Arthropod Diversity in a Tropical Forest

See allHide authors and affiliations

Science  14 Dec 2012:
Vol. 338, Issue 6113, pp. 1481-1484
DOI: 10.1126/science.1226727


Most eukaryotic organisms are arthropods. Yet, their diversity in rich terrestrial ecosystems is still unknown. Here we produce tangible estimates of the total species richness of arthropods in a tropical rainforest. Using a comprehensive range of structured protocols, we sampled the phylogenetic breadth of arthropod taxa from the soil to the forest canopy in the San Lorenzo forest, Panama. We collected 6144 arthropod species from 0.48 hectare and extrapolated total species richness to larger areas on the basis of competing models. The whole 6000-hectare forest reserve most likely sustains 25,000 arthropod species. Notably, just 1 hectare of rainforest yields >60% of the arthropod biodiversity held in the wider landscape. Models based on plant diversity fitted the accumulated species richness of both herbivore and nonherbivore taxa exceptionally well. This lends credence to global estimates of arthropod biodiversity developed from plant models.

Most eukaryote species are terrestrial arthropods (1), and most terrestrial arthropods occur in tropical rainforests (2). However, considerably greater sampling effort is required in tropical arthropod surveys to yield realistic estimates of global species richness (37). A basic hindrance to estimating global biodiversity lies in a lack of empirical data that establish local biodiversity, which can be scaled up to achieve a global estimate. Although many studies reported species richness for selected groups of well-studied insect taxa, no satisfactory estimate of total arthropod species richness exists for a single tropical rainforest location to date.

The unstructured collection and small-scale survey of tropical arthropods cannot yield convincing estimates of total species richness at a specific forest (79). Most studies either target few arthropod orders or trophic guilds, or use a limited array of sampling methods, or ignore the diverse upper canopy regions of tropical forests (1015). Moreover, sampling protocols have rarely been structured in such a way that, with increased sampling, incomplete data on local diversity (7) can be extrapolated to estimate total species richness across multiple spatial scales (16). Where such structured estimates are made, it is invariably for insect herbivores on their host plants (5). However, species accumulation rates may differ markedly for nonherbivore guilds, which include more than half of all described arthropod species (1, 17). As the degree of host specificity (effective specialization) of other guilds can be much lower than that of insect herbivores, or may be driven by different factors (18, 19), global estimates based on herbivores alone are questionable. Consequently, extensive cross-taxon surveys with structured protocols at reference sites may be the only effective approach toward estimating total arthropod species richness in tropical forests (3).

To provide a comprehensive estimate of total arthropod species richness in a tropical rainforest, we established a collaboration involving 102 researchers with expertise encompassing the full breadth of phylogenies and feeding modes present among arthropods (20). This consortium invested a total of 24,354 trap- (or person-) days sampling the San Lorenzo forest (SLPA) in Panama using structured protocols (fig. S1). We identified 129,494 arthropods representing 6144 focal species (Fig. 1 and table S1) from 0.48 ha of intensively sampled mature forest. This allowed us to extrapolate focal arthropod species richness to a larger forest area with unprecedented power, through a series of best-informed species richness estimates derived from six competing models for each of 18 focal data sets. Using taxon ratios to estimate the species richness of nonfocal taxa [see “Extrapolating results to nonfocal taxa” in materials and methods (20)], we then predicted the total species richness of the study area. We also evaluated differences in relative species accumulation rates among arthropod guilds, across spatial scales.

Fig. 1

Number of arthropod species estimated at SLPA (20). (A) Number of species (closed bars, log scale) and individuals (open bars, log scale) collected in 0.48 ha for each data set (three-letter guild code as in table S1) and number of species estimated in SLPA (best estimate, shaded boxes). Numbers above bars identify the best model used for calculation (a1 to a6, fig. S2) and the percentage of singletons. (B) Number of arthropod species estimated for SLPA (dots: all focal taxa; shaded boxes: focal and nonfocal taxa; table S4), as estimated by different methods: B+S: best estimates, including both biological and statistical arguments (table S2); B+Sloc: same as B+S but estimates calculated with local instead of global ratios (fig. S3); S: best estimates, including only statistical arguments; and models a1 to a5. ?: optimization algorithms did not converge to allow calculations of CL. Our estimates are robust to even moderate to large shifts in taxon ratios (table S5). (C) Plot of the number of species estimated in SLPA with the plant model against that estimated with the best model, for each data set. Line denotes unity. (D) Plot of the percentage error between all arthropod species observed and estimated by the plant model against cumulative number of sites. Shaded boxes indicate means and 95% CL.

