Local Species Richness of Leaf-Chewing Insects Feeding on Woody Plants from One Hectare of a Lowland Rainforest V NOVOTNY,* Y BASSET,t S. E. MILLER,^:!! R. L. KITCHING,? M. LAIDLAW,? P. DR02D,** AND L. CI2EK* 'Institute of Entomology, Czech Academy of Sciences and Biological Faculty, University of South Bohemia, Branisovska 31, 370 05 Ceske Budejovice, Czech Republic fSmithsonian Tropical Research Institute, Apartado 2072, Balboa, Ancon, Panama ^iDepartment of Systematic Biology, National Museum of Natural History, MRC 105, Smithsonian Institution, Washington, DC 20013-7012, U.S.A., einailiniller.scott@nmnh.si.edu ?Cooperative Research Centre tor Tropical Rainforest Ecology and Management, Griffith University, Nathan 4111, Brisbane, Australia **University of Ostrava, Department of Biology, 30. dubna 22, 701 03 Ostrava, Czech Republic Abstract: Local species diversity of insect herbivores feeding on rainforest vegetation remains poorly known. This ignorance limits evaluation of species extinction patterns following various deforestation scenarios. We studied leaf-chewing insects feeding on 59 species of woody plants from 39 genera and 18families in a lowland rainforest in Papua New Guinea and surveyed all plants with a stem diameter at breast height of>5 cm ina 1- haplot within the same area. We used two extrapolation methods, based on randomized species-accumulation curves, to combine these two data sets and estimate the number of species of leaf-chewing herbivores feeding on woody plants from the 1-ha area. We recorded 58,483 feeding individuals from 940 species of leaf-chewing insects. The extrapolation estimated that there were 1567-2559 species of leaf-chewing herbivores feeding on the 152 plant species from 97 genera and 45 families found in 1 ha of the forest. Most of the herbivore diversity was associated with plant diversity on the familial and generic levels. We predicted that, on average, the selection of 45 plant species each representing a different family supported 39% of all herbivore species, the 52 plant species each representing a different additional genus from these families supported another 39% of herbivore species, and the remaining 55 plant species from these genera supported 22%> of herbivore species. Lepidoptera was the most speciose taxon in the local fauna, followed by Cole?ptera and orthopteroids (Orthoptera and Phasmatodea). The ratio of herbivore to plant species and the estimated relative species richness of the Lepidoptera, Cole?ptera, and orthopteroids remained constant on the spatial scale from 0.25 to 1 ha. Tlowever, the utility of local taxon-to-taxon species ratios for extrapolations to geographic scales requires further study. Riqueza de Especies Locales de Insectos Masticadores de Hojas en Plantas Le?osas de una Hect?rea de Selva Lluviosa Resumen: Se conoce poco de la diversidad de especies locales de insectos herb?voros que se alimentan de la vegetaci?n de selvas lluviosas. Esta escasez de informaci?n limita la evaluaci?n de patrones de extinci?n de especies despu?s de varios escenarios de deforestaci?n. Estudiamos insectos masticadores de hojas que se alimentan de 59 especies de plantas le?osas de 39 g?neros y 18 familias en vestigios de selva lluviosa en Papua Nueva Guinea y examinamos todas las plantas con un di?metro > 5 cm a la altura del pecho en una ?ffAddress correspondence to S. E. Miller. Paper submitted August 14, 2002; revised manuscript accepted June 3, 2003- 227 Conservation Biology, Pages 227-237 Volume 18, No. 1, February 2004 228 Insect Diversity in Rainforests Novotnyetal. parcela de 1 ha en la misma zona. Utilizamos dos m?todos de extrapolaci?n, basado en curvas aleatorias de acumulaci?n de especies, para combinar estos dos conjuntos de datos y estimar el n?mero de especies de herb?voros masticadores de hojas que se alimentan de las plantas le?osas de tapar?ela de 1 ha. Registramos un total de 58,483 individuos de 940 especies de insectos masticadores de hojas. La extrapolaci?n estim? que hab?a 1567-2559 especies de herb?voros masticadores de hojas aliment?ndose de las 152 especies de plantas de 97 g?neros y 45 familias encontradas en 1 ha de bosque. La mayor parte de la diversidad de herb?voros estaba asociada con la diversidad de plantas a nivel de familia y g?nero. Predijimos que, en promedio, la selecci?n de 45 especies de plantas, cada una representando a una familia diferente, soportaba al 39% de todas las especies herb?voras, las 52 plantas, cada una representando a un g?nero adicional diferente de estas familias, soportaba a otro 39% de las especies herb?voras y las restantes 55 especies de plantas de estos g?neros soportaban al 22% de las especies herb?voras. El tax?n con m?s especies en la fauna local fue Lepidoptera, seguido por Cole?ptera y ortopter?dos (Orthoptera y Phasmatodea). La relaci?n herb?voros - planta, y la riqueza relativa estimada de especies de Lepidoptera, Cole?ptera y ortopter?dos permaneci? constante en la escala espacial de 0,25 a 1 ha. Sin embargo, la utilidad de proporciones de especies de tax?n - tax?n para extrapolaciones a escalas geogr?ficas requiere de mayor estudio. Introduction Plant ecologists have made significant progress in map- ping the local species richness of tropical rainforests, particularly when using standardized census protocols for 1-ha and 50-ha plots (Condit 1997). In contrast, in- sect ecologists have yet to accomplish even a baseline description of their subject. For example, despite Erwin's (1982) pioneering effort to formulate an estimate of the local species richness of insects in the tropics, they still do not knovs^ ^vhat and ho^v many insect species live in any single hectare of a tropical rainforest (Godfray et al. 1999). The value of cataloging and understanding species before attempting to conserve them is obvious, and our inability to do so is damaging the credibility of the con- servation movement (e.g., Mann 1991). One example of the problems faced by conservation biologists is the con- troversy surrounding Lomborg's (2001) optimistic esti- mates of \(y>N extinction rates caused by habitat destruc- tion. These estimates are hard to disprove because of the lack of data on tropical diversity patterns (Pimm & Harvey 2001; Lovejoy 2002). Numerous studies, using insecticide fogging, light trap- ping, and other mass-collection methods, have provided insect samples from rainforests that included large num- bers of species (Stork et al. 1997) but yielded only lim- ited information on the ecology of these