 (Dec., 2004), pp. 1695-1708 Published by: Ecological Society of America Stable URL: http://www.jstor.org/stable/4493684 Accessed: 29/09/2010 11:16

Your use of the JSTOR archive indicates your acceptance of JSTOR's Terms and Conditions of Use, available at http://links.jstor.org/page/info/about/policies/terms.jsp. JSTOR's Terms and Conditions of Use provides, in part, that unless you have obtained prior permission, you may not download an entire issue of a journal or multiple copies of articles, and you may use content in the JSTOR archive only for your personal, non-commercial use.

Please contact the publisher regarding any further use of this work. Publisher contact information may be obtained at http://links.jstor.org/action/showPublisher?publisherCode=esa.

Each copy of any part of a JSTOR transmission must contain the same copyright notice that appears on the screen or printed page of such transmission.

JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide range of content in a trusted digital archive. We use information technology and tools to increase productivity and facilitate new forms of scholarship. For more information about JSTOR, please contact [email protected].

Ecological Society of America is collaborating with JSTOR to digitize, preserve and extend access to Ecological Applications.

http://links.jstor.org Ecological Applications, 14(6), 2004, pp. 1695-1708 C 2004 by the Ecological Society of America

IMPORTANCE OF RESERVES, FRAGMENTS, AND PARKS FOR BUTTERFLY CONSERVATION IN A TROPICAL URBAN LANDSCAPE

LIAN PIN KOH AND NAVJOTS. SODHI1 Department of Biological Sciences, National University of , 14 Science Drive 4, Singapore 117543, Republic of Singapore

Abstract. We assessed the effectiveness of forest reserves (i.e., protected old secondary and primary forests), fragments (i.e., scattered ruderal vegetation), and urban parks (i.e., artificially revegetated habitats) in conserving butterfly diversity in a highly urbanized tropical landscape (i.e., Singapore), by testing the hypothesis that forest reserves have the highest butterfly species richness among these habitats. We investigated which environ- mental factors (e.g., canopy cover) affect the distribution of butterflies across the habitats; and also tested the hypothesis that butterfly communities of different habitats have distinct ecological traits. Further, we examined the important determinants (e.g., area) of butterfly richness in urban parks, by testing the hypothesis that the number of potential larval host plant species occurring in the park is the best predictor of butterfly species richness. Rar- efaction analyses showed that forest reserves had the highest number of species, number of unique species, density of species, and community evenness among the habitats, implying that, in urban landscapes, the least human-disturbed habitats should be given the highest conservation priority. Forest reserves and urban parks adjoining forests collectively ac- counted for 91% of all butterfly species recorded in this study, suggesting that their pres- ervation will likely achieve maximum complementarity for effective butterfly conservation. Ordination analyses revealed that different butterfly species responded differently to en- vironmental factors (e.g., canopy cover), highlighting the importance of maintaining en- vironmental heterogeneity for the conservation of different butterfly species. Classification tree analysis indicated that butterfly communities of different habitats (e.g., forests, urban parks) have distinct ecological traits (e.g., host plant specificity), whereby urban avoiders were 89% likely to be forest dependent and 63% likely to be monophagous, while urban adapters were 87% likely to be cosmopolitan and 67% likely to be oligo- or polyphagous. Regression analyses showed that the number of potential larval host plant species and isolation from forests were important determinants of butterfly species richness in urban parks, indicating that urban parks should be revegetated with a diversity of potential larval host plants and should be situated as near as possible to a forest, in order to maximize their conservation value. Key words: biodiversity; butterfly conservation; classification tree; conservation value; ordina- tion; rarefaction; reconciliation ecology; Southeast Asia; urbanization.

INTRODUCTION Human Settlements 1996), urban ecology has received little Forests are being lost at an alarming rate across the relatively emphasis from conservation biologists world's tropical regions (Achard et al. 2002). Among (McKinney 2002, Ormerod et al. 2002, Lim and Sodhi the anthropogenic causes of deforestation, urbanization 2004). Miller and Hobbs (2002) attribute this to the is expected to be a major determinant of species loss traditional focus of conservation research on "natural" (Kowarik 1995, Marzluff 2001, McKinney 2002, Mill- ecosystems (e.g., primary forests) to preserve and pro- er and Hobbs 2002). Unlike other forms of habitat dis- tect them from human settlements and activities. How- turbance (e.g., logging), where forests may regenerate ever, the economic and political constraints of pre- over time through ecological succession (Sheil and serving large tracts of pristine habitats, as well as the Burslem 2003), urbanization often irreversibly replaces rapidity and ubiquity of urban sprawl have galvanized natural habitats (e.g., primary forests) with persistent some ecologists to rethink the traditional conservation artificial ones (e.g., human cities), resulting in long- strategies of reserving and restoring natural habitats lasting negative impacts (e.g., species extinctions) on (Dale et al. 2000, Miller and Hobbs 2002, Daily 2003, the native biodiversity (Stein et al. 2000). Although Rosenzweig 2003). Instead, they argue that conserva- human settlement has become the globally prevailing tion planning should include reconciliatory measures source of land use Nations Centre for change (United that encompass a wider range of land uses (e.g., urban areas), so that human activities can with min- Manuscriptreceived 29 August 2003; revised 4 February proceed 2004; accepted9 April2004. CorrespondingEditor: J. A. Logan. imum displacement of the native species. For example, 1 Correspondingauthor. E-mail: [email protected] in addition to preserving as many remnant natural hab- 1695 1696 LIAN PIN KOH AND NAVJOT S. SODHI Ecological Applications Vol. 14, No. 6 itats (e.g., forest reserves) as possible, the creation and native and exotic cultivated flora (Teo et al. 2003). They maintenance of artificially revegetated habitats (e.g., function as "green spaces" in densely populated urban urban parks), where certain native species can persist residential and commercial areas. As forest reserves (Kendle and Forbes 1997), may be a viable conser- are the least human-disturbed of these habitats (Turner vation strategy in highly urbanized landscapes. et al. 1994, Corlett 1997), we test the hypothesis that Recent studies investigating the ecological impacts forest reserves are most effective in conserving but- of urbanization in mostly temperate regions have terfly species richness, by comparing the number of shown distinct changes in species richness and com- species, number of unique species, density of species, position along rural-urban gradients; local extinctions density of individuals, and community evenness among of native species; increases in the number of exotic the different habitats. Second, we examine how the species toward centers of urbanization; and the persis- distribution of butterflies may be affected by environ- tence of certain native species in urban areas (e.g., mental factors (e.g., canopy cover) and ecological traits Kowarik 1995, Blair and Launer 1997, Hardy and Den- (e.g., larval host plant specificity). We investigate nis 1999, Marzluff 2001). A current challenge is to which environmental factors (e.g., canopy cover) affect understand the differential response of individual spe- the distribution of butterflies across the habitats; and cies to urbanization, including their underlying mech- also test the hypothesis that butterfly communities of anisms, in order to identify vulnerable species and to different habitats have distinct ecological traits. Third, develop effective measures for their conservation. Fur- we examine the important determinants of butterfly ther, it is critical to evaluate the effectiveness of dif- species richness (e.g., area) in existing urban parks to ferent habitats (e.g., artificially revegetated urban improve future park design and management for ef- parks) in maintaining the diversity of native species in fective butterfly conservation. As previous studies have urban landscapes. shown butterflies to be closely associated with their With Southeast Asia's rapid economic development, larval host plants (e.g., Koh et al. 2004), we test the urban areas will likely be an important, if not dominant, hypothesis that the number of potential larval host plant feature of the regional landscape. Since Southeast Asia species occurring in the park is the best predictor of has one of the highest global concentrations of endemic butterfly species richness. We believe that our study species (Myers et al. 2000) and is predicted to lose up can contribute to the scientific basis for developing to 21% of its biodiversity over the next century (Brook effective conservation strategies in highly urbanized et al. 2003), it is critical to understand the patterns and tropical landscapes, toward the ultimate goal of rec- processes of species responses to urbanization in this onciling human activities and species conservation. region. The tropical island of Singapore epitomizes the METHODS ecological worst-case scenario for Southeast Asia, hav- ing lost >95% of its original vegetation, first to agri- Study sites culture, and subsequently to urbanization over the past The of is two centuries (Corlett 1991, 1992, Turner et al. 1994). Republic Singapore (103o50' E, 1o20' N) located off the southern of Peninsular As such, Singapore is an ideal case study for urban tip Malaysia (Fig. It has a climate with a mean ecological research in the tropics (Brook et al. 2003). 1). typical equatorial daily maximum of the We chose to study the effects of urbanization on but- temperature 30.6?C throughout year and a mean annual rainfall of 2375 mm and terflies because they are highly sensitive to habitat dis- (Chia Corlett turbance and have commonly been used as an indicator Foong 1991, 1992). Singapore's original veg- etation consisted of lowland rain forest taxon for ecological research (e.g., Ehrlich 1984, Kre- primary tropical and freshwater forest men 1994, Hill 1999, Hamer et al. 2003). Further, but- (82%), mangrove (13%), swamp Corlett Turner et al. A total terflies are among the best-studied taxa both in Sin- (5%; 1991, 1992, 1994). of 39 sites were selected and gapore (e.g., Khew and Neo 1997) and in Southeast study opportunistically classified into four habitat a forest Asia (e.g., D'Abrera 1982, 1985, 1986, Corbet and Pen- types priori (i.e., forest urban for- dlebury 1992). reserves, fragments, parks adjoining and isolated urban to their ef- Our study has three primary objectives. First, we ests, parks) investigate assess the conservation values of forest reserves, forest fectiveness in conserving butterfly diversity in Sin- fragments, and urban parks in Singapore. Forest re- gapore (Table 1, Fig. 1). serves are part of the Central Catchment Nature Re- Butterfly sampling serve of Singapore and they consist primarily of old secondary forest, primary lowland tropical rain forest, To compare butterfly species richness and compo- and freshwater swamp forest (Corlett 1997). Forest sition among the study sites, butterflies were surveyed fragments are patches of ruderal vegetation (e.g., aban- using the transect walk method (Pollard and Yates doned rubber plantation, Adinandra-dominated sec- 1993, Koh et al. 2002). This method is suitable for ondary forest) on degraded soils that are scattered conducting spatial comparisons of species composi- across the island. Urban parks are mostly land that has tion, as it does not suffer from interspecific inequalities been cleared, revegetated, and maintained with both in probability of attraction or capture, characteristic of December2004 URBAN BUTTERFLY CONSERVATION 1697