Although individual estimators adjusting for different aspects of sampling design offered slightly different estimates (Fig. 1B), the total species richness for the entire San Lorenzo forest (~6000 ha) was consistently quantified at between 18,000 and 44,000 species (including focal and nonfocal species). In particular, the most likely lower bound of species richness was estimated to be at least 21,833 species [95% confidence level (CL) = 18,665, 29,420; model a1 in Fig. 1B], and the biologically and statistically best-supported estimate of richness (criteria outlined in table S2) was 25,246 species (95% CL = 19,721, 33,181, model B+S in Fig. 1B). According to our estimates, a single hectare of rainforest will be inhabited by an average of 18,439 species (95% CL = 17,234, 18,575; Fig. 2B).

Fig. 2

Accumulation of species richness with area at SLPA (20). For all groups, a high proportion of overall species richness was detected at small spatial scales. (A) Partitioning of species richness within arthropod guilds at different spatial scales (α: single site of 0.04 ha; β3: three sites spaced apart totaling 0.12 ha; β6: six sites totaling 0.24 ha; β12: 12 sites totaling 0.48 ha; βha: 1.0 ha; βSLPA: 6000 ha; means ± SEM are shown for α, β3, and β6). (B) Species-area models for the main arthropod groups and large data sets (for all arthropods, including nonfocal species, values are indicated but the curve not plotted for the sake of clarity). Each curve is characterized by its function (Ex, Exponential; Lo, Lomolino; Po, Power; We, cumulative Weibull), its value for 1 ha (intersection with vertical line, shaded boxes with mean and 95% CL), and the percentage of the number of species present in 1 ha relative to the number of species estimated to occur in SLPA.

A relatively large proportion of the expected species richness of the forest was recovered for most of our focal taxonomic groups (Fig. 2). For example, high proportions of all ant species and of the parasitoid species targeted in our study were collected from our 12 sites, whereas fungal feeders would require more intensive sampling to achieve adequate coverage (Fig. 2). Beta diversity of all arthropods (in the broad sense of species turnover among sites) increased roughly linearly with cumulative area surveyed (F1,3 = 2422.5, P < 0.001). With increasing sampling effort, sample coverage [an unbiased measure of sample completeness, see (20)] was high and accumulated at significantly different rates across different arthropod orders and guilds, and across the various guilds comprised by beetles (Fig. 3). However, despite the high sample coverage values, we cannot discount the possibility that there were some vanishingly rare species that may not have been discovered with the sampling protocols used in this study.

Despite idiosyncrasies in the rate of increase in sample completeness across insects groups, the high proportion of overall species richness detected at small spatial scales (Figs. 2 and 3) has a remarkable consequence. Based on a general relationship between species numbers and area, we estimate that almost two-thirds (64%) of all species in SLPA occur in a single hectare of rainforest (Fig. 2). Our plant models predicted total arthropod richness in the San Lorenzo forest to a precision of 1% (correlation between richness estimates provided by the plant model and best estimates: r = 0.992, P < 0.0001, N = 18; Fig. 1C). Notably, small discrepancies between observed arthropod species richness and estimates derived from floristic diversity appeared not to be scale-dependent (Fig. 1D). Hence, even for arthropod guilds other than herbivores, plant diversity seems a powerful predictor of species richness across areas varying in size (at least within the limits of our study design and given the limited heterogeneity of the study area compared to larger geographical scales).

Fig. 3

Average sample coverage [±SEM; error bars, see methods (20)] plotted against the cumulative number of sites surveyed, for the main (A) arthropod guilds and orders and (B) beetle guilds. For the sake of clarity, SEMs are omitted in (A).