species. Thus, it remains unclear which species are genuine members of local food w^ebs and ^vhich are but transients, inflat- ing the species richness of the samples (Novotny & Bas- set 2000). A different approach focuses on sampling and studying Uve insects from selected plant species (Janzen 1988; Marquis 1991; Basset 1996; Barone 1998; 0degaard 2000a; Novotny et al. 2002?, 2002&). These studies pro- vide a novel source of information on the food w^ebs in rainforest communities but are often restricted to a small number of plant or herbivore taxa, limiting their useful- ness for inferring the size of the local pool of herbivore species in rainforests. One study relatively free of these limitations investigated a food w^eb including all locally coexisting species of macrolepidoptera in a dry forest in Costa Rica (Janzen 1988; Janzen & Gauld 1997). With such fragmentary information available at present, only indirect estimations of local herbivore di- versity in rainforests are feasible. Missa (personal com- munication) estimated the local species richness of w^ee- vils from an asymptotic species-accumulation curve, ob- tained from samples from diverse vegetation w^ithin a 1- km^ area of a lo^vland rainforest in Ne\^ Guinea. Novotny and Missa (2000) estimated the local species richness of several hemipteran families in a Ne^v Guinea rainforest from the overlap between a complete census of these taxa from a limited part of the vegetation (15 species of Ficus trees) and more comprehensive but incomplete samples from mixed vegetation. 0degaard (2003) extrap- olated data on herbivorous beetles from 50 species of trees and lianas sampled from a canopy crane in Panama to estimate the number of beetle species feeding on 500 woody plant species present in the rainforest ecosystem. Hammond et al. (1997) estimated the species richness of beetles in a 500-ha tract of rainforest in Sulaw^esi from samples collected by various sampling methods. Despite various spatial scales involved, a 1-ha standard is often used for extrapolation on the local scale (Erwin 1982; 0degaard 2003), in correspondence wth many botanical studies. We used novel methods based on species-accumulation curves (Novotny et al. 2002?) to extrapolate data on herbivore assemblages feeding on 59 species of woody plants to the assemblages occurring on the vegetation surveyed w^ithin a 1-ha area of a low^land rainforest in New Guinea. The estimate is based on a particularly de- tailed data set, including quantitative ecological informa- tion on 940 species of insect herbivores obtained through mass collection and rearing of insects by parataxonomists (Basset et al. 2000). Conservation Biology Volume 18, No. 1, Febriiary 2004 Novotny et al. Insect Diversity in Rainforests 11') Methods Study Area and Plant Census Our study area situated in the Madang Province is part of the low^Iands (0-400m above sea level) of Papua New Guinea that stretch from the coast to the slopes of the Adelbert Mountains. The study area has a humid tropi- cal climate w^ith average annual rainfall of 3558 mm, a moderate dry season from July to September, and mean air temperature of 26.5? C (McAlpine et al. 1983). Field work focused in a mosaic of primary and secondary \cr^- land forests near the villages of Baitabag, Ohu, and Mis (145?41-7'E, 5?08-14'S, approximately 50-200 m). At each site, the study area included 5-10 km^ of primary and secondary forests. The study sites w^ere <20 km from one another and had nearly identical vegetation (Laidla^v et al. 2003) and herbivore communities (Novotny et al. 2002c). The data from all sites w^ere therefore analyzed together All plants w^ith a diameter at breast height (dbh) of >5cm w^ere censused in a 1-ha, 100 x 100 m plot in a primary rainforest at our Baitabag site. The location of each plant above the threshold size was mapped and the plant identified if possible. Plant vouchers are deposited in Papua Ne^v Guinea's National Herbarium in Lae. Sampling of Insect Herbivores We selected 59 locally common species of trees and shrubs (13 species o? Ficus and 1 o? Artocarpus of the Moraceae, 6 species of Macaranga and 8 representing 8 other genera of Euphorbiaceae, 4 species of Psycho- tria and 12 representing 12 other genera of Rubiaceae, and 15 species representing 15 other families) for the study of their insect herbivores (listed in Novotny et al. 2002?). This selection included representatives of all ma- jor lineages of vascular plants (Angiosperm Phylogony Group 1998). Further, w^e included locally common plants from all main habitats, including early and late stages of forest succession and riverine habitats (Leps et al. 2001). The 3 families and 3 genera studied in greater detail are important general components of tropical flo- ras including low^land rainforests in Papua New Guinea (Corner 1965; Whitmore 1979; Sohmer 1988; Oatham & Beehler 1998). We studied the guild of externally feeding, leaf-chew^ing insects (Cole?ptera, Lepidoptera, and orthopteroids [i.e., Orthoptera and Phasmatodea]). Most of the Cole?ptera performed only maturation feeding on the leaves, so their larvae w^ere not sampled. Both immature individuals and adults ^vere feeding on the foliage in the orthopteroids. We collected insects from the foliage by hand. At each sampling occasion, a collector spent 1 day walk- ing throughout the study area and searching the foliage of the target tree species for caterpillars. The sampling w^as irregular because it included any tree from the target species encountered during the sampling ^valk w^ithin the study area. Numerous trees w^ere thus sampled on each sampling occasion, and many of the trees ^vere sampled repeatedly at different sampling occasions. The sampling included accessible branches from the forest canopy and understory, w^hich could be climbed or reached from the ground. Particularly poorly accessible trees w^ere not sam- pled. The approximate area of the foliage sampled was estimated visually and recorded. We kept sampling effort constant for all species at 1500 m^ of foliage sampled per species. We sampled each tree species for at least 1 year: Ficus from July 1994 to March 1996, Euphorbiaceae from Au- gust 1996 to August 1997, Rubiaceae from March 1998 to April 1999, and the remaining species from May 1999 to May 2000. Sampling was performed only in daytime. In the laboratory, each insect w^as provided w^ith fresh leaves of the plant species from w^hich it w^as collected until it fed or died. Only the individuals that fed w^ere