Contributed Paper Changes in Arthropod Assemblages along a Wide Gradient of Disturbance in Gabon YVES BASSET,? OLIVIER MISSA,? ALFONSO ALONSO,? SCOTT E. MILLER,? GIANFRANCO CURLETTI,?? MARC DE MEYER,?? CONNAL EARDLEY,?? OWEN T. LEWIS,?? MERVYN W. MANSELL,??? VOJTECH NOVOTNY,??? AND THOMAS WAGNER??? ?Smithsonian Tropical Research Institute, Apartado 0843-03092, Balboa, Ancon, Panama City, Republic of Panama, email bassety@si.edu ?Department of Biology, University of York, P.O. Box 373, York YO10 5YW, United Kingdom ?Smithsonian Institution/Monitoring and Assessment of Biodiversity Program, 1100 Jefferson Drive S.W., Suite 3123, Washington, D.C. 20560-0705, U.S.A. ?Department of Systematic Biology, National Museum of Natural History, Smithsonian Institution, Washington, D.C. 20560-0105, U.S.A. ??Museo Civico di Storia Naturale, Cas. Post. 89, 10022 Carmagnola TO, Italia ??Royal Museum for Central Africa, Leuvensesteenweg 13, 3080 Tervuren, Belgium ??Plant Protection Research Institute, Private Bag X134, 0121 Queenswood, Pretoria, South Africa ??Department of Zoology, University of Oxford, South Parks Road, Oxford OX1 3PS, United Kingdom ???Department of Zoology and Entomology, University of Pretoria, Pretoria 0002, South Africa ???Biology Center of the Czech Academy of Sciences and School of Biological Sciences, University of South Bohemia, Branisovska 31, 370 05 Ceske Budejovice, Czech Republic ???Universita?t Koblenz-Landau, Institut fu?r Integrierte Naturwissenschaften?Biologie, Universita?tsstr. 1, 56070 Koblenz, Germany Abstract: Searching for indicator taxa representative of diverse assemblages, such as arthropods, is an important objective of many conservation studies. We evaluated the impacts of a wide gradient of disturbance in Gabon on a range of arthropod assemblages representing different feeding guilds. We examined 4 ? 105 arthropod individuals fromwhich 21 focal taxawere separated into 1534morphospecies. Replication included the understory of 3 sites in each of 4 different stages of forest succession and land use (i.e., habitats) after logging (old and young forests, savanna, and gardens). We used 3 complementary sampling methods to survey sites throughout the year. Overall differences in arthropod abundance and diversity were greatest between forest and open habitats, and cleared forest invaded by savanna had the lowest abundance and diversity. The magnitude of faunal differences was much smaller between old and young forests. When considered at this local scale, anthropogenic modification of habitats did not result in a monotonous decline of diversity because many herbivore pests and their associated predators and parasitoids were abundant and diverse in gardens, where plant productivity was kept artificially high year-round through watering and crop rotation. We used a variety of response variables to measure the strength of correlations across survey locations among focal taxa. These could be ranked as follows in terms of decreasing number of significant correlations: species turnover > abundance > observed species richness > estimated species richness > percentage of site-specific species. The number of significant correlations was generally low and apparently unrelated to taxonomy or guild structure. Our results emphasize the value of reporting species turnover in conservation studies, as opposed to simply measuring species richness, and that the search for indicator taxa is elusive in the tropics. One promising alternative might be to consider ?predictor sets? of a small number of taxa representative of different functional groups, as identified in our study. Keywords: indicator taxa, predictor sets, rainforest, species loss, species turnover Paper submitted August 10, 2007; revised manuscript accepted March 26, 2008. 1552 Conservation Biology, Volume 22, No. 6, 1552?1563 C?2008 Society for Conservation Biology DOI: 10.1111/j.1523-1739.2008.01017.x Basset et al. 1553 Cambios en Ensambles de Artro?podos a lo largo de un Gradiente de Perturbacio?n Amplio en Gabo?n Resumen: La bu?squeda de taxa indicadores representativos de ensambles diversos, como lo artro?podos, es un objetivo importante de muchos estudios de conservacio?n. Evaluamos los impactos de un gradiente de perturbacio?n amplio en Gabo?n sobre ensambles de artro?podos representando gremios alimentarios diferentes. Examinamos 4 ? 105 individuos de artro?podos de los cuales 21 taxa focales fueron separados en 1534 morfoespecies. La replicacio?n incluyo? el sotobosque de tres sitios en cada una de cuatro etapas diferentes de la sucesio?n de bosques y de uso de suelo (i.e., ha?bitats) despue?s de ser talados (bosques jo?venes y maduros, sabana y jardines). Utilizamos tres me?todos de muestreo complementarios para estudiar los sitios a lo largo del an?o. Las diferencias generales en la abundancia y diversidad de artro?podos fueron mayores entre bosques y ha?bitats abiertos, y el bosque talado invadido por sabana tuvo la menor abundancia y diversidad. La magnitud de las diferencias fue mucho ma?s pequen?a entre los bosques maduros y viejos. Al considerarla en esta escala local, la modificacio?n antropoge?nica de ha?bitats no resulto? en una declinacio?n mono?tona de la diversidad porque muchas plagas de herb??voros y sus depredadores y parasitoides asociados fueron abundantes y diversas en los jardines, donde la productividad de plantas fue mantenida artificialmente a lo largo del an?o mediante riego y rotacio?n de cultivos. Utilizamos una variedad de variables de respuesta para medir la robustez de las correlaciones en los sitios de muestreo entre taxa focales. Estos pudieron ser clasificados, en te?rminos del nu?mero decreciente de correlaciones significativas, como sigue: renovacio?n de especies > abundancia > riqueza de especies observada > riqueza de especies estimada > porcentaje de especies espec??ficas de sitios. El nu?mero de correlaciones significativas generalmente fue bajo y aparentemente sin relacio?n con la taxonom??a ni la estructura del gremio. Nuestros resultados enfatizan el valor de reportar la renovacio?n de especies en los estudios de conservacio?n, en lugar de simplemente medir la riqueza de especies, y que la bu?squeda de taxa indicadores es elusiva en los tro?picos. Una alternativa prometedora pudiera ser la consideracio?n de ?conjuntos pronosticadores? de nu?meros pequen?os de taxa representativos de grupos funcionales diferentes, como los identificados en nuestro estudio. Palabras Clave: bosque lluviosos, conjuntos pronosticadores, pe?rdida de especies, renovacio?n de especies, taxa indicadores Introduction Understanding and maintaining the distribution of biodi- versity across habitat