insight review articles Getting the measure of biodiversity Andy Purvis*&Andy Hectort *Department of Biology and+NERC Centre for Population Biology,Imperial College,Silwood Park,Ascot,Berkshire SL5 7PY,UK The term 'biodiversity'is a simple contraction of'biological diversity',and at first sight the concept is simple too:biodiversity is the sum total of all biotic variation from the level of genes to ecosystems.The challenge comes in measuring such a broad concept in ways that are useful.We show that,although biodiversity can never be fully captured by a single number,study of particular facets has led to rapid,exciting and sometimes alarming discoveries.Phylogenetic and temporal analyses are shedding light on the ecological and evolutionary processes that have shaped current biodiversity.There is no doubt that humans are now destroying this diversity at an alarming rate.A vital question now being tackled is how badly this loss affects ecosystem functioning.Although current research efforts are impressive,they are tiny in comparison to the amount of unknown diversity and the urgency and importance of the task. o proceed very far with the study of biodiversity, and no single number can incorporate them both without we need to pin the concept down.We cannot loss of information.This should not be disappointing; even begin to look at how biodiversity is indeed we should probably be relieved that the variety distributed,or how fast it is disappearing, of life cannot be expressed along a single dimension. unless we can put units on it.However,any Rather,different facets of biodiversity can each be attempt to measure biodiversity quickly runs into the quantified (Box 1). problem that it is a fundamentally multidimensional Knowing the diversity(however measured)ofone place, concept:it cannot be reduced sensibly to a single group or time is in itselfmore-or-less useless.But,as we shall number2.A simple illustration can show this.Figure 1 discuss later,comparable measurements of diversity from shows samples from the insect fauna in each of two multiple places,groups or times can help us to answer habitats.Which sample is more diverse?At first sight it crucial questions about how the diversity arose and how we must be sample A,because it contains three species to might best act to maintain it.We shall see also how the sample B's two.But sample B is more diverse in that there usefulness of the answers depends critically on the selection is less chance in sample B that two randomly chosen of an appropriate diversity measure.No single measure will individuals will be of the same species.Neither of these always be appropriate (indeed,for some conservation ques- measures of diversity is 'wrong'-species richness and tions,no single measure can probably ever be appropriate). evenness are two(among many)of biodiversity's facets, The choice ofa good measure is complicated by the frequent Sample A Sample B Figure 1 Two samples of insects from different locations,illustrating two of the many different measures of diversity:species richness and species evenness Sample A could be described as being the more diverse as it contains three species to sample B's two.But there is less chance in sample B than in sample A that two randomly chosen individuals will be of the same species. 212 2000 Macmillan Magazines Ltd NATURE VOL 405|11 MAY 2000 www.nature.com
insight review articles 212 NATURE | VOL 405 | 11 MAY 2000 | www.nature.com T o proceed very far with the study of biodiversity, we need to pin the concept down. We cannot even begin to look at how biodiversity is distributed, or how fast it is disappearing, unless we can put units on it. However, any attempt to measure biodiversity quickly runs into the problem that it is a fundamentally multidimensional concept: it cannot be reduced sensibly to a single number1,2. A simple illustration can show this. Figure 1 shows samples from the insect fauna in each of two habitats. Which sample is more diverse? At first sight it must be sample A, because it contains three species to sample B’s two. But sample B is more diverse in that there is less chance in sample B that two randomly chosen individuals will be of the same species. Neither of these measures of diversity is ‘wrong’ — species richness and evenness are two (among many) of biodiversity’s facets, and no single number can incorporate them both without loss of information. This should not be disappointing; indeed we should probably be relieved that the variety of life cannot be expressed along a single dimension. Rather, different facets of biodiversity can each be quantified (Box 1). Knowing the diversity (however measured) of one place, group or time is in itself more-or-less useless. But, as we shall discuss later, comparable measurements of diversity from multiple places, groups or times can help us to answer crucial questions about how the diversity arose and how we might best act to maintain it. We shall see also how the usefulness of the answers depends critically on the selection of an appropriate diversity measure. No single measure will always be appropriate (indeed, for some conservation questions, no single measure can probably ever be appropriate). The choice of a good measure is complicated by the frequent Getting the measure of biodiversity Andy Purvis* & Andy Hector† *Department of Biology and †NERC Centre for Population Biology, Imperial College, Silwood Park, Ascot, Berkshire SL5 7PY, UK The term ‘biodiversity’ is a simple contraction of ‘biological diversity’, and