considered in the subsequent analyses. Caterpillars w^ere reared to adults w^henever possible (successful in ap- proximately 40% of individuals and 75% of species; cf Novotny et al. 2002c). AU insects w^ere assigned to mor- phospecies, w^hich w^ere subsequently verified by spe- cialist taxonomists and identified as far as possible. Thus our morphospecies correspond to species (Basset et al. 2000), w^hich have often been matched to named species and are the subject of further taxonomic research (e.g., HoUow^ay & Miller 2003). Voucher specimens are de- posited in the Smithsonian Institution (Washington), Bishop Museum (Honolulu), and National Agriculture Re- search Institute (Port Moresby). Extrapolation of Herbivore Species Richness We used two methods of extrapolation to estimate the number of herbivore species feeding on plants from the 1-ha study plot: (1) extrapolation from 18 plant species from different families studied to the 45 representatives of different families present in the plot, corrected for ad- ditional herbivore species feeding on confamUial plant genera and congeneric plant species and (2) extrapola- tion from subsets of the 59 studied plant species that had an identical taxonomic structure (i.e., the distribution of species among genera and genera among families) as ran- domly selected subsets of species from the 1-ha forest plot. Both methods rely on the extrapolation of the number of additional species of herbivores expected from further expansion of the sampling universe beyond plant species actually sampled from species-accumulation curves. The number of additional herbivore species (j) resulting from the addition of the xth plant species to the data (x = \, 2, 3- ? ? n, w^here n is the total number of plant species studied) is described by the power function y = ex*', Conservation Biology Volume 18, No. 1, Febmary 2004 230 Insect Diversity in Rainforests Novotnyetal. O Genera per family ?Species per genus 25 50 75 Rank of family or genus 100 Figure 1. Number of plant genera in each family and of species in each genus recorded in the 1-ha study plot in Baitabag. where c and k are constants (Novotny et al. 2002?). The k is inversely proportional to the overlap in species com- position among herbivore assemblages from individual plant species, and the c approximates the species rich- ness of an average assemblage on a single host species. We fitted this function, using ordinary least-squares lin- ear regression of log-transformed data, to the empirical species-accumulation curves generated by amalgamating data from x = 1, 2, 3- ? ? n host-plant species in a ran- domized sequence. Average values from 1000 random se- quences w^ere used for each such calculation. In method 1, we fitted the pow^er function to the data for 18 plant species, each representing a different family, and used it to estimate the number of herbivore species expected for a single representative of each of the 45 families present in the plot. Further, w^e used data for 9 plant species, each representing a different genus of Eu- phorbiaceae, to estimate the number of additional herbi- vore species expected due to the presence of confamilial plant genera. Only herbivore species feeding exclusively on Euphorbiaceae ^vere considered in this analysis. The number of additional herbivore species corresponding to 1-9 confamilial genera, present in each of the 22 fam- ilies in the plot (Fig. 1), was estimated on the basis of this relationship. Finally, w^e used data for 6 species of Macaranga to estimate the number of additional herbi- vore species expected due to the presence of congeneric plant species, foUow^ing the same procedure as that for genera of Euphorbiaceae. Only those herbivore species feeding exclusively on Macaranga w^ere used for the anal- ysis. We used this relationship to estimate the number of additional herbivore species due to the 1-11 congeneric species present in each of the 21 tree genera in the plot (Fig. 1). We performed the same analysis for the Rubiaceae, us- ing 13 genera from this family and 4 species of Psychotria, and for the Moraceae, using 2 genera from this family and 13 species of Ficus. These analyses produced three in- dependent estimates of the number of herbivore species feeding on additional confamilial genera and congeneric species of plants. Macaranga quadriglandulosa was chosen to repre- sent Macaranga and Euphorbiaceae, Psychotria micral- abastra representedftyc/boir/? and Rubiaceae, a.ndFicus wassa represented Ficus and Moraceae in the above anal- yses. These species ^vere selected because their succes- sional optimum and habitus w^ere representative of their respective genus (Leps et al. 2001). In method 2 we fitted the pow^er function to the data from a subset of the studied 59 plant species that exactly matched an equally sized, randomly selected subset of the 152 species from the 1-ha plot in taxonomic structure (i.e., distribution of genera among families and species among genera [cf Fig. 1]) but not necessarily in the iden- tity of plant species, genera, or families. At each step of the analysis, one species was randomly selected w^ithout replacement from the 152 species recorded in the 1-ha plot. The selection continued as long as the resulting set of selected species could be matched in taxonomic structure by an analogous set of species studied for herbivores, which was selected from the 59 species available. The selection from the 59 species studied for herbi- vores ^vas random w^ithin the constraints given by the taxonomic structure of the mirrored selection from the 152 species. For instance, a family represented by a sin- gle species in the random selection from the 152 species could be matched by a randomly selected family from all 18 families studied for herbivores, ^vhereas a family rep- resented by 10 different genera had to be al^vays matched by Rubiaceae, because no alternative was available among the families sampled for herbivores. The resulting set of species selected from the 59 species studied for herbivores w^as used for extrapolation of species richness of herbivores on 152 plant species. The process of random plant selection and species-richness extrapolation w^as repeated 50 times. Only 11-24 plant species w^ere used for each analysis because w^e could not match precisely the taxonomic structure of larger se- lections from the set of 59 plant species sampled for her- bivores w^ith randomly selected subsets of the 152 species from the 1-ha plot. The most important