A Forestreserves (FR) O Forestfragments (FF) Thailand 40 Urbanparks adjoining forests (PF) O Isolatedurban parks (IP)

sia Borneo Su tra

PF5 FF10 F PF74F7

A FF6 PF12 IP30 FF10 FF12 cF6P6Opi FF140 A FR3 FF3 PF3 PF1 FR2 lp O PF10FF130 FF4 PF13O1P50 INP PF8PF2

0 5 10 km IIII I I

FIG. 1. A map showing the location of Singapore and the 39 study sites. See Table 1 for site abbreviations. trapping and netting techniques (Walpole and Sheldon D'Abrera (1982, 1985, 1986), Fleming (1991), Corbet 1999). Although a potential limitation of their method and Pendlebury (1992), and Neo (1996). As with other is a bias toward the more conspicuous species, this similar studies in the tropics (e.g., Hill et al. 1995, source of error should not substantially affect the re- Ghazoul 2002), butterflies from the families of Hes- sults of our study because all butterflies were identified periidae and Lycaenidae were excluded from this study by the same observer (L. P. Koh), such that any bias due to the difficulty of identifying and capturing them was systematically applied to all sites surveyed. Fur- in the field. Each transect took an average of 12.77 ? ther, our transect surveys accounted for 16 out of 19 0.10 min (mean ? 1 SE). Butterfly surveys were stan- species (84%) of butterflies caught using banana-baited dardized by restricting recording to prescribed mini- traps (for details of this method, see DeVries and Walla mum weather conditions (i.e., no rain) and between the 2001) placed in all our study sites in a preliminary periods of 10:00 and 15:00 hours, corresponding to investigation. peak butterfly flight activity (Moore 1975, Pollard et Standard-sized transects (100 m) were randomly al. 1975, Pollard and Yates 1993). marked out on all sites. The number of transects at each Environmental variables site varied from one (e.g., Ang Mo Kio Garden East Park) to 14 (i.e., Nee Soon Forest), depending on the To determine the effects of environmental factors on area of the site (Table 1). Each butterfly transect was butterfly distribution, the following variables were surveyed three times during the course of a year (i.e., measured within circular plots of 5 m radius located at 17 June-23 August 2002; 3 December 2002-21 March the start and end points of each randomly selected but- 2003; 8 May-26 June 2003) to minimize any effect of terfly transect: the canopy cover of the forest, using a seasonal biases in butterfly richness. The surveyor spherical densiometer (Lemmon 1957); the number of walked at a constant pace and paused for a 1-min visual trees <30 cm diameter at breast height (dbh); the num- census at every 10-m interval along each transect. Ev- ber of trees >30 cm dbh; the number of dead trees; ery butterfly spotted within an imaginary box, 5 m to the estimated percentage shrub cover; the mean leaf each side, 5 m ahead, and 5 m above was recorded. litter depth; the estimated percentage flowering shrub Voucher specimens were collected for butterflies that cover; the number of trees with open flowers or fruits; could not be identified in the field, to be identified by the temperature and relative humidity using an Oakton comparisons with museum specimens. Butterflies that digital max/min thermohygrometer (Oakton Instru- were neither identified in the field nor caught were ments, Vernon Hills, Illinois, USA); and the light in- noted and excluded from the sample (1.9% of total tensity using a Topcon IM-2D illumination meter (Top- sample). Butterflies were identified to species using con, Tokyo, Japan). The mean value of each environ- 1698 LIAN PIN KOH AND NAVJOT S. SODHI EcologicalApplications Vol. 14, No. 6

TABLE 1. Summary information of 39 study sites in Singapore.