mosaics, including varying levels of disturbance, is a central issue for conservation biology and the related field of agrobiodiversity. Arthropods rep- resent a significant fraction of biodiversity and play major roles in ecosystem function, yet the theory and practice of assessing their response to fragmented habitats remain poorly developed (Kremen et al. 1993). Arthropods in- clude a variety of functional guilds (Moran & Southwood 1982); hence, their responses to anthropogenic distur- bance may vary greatly, even among congeneric species (Basset et al. 2001). Furthermore, the concepts of indica- tor and umbrella species appear untenable, especially in the tropics (Didham et al. 1996; Lawton et al. 1998). This has led to repeated pleas to consider multispecies and multiguild-assemblage responses to anthropogenic dis- turbance (Didham et al. 1996; Kotze & Samways 1999). The impacts of disturbance on multitaxic assemblages have rarely been studied in tropical rainforests, due prin- cipally to a lack of workforce, relevant expertise, and funding (reviewed in Lewis & Basset 2007). Studies in- cluding several taxa and guilds are uncommon but grow- ing in number (Didham et al. 1996; Lawton et al. 1998; Schulze et al. 2004; Pineda et al. 2005; Barlow et al. 2007). Still, there have been few studies dedicated to examining arthropod responses over a wide anthropogenic gradient of disturbance (as opposed to examining undisturbed vs. disturbed forests) and selecting a wide range of arthro- pod taxa that represent diverse taxonomic and functional guilds. In their seminal study, Lawton et al. (1998) surveyed 8 animal groups (1 vertebrate, 6 arthropod, and 1 inver- tebrate taxa) at 12 sites that ranged from near-primary forests and old-growth secondary forests to plantations and cleared farm fallows. Although sampling effort was not particularly large (56 trap days), it was deemed suf- ficient to characterize the arthropod fauna at these dif- ferent sites. The results of this study appear especially relevant to species-rich assemblages (1620 arthropod species considered) and to conservation biology in gen- eral: species richness generally declined with increasing disturbance; no single group served as a good indicator taxon of the species richness of others; the matrix of cor- relation coefficients among taxa appeared idiosyncratic; and these data indicated the huge scale of the biologi- cal effort required to measure the effect of tropical-forest disturbance. Similar studies in Sulawesi, Mexico, and in the Amazon surveyed 5, 3, and 15 groups, respectively (including 2, 1, and 8 arthropod taxa, respectively) at different sites ranging from old and young secondary forests to planta- tions and crop fields (Schulze et al. 2004; Pineda et al. Conservation Biology Volume 22, No. 6, 2008 1554 Tropical Arthropods and Disturbance 2005; Barlow et al. 2007). Although the general results of Schulze et al. (2004) were similar to those of Lawton et al. (1998), species richness was significantly correlated among taxonomic groups (and in particular between the 2 arthropod groups). Schulze et al. (2004) did not em- brace the concept of indicator groups for predicting lo- cal patterns of species richness; rather, they stressed that animal and plant groups may be affected similarly by dis- turbance. Species richness was rather weakly correlated among taxonomic groups in other studies (Pineda et al. 2005; Barlow et al. 2007). Most of the researchers cited earlier used observed species richness to calculate correlations among taxa. Barlow et al. (2007) noted that cross-taxon congruence in response patterns is stronger when evaluated with community similarity than with species richness data. In a previous analysis of the study system described later, species richness was not always the best variable to quan- tify the effect of anthropogenic disturbance on arthro- pods (Basset et al. 2008). Out of a range of variables tested, abundance, estimated species richness, percent- age of habitat-specific species, and species turnover were also interesting in this regard and sometimes of higher dis- criminating power than observed species richness. Thus, the concept of indicator taxa (e.g., Pearson & Cassola 1992) should not be discarded before considering this full set of variables. Furthermore, Lawton et al. (1998) commented on the high costs involved with biodiversity surveys in tropical systems, but did not consider including local paratax- onomists in their protocols (Janzen et al. 1993; Basset et al. 2004a). We do not dispute the huge taxonomic ef- fort involved in identifying tropical species, but at least the preparation and implementation of complex proto- cols, including adequate spatial and temporal statistical replicates, all aspects of specimen preparation and pre- sorting, and entering data into a database, can be handled by parataxonomists. Motivated by Lawton et al. (1998), we based our study on the work of trained parataxonomists in Gabon. We examined 4 ? 105 arthropod individuals, from which 21 focal taxa were separated into 1534 morphospecies (Basset et al. 2004b, 2008). Replication included 3 sites in each of 4 different stages of forest succession and land use (i.e., habitats). We evaluated the effects of this wide gradient of disturbance on a range of arthropod as- semblages that represented different feeding guilds, and contrasted our conclusions with observed species rich- ness and other variables. Our specific aims were to (1) evaluate the strength of correlation in richness among various arthropod taxa across study sites along a wide disturbance gradient, (2) examine whether conclusions are modified if alternative variables to observed species richness are used, and (3) examine whether correlation matrices among taxa appear idiosyncratic ormay be struc- tured by taxonomy or functional guilds. This information should be widely helpful to conservation biologists con- cerned with documenting the impacts of human modifi- cation of tropical forests on the arthropod assemblages they support because it will allow them to focus future sampling efforts on particularly informative sets of taxa and particular metrics of diversity. Methods Study Area and Sites The study area was in the Shell Gabon oil concession of Gamba, within the Gamba complex of protected areas in southeastern Gabon (see Alonso et al. [2006] for back- ground and botanical information). The Gamba oil field includes a mosaic of old-growth secondary rainforests, younger secondary rainforests, and savanna areas, result- ing mainly from anthropogenic action. The mean annual temperature in the area is 26 ?C and annual rainfall is 2093 mm per year, with