constraints lim- iting the number of species that could be used in the analysis included the low number of families ^vith multi- ple genera and the genera w^ith multiple species, sampled for herbivores. Conservation Biology Volume 18, No. 1, Febmary 2004 Novotny et al. Insect Diversity in Rainforests 231 Table 1. The most common plant species, genera, and families in the 1-ha study plot in Baitabag. Plant species, genus, and family Basal area (m^) Herbivores Species'^ Pometia pinnata 4.35 42 yes Pimelodendron 2.02 83 yes amboinicum Pterocarpus indicus 1.89 10 yes Neonauclea hagenii 1.80 6 no Intsia bifuga \ni 2 no others (147 species) 16.73 899 23 species Genus Pometia 4.35 42 yes Pimelodendron 2.02 83 yes Neonauclea 1.91 9 yes Pterocarpus 1.89 10 yes Celtis 1.89 43 yes others (92 genera) 16.50 855 15 genera Family Sapindaceae 4.67 55 yes Rubiaceae 2.93 60 yes Euphorbiaceae 2.42 140 yes Meliaceae 2.18 122 no Caesalpiniaceae 2.12 14 no others (40 families) 14.25 651 10 families '^Number of Individual plants with diameter at breast height >5 cm. ''Taxa sampled for herbivores. '^Plant species were ranlzed on the basis of basal area. Results Tree Flora The 1-ha plot in Baitabag contained 1042 plants w^ith a dbh of >5 cm, including 453 plants with a dbh of >10 cm. Their basal area was 28.6 m^/ha(26.4 m^/ha for plants w^ith a dbh of > 10 cm), and they represented 152 species (112 species w^ith a dbh of > 10 cm) from 97 genera and 45 families. Botanical results are described by Laidlavs? et al. (2003). Sapindaceae, Rubiaceae, and Euphorbiaceae ranked highest in basal area (Table 1). The most species-rich fam- ily was Moraceae (19 spp.), foUow^ed by Meliaceae and Rubiaceae (both 15 spp). Rubiaceae and Euphorbiaceae had the highest number of genera (10 and 7, respec- tively; Fig. 1). Twenty families w^ere locally monotypic. Ficus (12 spp.) was the most species-rich genus in the plot, w^hereas each of 20 other genera was represented by 2-6 species, and the remaining 76 genera w^ere locally monotypic (Fig. 1). The most abundant species, Pometia pinnata, Pimelodendron amboinicum, and Pterocarpus indicus, constituted 29% of the total basal area and 13% of all plants in the plot (Table 1). In contrast, the rarest 40 species (26%) w^ere each represented by only one individual. Twenty-six species (17%), 20 genera (21%), and 13 fam- ilies (29%) occurring in the 1-ha plot ^vere also sampled 150 125 0 "o Q. W M? O d 100 500 1000 1500 No. of individuals 2000 Figure 2. Species-accumulation curves for herbivores feeding on 18plant species, each representing a different family. Each curve was created by amalgamation of 1-month samples in the order they were collected during 1 year for their herbivores. Forty-seven plant species (31%) from the plot belonged to genera included in the insect study. The herbivore study also included a majority of the plant species, genera, and families most abundant in the plot (Table 1). Assemblage of Insect Herbivores The sampling of leaf-chewng assemblages on 59 plant species produced 58,483 feeding individuals belonging to 940 species: 452 Lepidoptera, 379 Cole?ptera, and 109 orthopteroid species. Their host-plant associations represented 4619 herbivore-host plant combinations. The number of leaf-chew^ing species feeding on a plant species ranged from 31 to 132. After 1 year of sampling, the species-accumulation curves for individual plants did not approach an asymptote, indicating that the total species richness of their herbivore assemblages had not been sam- pled (Fig. 2). Herbivorous assemblages w^ere dominated by 21 fam- ilies, each represented by >10 species (the number of species from each family is given in parentheses): Lepidoptera-Geometridae (68), Crambidae (62), Noctu- idae (54), Lymantriidae (41), Tortricidae (38), Uran?dae (22), Sphingidae (19), Choreutidae (18), Limacodidae (17), Psychidae (15), Lycaenidae (14), Thyrididae (10); Coleoptera-Chrysomelidae (126), Cerambycidae (112), Curculionidae (78), Elateridae (17), Brentidae (13); Orthoptera-Tettigoniidae (65), Acrididae (10); Phasmatodea- Heteronemiidae (12), and Phasmatidae (11). Conservation Biology Volume 18, No. 1, Febmary 2004 232 Insect Diversity in Rainforests Novotnyetal. Table 2. Equations {y = ex*) describing tbe number of additional species of berbivorous insects {y) resulting from tbe addition of the xtb plant species. Plant and Equation herbivore taxa k c n" R^" number Plant families, all -0.518 74.1 18 0.99 1 herbivores Euphorbiaceae -0.295 18.5 9 0.92 2 genera, all herbivores Rubiaceae genera, all -0.325 22.9 13 0.93 3 herbivores Moraceae genera, all -0.325 25.6 2 ? 4 herbivores Macaranga species. -0.613 13.9 6 0.99 5 all herbivores Psychotria species. -0.322 7.6 4 0.99 6 all herbivores Ficus species, all -0.592 27.0 13 0.99 7 herbivores Plant families, -0.378 28.3 18 0.96 8 Lepidoptera Plant families. -0.563 28.2 18 0.99 9 Cole?ptera Plant families. -0.802 18.3 18 0.99 10 Orthopteroids "Total number of plants studied (x = 1, 2, 3, ? ? -, n). *" Variance explained. Species Richness of Insect Herbivores The pow^er function (y = ex*) w^as a good descriptor of the relationship between herbivore and host-plant species di- versity in aU data sets analyzed, w^ith R^ > 0.9 in all cases (Table 2). The residuals did not show a systematic de- parture from the predicted values (the quadratic term of second-order polynomial regression of residuals on pre- dicted values w^as not significant,/? > 0.1, in any of the data sets in Table 2). The increase, for instance, in the number of herbivore species accompanying the expansion of sampling from 1 to 18 plant species from different families was described by the equation y- 74.lx~?'^^^(R^ = 0.99), (1) which predicts that, on average, 74 herbivore species w^ill occur on the first plant species sampled and 52 additional species on the second plant species, through to 17 new herbivore species obtained ^vhen the last (18th) plant species is included (Table 2; Fig. 3). We verified the ac- curacy of this extrapolation by using a randomly selected subset of 9 plant species (half of the original data set) to estimate the number of herbivore species expected in the entire set of 18 plant species from different families. We generated 50 random subsets of 9 plant species by sam- pling wthout replacement from the full set of 18 species each time w^e fitted a separate regression to each subset and used it to extrapolate species richness on 18 plant lUU n I k d. ^^^^ Q. ^"?"Tg^^ CO i k^ ^^3^ ? '^^"=^=:^^ "?