at first sight the concept is simple too: biodiversity is the sum total of all biotic variation from the level of genes to ecosystems. The challenge comes in measuring such a broad concept in ways that are useful. We show that, although biodiversity can never be fully captured by a single number, study of particular facets has led to rapid, exciting and sometimes alarming discoveries. Phylogenetic and temporal analyses are shedding light on the ecological and evolutionary processes that have shaped current biodiversity. There is no doubt that humans are now destroying this diversity at an alarming rate. A vital question now being tackled is how badly this loss affects ecosystem functioning. Although current research efforts are impressive, they are tiny in comparison to the amount of unknown diversity and the urgency and importance of the task. Figure 1 Two samples of insects from different locations, illustrating two of the many different measures of diversity: species richness and species evenness. Sample A could be described as being the more diverse as it contains three species to sample B’s two. But there is less chance in sample B than in sample A that two randomly chosen individuals will be of the same species. Sample A Sample B © 2000 Macmillan Magazines Ltd
insight review articles Box 1 Parts of the whole:numbers,evenness and difference Biodiversity has a multitude of facets that can be quantified.Here we Evenness classify some commonly used measures into three conceptually A site containing a thousand species might not seem particularly different(although not orthogonal)approaches. diverse if 99.9%of individuals that you find belong in the same species.Many diversity indices have been developed to convey the Numbers extent to which individuals are distributed evenly among species". The most commonly considered facet of biodiversity is species Most but not all combine evenness with species richness,losing richness-the number of species in a site,habitat or clade.Species information by reducing two dimensions to one.There are genetic are an obvious choice of unit when trying to measure diversity.Most analogues of these indices,such as heterozygosity,that people have an idea what'species'means and,although their ideas incorporate both allele number and relative frequencies. differ considerably (reviewed in ref.96),there is even less commonality about other levels in the taxonomic hierarchy(Fig.3). Difference Many other measures are less intuitive,and have arisen only through Some pairs of species(or alleles or populations)are very alike,whereas appreciation of limitations of measures of species richness.Species others are very different.Disparityt0 and character diversity are are also sensible units to choose from a biological perspective:they measures of phenotypic difference among the species in a sample,and keep their genes more or less to themselves,and to that extent have can be made independent of species number.Some phenotypic independent evolutionary trajectories and unique histories.The characteristics might be considered more important than others,for current 'best guess"7 is that there are around 14 million species,but instance the ecological diversity among species may be crucial for this is very much a provisional working figure.Regions with many ecosystem functioning.Genetic variability among populations can also species,especially endemic species,are sometimes called be measured in various ways.If populations within species differ hotspots7 enough either genetically or phenotypically,they may be considered to Species and regions differ in their number of populations. be subspecies,management units or evolutionarly significant units Populations of a given species,if defined on the basis of limited gene numbers of these therefore provide estimates of difference.All these flow among them,will evolve to an extent independently.Each kinds of difference are likely to be at least partly reflected by the population contributes additional diversity.The number of genetic phylogenetic diversity among organisms,which is estimated as the populations in the world has been estimated to lie between 1.1 and 6.6 sum total of the branch lengths in the phylogeny (evolutionary tree)linking billion them. Species or populations differ in the numbers of alleles they have at Sample in different places,and you will find different things.This given loci.For instance,Mauritius kestrels (Falco punctatus)have lost spatial turnover itself has many facets"(for example,beta diversity, over half of the alleles present historically at 12 sampled microsatellite gamma diversity and numbers of habitat types),and important locie consequences for any attempt to conserve overall diversity(see Moving above the species level,higher-taxon richness is often review by Margules and Pressey,pages 243-253,and refs 104,105). used in studies of biodiversity,usually as a less data-demanding Likewise,temporal turnovert is the extent to which what is found surrogate for species richness changes over time. need to use surrogates for the aspect in which we are most interest- below the surface;these subsurface lithoautotrophic microbial ed.Surrogacy is a pragmatic response to the frightening ignorance ecosystems(termed SLiMEs)may have persisted for millions ofyears about what is out there.Some recent discoveries highlight just how without any carbon from the surface.Controversy surrounds much we probably still do not know. another proposed discovery:whether or not the 100-nm-diameter