the major dry season from June to August (Alonso et al. 2006). The Gamba oil field has been active since 1967 and since then Gamba has grown from a small village in 1960 to a town of 8000 inhabitants. The earliest cultivated crop gardens of notable size were established near the town as recently as 1998. We considered 4 distinct habitats of increasing anthro- pogenic disturbance (i.e., increasing forest clearing and introduction of exotic vegetation) and selected 3 sites (replicates) within each habitat. The 4 habitat types were (1) understory of interior of old secondary rainforests (old forests), (2) understory of the edge of young secondary rainforests (young forests), (3) rainforest cleared to install oil rigs and subsequently invaded by savanna (savanna), and (4) cultivated crop gardens (gardens). At the time of the study, there were no substantial plantations in the area, and these 4 habitat types were predominant in the Gamba oil field. Salient characteristics of the study sites (coded A?L) are indicated in Table 1 (see also Fig. 1 and Basset et al. 2004b). Arthropod Collecting and Processing Each site was equipped with an identical set of traps rec- ommended for biological monitoring of flying and epi- gaeic arthropods of the understory and litter: 1 ground Malaise trap, 4 yellow pan traps on the ground, and 5 pitfall traps buried in the ground. Details about the traps, their emplacement, and mode of action are in Basset et al. (2004b). The 120 traps were operated for 3 days during each of the 38 survey periods from July 2001 to July 2002 (total 13,680 trap days). A team of 8 parataxonomists was trained and supervised by a professional entomolo- gist throughout the project (see Basset et al. [2004a] for a detailed discussion of this strategy). Conservation Biology Volume 22, No. 6, 2008 Basset et al. 1555 Table 1. Main characteristics of study sites within the Shell-Gabon Gamba oil field. Fragment Plant cover Code Habitat Coordinates size (ha) Physiognomy (level of cover)? A old forest 02?42?20? ?S 09?59?49? ?E 700 secondary forest, tallest trees 45 m, sandy soil Neochevalierodendron stephanii (dominant); Diospyros zenkeri, D. vermoeseni (common) B old forest 02?42?54? ?S 10?00?00? ?E 84 secondary forest, tallest trees 45 m, sandy soil N. stephanii (dominant); Diospyros zenkeri, D. vermoeseni, Palisota ambigua (common) C old forest 02?44?27? ?S 10?00?11? ?E 28 secondary forest, tallest trees 40 m, but many small trees 10?20 m tall, sandy soil D. vermoeseni, D. conocarpa (common); P. ambigua, Trichoscypha acuminata (less common) D young forest 02?45?38? ?S 10?01?37? ?E 12 secondary forest, tallest trees 20 m, many small trees and bushes, sandy soil P. ambigua, Aframomum sp., Rauvolfia sp. (common); Musanga cecropioides (present) E young forest 02?46?08? ?S 10?02?25? ?E 19 secondary forest, very open canopy, tallest trees 30 m, swampy soil Xylopia hypolampra, X. spp. (dominant) F young forest 02?47?32? ?S 10?03?45? ?E 166 secondary forest, plot at the edge of a thin tongue of forest connected to a large forested area; tallest trees 30 m, important regrowth in the understory, sandy soil Pachypodanthium staudtii, D. vermoeseni, P. ambigua, Leptactina mannii, Ouratea sulcata, Sacoglottis gabonensis, Bertiera subsessilis (present) G savanna 02?42?51? ?S 09?59?55? ?E 2.7 surrounded by forest; isolated bushes and trees, sandy soil, bare soil 50% Borreria verticillata, 2 unidentified Poaceae (dominant); Cyperus tenax, Dracaena sp. (present) H savanna 02?44?11? ?S 10?00?22? ?E 3.0 surrounded by forest, sandy soil, bare soil 25% B. verticillata, Dracaena sp., 1 unidentified Poaceae (dominant); Cyperus halpan, Heterotis decumbens (present) I savanna 02?48?23? ?S 10?03?21? ?E 2.5 surrounded by forest, sandy soil, bare soil 25% Merremia tridentata, C. tenax, 1 unidentified Poaceae (dominant) J garden 02?44?47? ?S 10?01?10? ?E 2 sandy soil fertilized with compost amaranth, aubergine, cabbage, carrot, lettuce, pepper, spinach, sweet pepper, tomato, water melon (present) K garden 02?43?36? ?S 10?02?06? ?E 0.5 clayish sand fertilized with compost aubergine, banana, maize, manioc, pepper, pineapple, spinach, sugar cane, taro (present) L garden 02?44?09? ?S 10?01?06? ?E 0.8 sandy soil fertilized with compost amaranth, aubergine, cabbage, cucumber, gombo, pepper, sorrel, spinach, tomato (present) ?For gardens, the main crops cultivated during the study period are listed. The material collected was first sorted into families or higher taxa by the parataxonomists. The material belonging to 21 focal taxa (Table 2)was isolated, and each individual was identified by a unique specimen number. Focal taxa were sorted to morphospecies (i.e., unnamed species diagnosed with standard taxonomic techniques) by the parataxonomists. Formal taxonomic study of this material is ongoing but subsamples of the material be- longing to 7 taxa have been examined by taxonomists (Table 2). We selected focal taxa so there would be good representation in the samples (so that much information was retained), because it would be workable taxonomi- cally, because taxonomists expressed interests in the ma- terial, and so there would be representation of a vari- ety of functional guilds and orders (Table 2). Specimens were stored at the Smithsonian Biodiversity Conserva- tion Center in Gamba, and vouchers were deposited at the National Museum of Natural History (Washington, D.C.) and with taxonomists who helped with species identification. Statistical Methods We calculated correlations among taxa collected at study sites with the following variables: abundance, observed and estimated species richness, percentage of habitat- specific species, and species turnover. Overall abundance and observed species richness were extracted from raw data for each focal taxa at each site because sampling effort was similar at each site (although not in terms of individual collected). To calculate estimated species rich- ness, we considered the results of consecutive surveys at each study site, pooling the results of all sampling meth- ods (n = 38 ? 