^_ ieaf-chewers ? S. +^\/>v? ?* JD \- ^~N^r~^"'?~~Q^o N, ^^*v,, ?^Sa? ^ 10 - Lepidoptera $ ?\ ^^tj- 0) c /iV M? O Cole?ptera o z orthopteroids 1 - 10 Serial no. of host plant genus 100 Figure 3- Number of new herbivore species obtained by sampling a new plant species. The average number of new herbivore species collected from first, second,..., eighteenth plant species, each from a different plant family, is depicted for all Ieaf-chewers, Lepidoptera, Cole?ptera, and orthopteroids and fitted by power functions (lines; Eqs. 1 and 8-10 from Table 2). species. These extrapolations provided the average (95% confidence interval) estimate of species richness at 511 (495-527) species, w^hereas the observed value w^as 520 species. According to extrapolation from Eq. 1 (Table 2), there w^ere 855 herbivore species feeding on 45 hosts, each representing a different family (Table 3). The aggregate number of additional herbivore species feeding on only 52 hosts representing confamUial tree genera w^as estimated at 700-938 (Table 3), based on Eqs. 2-4 (Table 2) and data on herbivores from three different plant families. The aggregate number of herbivore species feeding only on the other 55 hosts from these genera w^as estimated at 290-766 (Table 3), based on Eqs. 5-7 (Table 2) and data on herbivores from three plant genera. The total number of herbivore species feeding on plants from the 1-ha plot Table 3. Estimated number of berbivore species feeding on different families, confamilial genera, and congeneric species of plants from 1 ha of tbe forest* Euphorbiaceae Rubiaceae Moraceae Families 855 855 855 Confainilial genera 700 865 938 Congeneric species 385 290 766 Total 1940 2010 2559 * Estimates were calculated from data on tierbivores feeding on tt>ree plant families with equations from Table 2. Conservation Biology Volume 18, No. 1, Febmary 2004 Insect Diversity in Rainforests 233 ?9 900 - 300 0.6 0 5 w 1 0.4 SI ? ? 0-3 0) Q- 0.2 B (~^^^^^^ , u 0.1 Lep Col Taxon Ort 0.0 Figure 4. Number of species from Leptdoptera (Lep), Cole?ptera (Col), and orthopteroids (Ort) supported by 45 host species from different plant families (black), 52 hosts representing different additional genera from these families (white), and other 55 hosts from these genera (gray). was thus estimated by method 1 at 1940-2559 (Table 3). Separate analyses performed for Lepidoptera, Cole?ptera, and orthopteroids confirmed that most of the herbivore diversity was generated by plant diversity at the familial and generic levels in all these insect taxa (Fig. 4). Fifty estimates obtained w^ith method 2 ranged from 1335 to 2030 species. They w^ere normally distributed (Kolmogorov-Smirnov text, p > 0.15), w^ith an average (95% confidence interval) of 1567 (1332-1802) species. These estimates ^vere not mutually independent, ho-w- ever, because the same hosts ^vere used for numerous estimates. Differences between Herbivore Taxa and the Plant-Herbivore Ratio The three main herbivore groups, the Lepidoptera, Cole- ?ptera, and orthopteroids, w^ere each characterized by a different value of k in the y = cx^ relationship (Fig. 3; Eqs. 8-10 in Table 2). In particular, the orthopteroids had a low k, w^hich indicates a relatively large overlap among the assemblages from different plants and, accordingly, only a slo^v increase in the number of species ^vith in- creasing diversity of plants, w^hereas a high k for the Lepi- doptera reflected higher host specificity. Accordingly, the share of the Lepidoptera species in the compound herbiv- orous assemblage from diverse vegetation increased and that of the orthopteroid species decreased in compari- son to assemblages from single host species. An average herbivore assemblage from a single host, calculated from Cole?ptera ortliopteroids ?s A 1 1 1 1 1 1 1 1 host 0.01 0.125 0.25 0.5 0.75 1 Area (ha) Figure 5. Relative species richness (the proportion of species) of Lepidoptera, Cole?ptera, and orthopteroids in herbivorous assemblages on a single host and in assemblages estimated for plants from 0.01-1 ha of the forest vegetation. Number of herbivores on 1 host was calculated as an average from data for 18 hosts, each from a different family; number of herbivores on diverse vegetation was estimated by method 1 (see Methods). 18 assemblages from different plant families, ^vas likely composed of 38% Cole?ptera, 36% Lepidoptera, and 26% orthopteroid species, w^hereas a compound assemblage from 1 ha ^vould have 54% Lepidoptera, 37% Cole?ptera, and only 9% orthopteroid species. Species richness of these herbivore taxa w^as also es- timated for vegetation from areas of 0.01, 0.125, 0.25, 0.5, and 0.75 ha w^ith method 1. The relative species rich- ness of the Lepidoptera, Cole?ptera, and orthopteroids remained constant from 0.25 to 1 ha (Fig. 5). Likew^ise, the total herbivore and plant species richness increased in parallel from 0.25 to 1 ha so that the plant-herbivore species ratio changed only slightly from 15 to 14 (Fig. 6). Discussion Assemblages of Insect Herbivores Herbivorous insects w^ere not sampled exhaustively on any of the 59 plant species studied, as evidenced by their nonasymptotic species-accumulation curves. A complete census of locally feeding herbivores appears to be all but impossible for any single plant species in a highly di- verse tropical ecosystem (Price et al. 1995). Continuous Conservation Biology Volume 18, No. 1, Febmary 2004 234 Insect Diversity in Rainforests Novotnyetal. m w OJ c ? w 0 "o Q. CO 0.8 0.6 0.4 0.2 x - 40 ^ c 03 D. Plants - 30 ?3 Herbivores to Herb ./plant 0) - 20 p K X - 10 =^ 50 0.01 0.125 0.25 0.5 0.75 1 Area (ha) Figure 6. Species-area curves for plants and leaf-chewing herbivores. Species richness of plants and herbivores is expressed as the proportion of the species richness from. 1 ha (i.e., 152 plant and 2170 herbivore species). Number of herbivores was estimated by method 1 (see Methods); means from estimates based on three different families were used. Number of herbivore species per plant species is also represented. sampling tends to uncover additional species continually, many of them feeding on the target tree species only marginally and, accordingly, occurring at extremely low population densities on that species (Novotny & Basset 2000). High spatial diversity within the vegetation makes this "mass effect" (Shmida & Wilson 1985) particularly important in rainforests. For instance, an average 100-m^ subplot of the Baitabag plot contained 10.8 plant indi- viduals from 9-1 species; that is, almost