Butterflies recorded Site (abbreviation), No. transects by habitat type Area (ha) surveyed No. species No. individuals Forest reserves Lornie Forest (FR1) 54.26 4 8 25 Bukit Timah Nature Reserve (FR2) 245.29 11 22 207 MacRitchie Forest (FR3) 462.85 12 24 227 Nee Soon Forest (FR4) 1147.11 14 27 308 Forest fragments Woodlands East Forest (FF1) 2.46 1 2 3 Clementi Woods Forest (FF2) 3.32 1 1 3 Reservoir Forest (FF3) 4.22 1 3 3 Singapore Botanic Gardens Forest (FF4) 5.61 1 5 22 Labrador Forest (FF5) 8.55 2 8 48 Ang Mo Kio Garden West Forest (FF6) 10.72 1 5 8 Forest (FF7) 17.48 1 7 11 Telok Blangah Forest (FF8) 30.72 1 3 6 Kent Ridge Forest (FF9) 32.53 2 4 19 Town Forest (FF10) 34.34 2 2 9 Forest (FF 11) 35.28 2 7 26 Bukit Batok Nature Forest (FF12) 41.31 2 7 12 Holland Woods (FF13) 65.93 4 12 77 Bukit Batok West Woods (FF14) 72.85 3 6 10 Urban parks adjoining forests (PF1) 0.95 1 12 25 Telok Blangah Park (PF2) 1.78 1 13 92 (PF3) 2.42 1 7 11 Labrador Park (PF4) 4.35 1 8 44 Woodlands East Park (PF5) 4.89 1 6 20 Lower Peirce Reservoir Park (PF6) 5.03 1 11 19 (PF7) 5.09 1 4 9 (PF8) 7.70 2 8 58 Mount Faber Park (PF9) 8.33 2 14 89 Clementi Park (PFIO) 8.53 2 11 32 MacRitchie Reservoir Park (PFI 1) 11.84 2 8 25 Ang Mo Kio Garden West Park (PF12) 22.95 1 9 29 Singapore Botanic Gardens Park (PF13) 44.16 3 16 62 Park (PFl4) 52.85 2 3 9 Isolated urban parks Ang Mo Kio Garden East Park (IPI) 6.68 1 9 26 Sun Plaza Park (IP2) 12.24 2 8 24 Pungol Park (IP3) 13.07 2 8 62 (IP4) 14.88 2 12 65 Bishan Park II (IP5) 20.28 2 8 50 Bishan Park I (IP6) 22.81 2 14 61 (IP7) 29.49 4 9 62 mental variable for each transect was calculated for "monophagous" represents species with five or less subsequent analyses. genera of known host plants; "oligophagous" repre- sents species with 6-10 genera of known host plants; Ecological traits of species and "polyphagous" represents species with >10 gen- To compare ecological traits of butterfly species, the era of known host plants, using Robinson et al. (2001). following traits were selected based on their biological Adult body size.-Butterflies were classified into one relevance as shown in previous studies (e.g., McKinney of three categories according to their wingspan, where 1997, Purvis et al. 2000). "small" represents a wingspan of 20-30 mm; "me- Adult habitat specialization.-This was determined dium" represents a wingspan of 31 to 40 mm; and from Corbet and Pendlebury (1992), where "forest- "large" represents a wingspan of >40 mm, using dependent" represents species that have never been Fleming (1991). encountered outside a primary lowland forest in Pen- Sexual dichromatism.-This was estimated by "eye- insular Malaysia; and "nonforest-dependent" repre- balling" the percentage difference in pattern and color sents species that have been encountered in other hab- of the upperside of both pairs of wings between the itats (e.g., secondary forest). sexes of each species, where "monochromatic" rep- Larval host plant specificity.-Butterfly species were resents difference of <30%; "moderate" represents classified into one of three relative categories, where difference of 30-70%; and "dichromatic" represents December2004 URBAN BUTTERFLYCONSERVATION 1699

difference of >70%, using D'Abrera (1982, 1985, ban parks provided by NParks, and using Robinson et 1986), Fleming (1991), and Corbet and Pendlebury al. (2001). (1992). Statistical Adult conspicuousness.-This was estimated by eye- analyses balling the estimated proportion of black, brown, or We constructed sample-based rarefaction curves gray on the upperside of both pairs of wings, where (i.e., equivalent to smoothed accumulation curves) re- "inconspicuous" describes species with >70% black, scaled to the number of individuals to compare the brown, or gray; "moderate" describes species with 30- number of species and number of unique species (i.e., 70% black, brown, or gray; and "conspicuous" de- species not found in other habitats) among the different scribes species with <30% black, brown, or gray, using habitat types; and rescaled to the number of samples D'Abrera (1982, 1985, 1986), Fleming (1991), and (i.e., transects) to compare the density of species among Corbet and Pendlebury (1992). For adult conspicuous- habitats. Curves of the number of butterfly individuals ness and sexual dichromatism, all butterfly species against the number of samples were also constructed were scored by the same observer (L. P Koh) to min- to compare butterfly density among the habitats. The imize estimation bias. use of taxon sampling curves is a theoretically robust Geographical distribution.--This was based on the method for species diversity comparisons as it accounts maximum natural geographical range of each species for differences in sampling effort among study sites recorded in D'Abrera (1982, 1985, 1986), Fleming (Denslow 1995, Gotelli and Colwell 2001). Further, (1991), and Corbet and Pendlebury (1992), where when the data is inherently sample-based (e.g., from "Sundaland" represents species restricted to Sunda- random transects), the use of sample-based rarefaction land; "oriental" represents species restricted to the ori- can account for natural levels of sample heterogeneity ental region (Indomalayan); and "cosmopolitan" rep- (i.e., patchiness) in the data (Gotelli and Colwell 2001). resents species distributed across the oriental region All data sets were rarefied using EstimateS version and beyond. 6.0b.2 To compare community evenness among habi- tats, we fitted a linear regression model (i.e., geometric- Urban park variables series model) to a plot of logarithmic species abun- To identify the important determinants of butterfly dance against arithmetic species rank order for each species richness in urban parks, the following variables habitat type (Bazzaz 1975, Tokeshi 1993), using Min- were calculated. itab version 13.2 (Minitab 2000). The slopes of re- Site area.-Areas of urban parks were obtained us- gression indicate the relative butterfly community ing the Geographical Information System (GIS) data- evenness of the different habitats, whereby a regression base of the , Singapore (NParks). slope of zero represents a community where all species Site edge/area ratio.-Edge/area index of each park have equal abundance. Tokeshi (1993) recommended was calculated using Patton's (1975) formula, the geometric-series model as a simple, uniform frame- work to examine the evenness aspects of different com- e.dge/area = C/1/2 x flAII munities. All rarefaction and regression curves were where C and A are the circumference and area of the plotted using Sigmaplot version 8.02 (SPSS 2002). park, respectively, and were obtained from the GIS We adopted a two-step procedure to determine how database. different butterfly species respond to multiple environ- Site isolation.-Two measures of relative isolation mental variables (e.g., canopy cover). First, we per- from potential sources of butterfly populations in the formed indirect gradient analysis with nonmetric mul- vicinity were calculated for each park, using the GIS tidimensional scaling (NMS) to ordinate sample units database and 1:50 000 topographic maps of Singapore. (i.e., transects) in species space. Indirect gradient anal- Site isolation from forests was determined by measur- ysis reflects the environment the way the biotic com- ing the total area of forests (i.e., reserves and frag- munity interprets it, as opposed to other methods (e.g., ments) located no farther than 1, 2, or 3 km away; and canonical correspondence analysis; ter Braak 1986) site isolation from other parks was determined by mea- where ordinations are constrained by variables deter- suring the total area of parks located no farther than mined a priori, and are therefore subjected to the hy- 1, 2, or 3 km away. pothesized biological relevance of their selection Environmental variables.-Mean values of environ- (Beals 1984). NMS is a computational-intensive iter- mental variables (described previously) were calculat- ative optimization method that searches for the best ed by averaging their measured values over all transects positions of n entities (samples) on k dimensions (axes) in each park. The number of dead trees and mean leaf that minimizes the departure from monotonicity in the litter depth were excluded from this analysis because relationship between the original dissimilarity data of dead trees and leaf litter were not observed in urban the n samples and the reduced k-dimensional ordination parks. space of these samples (McCune and Grace 2002). The Number of potential larval host plant species.-This number was determined from plant species lists of ur- 2 (http://viceroy.eeb.uconn.edu/estimates) 1700 LIAN PIN KOH AND NAVJOT S. SODHI EcologicalApplications Vol. 14, No. 6

TABLE2. Summary information of butterfly species recorded, including their conservation status, community classification, and ecological traits.