12 = 456 samples). For ease of comparison with previous studies, we used EstimateS software to cal- culate Chao1 richness estimates (adequate when many species are rare) with 50 randomizations (Colwell 2005). To evaluate which species may be indicative of par- ticular sites and habitats, we used the indicator value index (Dufre?ne & Legendre 1997). We restricted the data set to the most abundant morphospecies (?12 Conservation Biology Volume 22, No. 6, 2008 1556 Tropical Arthropods and Disturbance Figure 1. Aerial photograph of the location of study sites within the Gamba oil field. Sites coded as in Table 2. The town of Gamba is located between sites B and K. Water bodies indicated as stippled areas. individuals; i.e., on average at least 1 individual col- lected per site; 227 morphospecies) and tested whether morphospecies were indicative of particular sites (site- specific species) or of particular habitats (habitat-specific species), particularly for old and young forests (133 mor- phospecies). The significance of the statistic was tested for each taxon by Monte Carlo randomization with 1000 permutations, performed with PC-ORD (McCune & Med- ford 1999). To quantify overall species turnover, we used de- trended correspondence analysis (DCA) with Hill?s scal- ing on untransformed data. The differences between the scores of any 2 sites on the first axis of the DCA represent a measure of species turnover between these 2 sites (ter Braak & Smilauer 1998). We again restricted our data set to the most abundant morphospecies (n = 227) and com- puted DCA with CANOCO (ter Braak & Smilauer 1998). To compare the species turnover of 2 taxa, we used a Pro- crustean randomization test (PROTEST) to compare the concordance between the raw data of 2 species-by-sites matrices (Peres-Neto & Jackson 2001; rationale detailed in Supporting Information). The significance of the m2 statistic computed between all matrix pairs was assessed byMonte Carlo randomizationwith 1000 permutations in ade4 package of R Project software (Chessel et al. 2004). Before calculating correlations among focal taxa, we checked for spatial autocorrelation and the indepen- dence of the data points (sites) with Mantel tests Conservation Biology Volume 22, No. 6, 2008 10'OTTE 10"Z0"E 10 4trE 2-44*03 2'4fO"S 2=4*"0"S- -2"44'(rS "Z401TS -r4Mrs ICTOTTE KT2XTE 10 4"(rE Basset et al. 1557 Table 2. Focal taxa considered in this study of tropical arthropods and disturbance. Focal taxa Ordera Guilda Individualsb Indmc Mor/Spp.d Mantele Pe Codea Mantodea Ma Pr 98 50 19 ?0.204 0.881 Man Acrididoideaf Or Lc 1129 360 40 ?0.071 0.664 Acr Fulgoroideag He Ss 4022 2345 233 0.010 0.409 Ful Membracidae He Ss 37 35 14 ?0.071 0.669 Mem Buprestidae Co Wo 115 91 16h ?0.003 0.428 Bup Scarabaeidae Co Lc, Sc 2240 1980 81 0.002 0.472 Sca Coccinellidae Co Pr 1409 1200 32 0.075 0.275 Coc Histeridae Co Pr 682 589 20 ?0.081 0.769 His Cleridae Co Pr 45 18 12 0.325 0.055 Cle Tenebrionidae Co Sc 839 605 54 0.199 0.052 Ten Cerambycidae Co Wo 278 79 51h 0.121 0.213 Cer Chrysomelidae Co Lc 2285 1761 157h ?0.049 0.634 Chr Neuropterai Ne Pr 235 133 25h 0.328 0.036 Neu Asilidae Di Pr 409 333 47 0.108 0.235 Asi Dolichopodidaej Di Pr 7339 2113 38 ?0.080 0.731 Dol Tephritidae Di Lck 535 426 34 0.150 0.175 Tep Syrphidae Di Pr, Sc 459 369 25h 0.126 0.388 Syr Pipunculidae Di Pa 123 97 22h 0.272 0.113 Pip Ichneumonidae Hy Pa 2302 1880 420 0.204 0.046 Ich Chalcidoideal Hy Pa 4577 1302 175 0.031 0.385 Cha Apoideam Hy Lcn 1239 1049 51h ?0.119 0.692 Apo aOrders: Co, Coleoptera; Di, Diptera; He, Hemiptera; Hy, Hymenoptera; Ma, Mantodea; Ne, Neuroptera; Or, Orthoptera. Guilds: Lc, leaf chewers; Pa, parasitoids; Pr, predators; Sc, Scavengers; Ss, sapsuckers; Wo, wood eaters (Moran & Southwood 1982). Code, abbreviations used in tables. bNumber of individuals collected. cNumber of individuals morphotyped by parataxonomists (some damaged or lost material could not be morphotyped; some material collected by flight-interception traps was not considered). dTotal number of morphospecies sorted by parataxonomists from Indm. eMantel statistic (r) testing for spatial autocorrelation and associated probability, p. f Including Acrididae, Pyrgomorphidae, and many juveniles, not morphotyped. gIncluding Achilidae, Cixiidae, Delphacidae, Derbidae, Dictyopharidae, Eurybrachidae, Flatidae, Fulgoridae, Issidae, Meenoplidae, Ricaniidae, Tettigometridae, and Tropiduchidae. hNumber of species sorted by taxonomists from a subsample of Indm. iIncluding Berothidae, Coniopterygidae, Chrysopidae, Dilaridae, Hemerobiidae, Mantispidae, Myrmeleontidae, and Osmylidae. jOnly morphotyped from July?December 2001, then kept unassigned in alcohol. kSubguild: fruit feeders. lOnly > 2 mm and including Agaonidae, Chalcididae, Elasmidae, Encyrtidae, Eucharitidae, Eulophidae, Eupelmidae, Eurytomidae, Leucospidae , Perilampidae, Pteromalidae, Tetracampidae, and Torymidae. mIncluding Apidae, Halictidae, and Megachilidae. nSubguild: pollinators. appropriate for this purpose. We tested the indepen- dence between the distance matrix of study sites (carte- sian coordinates; euclidean distance) and the dissimilar- ity matrix of each focal taxa (sites ? species; abundance data; distance equivalent of theMorisita-Horn index)with the vegan package of R Project software (Oksanen et al. 2006). For ease of comparison with previous studies, we computed correlations for insect abundance and species richness variables between each focal taxa with Pearson coefficients. We are aware of the potential problem of double zeros in these calculations. This concerned a mi- nority of comparisons for most insect variables but not for the percentage of site-specific species. To reduce the occurrence of double zeros, we calculated correlations only for focal taxa with site-specific species present at least at half of the study sites (9 taxa). For correlation analyses, we used the false discovery rate method to cor- rect for multiple tests (Garc??a 2004). For species turnover (PROTESTprocedure),we did not adjust p values because Monte Carlo simulations provided strong control of type I error rates (Peres-Neto 1999). Results In total 400,404 arthropods were collected with all col- lecting methods during the 38 sampling events. Thirty- one orders and at least 218 families were represented (see Supporting Information). The 21 focal taxa represented 16,855 individuals and 1,534 morphospecies (Table 2). Furthermore, 347 species were recognized from the 7 fo- cal taxa that to date have been examined by taxonomists (Table 2). Generally, there was good correspondence Conservation Biology Volume 22, No. 6, 2008 1558 Tropical Arthropods and Disturbance between the number of species sorted by taxonomists and the number of morphospecies sorted by paratax- onomists (r = 0.96, p < 0.01, n = 7), but these results will be discussed in detail elsewhere. Comparisons of mean insect variables among habitats are also presented elsewhere (Basset et al. 2008). When the mean abundance of all arthropods col- lected among habitats was considered, abundance was significantly higher in forests and gardens than in sa- vanna (Fig. 2a). Nevertheless, this pattern was differ- ent when we excluded ants from analyses (Fig. 2a): arthropod abundance was significantly higher in