none of any two adjacent trees w^ere conspeciflc. Such spatial heterogene- ity provides ample opportunity for colonization of each tree by polyphagous herbivores from neighboring trees (Basset 1999). The nonasymptotic increase in species w^ith sampling effort suggests that no particular magnitude of species richness could properly be used to characterize a com- munity per se, w^ithout reference to a particular sample size or sampled area (cf. Gotelli & Colwell 2001). The lo- cal species richness of herbivores thus depends not only on the species richness of plants but also on the abun- dance of individual plant species. The latter factor w^as not included in the present analysis because herbivore di- versity supported by each host w^as assessed on the basis of equal sample size, corresponding to 1500 m^ of foliage for each tree species studied. This sample size, set by lo- gistic constraints, is arbitrary but probably high enough to include all regular members of the herbivorous assem- blages studied (Novotny et al. 2002a). A more sophisticated estimate should combine species- accumulation curves, describing an increase in species richness of herbivores w^ith foliage area for each plant species, ^vith data on the foliage area of plant species in the studied area of forest. Unfortunately, such data w^ere not available. Local Species Richness of Insect Herbivores Our estimate of the number of leaf-chew^ing species feed- ing on plants from a 1-ha area of the forest is necessar- ily only approximate, and its accuracy is compromised by several methodological problems. The vegetation sur- vey w^as incomplete, not including lianas, epiphytes, and plant species w^ith a dbh of <5 cm. Liana species in partic- ular may have a rich herbivore fauna (0degaard 2000?). Further, w^e used 1-ha plots because they have become standard units for quantitative vegetation analysis in trop- ical rainforests, despite the fact that they do not represent an adequate sample of the local flora (e.g., Condit 1997; Oatham & Beehler 1998). Our sampling probably also missed some of the herbivore species limited to poorly accessible parts of the forest canopy, ^vhich may be an important bias (Basset et al. 2001). A further caveat to our study is that conclusions are valid only for the leaf- chew^ing taxa and stages we studied: larval Lepidoptera, mostly adult Cole?ptera, and both immature and adult Orthoptera and Phasmatodea. We probably underestimated the host specificity of the taxa studied only in the adult stage because immature individuals tend to be more host-specific than adults. Species-richness estimates made by methods 1 and 2 w^ere close to each other, w^hich is not surprising because both methods are theoretically equally sound and both w^ere based on the same data set. This also means that both estimates may be biased by limitations of available data, particularly those on the overlap among herbivore assemblages on congeneric plant species and confamilial plant genera. In both cases, the extrapolation w^as based on data sets from only three families from the 45 present, which may not be representative of other taxa. The esti- mates based on data from Moraceae ^vere particularly sus- ceptible to error because they w^ere based on the study of only tw^o confamilial genera. How^ever, the three families used for extrapolation were prominent in the vegetation because they included 20 from the 97 genera and 42 from the 152 species present in the 1-ha plot. Further, the dom- inant position o? Ficus and paucity of species from other genera are characteristic of the family Moraceae not only in our study design but also in Ne^v Gui?ean flora (H?ft 1992). The estimates of herbivore species richness supported by congeneric host species and confamilial host genera w^ere obtained by the analysis of three different plant fam- ilies and were thus mutually independent. The highest estimate ^vas 160% of the low^est one; w^e consider this Conservation Biology Volume 18, No. 1, Febmary 2004 Novotny et al. Insect Diversity in Rainforests 235 variation acceptable, given the exploratory and rather preliminary nature of insect species-richness estimates in tropical forests, including those in the present study. Further, any approach based on taxonomic rank is po- tentially misleading because the level at w^hich supraspe- cific taxa are recognized is a convention (Stevens 1998). Ne\^ approaches based on intertaxon distances calcu- lated from higher-level molecular phylogenies may help resolve this inadequacy in the future (Kitching et al. 2003). An approach similar to ours w^as adopted by 0degaard (2003), \s^ho sampled herbivorous beetles from 50 species of trees and lianas in a rainforest in Panama and then ex- trapolated the results to 500 plant species present locally. His estimate for 150 plant species w^as approximately 1250 species of herbivorous beetles, including species feeding on flow^ers and w^ood. Because leaf-chewng bee- tles represent 46% of all species effectively specialized (sensu May 1990) to an average plant species, the es- timate for leaf-chew^ing beetles on 150 plant species in Panama is 571 species?not far from the 788 species of leaf-chew^ing beetles estimated for 152 plant species by method 1 in our study. Janzen (1988) relied on mass light-trapping of adults and collecting of caterpillars from diverse vegetation of a dry forest in Costa Rica rather than extrapolation. He found 3140 species of Lepidoptera in an ecosystem with 725 species of vascular plants. No plant species ^vas at- tacked by more than 20 species of caterpillars. Our study found, on average, 29 caterpillar species per plant species, but our data were too limited for prediction of species richness over the entire local vegetation. We are not aw^are of any other study quantifying herbi- vore diversity supported by different levels of taxonomic diversity of the vegetation?by single representatives of all plant families, confamilial genera, and congeneric species present in the study area. The crucial role of the familial and generic diversity of plants and the lesser role of congeneric plant species in supporting herbivorous di- versity reflect a lo^v host specificity of herbivores with regard to congeneric hosts (Novotny et al. 2002?). Extrapolation to Other Taxa and beyond the Local Scale The extrapolation methods w^e applied are suited only to estimates of local species richness because they take no account of beta diversity. It is alw^ays tempting to extrap- olate such data, even