Dichro- Conspicu- Distribu- Species Abbrev. Status Community HabSp HostSp Size matism ous tion Amathusia (Spl) rare urban NFD oligopha- large monochro- conspic- oriental phidippus avoiders gous matic uous Appias (Sp2) common urban NFD oligopha- small dichro- conspic- oriental libythea adapters gous matic uous Catopsilia (Sp3) common urban NFD oligopha- medium monochro- monspic- cosmo- pomona adapters gous matic uous politan Catopsilia (Sp4) rare urban NFD oligopha- medium monochro- conspic- oriental pyranthe avoiders gous matic uous Catopsilia (Sp5) common urban NFD monopha- small monochro- conspic- cosmo- scylla adapters gous matic uous politan Cethosia (Sp6) common urban NFD monopha- large monochro- conspic- oriental hypsia avoiders gous matic uous Chilasa (Sp7) rare urban NFD oligopha- large monochro- moderate cosmo- clytia adapters gous matic politan Cirrochroa (Sp8) rare urban NFD NA small monochro- conspic- oriental orissa adapters matic uous Cupha (Sp9) rare urban NFD oligopha- small monochro- conspic- oriental erymanthis adapters gous matic uous Danaus (Spl0) rare urban NFD polypha- medium monochro- conspic- cosmo- chrysippus adapters gous matic uous politan Danaus (Sp 11) rare urban NFD oligopha- large monochro- conspic- oriental melanippus adapters gous matic uous Delias (Sp12) common urban NFD polypha- medium monochro- conspic- oriental hyparete adapters gous matic uous Doleschallia (Sp13) NA urban FD oligopha- medium monochro- conspic- cosmo- bisaltide avoiders gous matic uous politan Elyminas (Sp14) common urban NFD polypha- medium monochro- inconspic- oriental hypermnestra adapters gous matic uous Elymnias (Sp15) rare urban FD monopha- medium monochro- inconspic- Sunda- panthera avoiders gous matic uous land Eulaceura (Spl6) common urban FD monopha- small dichro- moderate oriental osteria avoiders gous matic Euploea (Spl7) common urban NFD oligopha- large moderate moderate oriental mulciber adapters gous Euploea (Sp1 8) common urban NFD monopha- large monochro- moderate cosmo- phaenareta adapters- gous matic politan Euploea (Spl9) common urban NFD monopha- medium monochro- moderate oriental radamanthus avoiders gous matic Eurema (Sp20) common urban NFD polypha- small monochro- conspic- cosmo- hecabe adapters gous matic uous politan Euthalia (Sp21) rare urban NFD oligopha- medium monochro- inconspic- oriental aconthea adapters gous matic uous Euthalia (Sp22) common urban FD oligopha- medium dichro- moderate oriental monina avoiders gous matic Faunis (Sp23) common urban NFD oligopha- small monochro- inconspic- oriental canens avoiders gous matic uous Graphium (Sp24) common urban NFD polypha- large monochro- moderate cosmo- agamemnon avoiders gous matic politan Graphium (Sp25) common urban FD NA medium monochro- moderate oriental evemon avoiders matic Graphium (Sp26) common urban NFD polypha- medium monochro- moderate cosmo- sarpedon adapters gous matic politan Hypolimnas (Sp27) common urban FD monopha- medium monochro- inconspic- cosmo- anomala adapters gous matic uous politan Hypolimnas (Sp28) common urban NFD polypha- large moderate moderate cosmo- bolina adapters gous politan Idea stolli (Sp29) common urban FD monopha- large monochro- conspic- oriental avoiders gous matic uous Ideopsis (Sp30) common urban NFD monopha- medium monochro- moderate oriental vulgaris adapters gous matic Junonia (Sp31) rare urban NFD polypha- small monochro- conspic- oriental almana adapters gous matic uous Junonia (Sp32) common urban NFD monopha- small monochro- inconspic- cosmo- hedonia adapters gous matic uous politan Junonia (Sp33) common urban NFD polypha- small dichro- conspic- cosmo- orithya adapters gous matic uous politan Lasippa (Sp34) common urban NFD monopha- small monochro- conspic- oriental tiga avoiders gous matic uous December 2004 URBAN BUTTERFLY CONSERVATION 1701

TABLE 2. (Continued)

Dichro- Conspicu- Distribu- Species Abbrev. Status Community HabSp HostSp Size matism ous tion Lebadea (Sp35) rare urban NFD NA small monochro- moderate oriental martha avoiders matic Leptosia (Sp36) rare urban NFD monopha- small monochro- conspic- cosmo- nina adapters gous matic uous politan Lexias (Sp37) common urban NFD NA large dichro- moderate oriental pardalis avoiders matic Moduza (Sp38) common urban NFD polypha- medium monochro- moderate oriental procris avoiders gous matic Mycalesis sp. (Sp39) common NA NA NA NA NA NA NA Papilio (Sp40) common urban NFD polypha- large monochro- conspic- cosmo- demoleus adapters gous matic uous politan Papilio (Sp41) rare urban NFD monopha- large monochro- moderate oriental demolion avoiders gous matic Papilio (Sp42) rare urban FD NA large monochro- moderate oriental iswara avoiders matic Papilio (Sp43) common urban NFD polypha- large moderate moderate oriental memnon avoiders gous Papilio (Sp44) common urban NFD polypha- large dichro- moderate cosmo- polytes adapters gous matic politan Parantica (Sp45) common urban NFD monopha- medium monochro- moderate oriental agleoides adapters gous matic Pathysa (Sp46) rare urban FD oligopha- large monochro- conspic- oriental antiphates avoiders gous matic uous Phaedyma (Sp47) common urban NFD oligopha- medium monochro- moderate oriental columella adapters gous matic Phalanta (Sp48) common urban NFD polypha- small monochro- conspic- oriental phalantha adapters gous matic uous Polyura (Sp49) rare urban NA oligopha- medium monochro- conspic- oriental hebe adapters gous matic uous Tanaecia (Sp50) common urban FD monopha- medium dichro- moderate oriental iapis avoiders gous matic Tanaecia (Sp51) common urban FD monopha- medium monochro- moderate Sunda- pelea avoiders gous matic land Thaumantis (Sp52) rare urban FD NA large moderate conspic- Sunda- klugius avoiders uous land Troides (Sp53) rare urban NFD monopha- large monochro- moderate oriental helena adapters gous matic Vindula (Sp54) common urban NFD monopha- large dichro- conspic- cosmo- dejone avoiders gous matic uous politan Ypthima sp. (Sp55) common NA NA NA NA NA NA NA Zeuxidia (Sp56) common urban NFD monopha- large dichro- moderate oriental amethystus avoiders gous matic Notes: HabSp, HostSp, Size, Dichromatism, Conspicuous, and Distribution represent adult habitat specialization, larval host plant specificity, adult body size, sexual dichromatism, adult conspicuousness, and geographical distribution, respectively. For HabSp, FD and NFD represent forest dependent and nonforest-dependent species, respectively. NA represents a missing value. Refer to the Methods for definitions of classes of conservation status, community classification, and ecological traits.

NMS procedure was performed using the "autopilot eraging of the abundances of each species in each sam- (slow and thorough)" mode in PC-ORD version 4.14 ple unit) was also plotted. (McCune and Mefford 1999) with Sorensen distance We used univariate classification trees (UCT) to as a dissimilarity measure. Next, the NMS ordination compare the ecological traits of butterflies in different is related to the measured environmental variables habitats (Table 2). We defined butterfly community as (e.g., canopy cover) by correlating each variable to the a response variable for the UCT model, whereby but- two axes of the final optimum two-dimensional ordi- terfly species with >80% of individuals occurring in nation space. The Spearman correlation coefficient in- forest habitats (i.e., forest reserves or fragments) were dicates how well each variable explains the position of classified as urban avoiders, while the remaining spe- samples along that ordination axis. For interpretability, cies were classified as urban adapters. UCT repetitively the significantly correlated (i.e., R > 0.50) environ- partitions the data set into mutually exclusive groups, mental variables were plotted as vectors on a joint plot each of which is as homogeneous as possible, to pro- to show their relationships with the sample scores (see vide a final tree-like classification and an associated McCune and Grace 2002), using Sigmaplot version dichotomous key that can be used to classify unknown 8.02 (SPSS 2002). A joint plot of environmental var- samples into the groups (De'ath and Fabricius 2000, iables and species scores (calculated by weighted av- De'ath 2002, McCune and Grace 2002). As a non- 1702 LIAN PIN KOH AND NAVJOT S. SODHI Ecological Applications Vol. 14, No. 6

50 50 a c FR FR 40 u) 40 a)