gar- dens than in savanna and intermediate in forests. Fi- nally, when we considered only the abundance of our Figure 2. Mean (SE) per site of insect variables compared across habitats (Olf, old forests; Yof, young forests; Sav, savanna; Gar, gardens): (a) abundance of all arthropods (black bars), all arthropods without ants (gray bars), all focal taxa (white bars;); (b) observed species richness (black bars) and estimated species richness (white bars); (c) percentage of site-specific species (gray bars); (d) species turnover: plot of 227 morphospecies (small, open circles) and sites (coded A?L, large closed circles) in the plane formed by axes 1 and 2 of the DCA; (e) detail of guild composition: percent individual leaf chewers (Lc, black bars), sapsuckers (Ss, light stippled bars), scavengers (Sc, dark stippled bars), parasitoids (Pa, gray bars), and predators (Pr, white bars) (wood eaters represent <2% of individuals and are not figured in here). In (a), (b), (c) and (e) for each variable, different letters denote different means (Tukey tests, p < 0.05). Figures 2b?e refer to all focal taxa. focal taxa (Fig. 2a), abundance in gardens was signif- icantly higher than in other habitats. The abundance of 14 focal taxa (67% of taxa tested) was significantly different among habitats. Many taxa were more abun- dant in unforested habitats (Supporting Information) and particularly in gardens. Nevertheless, a certain number of taxa not sorted to species were notably more abun- dant in forests than in gardens and particularly abun- dant in old forests (Supporting Information: Gryllidae, Endomychidae, Mycetophilidae, Phoridae, and Formici- dae). In addition, Blattodea, Curculionidae, Staphylin- idae, Cecidomyiidae, and Isoptera were either equally well represented in both old and young forests or better represented in the latter. Collectively, these Conservation Biology Volume 22, No. 6, 2008 (b)Observed and estimated species richness ?b b (C) Percentage of slte-specUlc species i 00 Y*J* *#v 0#f OCA A.t* 1 YW ##* (e) Guilds Basset et al. 1559 observations explained most of the differences in abun- dance pattern between habitats (Fig. 2a). When all focal taxa were considered, observed arthro- pod species richness was significantly higher in gardens than in savanna and intermediate in forests (Fig. 2b). This trend was similar for estimated species richness (Fig. 2b). Most of the species tested were significantly habitat or site-specific (92% and 82% of species tested, respec- tively). The percentage of site-specific species was signif- icantly higher in gardens than in other habitats (Fig. 2c). When we restricted our comparison to old and young forests, 29% of species were significantly habitat spe- cific, including 21 species for old forests and 17 for young forests. Considering all focal taxa, themean scores of sites on axis 1 of the DCA were significantly different among habitats. The DCA revealed large faunal differences be- tween forested and unforested habitats (Tukey tests, p < 0.05) and less of a difference between old and young forests and savanna and gardens (Tukey tests, p > 0.05; Fig. 2d). Of the 21 focal taxa, 14 differed significantly in their DCA scores among habitats. Most of these differ- ences concerned again forested and unforested habitats (Tukey tests, p < 0.05). Old and young forests supported significantly higher percentages of sapsuckers and para- sitoids than savanna and gardens (Fig. 2e). On the other hand, gardens and savanna supported higher percentages of leaf chewers and predators than forests. The percent- age of scavengers was also significantly higher in savanna than in gardens (Fig. 2e). Only 2 focal taxa were weakly spatially autocorrelated (Table 2). Therefore, we considered the data points for correlations as independent overall and did not adjust the degree of freedom in our calculations. Correlations of abundance among focal taxa were rather low (aver- age [SE] r = 0.176 [0.029], n = 210). Of 210 possi- ble combinations, only 50 correlations were significant with p < 0.05, including only 21 correlations significant after correction for multiple tests (Supporting Informa- tion). There was no obvious structure in the matrix of correlation, either when grouping correlations by func- tional guilds (Supporting Information) or by insect or- ders. In particular, there was no significant difference between the average values of correlation coefficients re- sulting from intraguild and interguild comparisons (t = 1.298, p = 0.222). Similarly, there was no significant dif- ference between the average values of correlation co- efficients resulting from intraorder and interorder com- parisons (t = ?0.835, p = 0.405). Although there was a trend for correlations between leaf chewers and par- asitoids to be rather high, they were not significantly higher than other correlations (1-sample t test, t = 1.637, p = 0.130, n = 12). Similarly, correlations between Diptera and Hymenoptera appeared higher than other correlations, but not significantly so (t = 1.623, p = 0.124, n = 15). The 4 best indicator taxa were only each significantly correlated with 4 other taxa (Bupresti- dae, Chrysomelidae, Syrphidae, and Apoidea, Supporting Information). Correlations of observed species richness between fo- cal taxa followed similar trends as for correlations of abundance (Table 3). They were also rather low (average [SE] r = 0.193 [0.029]; 48 correlations significant with p < 0.05, including only 18 correlations significant after correction for multiple tests). There was no significant difference between the average values of correlation co- efficients resulting from either intraguild and interguild comparisons and intraorder and interorder comparisons (t = ?0.914, p = 0.362 and t = ?0.356, p = 0.722, respec- tively). Correlations either between leaf chewers and par- asitoids or between Diptera and Hymenoptera were not significantly higher than other correlations (t = 1.724, p = 0.113 and t = ?0.939, p = 0.364, respectively). There was a sharp decrease in the number of signif- icant correlations derived from estimated species rich- ness (average [SE] r = 0.054 [0.024]; Supporting Infor- mation). Only a single correlation could be considered significant after correction for multiple tests. For taxa amenable to analysis, no correlation was significant af- ter correction for multiple tests within the matrix on the basis of percentage of site-specific species (average r = 0.202 [0.052], n = 36; Supporting Information). Never- theless, comparisons of species-by-sites matrices (species turnover) provided the highest share of significant com- parisons (42 of 210 possible comparisons, 20%; Support- ing