Because this study targeted the full spectrum of arthropods, it offers a comprehensive test of previous estimates of species richness based only on selected guilds or taxa. Reassuringly, our well-resolved estimates of tropical arthropod species richness are of the same order of magnitude as prior estimates (table S3), adding credence to recent estimates of tropical arthropod diversity (5, 21). Although the scope for direct comparison is limited because of regional differences in sampling effort, lowland tropical forest in Panama seems to support 2.1 to 8.4 times as many arthropod species as observed in temperate forests (table S3). While this supports the obvious truism that tropical arthropods are indeed more diverse than their temperate counterparts (22), the magnitude of that difference is far lower than many previous estimates would suggest (2).

The implications of the results observed on a local scale are clear. For every species in the well-known vascular flora (1294 species), avifauna (306 species), and mammalian fauna (81 species) of SLPA, we estimate that there will be a minimum of 17, 71, and 270 arthropod species, respectively (based on lower bound of species richness) and most likely as high as 20, 83, and 312 arthropod species, respectively (table S3). Based on the dominance of arthropod species in the tropical fauna (table S3), we may then argue that conservation planning for biodiversity should be largely determined by the spatial scaling of arthropod diversity. In this context, the association between the species richness of plants and arthropods detected across spatial scales suggests that conservation efforts targeted at floristically diverse sites may also serve to conserve arthropod diversity across both taxonomic lineages and trophic guilds. As arthropods are notoriously labor-intensive to survey, such an “umbrella” approach may be an efficient way forward.

Nonetheless, our findings also suggest that large-scale, region-wide understanding of tropical arthropod richness may actually be more achievable than previously assumed. Our data indicate that a thorough sampling of 1 ha of rainforest may reveal nearly two-thirds of all arthropod species present in a much larger area (6000 ha in our case; Fig. 2B), consistent with reports of relatively low beta diversity of insect herbivores in tropical rainforests (23). Hence, to determine the species diversity of a tropical rainforest, the total area sampled need not be overly large—provided that the sampling design adequately covers both microhabitats and plant species. However, this does not imply that most arthropod species have self-supporting populations in small forest areas or fragments.

On a global scale, our results have implications for current estimates of total species richness, which have been weakened by the lack of knowledge regarding the strength of association between vascular plant species and nonherbivore guilds (5). Based on the close association observed here between floristic diversity and both herbivore and nonherbivore species richness, we tentatively conclude that the most recent estimate of global tropical arthropod species [6.1 million arthropod species (24)] does not require drastic correction to account for differential scaling relationships of nonherbivore taxa. The robust estimates of local arthropod diversity derived in our study thus support previous estimates of global species richness. They also show how stratified sampling designs and broad scientific cooperation may be developed to formulate efficient estimates of tropical arthropod diversity. Similar initiatives have recently been implemented in other tropical locations around the world, using the current template as a foundation (25).

Supplementary Materials

Materials and Methods

Supplementary Text

Figs. S1 to S3

Tables S1 to S5

References (26151)

References and Notes

  1. Materials and methods are available as supplementary materials on Science Online.
  2. Acknowledgments: IBISCA-Panama is an initiative of Pro-Natura International, Océan Vert, the universities Blaise Pascal and of Panama, and the Smithsonian Tropical Research Institute (STRI), with core funding from SolVin-Solvay SA, STRI, the United Nations Environment Programme, the Smithsonian Institution (Walcott Fund), the European Science Foundation, and the Global Canopy Programme. J. Herrera, E. Andrade, M. Samaniego, S. J. Wright, N. Baiben, S. Bechet, J. Belleguic, T. Aubert, K. Jordan, G. Ebersolt, D. Cleyet-Marrel, L. Pyot, O. Pascal, P. Basset, and E. Bauhaus helped with logistics in the field. A. Barba, R. Cabrera, A. Cornejo, I. Díaz, A. F. R. do Carmo, I. C. do Nascimento, E. A. dos Santos, M. González, A. Hernandez, M. Manumbor, M. Mogia, S. Pinzón, B. Pérez, L. S. Ramos-Lacau, and O. Valdez helped with initial sorting of the arthropod and plant material. Data (as of 10 May 2012) have been deposited in the Dryad repository:
View Abstract

Navigate This Article