30, F 0 30PF c 5 FF cn20 IP 20 IP FF

0

10 i z 10

010 0

16 1000 b d 14 FR d FR 800 12

S )1010-cu 600 PF

u) . 400 6 6 Z 4 PF z

FF 2FF 0 200 400 600 800 1000 0 10 20 30 40 50 No. individuals No. transects FIG.2. Sample-basedrarefaction curves of the butterflycommunities of forest reserves (FR), forest fragments(FF), urban parks adjoining forests (PF), and isolated urbanparks (IP). Curves were rescaled to the numberof individuals to compare (a) the numberof species and (b) the number of unique species among habitats. Curves were rescaled to the number of transects to compare (c) the density of species among habitats. Curves of the number of individuals versus the numberof transects were plotted to compare (d) butterfly density among habitats. Error bars representstandard errors of the mean numberof species or individuals resampledfor 1000 simulated runs of an increasing sample size.

parametric method, UCT is robust to many of the prob- among the predictors. A subset of predictors was then lems that plague parametric models (e.g., assumptions selected based on their independence (Pearson R < of data normality and homoscedasticity; McCune and 0.70) from other predictors, as well as their biological Grace 2002). Further, UCT is amenable to analyzing relevance. These predictors were entered for STEP- complex ecological data that have missing values or WISE multiple regressions, using P < 0.05 as the cri- nonlinear relationships between variables (De'ath terion to add or remove variables. All variables were 2002). The response (i.e., butterfly community) and suitably transformed prior to regression analyses to sat- explanatory (i.e., adult habitat specialization, larval isfy the assumptions of parametric analyses (Zar 1999). host plant specificity, adult body size, sexual dichro- All regression analyses were performed using Minitab matism, adult conspicuousness, geographical distri- version 13.2 (Minitab 2000). bution) variables were used to grow an overlarge tree, RESULTS which is subsequently "pruned" to optimize its clas- sification efficacy by a 10-fold cross validation process, A total of 56 species and 1898 individuals of but- repeated 500 times using the tree function in S-Plus terflies were recorded. The number of butterfly species version 6.1 (Insightful 2002; Rejwan et al. 1999, at the study sites ranged from one (i.e., Clementi Woods McCune and Grace 2002, Venables and Ripley 2002, Forest) to 27 (i.e., Nee Soon Forest), while the number Steele et al. 2003). of butterfly individuals ranged from three (e.g., Wood- To identify the important determinants of butterfly lands East Forest) to 308 (i.e., Nee Soon Forest; Table species richness in urban parks, simple linear regres- 1). sions were first performed between each predictor var- and habitats iable (e.g., area) and the total number of butterfly spe- Species diversity composition among cies recorded in each park. Next, to satisfy the as- Forest reserves (FR) had the highest number of spe- sumption of multiple regression analysis that predictor cies (Fig. 2a), number of unique species (Fig. 2b), den- variables are not strongly intercorrelated, a simple cor- sity of species (Fig. 2c), and community evenness (Fig. relation matrix was used to check for collinearity 3) among the habitats. Urban parks adjoining forests December 2004 URBAN BUTTERFLY CONSERVATION 1703

3.0 to environmental * Forest reserves (FR) Species responses factors SParks adjoiningforests (PF) 0 2.5- Isolated parks (IP) The distance between two transects in the ordination A Forest fragments (FF) scores reflects - of sample the relative dissimilarity in 2.0 -C- their species compositions (Fig. 5a). A graphical over- CU 1.5 lay of habitat type on this ordination clearly distin- U) a) guished forest (i.e., FR and FF) from park transects (i.e., PF and IP), reflecting their species compositional differences. The ordination of species scores identifies o00.5 0 the butterfly species separating forest from park tran- sects For forest 0.0 - (Fig. 5b). example, transects were IP closely associated with Zeuxidia amethystus (Sp56), (-0.106)R FF (-0.095) PF (-0.068) FR(-0.049) Thaumantis klugius (Sp52), Amathusia phidippus 0 10 20 30 40 50 and Faunis canens while transects rankorder (Spl), (Sp23), park Species were associated with Catopsilia pomona (Sp3), Papilio FIG.3. Fittedlinear regression models of logarithmicspe- demoleus (Sp40), Appias libythea (Sp2), and Junonia cies abundanceagainst arithmeticspecies rank order.Num- hedonia (Sp32). Vector plots of environmental vari- bers in are of rel- parentheses slopes regression, indicating ables (e.g., canopy cover) showed that transects ative communityevenness of the differenthabitats (Tokeshi (e.g., 1993). in forests) and species (e.g., Zeuxidia amethystus) in the upper right quadrant of the ordinations were pos- itively associated with canopy cover, mean leaf litter depth, number of dead trees, and number of trees <30 (PF) had a higher number of species (Fig. 2a), density cm dbh, while transects (e.g., in urban parks) and spe- of species (Fig. 2c), butterfly density (Fig. 2d), and cies (e.g., Catopsilia pomona) in the lower left quadrant community evenness (Fig. 3) than forest fragments of the ordinations were positively associated with light (FF) and isolated urban parks (IP). Generally, FF and intensity (Fig. 5). IP had lower number of species (Fig. 2a), number of traits unique species (Fig. 2b), density of species (Fig. 2c), Ecological of butterfly species butterfly density (Fig. 2d), and community evenness The final and optimum tree from our UCT analysis (Fig. 3) than either FR or PE Different habitats also has four terminal nodes and a misclassification rate of differed in the composition and relative abundance of 0.23 (Fig. 6). The variables actually used in tree con- their species (Fig. 4). For example, while Catopsilia struction were adult habitat specialization, geographi- pomona (Sp3) was the dominant species in PF (30.9%) cal distribution, and larval host plant specificity, in and IP (48.9%), it was a relatively uncommon species decreasing order of importance in explaining variation in FR (0.002%) and FF (0.08%). Further, there were in the data. Our findings indicate that urban avoiders more "rare" species (i.e., 20 or less individuals ob- were 89% likely to be forest dependent and 63% likely served per year; Khew and Neo 1997) recorded from to be monophagous, whereas urban adapters were 87% FR (13) and PF (11) than FF (three) and IP (three) likely to be cosmopolitan and 67% likely to be oligo- (Table 2, Fig. 4). or polyphagous.

50 o, 45- ) 40-

-(l 35-

=33 -V, m 25- i .: " ''i - FR S20- PF CL 15- aiiiliu,10 FF

u 5-

O Cm Tr- r~--(m'moC NIa Lo o o-q-to L toIT M c(0oT-- r- . cTN WW-M c LO LWMC) LOe)O QC) eC\) C-N-.P 00 1,-Or- - L C14CY)0 00 C)M " Q C "T V"v- " T- C) N C%4CL 04 U T - - N aCO 0- C Tt r a 0- v- C%4 MO " i O LO LtO )- M"3U) 3 't Ce 000000aM- a a, 0000000a a a a a .00 0003 000.0 , 0000a a ", 0)U) CO'IT'-O V'-T-00000C 0)00000" ?O O "-a0-0 a-0)- C-04IT a a o nu 0.U)O -- 00a0 a 00a a U)U)U)U)U(1) U) U) U) U)UU0U)(n U)(1 U)(1) U) U)a) U) U)UU) U) (nUco ()U UU) U) U)CO U))O U)Uo)nU)-•e. U) f) U) 1) U) U) U) FIG.4. A bar graph showing relative abundanceof species and species deletions among differenthabitats. FR, FF, PF, and IP represent forest reserves, forest fragments, urban parks adjoining forests, and isolated urban parks, respectively. See Table 2 for species abbreviations. 1704 LIAN PIN KOH AND NAVJOT S. SODHI Ecological Applications Vol. 14, No. 6

Variable RAxis1 RAxis 2 a Canopy 0.63 0.46 Litter 0.64 0.57 Dead tree 0.54 0.44 Tree <30 0.67 0.63 o Tree <30 Lightintensity -0.66 -0.43 o Litter o o0 0 0000Litter Soo o o Canopy o ea A AA AO 00 A

A 0A Axis1 Lightintensity A A A A

A AoFR A FF 04? PF - IP ?

b

43 410 16 1937 56,1 Tree <30 25 0 22 0 o0. 052 Litter ?27529 26 o 420o38 270 36 6 47 7 4 0 ne0o 1R 440 13 Axis 1 21 (3 3 Lightintensity k- 01 2 45 2 310 02 28,30 5 11 48 7o 33 18

10

C4 .•_

FIG. 5. NMS ordination joint plot of (a) sample scores and (b) species scores with the significantly correlated (R > 0.50) environmentalvariables, numberof trees <30 cm dbh ("trees < 30"); mean leaf litter depth ("litter"); numberof dead trees ("dead tree"); canopy cover ("canopy"); and light intensity. FR, FF, PF, and IP representforest reserves, forest fragments, urbanparks adjoining forests, and isolated urbanparks, respectively. See Table 2 for species abbreviations.