Information). In particular, distribution patterns of Dolichopodidae were correlated with distribution pat- terns of 9 other taxa (at least 1 in each major guild), that of Pipunculidae and Syrphidae with 8 other taxa, and that of Apoidea with 7 taxa. Still, there was no obvious structure when considering significant statistics within the matrix (Supporting Information). Discussion We examined spatial congruence among arthropods, the most important component of terrestrial biodiversity. Our sample sizewas unusually large (in terms of trap days, number and variety of arthropod taxa considered, individ- uals collected, and number of species andmorphospecies analyzed) andwas obtainedwith adequate spatial and sea- sonal replicates. In addition, we used different methods to collect specimens. As far as we know, this represents one of the best tropical data sets of its kind, one that is ad- equate to evaluate arthropod responses to large-scale dis- turbance within an area of approximately 70 km2 (Fig. 1). We further examined arthropod responses and possible correlations among these responses in our consideration of several variables (abundance, observed and estimated species richness, percentage of site-specific species, and species turnover), whereas other researchers focused Conservation Biology Volume 22, No. 6, 2008 1560 Tropical Arthropods and Disturbance Ta bl e 3. Lo w er co rr el at io n m at ri x (P ea rs on co ef fic ie nt ) of ob se rv ed sp ec ie s ri ch ne ss be tw ee n fo ca lt ax a at th e di ffe re nt st ud y si te s (n = 12 ), or de re d by fu nc tio na lg ui ld s. a Ss Lc W o P r P a Sc M ix ed b G u il d Ta xa Fu l M em A cr C h r Te p A po B u p C er Te n M a n H is C oc C le N eu A si D ol P ip Ic h C h a Sc a Ss M em 0. 62 c Lc A cr 0. 51 0. 13 C h r 0. 85 d 0. 20 0. 54 T ep 0. 23 0. 15 0. 75 c 0. 29 A p o 0. 46 0. 24 0. 97 d 0. 44 0. 79 d W o B up 0. 44 0. 46 0. 57 0. 28 0. 61 c 0. 66 c C er ? 0. 17 0. 21 ? 0. 67 c ? 0. 24 ? 0. 50 ? 0. 59 c ? 0. 07 Sc T en 0. 71 c 0. 20 0. 43 0. 82 d 0. 38 0. 40 0. 51 0. 03 Pr M an ? 0. 14 0. 10 0. 11 ? 0. 24 ? 0. 01 0. 25 0. 34 0. 13 ? 0. 19 H is ? 0. 35 ? 0. 35 ? 0. 40 ? 0. 05 ? 0. 27 ? 0. 41 ? 0. 27 0. 60 c 0. 12 0. 00 C oc 0. 14 ? 0. 16 0. 80 d 0. 23 0. 73 c 0. 76 d 0. 57 ? 0. 68 c 0. 19 0. 27 ? 0. 36 C le ? 0. 03 0. 11 ? 0. 69 c ? 0. 02 ? 0. 65 c ? 0. 69 c ? 0. 54 0. 71 c ? 0. 11 ? 0. 21 0. 36 ? 0. 87 d N eu 0. 44 0. 24 0. 64 c 0. 43 0. 40 0. 67 c 0. 44 ? 0. 59 c 0. 16 0. 39 ? 0. 34 0. 55 ? 0. 54 A si 0. 48 0. 67 c 0. 03 0. 23 0. 16 0. 08 0. 06 ? 0. 01 0. 04 ? 0. 25 ? 0. 57 ? 0. 24 0. 35 0. 04 D ol ? 0. 02 ? 0. 13 0. 35 0. 17 0. 56 0. 33 ? 0. 01 ? 0. 32 ? 0. 01 0. 13 ? 0. 04 0. 49 ? 0. 18 0. 16 0. 13 Pa Pi p 0. 37 0. 06 0. 90 d 0. 43 0. 83 d 0. 91 d 0. 51 ? 0. 62 c 0. 45 ? 0. 07 ? 0. 29 0. 66 c ? 0. 63 c 0. 49 0. 03 0. 32 Ic h 0. 41 0. 54 ? 0. 36 0. 23 ? 0. 33 ? 0. 28 0. 21 0. 71 c 0. 47 0. 08 0. 21 ? 0. 57 0. 51 ? 0. 23 0. 27 ? 0. 46 ? 0. 38 C h a 0. 57 0. 31 0. 95 d 0. 50 0. 68 c 0. 96 d 0. 62 c ? 0. 50 0. 42 0. 30 ? 0. 43 0. 70 c ? 0. 56 0. 61 c 0. 15 0. 36 0. 82 d ? 0. 19 M ix ed Sc a 0. 46 0. 37 0. 31 0. 43 0. 35 0. 32 0. 60 c 0. 24 0. 53 ? 0. 01 ? 0. 11 0. 22 0. 02 ? 0. 05 0. 34 0. 05 0. 16 0. 40 0. 38 Sy r 0. 52 0. 28 0. 82 d 0. 57 0. 94 d 0. 83 d 0. 67 c ? 0. 49 0. 57 ? 0. 02 ? 0. 28 0. 73 c ? 0. 59 0. 53 0. 23 0. 53 0. 81 d ? 0. 17 0. 77 d 0. 48 a A bb re vi a ti on s of fo ca lt a xa a n d gu il ds a s in Ta bl e 2. b Th es e ta xa in cl u de re pr es en ta ti ve s of di ff er en t gu il ds . c C oe ff ic ie n ts si gn if ic a n t w it h p < 0. 05 . d C oe ff ic ie n ts si gn if ic a n t a ft er co rr ec ti on fo r m u lt ip le te st s w it h th e fa ls e de te ct io n ra te m et ho d (p ? 0. 00 43 ). Conservation Biology Volume 22, No. 6, 2008 Basset et al. 1561 mainly on observed species richness (but see Barlow et al. 2007). The variety of assemblages we considered also al- lowed us to test whether taxa belonging to the same order or guild showed similar responses. The problem that remains is the choice of focal taxa. The majority of conservation studies on arthropods ex- amine only a small amount of local terrestrial biodiver- sity, usually 1 or 2 taxa that are not particularly speciose (Lewis & Basset 2007). Despite a massive collecting effort rarely equaled in the literature, we could only evaluate the influence of forest disturbance for 4% of the material collected. This raises serious questions about the choice of focal taxa and their representativeness. Had we con- sidered taxa abundant in forests such as Formicidae and a few others (Supporting Information), our conclusions regarding relative arthropod abundance along the distur- bance gradient would have been different. In fact, overall arthropod abundance was not lower in forests than in gardens (Fig. 2a). We also strongly suspect that, overall, arthropod species richness is higher in forests than in gardens be- cause (1) we targeted understory arthropods in forests, and most tropical forests harbor a significant and differ- ent canopy fauna, which may sometimes be as speciose as the understory fauna (Basset et al. 2003), (2) esti- mated species richness was lower in forests than in gar- dens but not significantly so, and it is notoriously diffi- cult to survey all local rainforest species (Longino et al. 2002), (3) seasonal turnover was higher in forests than in gardens (Basset et al. 2008), and (4) regional species richness of garden assemblages is probably much lower than that of forest assemblages. We conclude that a mas- sive project considering an even wider choice of focal taxa than ours would probably emphasize the sheer di- versity of arthropod life histories and concomitant vari- ety of responses to disturbance, decreasing further the proportion of significant correlations among arthropod taxa. Overall arthropod responses to disturbance empha- sized differences between forests and open habitats, with cleared forest subsequently invaded by savanna being the least populated and diverse habitat for arthropods. Few faunal differences were apparent between old and young forests, but this observation needs to be consid- ered in the context of the wide disturbance gradient at Gamba. Despite this, 29% of species tested could be considered as habitat specific for either old or young forests, when comparing these 2 forest types. Hence, our data provide information on the biodiversity value of the wider, human-degraded landscape, indicating that many, but not all, taxa persist in these habitats. Thus, conservation strategies that promote secondary forests may maintain reasonably high species richness (Wright &Muller-Landau 2006), but a proportion of species