nanobacteria found in,among other places,kidney stones are living The growing biosphere organisms".At an even smaller scale,genomes provide fossils Technological advances and the sense ofurgency imparted by the rate that indicate great past retroviral diversity 2.Genomes have also ofhabitat loss are combining to yield discoveries at an incredible rate. been found to provide habitats for many kinds of genetic entity- This may seem surprising,given that expedition accounts of natural transposable elements-that can move around and replicate historians from the 18th and 19th centuries conjure up images ofdis- themselves.Such elements can provide important genetic variation covery on a grand scale that seemingly cannot be matched today- to their hosts,can makeup more than halfofthe host's genome,and look in the rocks...a new fossil mammal;look in the lake...a new have life histories of their own'4. fish genus;look on the dinner plate...a new species of bird.Finding There are two other ways in which the biosphere can perhaps be new large vertebrates nowadays is indeed newsworthy,but a new said to be growing.The first is that the rate at which taxonomists split species oflarge mammal is still discovered roughly every three years one previously recognized species into two or more exceeds the rate and a new large vertebrate from the open ocean every five years.And at which they lump different species together,especially in taxa that most organisms are much smaller than these are.An average day sees are of particular concern to conservationists (for example, the formal description of around 300 new species across the whole platyrrhine primates5).Part of the reason is the growing popularity range oflife,and there is no slowdown in sight.Based on rates of dis- of one way of delimiting species-the phylogenetic species concept covery and geographical scaling-up,it seems that the roughly 1.75 (PSC)- under which taxa are separate species if they can be diag- million described species oforganismmaybeonly around 10%ofthe nosed as distinct,whether on the basis of phenotype or genotype.If total'. the PSCbecomes widelyapplied-which is acontroversial issue It is not only new species that are discovered.Cycliophora and then the numbers of'species'in many groups are sure to increase Loricifera are animal phyla(the level just below kingdom in the taxo- greatly(although theamount ofdisparity willbarely increaseatall). nomic hierarchy)that are new to science in the past 20 years.Within Asecond way in which the catalogue of diversity is growing is that the Archaea,the discovery of new phylum-level groups proceeds at computer databases and the Internet are making the process ofinfor- the rate of more than one a month.The physical limits of the bios- mation gathering more truly cumulative than perhaps ever before. phere have been pushed back by the recent discovery of microbial Some existing sites serve to provide examples of the information communities in sedimentary and even igneous rocks over 2 km already available:not just species lists(http://www.sp2000.org/),but NATURE|VOL 40511 MAY 2000www.nature.com 2000 Macmillan Magazines Ltd 213
need to use surrogates for the aspect in which we are most interested3,4. Surrogacy is a pragmatic response to the frightening ignorance about what is out there. Some recent discoveries highlight just how much we probably still do not know. The growing biosphere Technological advances and the sense of urgency imparted by the rate of habitat loss are combining to yield discoveries at an incredible rate. This may seem surprising, given that expedition accounts of natural historians from the 18th and 19th centuries conjure up images of discovery on a grand scale that seemingly cannot be matched today — look in the rocks … a new fossil mammal; look in the lake … a new fish genus; look on the dinner plate … a new species of bird. Finding new large vertebrates nowadays is indeed newsworthy, but a new species of large mammal is still discovered roughly every three years5 and a new large vertebrate from the open ocean every five years6 . And most organisms are much smaller than these are. An average day sees the formal description of around 300 new species across the whole range of life, and there is no slowdown in sight. Based on rates of discovery and geographical scaling-up, it seems that the roughly 1.75 million described species of organism may be only around 10% of the total7 . It is not only new species that are discovered. Cycliophora and Loricifera are animal phyla (the level just below kingdom in the taxonomic hierarchy) that are new to science in the past 20 years8 . Within the Archaea, the discovery of new phylum-level groups proceeds at the rate of more than one a month9 . The physical limits of the biosphere have been pushed back by the recent discovery of microbial communities in sedimentary and even igneous rocks over 2 km below the surface; these subsurface lithoautotrophic microbial ecosystems (termed SLiMEs) may have persisted for millions of years without any carbon from the surface10. Controversy surrounds another proposed discovery: whether or not the 100-nm-diameter nanobacteria found in, among other places, kidney stones are living organisms11. At an even smaller scale, genomes provide fossils that indicate great past retroviral diversity12. Genomes have also been found to provide habitats for many kinds of genetic entity — transposable elements — that can move around and replicate themselves. Such elements can provide