Species richness in urban parks isolation from forest (within 2 km) are the important factors richness in urban Simple linear regressions revealed that the number affecting butterfly species of potential larval host plant species and total forest parks. area within 2 km were significant predictors (P < 0.05) DISCUSSION of butterfly species richness in urban parks (Table 3). and habitats A total of 14 predictors were retained for multiple re- Species diversity composition among gression analysis (Table 3). The final multiple regres- The relatively larger areas of forest reserves (Table sion model included only the number of potential larval 1) may enhance their butterfly diversities either by sus- host plant species (HP) and log10 total forest area within taining larger and more viable populations of species 2 km (for2km) as statistically significant (P < 0.05) with lower risks of extinction, or by having greater and biologically relevant predictors, with the regres- diversities of microhabitats with myriad ecological sion equation: no. butterfly species = 4.20 + 0.14 HP niches that can support more species (MacArthur and + 1.33 for2km (n = 14; R2 = 65.17). This final model Wilson 1963, 1967, Simberloff 1974, Laurance et al. implies that the number of larval host plant species and 2002). Further, the primary and old secondary vege- December2004 URBAN BUTTERFLY CONSERVATION 1705

HabSp:FD NFD I HabSp: butterfly diversities among the habitats in Singapore, possibly due to their small areas or impoverished floras (Tables 1 and 3).

Species responses to environmental factors Previous studies showed that changes in canopy cov- er Distribution:cosmopolitan Distribution:oriental and light intensity can affect the diversity and com- Urbanavoiders position of tropical forest butterflies (e.g., Hill et al. 0.89 HostSp: Oligophagous, 2001, Laurance et al. 2002, Hamer et al. 2003), either (9) HostSp:monophagous polyphagous, directly through microclimatic effects (e.g., moisture Urbanadapters availability) on survival, or indirectly by af- 0.87 butterfly (15) Urbanavoiders Urbanadapters fecting the productivity and quality of larval host plants 0.63 0.67 Basset et al. The other environ- (8) (15) (Blau 1980, 2001). Misclassificationrate = 0.23 mental variables (e.g., leaf litter depth), which corre- lated with sample scores in our analyses (Fig. 5), are FIG. 6. Univariateclassification tree. Numbersat terminal indicative of the level of habitat disturbance nodes representthe probabilityof correctly classifying but- generally terfly species, and numbersin parenthesesrepresent sample in tropical forests (Liow et al. 2001), but are less likely size. See Table 2 for ecological trait abbreviations. to have direct impacts on butterfly communities. Ecological traits of butterfly species tation in forest reserves may provide the unique mi- The goodness of fit for a classification tree may be croclimatic conditions (e.g., closed canopy) or specific assessed by comparing its misclassification rate to that larval host plants (e.g., Gironniera subaequalis) vital of a null model where all species are classified into the to the persistence of specialist butterfly species (e.g., majority group (i.e., urban adapters for our data; De'ath Eulaceura osteria; Khew and Neo 1997). Urban parks and Fabricius 2000). The misclassification rate for our adjoining forests were the second most diverse habitat tree (23%) is considerably lower than that for the null in our study. The prevalence of cultivated flowering model (43%), indicating that it is more reliable to clas- plants in these urban parks (e.g., Cassia fistula, Bau- sify an unknown species using our tree than it is to hinia blakeana) likely support resident butterfly species "guess with the majority" (De'ath and Fabricius 2000). that are adapted to an open canopy (e.g., Catopsilia Our classification tree shows that adult habitat spe- pomona), as well as species from adjacent forests that cialization and larval host plant specificity were im- forage in these parks (e.g., Euploea phaenareta). Forest portant ecological traits defining butterfly species as fragments and isolated urban parks recorded the lowest either an urban avoider or adapter (Fig. 6). This is

TABLE3. Simple linear regressions of total numberof butterfly species on urbanpark vari- ables.

