When it comes to understanding patterns of biodiversity, ours is a little-known planet. Large-scale sampling projects, as carried out in two investigations of insect diversity, show a way forward.
To a first approximation, all multicellular species on Earth are insects1, and yet explanations for terrestrial biodiversity are largely based on birds, large mammals and plants. Studies of insect diversity by Novotny et al.2 and Dyer et al.3 (pages 692 and 696 of this issue) help to redress this imbalance, and provide an improved understanding of the distribution of global diversity.
Some 80–95% of insect species have yet to be collected, named and described, most of them living in the tropics. Even for the 850,000-plus species that have been named, we know little about how they are distributed or what they feed on4. Yet this information is essential for understanding the relationship between biodiversity and the functioning of global ecosystems. One reason is that a massive effort would be required to provide the field-based data for an analysis of patterns that might be applied generally at the global scale.
With the help of a team of locally trained parataxonomists, Novotny et al.2 have compiled such a database of records for three groups of rainforest insects: those that feed on foliage (Fig. 1), wood and fruit. They show that there is a low rate of change in species composition, or 'β diversity', across 75,000 km2 (an area equivalent to that of South Carolina or Ireland) of continuous lowland rainforest in Papua New Guinea. This contrasts with the previous evidence, as discussed by Novotny et al., of high β diversity for insects in the forest canopy and with changes in β diversity with latitude, altitude and climatic gradients.
Novotny et al.2 also show that insect species on host trees of the same genus, but separated by as much as 500 km, are remarkably similar, and that there do not seem to be barriers to their dispersal. The authors conclude that large, lowland areas of tropical forest, such as the Amazon and Congo, where there is low β diversity of vegetation, should also have low β diversity of insect herbivores.
In a previous paper, Novotny and colleagues5 had compared their Papua New Guinea database of feeding records for the caterpillars of moths and butterflies, adult beetles and adult grasshoppers with similar records for taxa in temperate regions of Europe. They controlled for the relatedness of host trees, and concluded that the insect herbivores show similar levels of host specificity in both climatic regions.
In the second new paper discussed here, Dyer et al.3 describe how they carried out an equivalent analysis in the New World and have come to a different conclusion. Their approach required examination of hundreds of thousands of host-specificity feeding records for butterfly and moth caterpillars, from as far back as 1936 and from areas ranging from Canada to Brazil. In contrast to Novotny and colleagues5, they find that, on average, the number of tree species on which an insect species feeds is fewer in the tropics than in temperate parts of the New World. They suggest that higher specialization in the tropics might be because of more intense interactions between an insect and its food source, as might be caused by more distinct secondary chemicals in tropical plants than in temperate plants.
Dyer et al.3 suggest that the difference between their results and those of Novotny et al.5 may be due to true biological differences between the continents, or because Novotny et al. used only 8–14 focal host-tree species in the study as opposed to the large number of host trees in the Dyer et al. study. Other reasons may be in the way Dyer and colleagues' data sets were compiled, particularly differences between the older and much larger Canadian data set and the smaller, more recent data sets, and in the considerable differences in the sample sizes in the temperate and tropical data sets. Dyer and colleagues also suggest that there may be real differences in host specificity between the Americas, Europe and tropical Asia, but this seems unlikely. The question of which of these contrasting conclusions is correct will remain unresolved until further comparative studies take these sampling and geographical issues into account.
There has been an understandable bias towards the herbivorous insects in ecological studies6, because insects have coevolved with the plants and trees on which they feed. Indeed, tree species richness may serve as the best proxy for overall biodiversity in tropical forests, as Terry Erwin inferred in his famous calculation7 that raised estimates of tropical insect species tenfold to 30 million. Crucial suppositions he made were that each of the 50,000 tree species or groups of species in the world would have 165 host-specific beetle species, that beetles represent 40% of all insect species, and that the canopy is twice as rich in insect species as the ground, with the inference that species are stratum specific. His calculation implied that 84% of tropical insects are herbivores. The number of insect species that are specific to a particular tree species has since been carefully re-examined, however, and reduced by a factor of four or five8.
But what of the insects that have less glamorous and obvious lifestyles than the herbivores: those that feed on dead and decaying material, or on the bacteria and fungi that break down organic material; or the predators and parasites that feed on living plants and animals? The proportion of insect biodiversity that these 'feeding guilds' comprise is uncertain, but could be as high as 50–70%, and not 16% as Erwin proposed.
Looking beyond insects and setting aside microorganisms, what about fungi, other invertebrates and most marine life? These groups, too, are often poorly understood because of their taxonomic intractability or because they are so infrequently collected. The apparent rarity of many species in most samples of invertebrates and fungi is probably due to our low level of sampling rather than representing biological rarity. Making sense of such communities is almost impossible without the scale of sampling shown by Novotny and Dyer and their teams. Answers to such fundamental questions as how many species there are, how they are distributed, and how many are being lost through extinction will remain elusive without similar collaborative and large-scale enterprises. Of course, documenting how communities of organisms and their interactions change along ecological gradients is fundamentally more important than merely counting species.
So how much nearer are we to a model or group of models that might predict and explain the distribution of biodiversity on a global or even a regional scale? Roger Kitching9 talked about “crafting the pieces of the diversity jigsaw puzzle”, and these two new papers2,3 help to identify a few more pieces of this puzzle. But we are still a long way from being able to explain the distribution of global biodiversity. Perhaps the nearest functional model is the mid-domain theory10,11, which attempts to model the distribution of species and shows that species richness is greatest at the centre of a spatial, temporal or functional domain. But whether that theory can be expanded and modified remains to be seen.