across all taxa is likely to be lost with unknown long-term func- tional consequences (Brook et al. 2006). Detailed analysis of a major herbivore group (Chrysomelidae) indicated that gardens were very dis- tinct from other habitats and were invaded by a pest fauna mostly associated with crops (Basset et al. unpub- lished data). Sandy soils in gardens are fertilized by com- post, which also increases water retention. Plant pro- ductivity is kept artificially high in gardens year-round through watering and crop rotation. Arthropods prob- ably responded to these favorable conditions, notably leaf-chewing species and associated enemies. Hence, our results emphasize that anthropogenic modification of habitats, when considered at a local scale, does not necessarily result in a monotonous decline of diversity. This is in agreement with the results of many studies in which old-growth forests were compared with for- est plantations in the tropics (Speight et al. 2003). Al- though garden assemblages had high local alpha diver- sity (Basset et al. 2008) and may be considered quite habitat specific at the meso- or local scale, they consisted mostly of pests and generalist species of lower conser- vation value, which are unlikely to show high spatial beta diversity and overall species richness at the regional scale. Insect response variables could be ranked as follows in terms of decreasing number of significant correla- tions among focal taxa: species turnover > abundance > observed species richness > estimated species rich- ness > percentage of site-specific species. As discussed elsewhere, conservation studies should as a matter of pri- ority report species turnover alongwith observed species richness as a more accurate assessment of disturbance ef- fects on faunal assemblages (Barlow et al. 2007; Tylianakis et al. 2007; Basset et al. 2008). Misleading conclusions are likely to be drawn if focusing solely on species richness (or a species diversity index) to assess human impacts, even when considering multiple taxa or guilds. That said, our matrix of concordance accounting for species turnover included only 20% of significant con- cordances, a rather low figure if one is interested in identifying potential indicator taxa. Worse, there was no apparent structure in this matrix in terms of either tax- onomy or functional guilds. In particular, concordance between taxa belonging to the same guilds was not no- ticeably higher than concordance between taxa belong- ing to different guilds. On the basis of species turnover, the best indicator taxon was Dolichopodidae. These flies are reasonably good fliers whose adults prey on a wide array of arthropods and respond quickly to habitat modi- fication (Couturier & Duviard 1976). Nevertheless, their abundance and species richness was poorly correlated with that of other taxa. Furthermore, a reasonably good indicator for a particular disturbance gradient may not necessarily be good for narrower or wider gradients. For example, of the 38 arthropod species that were habitat specific for either old or young forests, only one was a dolichopodid. Because strong concordances between Conservation Biology Volume 22, No. 6, 2008 1562 Tropical Arthropods and Disturbance taxa and related clear biological arguments to explain them are lacking, we believe correlation matrices mostly reflect idiosyncrasies related to the variety of arthropods examined (Lawton et al. 1998; Barlow et al. 2007). The concept of indicator species represents one of the perennial issues in conservation-oriented literature. Sim- ilar to other arthropod-oriented studies (Didham et al. 1996; Lawton et al. 1998), our results emphasize the fu- tility of using one taxon as an indicator for studies of an- thropogenic disturbance, despite the ubiquity of single- taxon studies in the literature. Different taxa have dif- ferent ecological requirements, and their responses to anthropogenic disturbance may therefore be different (Lawton et al. 1998), even at the congeneric level (Bas- set et al. 2001). This severely limits any generalization of studies derived from a selected group of taxa, both for biodiversity assessments and analyses of anthropogenic disturbance. One depressing option may be to admit that present resources will never be able to cope with the magnitude of effort needed to properly survey tropical habitats (Lawton et al. 1998) and, consequently, we will never be able to grasp the full extent and consequences of anthropogenic changes that are being induced on this planet. A more stimulating alternative may be to standardize the choice of focal taxa, because only patterns for match- ing taxa can be meaningfully compared across distur- bance gradients, and then use ?predictor sets,? which are composed of a small number of taxa representative of different functional groups (Kitching 1993; Didham et al. 1996; Basset et al. 2004b). Such predictor sets are properly selected only following statistical analysis of a larger, relatively complete, data set including all taxa and the catches from several complementary sampling methods. We suggest baseline surveys be run in different habitats, taxa selection be refined by analyzing species turnover between these habitats, and performing such surveys at a few tropical sites to determine how general patterns may be. For example, selection of Dolichopo- didae, Pipunculidae, Apoidea, Fulgoroidea, and Bupresti- dae, which recruit from 5 major guilds and account to- gether for 71% of significant concordances in species turnover, may represent a pertinent strategy to monitor the effects of anthropogenic disturbance on arthropods within the Gamba area. In the context of long-term man- agement of tropical forests, similar information could be used by conservationists to assess to what extent arthro- pod communities of secondary forests approach those of undisturbed forests as time progresses. Acknowledgments F. Dallmeier, J. Comiskey, M. Lee, J. Mavoungou, and J. B. Mikissa helped implement the project. Parataxonomists B. Amvame, N. Koumba, S. Mboumba Ditona, G. Mous- savou, P. Ngoma, J. Syssou, L. Tchignoumba, and E. Tobi collected, processed, sorted, and collated most of the in- sect material with great competence. The project was funded by the Smithsonian Institution, National Zoolog- ical Park, and Conservation and Research Center/MAB Program through grants from the Shell Foundation and Shell Gabon. This is contribution 104 of the Gabon Bio- diversity Program. Supporting Information Six tables are available as part of the on-line article (Ap- pendix S1). The author is responsible for the content and functionality of these materials. Queries (other than absence of the material) should be directed to the corre- sponding author. Literature Cited Alonso, A., M. E. Lee, P. Campbell, O. S. G. Pauwels, and F. Dallmeier, editors. 