Predictor Coefficient P R2 Logl0 (site area)t -0.215 0.900 0 Edge/areaindext 0.990 0.878 0 Log10(total forest area) Within 1 km 1.066 0.166 0.10 Within 2 kmt 1.237 0.049 0.19 Within 3 km 1.091 0.054 0.18 Log10(total park area) Within 1 kmt 1.014 0.301 0.06 Within 2 km 0.385 0.701 0.01 Within 3 kmt 0.825 0.406 0.04 Mean canopy covert 0.026 0.549 0.02 Mean shrub covert 0 0.903 0 Mean (no. living trees) <30 cm dbht -0.336 0.720 0.01 Logl0 (mean no. living trees) >30 cm dbht 0.262 0.959 0 Log1o(mean flowering shrub cover)t 0.508 0.818 0.30 Logl0 (mean no. trees with open flowers or fruits)t 4.200 0.326 5.10 Mean temperature 0.245 0.682 0.90 Mean relative humidityt 0.071 0.456 3.00 Logl0 (mean light intensity)t -3.408 0.550 1.90 No. potential larval host plant speciest 0.171 0.003 0.53 Note: Twenty-one samples were analyzed in the regression for all predictors, except "no. of potential larval host plant species" (n = 14). t Predictorretained for multiple regression analysis. 1706 LIAN PIN KOH AND NAVJOT S. SODHI EcologicalApplications Vol. 14, No. 6 consistent with the theory that specialist species with greatly enhance butterfly species richness, perhaps due narrow ecological niches (e.g., forest dependent, mo- to the availability of varied microclimatic conditions nophagous) are less resilient to environmental pertur- and ecological niches in these two different habitat bations, such as those associated with habitat destruc- types. Third, larval resource availability and isolation tion and fragmentation, than more generalist species from forests were important determinants of butterfly (McKinney 1997, Purvis et al. 2000). Further, our re- richness in urban parks, indicating that urban parks sults also indicate that cosmopolitan butterfly species should be revegetated with a diversity of potential lar- are more likely to persist in urban parks than those val host plants and should be situated as near as pos- restricted to the oriental region (Fig. 6), possibly be- sible to a forest, in order to maximize their conservation cause widely distributed species are inherently more value. Fourth, the differential response of butterfly spe- adaptable and better able to exploit a wider range of cies to environmental factors (e.g., canopy cover) high- ecological niches than narrowly distributed species lights the importance of maintaining environmental (Jones et al. 2001, Harcourt et al. 2002). Our findings heterogeneity for the effective conservation of different indicate that distinct ecological traits of butterfly spe- butterfly species (e.g., specialist species that are de- cies may affect their distribution across different hab- pendent on a closed canopy forest; Hamer et al. 2003). itats. Urban landscapes represent an extreme on the con- tinuum of desirable environmental conditions, one that richness in urban Species parks conservation biologists try to avoid in ecosystem man- Koh et al. (2004) recently reported the coextinctions agement (Hunter 1996, McIntyre and Hobbs 1999, of butterflies and their host plants from Singapore, and Miller and Hobbs 2002). However, bearing the unlikely predicted that the number of extinct butterflies would scenario that urban sprawl will come to a halt, we are increase exponentially with that of extinct host plants. increasingly faced with the task of conserving species Our results suggest that the reverse is also true, where- in largely "unnatural" human-dominated environ- by the number of butterfly species persisting in urban ments. Hence, it is critical that more research be fo- parks is dependent on the number of potential host cused on developing viable strategies for the effective plants occurring in them. The proximate mechanism conservation of biodiversity in urban landscapes. is underlying butterfly-host plant specificity primarily ACKNOWLEDGMENTS chemical in nature (Ehrlich and Raven 1964), although We are to H. C. Lim and T. M. Lee for discussions. their and bases are less well grateful evolutionary ecological We thank Claire Kremen and an anonymous reviewer for understood (see Gilbert 1984). The isolation of parks comments. We also thank the National Parks Board, Singa- from potential sources of butterfly populations in for- pore for the use of the GIS data and for providing species lists of of urban This was ests within 2 km is also important in determining but- plants parks. study supportedby the National University of Singapore (R-154-000-144-112). terfly species richness in urban parks. This is consistent with our earlier findings indicating that urban parks LITERATURE CITED adjoining forests have higher butterfly species richness Achard,E, H. D. Eva, H.-J. Stibig, P. Mayaux,J. Gallego, T. than those that are isolated (Figs. 2 and 3). Richards, and J. P. Malingreau. 2002. Determinationof deforestation rates of the world's humid tropical forests. Conservation implications Science 297:999-1002. Basset, Y., E. Charles, D. S. Hammond,and V. K. Brown. Several implications for the conservation of tropical 2001. Short-termeffects of canopy openness on insect her- butterflies may be drawn from our study. First, forest bivores in a rain forest in Guyana.Journal of Applied Ecol- reserves recorded the highest butterfly species richness ogy 38:1045-1058. Bazzaz, E A. 1975. Plant in old-field suc- among the habitats in Singapore (Figs. 2-4), implying species diversity cessional ecosystems in southernIllinois. Ecology 56:485- that in highly urbanized tropical landscapes, the least 488. human-disturbed habitats are the most valuable for pre- Beals, E. W. 1984. Bray-Curtisordination: an effective strat- serving the native butterfly species richness, and should egy for analysis of multivariateecological data. Advances therefore be the conservation in Ecological Research 14:1-55. given highest priority. R. and A. E. Launer.1997. and forest reserves and urban for- Blair, B., Butterflydiversity Second, parks adjoining human land use: species assemblages along an urbangra- ests collectively accounted for 91% of all butterfly spe- dient. Biological Conservation80:113-125. cies recorded in this study (Fig. 4), suggesting that the Blau, W. S. 1980. The effect of environmentaldisturbance preservation of these two habitats will likely achieve on a tropical butterflypopulation. Ecology 61:1005-1012. Brook, B. W., N. S. Sodhi, andP. K. L. Ng. 2003. maximum for butterfly conservation, Catastrophic complementarity extinctions follow deforestationin Singapore.Nature 424: especially when it is impractical to preserve large tracts 420-423. of natural habitats. Also, since 12 of the 14 nonisolated Chia, L. S., and S. E Foong. 1991. Climate and weather. urban parks were located next to forest fragments, and Pages 13-49 in L. S. Chia, A. Rahman,and D. B. H. Tay, contained similar richness and composition of butterfly editors. The biophysical environmentof Singapore. Sin- gapore University Press, Singapore. species as the two nonisolated parks located next to Corbet, A. S., and H. M. Pendlebury. 1992. The butterflies forest reserves, it suggests that the preservation of a of the Malay Peninsula. Fourth edition. Malayan Nature combination of urban parks and forest fragments could Society, Kuala Lumpur,Malaysia. December 2004 URBAN BUTTERFLY CONSERVATION 1707

Corlett, R. T. 1991. Vegetation. Pages 134-154 in L. S. Chia, terflies on Buru, Indonesia. Journal of Applied Ecology 32: A. Rahman, and D. B. H. Tay, editors. The biophysical 754-760. environment of Singapore. Singapore University Press, Hill, J. K., K. C. Hamer, J. Tangah, and M. Dawood. 2001. Singapore. Ecology of tropical butterflies in rainforest gaps. Oecologia Corlett, R. T. 1992. The ecological transformation of Sin- 128:294-302. gapore 1819-1990. Journal of Biogeography 19:411-420. Hunter, M. L., Jr. 1996. Benchmarks for managing ecosys- Corlett, R. T. 1997. The vegetation in the Nature Reserves tems: are human activities natural? Conservation Biology of Singapore. Gardens' Bulletin Singapore 49:147-159. 10:659-697. D'Abrera, B. 1982. Butterflies of the Oriental Region. Part Insightful. 2002. S-Plus 6.1 for Windows. Insightful, Seattle, I. Hill House, Melbourne, Australia. Washington, USA. D'Abrera, B. 1985. Butterflies of the Oriental Region. Part Jones, M. J., M. S. Sullivan, S. J. Marsden, and M. D. Linsley. II. Hill House, Melbourne, Australia. 2001. Correlates of extinction risk of birds from two In- D'Abrera, B. 1986. Butterflies of the Oriental Region. Part donesian islands. Biological Journal of the Linnean Society III. Hill House, Melbourne, Australia. 73:65-79. and S. Forbes. 1997. Urban nature conservation. Daily, G. C. 2003. Time to rethink conservation strategy. Kendle, T., Science 300:1508-1509. Chapman and Hall, London, UK. S. and S. S. H. Neo. 1997. Dale, V. H., S. Brown, R. A. Haeuber, N. T. Hobbs, N. Huntly, Khew, K., Butterfly biodiversity in with reference the Catch- R. J. Naiman, W. E. Riebsame, M. G. Turner, and T. J. Singapore particular to Central ment Nature Reserve. Gardens' Bulletin 49:273- Valone. 2000. Ecological and for Singapore principles guidelines 296. managing the use of land. Ecological Applications 10:639- 670. Koh, L. P., N. S. Sodhi, and B. W. Brook. 2004. Co-extinc- tions of butterflies and their De'ath, G. 2002. Multivariate trees: a new tech- tropical hostplants. Biotropica, regression 36:272-274. nique modeling species-environment relationships. for Koh, L. P, N. S. Sodhi, H. T. W. and K. S.-H. Peh. 2002. 83:1105-1117. Tan, Ecology Factors the distribution of vascular and K. E. Fabiicius. 2000. and affecting plants, spring- D•'ath, G., Classification butterflies and birds on small islands. Journal trees: a for the tails, tropical regression powerful yet simple technique of 29:93-108. of data. 81:3178- Biogeography analysis complex ecological Ecology I. 1995. On the role of alien in urban flora 3192. Kowarik, species and vegetation. Pages 85-103 in P Pysek, K. Prach, M. Denslow, J. 1995. Disturbance and in rain diversity tropical Rejmanek, and P. M. Wade, editors. Plant invasions: general forests: the effect. 5:962- density Ecological Applications aspects and special problems. SPB Academic, Amsterdam, 968. The Netherlands. P. J., and T. R. Walla. 2001. and DeVries, Species diversity Kremen, C. 1994. Biological inventory using target taxa: a structure in butter- community neotropical fruit-feeding case study of the butterflies of Madagascar. Ecological Ap- flies. Journal of the Linnean 74:1-15. Biological Society plications 4:407-422. R. 1984. The structure and of Ehrlich, P. dynamics butterfly Laurance, W. E, T. E. Lovejoy, H. L. Vasconcelos, E. M. populations. Pages 25-40 in R. I. Vane-Wright and P. R. Bruna, R. K. Didham, P. C. Stouffer, C. Gascon, R. O. Ackery, editors. The biology of butterflies. Academic Bierregaard, S. G. Laurance, and E. Sampaio. 2002. Eco- Press, London, UK. system decay of Amazonian fragments: a 22-year investi- Ehrlich, P. R., and P. H. Raven. 1964. Butterflies and plants: gation. Conservation Biology 16:605-618. a study in coevolution. Evolution 18:586-608. Lemmon, P. E. 1957. A new instrument for measuring forest Fleming, W. A. 1991. Butterflies of West Malaysia and Sin- overstorey density. Journal of Forestry 55:667-668. gapore. Second edition. Longman Malaysia, Kuala Lumpur, Lim, H. C., and N. S. Sodhi. 2004. Responses of avian guilds Malaysia. to urbanization in a tropical city. Landscape and Urban Ghazoul, J. 2002. Impact of logging on the richness and Planning 66:199-215. diversity of forest butterflies in a tropical dry forest in Liow, L. H., N. S. Sodhi, and T. Elmqvist. 2001. Bee diversity Thailand. Biodiversity and Conservation 11:521-541. along a disturbance gradient in tropical lowland forests of Gilbert, L. E. 1984. The biology of butterfly communities. south-east Asia. Journal of Applied Ecology 38:180-192. Pages 41-54 in R. I. Vane-Wright and P. R. Ackery, editors. MacArthur, R. H., and E. O. Wilson. 1963. An equilibrium The biology of butterflies. Academic Press, London, UK. theory of insular zoogeography. Evolution 17:373-387. Gotelli, N. J., and R. K. Colwell. 2001. Quantifying biodi- MacArthur, R. H., and E. O. Wilson. 1967. The theory of versity: procedures and pitfalls in the measurement and island biogeography. Princeton University Press, Princeton, New USA. comparison of species richness. Ecology Letters 4:379- Jersey, 391. Marzluff, J. M. 2001. Worldwide urbanization and its effects on birds. 19-47 in J. M. R. and Hamer, K. C., J. K. Hill, S. Benedick, N. Mustaffa, T. N. Pages Marzluff, Bowman, R. editors. Avian and conservation in Sherratt, M. and V. K. 2003. of Donnelly, ecology Maryati, Chey. Ecology an butterflies in natural and forests of north- urbanizing world. Kluwer, Norwell, Massachusetts, selectively logged USA. Borneo: the of habitat Jour- ern importance heterogeneity. and J. B. Grace. 2002. of nal of Applied Ecology 40:150-162. McCune, B., Analysis ecological communities. MjM Software, Glenedon Beach, Oregon, Harcourt, A. H., S. A. Coppeto, and S. A. Parks. 2002. Rarity, USA. specialization and extinction in primates. Journal of Bio- McCune, B., and M. J. Mefford. 1999. Multivariate analysis geography 29:445-456. of ecological data. Version 4.14. MjM Software, Glenedon P. and R. L. H. Dennis. Hardy, B., 1999. The impact of urban Beach, Oregon, USA. development on butterflies within a city region. Biodiver- McIntyre, S., and R. Hobbs. 1999. A framework for con- sity and Conservation 8:1261-1279. ceptualizing human effects on landscapes and its relevance Hill, J. K. 1999. Butterfly spatial distribution and habitat to management and research models. Conservation Biology requirements in a tropical forest: impacts of selective log- 13:1282-1292. ging. Journal of Applied Ecology 36:564-572. McKinney, M. L. 1997. Extinction vulnerability and selec- Hill, J. K., K. C. Hamer, L. A. Lace, and W. M. T. Banham. tivity: combining ecological and paleontological views. 1995. Effects of selective logging on tropical forest but- Annual Review of Ecology and Systematics 28:495-516. 1708 LIAN PIN KOH AND NAVJOT S. SODHI Ecological Applications Vol. 14, No. 6