to an estimate of the global species richness of all biota, but there are problems associated wth long-range extrapolations (Stork 1988; Basset et al. 1996; 0degaard 2000&; Gotelli & ColweU 2001; Novotny et al. 20020). One frequently used approach is extrapola- tion based on taxon-to-taxon (e.g., plant-to-insect) ratios of species richness (Erwin 1982; May 1990; Gaston 1992). It relies on the often relatively accurate estimates of local and regional species richness of one taxon and its locally estimated ratio to another taxon, w^hich is then assumed to remain constant on a regional scale, thus allowing for calculation of its regional species richness. This approach is based on an assumption that the beta-diversity of both taxa remains approximately the same. Our data confirm this assumption for taxon-to-taxon ratios involving Lepidoptera, Cole?ptera, and orthop- teroids, as w^ell as for a plant-to-herbivore species ratio, but only on a limited spatial scale from 0.25 to 1 ha. The generality of this result requires further confirmation. We show^ed that herbivore diversity is associated particularly wth the diversity of vegetation on familial and generic levels. The species-to-genus and species-to-family ratios tend to change wth spatial scale (Gotelli & Colw^eU 2001), w^hich may also affect plant-to-herbivore species ratios. Our tentative estimate of local species richness of rainforest leaf-chew^ing insects feeding on a single plant species and on plants from a 1-ha area of a lo^vland for- est is in broad agreement w^ith similarly lo^v estimates for other tropical host trees (Janzen 1988; Marquis 1991; Basset 1996; Barone 1998) and to a community-w^ide es- timate for herbivorous beetles from a Neotropical forest (0degaard 2003). In contrast, mass collecting methods, particularly insecticide fogging and light-trapping, yield samples that often include very large numbers of herbivo- rous species. For instance, Floren and Linsenmair (1998) obtained 1063 herbivorous beetles by fogging 19 trees from three species in Borneo; Missa (personal communi- cation) collected 1168 species of weevils from 1 km^ of a rainforest in Ne^v Guinea; and Barlow and Woiwod (1990) obtained 1520 species of macrolepidoptera and pyralids by light-trapping at a single site in Sulaw^esi. These data are difficult to compare w^ith ours because they include her- bivores other than externally feeding leaf-chew^ers and, more important, numerous tourist species. Reconciling these tw^o methodological approaches through more de- tailed study of the role of species in communities and bet- ter comprehensive sampling programs remains an impor- tant step toward understanding the organization of insect communities in tropical rainforests and tow^ard resolving the current debate over the magnitude of biodiversity on large geographical scales (Erwin 1982; Stork 1988; Basset et al. 1996). Acknowledgments Parataxonomists J. Auga, W Boen, C. Dal, S. Hiuk, B. Isua, M. Kasbal, R. Kutil, M. Manumbor, K. Molem, and K. Darrow assisted with technical aspects of the project. K. Damas and R. Kiapranis identified all trees recorded on the 1-ha plot. J. HoUow^ay, G. A. Samuelson, D. Perez, and H. C. H. van Herwaarden provided vital taxonomic assistance with insects. Numerous assistants w^ho col- lected many of the insect specimens and taxonomists Conservation Biology Volume 18, No. 1, Febmary 2004 236 Insect Diversity in Rainforests Novotnyetal. who identified plants and insects are acknowledged else- where. J. Leps commented on the manuscript. The Bishop Museum (Honolulu), the Smithsonian Institution (Wash- ington, D.C.) and the Natural History Museum (London) kindly provided access to their collections and facilities. The project was handed by the U.S. National Science Foundation (DEB-94-07297, 96-28840, 97-07928, and 02-11591), the Christensen Fund (Palo Alto, California), the Czech Academy of Sciences (A6007106, Z 5007907), the Czech Ministry of Education (ES 041), the Czech Grant Agency (206/99/1115, B6187001), ACIAR (Papua New Guinea Fruit Fly Project), the Cooperative Research Centre for Tropical Rainforest Ecology and Management, the National Geographic Society (5398-94), the Fulbright Commission, the Otto Kinne Foundation, and the Papua New Guinea Biological Foundation. Literature Cited Angiosperm Phylogeny Group. 1998. An ordinal classification for the families of flowering plants. Annals of the Missouri Botanical Gar- dens 85:531-553. Barlow, H. S., and I. E Woiw^od. 1990. Seasonality and diversity of Macrolepidoptera in two low^land sites in the Dumoga-Bone Na- tional Park, Sulawesi Utara. Pages 167-172 in W J. Knight and J. D. HoUow^ay, editors. Insects and the rain forests of South East Asia (Wallacea). Royal Entomological Society of London, London. Barone, J. A. 1998. Host-specificity of folivorous insects in a moist trop- ical forest. Journal of Animal Ecology 67:400-409. Basset, Y. 1996. Local communities of arboreal herbivores in Papua New^ Guinea: predictors of insect variables. Ecology 77:1906-1919. Basset, Y. 1999. Diversity and abundance of insect herbivores collected on Castanopsis acuminatissima (Fagaceae) in New Guinea: rela- tionships with leaf production and surrounding vegetation. Euro- pean Journal of Entomology 96:381-391. Basset, Y, et al. 2001. Stratification and diel activity of arthropods in a low^land rain forest in Gabon. Biological Journal of the Linnean Society'72:585-607. Basset, Y, G. A. Samuelson, A. Allison, and S. E. Miller. 1996. How many species of host-specific insects feed on a species of tropical tree? Biological Journal of the Linnean Society 59:201-216. Basset, Y, V. Novotny, S. E. Miller, and R. Pyle. 2000. Quantifying biodi- versity': experience with parataxonomists and digital photography in Papua New Guinea and Guyana. BioScience 50:899-908. Condit, R. 1997. Tropical forest census plots: methods and results from Barro Colorado Island, Panama and a comparison with other plots. Springer Verlag, New York. Corner, E. J. H. 1965. Check-list o? Ficus in Asia and Australasia with keys to identification. The Garden's Bulletin Singapore 21:1-186. Erwin, T. L. 1982. Tropical forests: their richness in Cole?ptera and other arthropod species. The Coleopterist's Bulletin 36:74-75. Floren, A., and K. E. Linsenmair. 1998. Non-equilibrium communities of Cole?ptera in trees in a lowland rain forest of Borneo. Ecotropica 4:55-67. Gaston, K. 1992. Regional numbers of insect and plant species. Func- tional Ecology 6:243-247. Godfray, H. C. J., O. T. Lewis, and J. Memmott. 