important genetic variation to their hosts, can make up more than half of the host’s genome13, and have life histories of their own14. There are two other ways in which the biosphere can perhaps be said to be growing. The first is that the rate at which taxonomists split one previously recognized species into two or more exceeds the rate at which they lump different species together, especially in taxa that are of particular concern to conservationists (for example, platyrrhine primates15). Part of the reason is the growing popularity of one way of delimiting species — the phylogenetic species concept (PSC)16 — under which taxa are separate species if they can be diagnosed as distinct, whether on the basis of phenotype or genotype. If the PSC becomes widely applied — which is a controversial issue17— then the numbers of ‘species’ in many groups are sure to increase greatly18 (although the amount of disparity will barely increase at all). A second way in which the catalogue of diversity is growing is that computer databases and the Internet are making the process of information gathering more truly cumulative than perhaps ever before. Some existing sites serve to provide examples of the information already available: not just species lists (http://www.sp2000.org/), but insight review articles NATURE | VOL 405 | 11 MAY 2000 | www.nature.com 213 Biodiversity has a multitude of facets that can be quantified. Here we classify some commonly used measures into three conceptually different (although not orthogonal) approaches. Numbers The most commonly considered facet of biodiversity is species richness — the number of species in a site, habitat or clade. Species are an obvious choice of unit when trying to measure diversity. Most people have an idea what ‘species’ means and, although their ideas differ considerably (reviewed in ref. 96), there is even less commonality about other levels in the taxonomic hierarchy30 (Fig. 3). Many other measures are less intuitive, and have arisen only through appreciation of limitations of measures of species richness. Species are also sensible units to choose from a biological perspective: they keep their genes more or less to themselves, and to that extent have independent evolutionary trajectories and unique histories. The current ‘best guess’7 is that there are around 14 million species, but this is very much a provisional working figure. Regions with many species, especially endemic species, are sometimes called hotspots97. Species and regions differ in their number of populations. Populations of a given species, if defined on the basis of limited gene flow among them, will evolve to an extent independently. Each population contributes additional diversity. The number of genetic populations in the world has been estimated to lie between 1.1 and 6.6 billion66. Species or populations differ in the numbers of alleles they have at given loci. For instance, Mauritius kestrels (Falco punctatus) have lost over half of the alleles present historically at 12 sampled microsatellite loci98. Moving above the species level, higher-taxon richness is often used in studies of biodiversity, usually as a less data-demanding surrogate for species richness99. Evenness A site containing a thousand species might not seem particularly diverse if 99.9% of individuals that you find belong in the same species. Many diversity indices have been developed to convey the extent to which individuals are distributed evenly among species2 . Most but not all combine evenness with species richness, losing information by reducing two dimensions to one. There are genetic analogues of these indices100, such as heterozygosity, that incorporate both allele number and relative frequencies. Difference Some pairs of species (or alleles or populations) are very alike, whereas others are very different. Disparity101 and character diversity93 are measures of phenotypic difference among the species in a sample, and can be made independent of species number. Some phenotypic characteristics might be considered more important than others, for instance the ecological diversity among species may be crucial for ecosystem functioning. Genetic variability among populations can also be measured in various ways100. If populations within species differ enough either genetically or phenotypically, they may be considered to be subspecies, management units or evolutionarily significant units102; numbers of these therefore provide estimates of difference. All these kinds of difference are likely to be at least partly reflected by the phylogenetic diversity103 among organisms, which is estimated as the sum total of the branch lengths in the phylogeny (evolutionary tree) linking them. Sample in different places, and you will find different things. This spatial turnover itself has many facets2 (for example, beta diversity, gamma diversity and numbers of habitat types), and important consequences for any attempt to conserve overall diversity (see review by Margules and Pressey, pages 243–253, and refs 104, 105). Likewise, temporal turnover106 is the extent to which what is found changes over time. Box 1 Parts of the whole: numbers, evenness and difference © 2000 Macmillan Magazines Ltd