McKinney, M. L. 2002. Urbanization, biodiversity, and con- Sheil, D., and D. E R. P. Burslem. 2003. Disturbing hypoth- servation. BioScience 52:883-890. eses in tropical forests. Trends in Ecology and Evolution Miller, J. R., and R. J. Hobbs. 2002. Conservation where 18:18-26. people live and work. Conservation Biology 16:330-337. Simberloff, D. S. 1974. Equilibrium theory of island bio- Minitab. 2000. Minitab Release 13.2. Minitab, State College, geography and ecology. Annual Review of Ecology and Pennsylvania, USA. Systematics 5:161-182. SPSS. 2002. 2002 for Windows. Version 8.02. Moore, N. W. 1975. Butterfly transects in a linear habitat Sigmaplot USA. 1964-1973. Entomologist's Gazette 26:71-78. SPSS, Chicago, Illinois, C. E. D. and J. G. MacCracken. 2003. N., R. A. Mittermeier, C. G. Mittermeier, G. A. B. da Steele, A., Brodie, Jr., Myers, between abundance of cascade torrent sala- Fonseca, and J. Kent. 2000. for con- Relationships Biodiversity hotspots manders and forest Journal of Wildlife servation Nature 403:853-858. age. Management priorities. 67:447-453. Neo, S. S. H. 1996. A to common butterflies of Sin- guide Stein, B. A., L. Kutner, and J. Adams. 2000. Precious heri- Science Centre, gapore. Singapore Singapore. tage. Oxford University Press, Oxford, UK. Ormerod, S. J., N. D. Barlow, E. J. P. Marshall, and G. Kerby. Teo, D. H. L., H. T. W. Tan, R. T. Corlett, C. M. Wong, and 2002. The of Journal of uptake applied ecology. Applied S. K. Y. Lum. 2003. Continental rain forest fragments in 39:1-7. Ecology Singapore resist invasion by exotic plants. Journal of Bio- habitat Patton, D. R. 1975. A diversity index for quantifying geography 30:305-310. 'edges'. Wildlife Society Bulletin 3:171-173. ter Braak, C. J. E 1986. Canonical correspondence analysis: Pollard, E., D. O. Elias, M. J. Skelton, and J. A. Thomas. a new eigenvector technique for multivariate direct gradient 1975. A method of assessing the abundance of butterflies analysis. Ecology 67:1167-1179. in Monk's Wood national nature reserve in 1973. Ento- Tokeshi, M. 1993. Species abundance patterns and commu- mologist's Gazette 26:79-88. nity structure. Advances in Ecological Research 24:111- Pollard, E., and T. J. Yates. 1993. Monitoring butterflies for 186. ecology and conservation. Conservation biology series. Turner, I. M., H. T. W. Tan, Y. C. Wee, Ali Bin Ibrahim, P. Volume 1. Chapman and Hall, London, UK. T. Chew, and R. T. Corlett. 1994. A study of plant species Purvis, A., K. E. Jones, and G. M. Mace. 2000. Extinction. extinction in Singapore: lessons for the conservation of Conservation 8:705-712. BioEssays 22:1123-1133. tropical biodiversity. Biology United Nations Centre for Human Settlements. 1996. An Rejwan, C., N. C. Collins, L. J. Brunner, B. J. Shuter, and urbanizing world: global report on human settlements 1996. M. S. Ridgeway. 1999. Tree regression analysis on the Oxford University Press, Oxford, UK. nesting habitat of smallmouth bass. Ecology 80:341-348. Venables, W. N., and B. N. Ripley. 2002. Modern applied Robinson, G. S., P. R. I. J. G. W. Beccaloni, Ackery, Kitching, statistics with S. Fourth edition. Springer, New York, New and L. M. Hernandez. 2001. Hostplants of the moth and York, USA. of the Oriental United Selan- butterfly caterpillars Region. Walpole, M. J., and I. R. Sheldon. 1999. Sampling butterflies gor Press, Kuala Lumpur, Malaysia. in tropical rainforest: an evaluation of a transect walk meth- Rosenzweig, M. L. 2003. Win-win ecology: how the Earth's od. Biological Conservation 87:85-91. species can survive in the midst of human enterprise. Ox- Zar, J. H. 1999. Biostatistical analysis. Fourth edition. Pren- ford University Press, New York, New York, USA. tice-Hall, Englewood Cliffs, New Jersey, USA.