2006. Gamba, Gabon: Biodiversite? d?une fore?t e?quatoriale africaine/Gamba, Gabon: biodiversity of an equatorial African rain- forest. Bulletin of the Biological Society of Washington 12:1?436. Barlow, J., et al. 2007. Quantifying the biodiversity value of tropical pri- mary, secondary, and plantation forests. Proceedings of the National Academy of Sciences of the United States of America 104:18555? 18560. Basset, Y., E. Charles, D. S. Hammond, and V. K. Brown. 2001. Short- term effects of canopy openness on insect herbivores in a rain forest in Guyana. Journal of Applied Ecology 38:1045?1058. Basset, Y., J. F. Mavoungou, J. B. Mikissa, O. Missa, S. E. Miller, R. L. Kitching, and A. Alonso. 2004b. Discriminatory power of dif- ferent arthropod data sets for the biological monitoring of anthro- pogenic disturbance in tropical forests. Biodiversity and Conserva- tion 13:709?732. Basset, Y., et al. 2008. Choice of metrics for studying arthropod re- sponses to habitat disturbance: one example from Gabon. Insect Conservation and Diversity 1:55?66. Basset, Y., V. Novotny, S. E. Miller, and R. L. Kitching, editors. 2003. Arthropods of tropical forests. Spatio-temporal dynamics and re- source use in the canopy. Cambridge University Press, Cambridge, United Kingdom. Basset, Y., V. Novotny, S. E. Miller, G. D. Weiblen, O. Missa, and A. J. A. Stewart. 2004a. Conservation and biological monitoring of tropical forests: the role of parataxonomists. Journal of Applied Ecology 41:163?174. Brook, B. W., C. J. A. Bradshaw, L. P. Koh, and N. S. Sodhi. 2006. Mo- mentum drives the crash: mass extinction in the tropics. Biotropica 38:302?305. Chessel, D., A.-B. Dufour, and J. Thioulouse. 2004. The ade4 package- I?one-table methods. R News 4:5?10. Colwell, R. K. 2005. EstimateS: statistical estimation of species rich- ness and shared species from samples. Version 7.5. User?s Guide and application. University of Connecticut, Storrs. Available from http://purl.oclc.org/estimates (accessed March 2008). Couturier, G., and D. Duviard. 1976. Ethologie des peuplements de Dolichopodidae (Diptera) des savanes vierges et cultive?es de Co?te d?Ivoire Centrale. Cahiers de l? ORSTOM, se?rie Biologie 11:97? 113. Didham, R. K., J. Ghazoul, N. E. Stork, and A. J. Davis. 1996. Insects in fragmented forests: a functional approach. Trends in Ecology & Evolution 11:255?260. Conservation Biology Volume 22, No. 6, 2008 Basset et al. 1563 Dufre?ne M., and P. Legendre. 1997. Species assemblages and indicator species: the need for a flexible asymmetrical approach. Ecological Monographs 67:345?366. Garc??a, L. V. 2004. Escaping the Bonferroni iron claw in ecological studies. Oikos 105:657?663. Janzen, D. H.,W. Hallwachs, J. Jimenez, and R. Gamez. 1993. The role of the parataxonomists, inventory managers, and taxonomists in Costa Rica?s national biodiversity inventory. Pages 223?254 in W. V. Reid, S. A. Laird, C. A. Meyer, R. Gamez, A. Sittenfeld, D. H. Janzen, M. A. Gollin, and C. Juma, editors. Biodiversity prospecting: using generic resources for sustainable development. World Resources Institute, Washington, D.C. Kitching, R. L. 1993. Towards rapid biodiversity assessment?lessons following studies of arthropods of rainforest canopies. Pages 26?30 in A. J. Beattie, editor. Rapid biodiversity assessment. Research Unit for Biodiversity and Bioresources, Macquarie University, Sydney, Australia. Kotze, D. J., and M. J. Samways. 1999. Support for the multi-taxa ap- proach in biodiversity assessment, as shown by epigaeic inverte- brates in an Afromontane forest archipelago. Journal of Insect Con- servation 3:125?143. Kremen, C., R. K. Colwell, T. L. Erwin, D. D. Murphy, R. F. Noss, and M. A. Sanjayan. 1993. Terrestrial arthropod assemblages: their use in conservation planning. Conservation Biology 7:796? 808. Lawton, J. H., et al. 1998. Biodiversity inventories, indicator taxa and effects of habitat modification in tropical forest. Nature 391:72? 76. Lewis, O. T., and Y. Basset. 2007. Insect conservation in tropical forests. Pages 34?56 in A. J. A. Stewart, T. R. New, and O. T. Lewis, editors. Insect conservation biology. The Royal Entomological Society and CABI, Wallingford, United Kingdom. Longino, J. T., J. Coddington, and R. K. Colwell. 2002. The ant fauna of a tropical rain forest: estimating species richness three different ways. Ecology 83:689?702. McCune, B., and M. J. Medford. 1999. Multivariate analysis of ecological data version 4.10. MjM Software, Gleneden Beach, Oregon. Moran, C. V., and T. R. E. Southwood. 1982. The guild composition of arthropod communities in trees. Journal of Animal Ecology 51:289? 306. Oksanen, J., R. Kindt, P. Legendre, and B. O?Hara. 2006. The vegan package. Version 1.8-2. R Foundation for Statistical Computing, Wirtschaftuniversita?t Wien, Vienna. Available from http://www.R- project.org (accessed March 2008). Pearson, D. L., and F. Cassola. 1992. World-wide species richness pat- terns of tiger beetles (Coleoptera: Cicindelidae): indicator taxon for biodiversity and conservation studies. Conservation Biology 6:376? 391. Peres-Neto, P. R. 1999. How many statistical tests are too many? The problem of conducting multiple ecological inferences revisited. Ma- rine Ecology Progress Series 176:303?306. Peres-Neto, P. R., and D. A. Jackson. 2001. How well do multivariate data sets match? The advantages of a Procrustean superimposition approach over the Mantel test. Oecologia 129:169?178. Pineda, E., C. Moreno, F. Escobar, and G. Halffter. 2005. Frog, bat, and dung beetle diversity in the cloud forest and coffee agroecosystems of Veracruz, Mexico. Conservation Biology 19:400?410. Schulze, C. H., et al. 2004. Biodiversity indicator groups of tropical land-use systems: comparing plant, birds, and insects. Ecological Applications 14:1321?1333. Speight, M. R., J. Intachat, C. Vun Khen, and A. Y. C. Chung. 2003. Influ- ences of forest management on insects. Pages 380?393 in Y. Basset, V. 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. ter Braak, C. J. F., and P. Smilauer. 1998. CANOCO reference manual and user?s guide to Canoco for Windows: software for Canonical community ordination (version 4). Mirocomputer Power, Ithaca, New York. Tylianakis, J. M., T. Tscharntke, and O. T. Lewis. 2007. Habitat mod- ification alters the structure of tropical host-parasitoid food webs. Nature 455:202?205. Wright, S. J., and H. C. Muller-Landau. 2006. The future of tropical forest species. Biotropica 38:207?301. Conservation Biology Volume 22, No. 6, 2008