insight review articles known fossils.The palaeontological record indicates a Cambrian explosion of phyla around 540 million years (Myr)ago,but sequences suggest a more gradual series of splits around twice as old2 Likewise,many orders of mammals and birds are now thought to have originated long before the end-Cretaceous extinction25 which occurred 65 Myr ago and which was thought previously to have been the signal for their radiation.If the new timescale can be trusted,these findings present a puzzle and a warning.The puzzle is the absence of fossils.Why have we not found traces of these lineages in their first tens or even hundreds ofmillions ofyears?It seems likely Hominida that the animals were too small or too rare,with the sudden appear- ance in the rocks corresponding to an increase in size and rise to ecological dominance".The warning is that current biodiversity is in a sense greater than we had realized.Major lineages alive today represent more unique evolutionary history than previously suspected-history that would be lost with their extinction. Old-World monkeys Analysis of the shape of phylogenies has shown that lineages have New-World monkeys differed in their potential for diversification.Darwin had noted that species in species-rich genera had more subspecific varieties,and subtaxa within taxa are often distributed very unevenly29,as Fig.2 illustrates for eutherian species.But these taxonomic patterns can be taken at face value only iftaxa are comparable,which they may not be. For example,species-rich groups may simply be older,and it is clear that workers on different groups currently place taxonomic bound- aries in very different places(Fig.3).Phylogenies allow comparison of sister clades-each other's closest relatives-which by definition are the same age.Time and again,species are distributed too uneven- ly for simple null models to be tested in which all species have the same chances ofdiversifying genus Scaptomyza What are the species-rich groups 'doing right'?Many explana- tions fall broadly into two types.Key innovation hypotheses"posit 40 30 20 10 the evolution of some trait that permits its bearers to gain access to Millions of years ago more resources or be more competitive than non-bearers.Examples include phytophagy in insects"and high reproductive rate in mam- malss.Other hypotheses focus on traits that facilitate the evolution Figure 2 Taxonomic boundaries are not comparable among major groups.a,Fourteen of reproductive isolation-speciation-without necessarily species in nine genera representative of cichlid fish in Lake Victoria.b,Seven species increasing the fitness of bearers.Sexual selection%and range representative of several families in anthropoid primates.c.Thirteen species fragmentation"are examples of this kind.These two types can be representative of a single genus,Drosophila.Figure reproduced from ref 30,with contrasted as 'bigger cake'and 'thinner slices'explanations, permission. although some traits may act in both ways(for example,body size);another way to split them is to view diversity as'demand- driven'(niches are waiting to be filled,and differentiation leads to also maps of the geographical ranges of species (http://www. speciation)or 'supply-driven'(speciation occurs unbidden,with gisbau.uniromal.it/amd/homepage.html),information on conser- differentiation arising through character displacement).Statistical vation status of species (http://www.wcmc.org.uk),bibliographies testing of many key innovation hypotheses is hampered by a lack of (http://eteweb.lscf.ucsb.edu/bfv/bfv_form.html),data on molecular replication-often,the trait in question is unique,and all that can sequence (http://www.ebi.ac.uk/and http://www.ncbi.nlm.nih.gov/ be done is to model the trait's evolution to assess how well it fits the Genbank/GenbankOverview.html),data on phylogenetic position scenario.When characters have evolved multiple times in (http://phylogeny.arizona.edu/tree/phylogeny.html and http:// independent lineages,sister clades provide automatic matched pairs herbaria.harvard.edu/treebase/),information on the stratigraphic for hypothesis testing(although other phylogenetic approaches are range of species (http://ibs.uel.ac.uk/ibs/palaeo/benton/and also available4).Comparing sister clades(the procedure used in http://www.nceas.ucsb.edu/-alroy/nafmtd.html)and much more. most of the examples above)avoids two problems that otherwise Although the terabytes of information already stored constitute only cloud the issue.First,taxa may not be comparable (Fig.3),and a small drop in the ocean,the next two sections show how much can second,they are not statistically independent-related clades be seen in that droplet about the distribution of biodiversity among inherit their traits from common ancestors,so are pseudorepli- evolutionary lineages and through time. cates.Nonetheless,there is ongoing debate about the role and limitations of phylogenetic tests for correlates of species Learning from the tree of life richness The ongoing explosion ofphylogenetic studies not only provides an ever-clearer snapshot ofbiodiversity today,but also allows us to make Temporal patterns in biodiversity inferences about how the diversity has come about2.(For an Is biodiversity typically at some equilibrium level,with competi- ecological perspective,see review by Gaston,pages 220-227.)Phylo- tion setting an upper limit,or do mass extinctions occur so regular- genies give key information that is not available from species lists or ly that equilibrium is never reached?And,with one eye on the taxonomies.They detail the pattern of nested relationships among future prospects for biodiversity,how quickly does diversity species,and increasingly provide at least a rough timescale even recover from mass extinctions?Palaeontologists have addressed without reliance on a molecular clock2.These new phylogenies are these questions at many scales,from local to global.For the global pushing back the origins of many groups to long before their earliest view,the data come from huge compendia ofstratigraphic ranges of 214 ☆©20o0 Macmillan Magazines Ltd NATURE VOL 40511 MAY 2000 www.nature.com
