Ecological determinants
of tropical-temperate trends in insect diversity

Ecological determinants of tropical-temperate trends in insect diversity

Vojtech Novotny, Scott Miller, George Weiblen, Yves Basset, Pavel Drozd, Tom Fayle,
Masashi Murakami, Naoto Kamata, Martin Volf

The study examines one of the most fundamental, yet poorly understood patterns of global biodiversity distribution:
How can so many species coexist in a tropical forest?

This key question of current ecology will be studied using quantitative surveys of plant-herbivore-parasitoid food webs within paired sets of tropical and temperate forests from six continents, in Papua New Guinea (PNG), Gabon, Panama, the Czech Republic, Japan, and USA, sampled using canopy cranes, truck-mounted elevated platforms and forest felling.

This novel type of data will be analysed using a new rarefaction method, developed to test mechanistic explanations for biodiversity patterns along ecological gradients. It will evaluate competing hypotheses explaining latitudinal trends in insect herbivore diversity by the variation in either phylogenetic or functional diversity of plants, the host specificity of herbivores, or the diversity and specificity of their parasitoids and predators.

The study will thus examine the importance of bottom-up (plants) and top-down (enemies) drivers of latitudinal trends in herbivore food webs, central to ecological theory that postulates the role of specialized herbivores as density-dependent agents of mortality involved in maintaining high tropical plant diversity. The project will build a canopy crane in PNG and link researchers and infrastructure from several countries.


Why are there so many species in tropical forests? This may be one of the most obvious questions asked by biologists concerned with the global distribution of biodiversity. The simplicity of the question is deceptive given that the increase in diversity towards the tropics appears to be a complex result of evolutionary history and contemporary ecological interactions (Schemske et al. 2009, Agrawal et al. 2010).

The tropics appear to be characterized by both higher speciation and lower extinction rates than temperate regions (McKenna & Farrell 2006). Such evolutionary dynamics can generate larger regional species pools in the tropics than in the temperate zone. In insects, tropical and global numbers of species remain unknown but our research group estimated ~six million species worldwide (Hamilton et al. 2013).

May (1988) noted at the time that we did not know how many insect species lived in a 'representative hectare' of a tropical forest. A quarter of century later, we have the first estimates, provided by our research team: ~9,500 herbivore species feeding on 200 tree species and involved in ~50,000 trophic interactions between particular species in a New Guinea rainforest (Novotny et al. 2010), and a total diversity of ~25,000 arthropod species in a 6,000 ha Panamanian rainforest (Basset et al. 2012).

The large diversity of herbivores could be attributed to the distribution of plant resources (bottom-up effects) or the effects of natural enemies including predators, parasitoids and pathogens (top-down effects).Herbivores may in turn act as density-dependent mortality agents maintaining the diversity of plants (Leigh et al. 2004). Twenty years of continuous research has enabled us to assemble host records for >250,000 herbivorous insects including ~2,500 species and 11 feeding guilds in Papua New Guinea (PNG) (Novotny et al. 2010, 2012a). This data set, among the largest for tropical ecosystems and “likely to be as good as it gets” (Lewinsohn 2010), allowed us to challenge the widely held assumption that herbivore species coexistence in the tropics is a consequence of finely divided plant resources (Novotny et al. 2002b, 2012a).

We compared herbivores between sets of tropical and temperate tree species with equivalent phylogenetic diversity (Novotny et al. 2006) and found no difference in host specificity between folivores in C. Europe and New Guinea (Fig. 1). In response to this study, Dyer et al. (2007) let the phylogenetic diversity of their study plants vary with latitude and found higher host specificity in the American tropics. These findings stimulated a lively debate on the latitudinal trends in ecological specialization (Kitching 2006; Dyer et al. 2007; Stork 2007, Lewinsohn & Roslin 2008, Lewinsohn 2010) that remains presently unresolved.

We need to progress from the description of latitudinal trends to their explanation, based on quantitative analysis of food webs, rigorously comparable between tropical and temperate forests (Lewinsohn et al. 2005). We propose to obtain new data on plant-herbivore-parasitoid food webs for tropical and temperate forests and use a rarefaction method that we originally developed to examine diversity through forest succession (Klimes et al. 2012) as a new analytical tool to investigate mechanistic explanations for latitudinal patterns in the species diversity and specificity of herbivores. In particular, we will evaluate competing hypotheses that attribute elevated diversity in the tropics to such factors as latitudinal trends in plant species richness, phylogenetic and functional diversity, herbivore host specificity, and the diversity and host specificity of associated parasitoids.


Fig. 1. Host specificity of folivorous herbivores on temperate and tropical sets of 14 tree species standardized for phylogenetic diversity (Novotny et al. 2006).