1999. Studying insect di- versity in the tropics. Philosophical Transactions of the Royal Society of London Series B 354:1811-1824. Gotelli, N. J., and R. K. Colwe?. 2001. Quantifying biodiversity: proce- dures and pitfalls in the measurement and comparison of species richness. Ecology Letters 4:379-391- Hammond, P M., N. E. Stork, and M. J. D. Brendell. 1997. Tree-crown beetles in context: a comparison of canopy and other ecotone as- semblages in a lowland tropical forest in Sulawesi. Pages 184-223 in N. E. Stork, J. A. Adis, and R. K. Didham, editors. Canopy arthropods. Chapman c& Hau, London. H?ft, R. 1992. Plants of New Guinea and the Solomon Islands: dictionary of the genera and families of f low^ering plants and ferns. Handbook no. 13. Wau Ecology Institute, Wau, Papua New Guinea. HoUow^ay, J. D., and S. E. Miller. 2003. The composition, generic place- ment and host plant relationships of the joviana-group in the Paral- lelia generic complex (Lepidoptera: Noctuidae, Catocalinae). Inver- tebrate Systematics 17:111-128. Janzen, D. H. 1988. Ecological characterization of a Costa Rican dry forest caterpillar fauna. Biotropica 20:120-135. Janzen, D. H., and I. D. Gauld. 1997. Patterns of use of large moth cater- pillars (Lepidoptera: Saturniidae and Sphingidae) by ichneumonid parasitoids (Hymenoptera) in Costa Rican dry forest. Pages 251- 271 in A. D. Watt, N. E. Stork, and M. D. Hunter, editors. Forests and insects. Chapman & Hall, London. Kitching, R. L., K. Hurley, and L. Thalib. 2003. Tree relatedness and the similarity of canopy insect assemblages: pushing the limits? Pages 329-340 in Y Basset, Y Novotny, S. E. Miller, and R. L. Kitching, editors. Arthropods of tropical forests: spatio-temporal dynamics and resource use in the canopy. Cambridge University Press, Cambridge, United Kingdom. Laidlaw, M. J., R. L. Kitching, K. Damas, andR. Kiapranis. 2003. Structure and floristics of lowland rainforest plots in northern Papua New Guinea. Biotropica 35: in press. Leps, J., V Novotny, and Y Basset. 2001. Habitat and successional opti- mum of plants and the composition of their leaf-chewing herbivores in Papua New Guinea. Journal of Ecology 89:186-199. Lomborg, B. 2001. The skeptical environmentalist: measuring the real state of the w^orld. Cambridge University Press, Cambridge, United Kingdom. Lovejoy, T 2002. Biodiversity: dismissing scientific process. Scientific American 286:69-71. Mann, C. C. 1991. Extinction: are ecologists crying woU? Science 253:736-738. Marquis, R. J. 1991. Herbivore fauna o? Piper (Piperaceae) in a Costa Rican wet forest: diversity, specificity and impact. Pages 179-208 in P W. Price, T. M. Lewinsohn, G. W. Fernandes, and W. W. Benson, editors. Plant-animal interactions: evolutionary ecology in tropical and temperate regions. WUey, London. May, R. M. 1990. How many species? Philosophical Transactions of the Royal Society of London Series B 330:293-304. McAlpine, J. R., G. Keig, and R. Falls. 1983. Climate of Papua New Guinea. Commonw^ealth Scientific, Industrial, and Research Organization, and Australian National University Press, Canberra. Novotny, V., and Y Basset. 2000. Ecological characteristics of rare species in communities of tropical insect herbivores: pondering the mystery of singletons. Oikos 89:564-572. Novotny, V., and O. Missa. 2000. Local versus regional species richness in tropical insects: one lowland site compared with the island of New Guinea. Ecological Entomology 25:445-451. Novotny, V., Y Basset, S. E. Miller, P Drozd, and L. Cizek. 2002?. Host specialisation of leaf chewing in a New Guinea rainforest. Journal of Animal Ecology 71:400-412. Novotny, Y, Y Basset, S. E. Miller, G. D. Weiblen, B. Bremer, L. Cizek, and P Drozd. 2002?. Low host specificity of herbivorous insects in a tropical forest. Nature 416:841-844. Novotny, V, S. E. Miller, Y Basset, L. Cizek, P. Drozd, K. Darrow, and J. Leps. 2002c. Predictably simple: communities of caterpillars (Lepi- doptera) feeding on rainforest trees in Papua New Guinea. Proceed- ings of the Royal Society of London Series B 269:2337-2344. Oatham, M., and B. M. Beehler. 1998. Richness, taxonomic composition and species patchiness in three lowland forest tree plots in Papua Newr Guinea. Pages 613-631 in F Dallmeier and J. A. Comiskey, Conservation Biology Volume 18, No. 1, Febniary 2004 Novotny et al. Insect Diversity in Rainforests 237 editors. Forest biodiversity' research, monitoring and modelling: con- ceptual bacliground and Old World case studies. United Nations Ed- ucational, Scientific, and Cultural Organization and Parthenon Pub- lishing, Paris. 0degaard, F. 2000?. The relative importance of trees versus lianas as hosts for phytophagous beetles (Cole?ptera) in tropical forests. Jour- nal of Biogeography 27:283-296. 0degaard, F 2000&. How many species of arthropods? Erwin's estimate revised. Biological Journal of the Linnean Society' 71:583-597. 0degaard, F 2003. Taxonomic composition and host specificity of phy- tophagous beetles in a dry forest in Panama. Pages 220-236 in Y. Basset, V. Novotny, S. E. Miller, and R. L. Kitching, editors. Arthro- pods of tropical forests. Spatio-temporal dynamics and resource use in the canopy. Cambridge University Press, Cambridge, United King- dom. Pimm, S., and J. Harvey. 2001. No need to worry about the future. Nature 414:149-150. Price, P W, I. R. Diniz, H. C. Morals, and E. S. A. Marques. 1995. The abundance of insect herbivore species in the tropics: the high local richness of rare species. Biotropica 27:468-478. Shmida, A., and M. R. Wilson. 1985. Biological determinants of species diversity. Journal of Biogeography 12:1-20. Sohmer, S. H. 1988. The nonclimbing species of the genus Psychotria (Rubiaceae) in the New Guinea and the Bismarck Archipelago. Bul- letin in botany 1. Bishop Museum, Honolulu, Hawaii. Stevens, P. F 1998. What kind of classification should the practising taxonomist use to be saved? Pages 295-319 in J. Dransfield, M. J. E. Coode, and D. A. Simpson, editors. Plant diversity in Malesia III. Royal Botanic Gardens, Kew, United Kingdom. Stork, N. E. 1988. Insect diversity: facts, fiction and speculation. Biolog- ical Journal of the Linnean Society 35:321-337. Stork, N. E., J. A. Adis, and R. K. Didham. editors. 1997. Canopy arthro- pods. Chapman c& Hall, London. Whitmore, T C. 1979. Studies in Macaranga. X: potentially commercial species in New Guinea. Commonw^ealth Forestry Review 58:271- 272. Conservation Biology Volume 18, No. 1, Febniary 2004