also maps of the geographical ranges of species (http://www. gisbau.uniroma1.it/amd/homepage.html), information on conservation status of species (http://www.wcmc.org.uk), bibliographies (http://eteweb.lscf.ucsb.edu/bfv/bfv_form.html), data on molecular sequence (http://www.ebi.ac.uk/ and http://www.ncbi.nlm.nih.gov/ Genbank/GenbankOverview.html), data on phylogenetic position (http://phylogeny.arizona.edu/tree/phylogeny.html and http:// herbaria.harvard.edu/treebase/), information on the stratigraphic range of species (http://ibs.uel.ac.uk/ibs/palaeo/benton/ and http://www.nceas.ucsb.edu/~alroy/nafmtd.html) and much more. Although the terabytes of information already stored constitute only a small drop in the ocean, the next two sections show how much can be seen in that droplet about the distribution of biodiversity among evolutionary lineages and through time. Learning from the tree of life The ongoing explosion of phylogenetic studies not only provides an ever-clearer snapshot of biodiversity today, but also allows us to make inferences about how the diversity has come about19–21. (For an ecological perspective, see review by Gaston, pages 220–227.) Phylogenies give key information that is not available from species lists or taxonomies. They detail the pattern of nested relationships among species, and increasingly provide at least a rough timescale even without reliance on a molecular clock22. These new phylogenies are pushing back the origins of many groups to long before their earliest known fossils. The palaeontological record indicates a Cambrian explosion of phyla around 540 million years (Myr) ago, but sequences suggest a more gradual series of splits around twice as old23. Likewise, many orders of mammals and birds are now thought to have originated long before the end-Cretaceous extinction24,25, which occurred 65 Myr ago and which was thought previously to have been the signal for their radiation. If the new timescale can be trusted26, these findings present a puzzle and a warning. The puzzle is the absence of fossils. Why have we not found traces of these lineages in their first tens or even hundreds of millions of years? It seems likely that the animals were too small or too rare, with the sudden appearance in the rocks corresponding to an increase in size and rise to ecological dominance27. The warning is that current biodiversity is in a sense greater than we had realized. Major lineages alive today represent more unique evolutionary history than previously suspected — history that would be lost with their extinction. Analysis of the shape of phylogenies has shown that lineages have differed in their potential for diversification. Darwin28had noted that species in species-rich genera had more subspecific varieties, and subtaxa within taxa are often distributed very unevenly29, as Fig. 2 illustrates for eutherian species. But these taxonomic patterns can be taken at face value only if taxa are comparable, which they may not be. For example, species-rich groups may simply be older, and it is clear that workers on different groups currently place taxonomic boundaries in very different places30 (Fig. 3). Phylogenies allow comparison of sister clades — each other’s closest relatives — which by definition are the same age. Time and again, species are distributed too unevenly for simple null models to be tested in which all species have the same chances of diversifying31,32. What are the species-rich groups ‘doing right’? Many explanations fall broadly into two types. Key innovation hypotheses33 posit the evolution of some trait that permits its bearers to gain access to more resources or be more competitive than non-bearers. Examples include phytophagy in insects34 and high reproductive rate in mammals35. Other hypotheses focus on traits that facilitate the evolution of reproductive isolation — speciation — without necessarily increasing the fitness of bearers. Sexual selection36 and range fragmentation37 are examples of this kind. These two types can be contrasted as ‘bigger cake’ and ‘thinner slices’ explanations, although some traits may act in both ways (for example, body size38,39); another way to split them is to view diversity as ‘demanddriven’ (niches are waiting to be filled, and differentiation leads to speciation) or ‘supply-driven’ (speciation occurs unbidden, with differentiation arising through character displacement). Statistical testing of many key innovation hypotheses is hampered by a lack of replication — often, the trait in question is unique, and all that can be done is to model the trait’s evolution to assess how well it fits the scenario40. When characters have evolved multiple times in independent lineages, sister clades provide automatic matched pairs for hypothesis testing (although other phylogenetic approaches are also available41,42). Comparing sister clades (the procedure used in most of the examples above) avoids two problems that otherwise cloud the issue. First, taxa may not be comparable (Fig. 3), and second, they are not statistically independent — related clades inherit their traits from common ancestors, so are pseudoreplicates43. Nonetheless, there is ongoing debate about the