Objectives and outline of research

We propose to survey plant-herbivore-parasitoid food webs including all woody plants >5 cm in diameter, insect herbivores from five folivorous guilds (semi-concealed and exposed chewers, miners, gallers, and mesophyll suckers) and parasitoids within 0.1 ha plots at paired tropical and temperate sites on six continents: in Panama vs. USA, Gabon vs. the Czech Republic, and Papua New Guinea vs. Japan, in order to address the following questions:

1. Explaining latitudinal trends in insect herbivore diversity. What is the role of latitudinal variation in the composition of vegetation?

We will test the null hypothesis that the rate of increase in herbivore species richness towards the tropics is driven by the composition, structure, and phylogenetic distribution of the vegetation (Fig. 2), expecting in particular that (i) plant species richness is a first-order predictor of herbivore diversity and affects the distribution of specialists more than generalists, (ii) plant phylogenetic diversity is associated with latitudinal gradients in herbivore diversity, as the generally smaller plant lineages comprising temperate vegetation host fewer herbivore species than the plant lineages of the tropics, and (iii) plant functional diversity of anti-herbivore defenses, which appears to be elevated in the tropics, explains residual variability in herbivore diversity beyond factors (i) and (ii).

Fig. 2. Hypothetical latitudinal trend (tropical to temperate) in herbivore species richness where specialists and generalists respond uniformly to latitude (A), generated entirely by trends in vegetation composition (B), leads to decreasing host specificity along latitude (C) but only on non-standardized vegetation (D). An alternative with latitudinal trends steeper in specialists than generalists (E), and only partly explained by vegetation (which is more important for specialists than generalists) (F), leads to a latitudinal trend in host specificity (G), only partly explained by vegetation (H).

2. Searching for latitudinal variation in the host specificity of herbivorous insects. Is there a trend?

A uniform response of generalists and specialists to latitudinal differences (Fig. 2A, B) implies no latitudinal trend in host specificity (Figs. 2C, D). An alternative hypothesis (Dyer et al. 2007) postulates a steeper latitudinal diversity gradient for specialists than for generalists (Fig. 2E, G). This explanation assumes that specialization facilitates coexistence by niche partitioning to account for elevated tropical diversity.

3. Extending to the next trophic level. What is the relative importance of vegetation and herbivorous hosts for latitudinal trends in the species diversity and host specificity of parasitoids?

The species richness trends of parasitoids may be generated entirely by latitudinal change in the species richness and phylogenetic diversity of their herbivorous hosts, which would imply no role of parasitoids in generating latitudinal trends in herbivorous communities. Alternately, there could be latitudinal differences in the number of parasitoid species per host species and potential influence of parasitoids on herbivores (Fig. 2).

4. Synthesis. Searching for latitudinal trends in the structure of tri-trophic food webs.

We will test the hypothesis that latitudinal differences in vegetation composition, species richness, phylogenetic diversity, functional diversity, and biomass distribution are associated with latitudinal trends in herbivore and parasitoid communities. Further, we will test the null hypothesis that linkage density, connectance, generality, vulnerability and network specialization in tri-trophic food webs do not vary between latitudes.

Research approach

Plant and insect sampling Plant-herbivore food webs will be sampled at six forests comprising three pairs of tropical-temperate contrasts from six continents We will obtain a complete census of plants with DBH ≥1 cm and their focal herbivore guilds and their parasitoids (Fig. 3) within 0.1 ha plots. A plant community phylogeny for each study plot will be estimated through a combination of our own sequencing with the existing phylogenies (Weiblen et al. 2006, Stevens 2012).

Fig. 3. Quantitative food webs for the focal guilds from 38 plant species in PNG.

Five focal guilds of folivores are targeted for sampling (Fig. 3), including exposed chewers (exposed holometabolous larvae), semi-concealed chewers (leaf rolling and tying holometabolous larvae), miners , gallers and mesophyll suckers (sucking individual mesophyll cells). The focal herbivore taxa will be, as far as possible, sampled from the entire leaf area within each 0.1 ha plot using forest felling, canopy cranes, or truck-mounted platforms (Fig. 4).

Fig. 4. Canopy crane at San Lorenzo (A, B), sampling from felled trees in Wanang (C), and a truck-mounted platform in Mikulcice (D, E).

Food webs will be characterized by quantitative metrics based on information theory including linkage density, connectance, generality, vulnerability and H2’ network specialization (Bersier et al. 2002). We will use DNA sequence clusters using the BIN algorithm implemented in BOLD (Ratnasingham & Hebert 2013) in our taxonomy. Species level phylogenies will then be constructed for the analysis of phylogenetic signal in plant-herbivore food webs focusing on Geometridae and Pyraloidea.