role and limitations of phylogenetic tests for correlates of species richness44,45. Temporal patterns in biodiversity Is biodiversity typically at some equilibrium level, with competition setting an upper limit, or do mass extinctions occur so regularly that equilibrium is never reached? And, with one eye on the future prospects for biodiversity, how quickly does diversity recover from mass extinctions? Palaeontologists have addressed these questions at many scales, from local to global. For the global view, the data come from huge compendia of stratigraphic ranges of insight review articles 214 NATURE | VOL 405 | 11 MAY 2000 | www.nature.com Figure 2 Taxonomic boundaries are not comparable among major groups. a, Fourteen species in nine genera representative of cichlid fish in Lake Victoria. b, Seven species representative of several families in anthropoid primates. c, Thirteen species representative of a single genus, Drosophila. Figure reproduced from ref 30, with permission. Hominidae Pongidae Old-World monkeys New-World monkeys melanogaster subgroup genus Scaptomyza 40 30 20 10 0 Millions of years ago a b c © 2000 Macmillan Magazines Ltd
insight review articles taxonomic families (see,for example,refs 46,47),led by Sepkoski's highlights the difficulties of taking taxonomic patterns at face ground-breaking efforts,and made possible by the development of value.Neontologists may face much the same problem with species: computer databases.There are more families now than ever before, taxonomists tend to recognize bird lineages as species if they are and a model ofexponential growth provides a good overall fit to the older than 2.8 Myr but not ifthey are younger than 1.1 Myr(ref.52), numbers of families through time,suggesting expansion without so apparent logistic growth in species numbers through time limit and no major role for competition in limiting diversitys.But a within bird generas might be expected even without a slow-down significantly better fit is provided by a set of three logistic curves, ofcladogenesis. each with a different carrying capacity,punctuated by mass extinc- The patchy nature of the known fossil record means that some tion events.Leaving aside the thorny issue of multiplicity of tests taxa in some places at some times can be studied in much greater and the big question of why the three carrying capacities are differ- detail than is possible for the biota as a whole.Studies at these ent,there may be a perceptual problem at play here.Families do not smaller scales can analyse the record at the species level,within a arise overnight:they are the result of speciation and a lot of time. region or biome,and can better control for problems such as incom- Consequently,exponential growth at the species level might appear plete and uneven samplings Such studies find a range ofanswers: like logistic growth at higher levels.This problem of perception is communities may show an equilibrium diversityss. "an increasing a recurrent one in palaeontology.For instance,good evidence that geographical turnovers7,or radiation punctuated by mass extinc- biodiversity is often near equilibrium comes from the fact that tions.This may be a more appropriate spatial scale at which tolook extinction events are commonly followed by higher than normal for equilibrium,as the units have a greater chance ofinteracting rates of diversification'.However,the peak of origination rates of The temporal pattern of disparity is also of great interest.Does genera and families is not straight after the extinction peak.Instead, difference accumulate gradually and evenly as lineages evolve their there is a 10-Myr time-lag throughout the fossil record,implying a separate ways,or is evolutionary change more rapid early in a lagphase before diversificationoccursBut could thesame pattern group's history,as it stakes its claim to a new niche?Information arise if speciation rates rose immediately in response to the extinc- from living and fossil species and phylogenies can be combined with tion,but the new lineages are given generic or familial rank only statistical modesto answer this question,although so far rela- after being around for some time?This scenario would predict tively little work has combined palaeontological and neontological (incorrectly)that family diversification rates would take longer to data.Rates of morphological and taxic diversification are often respond than generic rates,so cannot be the whole story,but it incongruent,or even uncoupleds,again highlighting that there is 2.000 1500 1.500- 1.000 1.000 500 Family n0o0oooo Figure 3 Subtaxa within taxa are often distributed unevenly.Uneven distribution of Genus species among:a,eutherian orders,with rodents being the dominant group;b,rodent families,with murids being dominant;and c,murid genera.Data from ref.95. NATURE VOL 405 11 MAY 2000www.nature.com 2000 Macmillan Magazines Ltd 215