Analytical methods for plot-based food webs

Our approach will be based on the method we developed to associate insect diversity to components of plot-based vegetation data (Klimes et al. 2012). We will develop this algorithm into a general analytical tool, Trait-Based Rarefaction of Interaction Networks (TRIN) that could predict arthropod diversity according to plant traits, species composition, and community phylogenetic structure. At each step, replicate draws of individual trees, and their associated insects, from the source vegetation, such as tropical forest plot, are performed to match selected vegetation traits in the target vegetation, such as temperate forest plot, and the resulting herbivore communities from the sets of selected trees are analyzed (Fig. 5). This approach allows us to directly compare predicted plot-level insect communities based on different aspects of the vegetation with observed communities and thereby quantify the degree to which changes in insect diversity and food web structure are explained by differences in vegetation structure.

Fig. 5. TRIN uses repeated sampling from a large source vegetation plot (e.g., a tropical forest) to match increasing number of vegetation traits in a smaller target vegetation plot (e.g., a temperate zone forest). The three samples (comprising the highlighted trees) match the target vegetation in numbers of trees, the number and size of trees, and the number, size, and abundance of trees among species.

Research Team and Calendar

This is a highly collaborative project lead by the Biology Centre of the Czech Academy of Sciences (Vojtech Novotny lab), with partners from the Smithsonian Institution (Scott Miller), University of Minnesota (George Weiblen), Smithsonian Tropical Research Institute (Yves Basset), New Guinea Binatang Research Center (Pagi Toko), University of South Bohemia (Martin Volf), Ostrava University (Pavel Drozd), Chiba University (Masashi Murakami), Tokyo University (Naoto Kamata), and many others. The project will be active from October 2015 to September 2020.

Fig. 6. Study sites and collaborating institutions.

Broader Impacts

(i) We will examine ecological explanations for (a) high tropical biodiversity and (b) latitudinal gradients in diversity, which are fundamentally important for both theory and conservation, using novel analytical approaches applied to an extensive set of plot-based plant-insect food web data collected using uniform methods in tropical and temperate forests on six continents.

(ii) Our new rarefaction algorithm has potential for broader application in the analysis of ecological communities and food webs in spatially explicit and phylogenetic contexts.

(iii) The project links researchers and infrastructure among the Czech Republic, USA, Japan, Gabon, Panama and Papua New Guinea in a major collaborative effort.

(iv) The project will build a canopy crane in Wanang – a major facility for tropical rainforest research, which will, in combination with the largest teams of paraecologists in the tropics create a research centre for site rainforest ecology and conservation research.

(v) Our efforts in graduate student and paraecologist training in low income countries (the PNG program being among the most prominent in the tropics)will be strengthened. We will involve students and paraecologists in our field research, providing them with a unique opportunity to experience and contribute to collaborative international research.


Agrawal, A. A. et al. 2010. PNAS 106, 18067

Basset, Y. et al. 2012. Science 338, 1481

Bersier, L.-F. et al. 2002. Ecology 83, 2394

Craft, K. J. et al. 2010. PNAS 107, 5041

Dyer, L. A. et al. P. D. 2007. Nature 448, 696

Gentry, A. H. 2012 .

Hamilton, A. J. et al. 2013. Oecologia 171, 357

Kitching, R. L. 2006. Science 313, 1055

Klimes, P. et al. 2012. J. Anim. Ecol. 81, 1103

Leigh, E. G. J. et al. 2004. Biotropica 36, 447

Lewinsohn, T. 2010. J. Anim. Ecol. 79, 1143

Lewinsohn, T. M. et al. 2005. AREES 36 , 597

Lewinsohn, T. M. & Roslin, T. 2008. Ecol. Let. 11, 398

May, R. M. 1988. Science 241, 1441

McKenna, D. D. & Farrell, B. D. 2006. PNAS 103, 10947

Morris, R. J. et al. 2014. Ecol. Let. 17, 340

Novotny, V. 2010. Biotropica , 42, 546

Novotny, V. et al. 2002b. Nature 416, 841

Novotny, V. et al. 2002c. Proc. R. Soc., Lond., Biol. Sci. 269, 2337

Novotny, V. et al. 2006. Science 313 , 1115

Novotny, V. et al. 2010. J. Anim. Ecol. 79, 1193

Novotny, V. et al. 2012a. Am. Nat. 179, 351

Novotny, V. et al. 2007 Nature 448, 692

Novotny, V. et al. 2012b. Pp. 154-157 in M. D. Lowman et al. (eds.) Methods in forest canopy research. Univ. of California Press, Berkeley.

Ratnasingham, S. & Hebert, P. D. N. 2013. PloS ONE 8, e66213

Ricklefs, R. E. & Marquis, R. J. 2012. Oecologia 168, 213

Schemske, D. W. et al. 2009. AREES 40, 245

Shearman, P. L. et al. 2009. Biotropica 41 , 379

Simons, C. 2011. Science 332, 298

Stevens, P. F. 2012.

Stork, N. E. 2007. Nature 448, 657

Tvardikova, K. & Novotny, V. 2012. J. Trop. Ecol. 28, 331

Weiblen, G. D. et al. 2006. Ecology 87 , S62

This page as a pdf file: [Novotny_ERC_WebInfo.pdf]