taxonomic families (see, for example, refs 46, 47), led by Sepkoski’s ground-breaking efforts, and made possible by the development of computer databases. There are more families now than ever before, and a model of exponential growth provides a good overall fit to the numbers of families through time, suggesting expansion without limit and no major role for competition in limiting diversity48. But a significantly better fit is provided by a set of three logistic curves, each with a different carrying capacity, punctuated by mass extinction events49. Leaving aside the thorny issue of multiplicity of tests and the big question of why the three carrying capacities are different, there may be a perceptual problem at play here. Families do not arise overnight: they are the result of speciation and a lot of time. Consequently, exponential growth at the species level might appear like logistic growth at higher levels50. This problem of perception is a recurrent one in palaeontology. For instance, good evidence that biodiversity is often near equilibrium comes from the fact that extinction events are commonly followed by higher than normal rates of diversification4 . However, the peak of origination rates of genera and families is not straight after the extinction peak. Instead, there is a 10-Myr time-lag throughout the fossil record, implying a lag phase before diversification occurs51. But could the same pattern arise if speciation rates rose immediately in response to the extinction, but the new lineages are given generic or familial rank only after being around for some time? This scenario would predict (incorrectly) that family diversification rates would take longer to respond than generic rates, so cannot be the whole story, but it highlights the difficulties of taking taxonomic patterns at face value. Neontologists may face much the same problem with species: taxonomists tend to recognize bird lineages as species if they are older than 2.8 Myr but not if they are younger than 1.1 Myr (ref. 52), so apparent logistic growth in species numbers through time within bird genera53 might be expected even without a slow-down of cladogenesis. The patchy nature of the known fossil record means that some taxa in some places at some times can be studied in much greater detail than is possible for the biota as a whole. Studies at these smaller scales can analyse the record at the species level, within a region or biome, and can better control for problems such as incomplete and uneven sampling54,55. Such studies find a range of answers: communities may show an equilibrium diversity55,56, an increasing geographical turnover57, or radiation punctuated by mass extinction58. This may be a more appropriate spatial scale at which to look for equilibrium, as the units have a greater chance of interacting59. The temporal pattern of disparity is also of great interest. Does difference accumulate gradually and evenly as lineages evolve their separate ways, or is evolutionary change more rapid early in a group’s history, as it stakes its claim to a new niche? Information from living and fossil species and phylogenies can be combined with statistical models41,60,61 to answer this question, although so far relatively little work has combined palaeontological and neontological data. Rates of morphological and taxic diversification are often incongruent, or even uncoupled61, again highlighting that there is insight review articles NATURE | VOL 405 | 11 MAY 2000 | www.nature.com 215 60 50 40 30 20 10 0 Species richness Rodentia Chiroptera Insectivora Carnivora Primates Artiodactyla Cetacea Lagomorpha Pinnipedia Xenarthra Scandentia Perissodactyla Macroscelidea Hyracoidea Pholidota Sirenia Dermoptera Proboscidea Tubulidentata 2,000 1,500 1,000 500 0 Species richness Muridae Sciuridae Echimyidae Heteromyidae Ctenomyidae Geomyidae Dipodidae Gliridae Zapodidae Caviidae Dasyproctidae Capromyidae Hystricidae Bathyergidae Erethizontidae Octodontidae Anomaluridae Chinchillidae Ctenodactylidae Castoridae Abrocomidae Thryonomyidae Aplodontidae Pedetidae Seleviniidae Hydrochaeridae Dinomyidae Myocastoridae Petromuridae 1,500 1,000 500 0 Species richness Genus Family Order a b c Figure 3 Subtaxa within taxa are often distributed unevenly. Uneven distribution of species among: a, eutherian orders, with rodents being the dominant group; b, rodent families, with murids being dominant; and c, murid genera. Data from ref. 95. © 2000 Macmillan Magazines Ltd
insight review articles Figure 4 Species richness in major groups of organisms.The main'pie' ▣Chordates shows the species estimated to exist in each group;the hatched area within each slice shows the proportion that have been formally described.Data Plants from ref.7. Molluscs Crustaceans 'Protozoa Insects Arachnids Nematodes Fungi Viruses Bacteria Others more to biodiversity than numbers of taxa.At present,it is hard to extinctions,is highly non-random,with related twigson thetree tell under what circumstances disparity precedes,or perhaps drives, tending to share the same fate.This selectivity greatly reduces the species richness,and when the reverse applies.Different modelscan ability ofthe phylogenetic hierarchy to retain structure in the face of give very similar patterns of diversity and disparity over time,and agiven severity of species extinction detailed studies at smaller scale may provide the greatest chance Buthow much structureis needed?Imagine ifthe only function of ofan answer. this article was the transfer of information.Many of the words could be deleted and you would still get the message.It would(we hope)be The shrinking biosphere less pleasant to read.Similarly,for many people we need biodiversity What about human impacts on biodiversity?A simple calculation because we like it;it should be conserved just as we conserve Mozart shows that recent rates of species losses are unsustainable.Ifthere are concertos and Van Gogh paintings?.But how many words could you 14 million species at present,then each year the tree of life grows by delete before the meaning starts to get lost?Recently,ecologists have an extra 14 Myr of branch length.The average age ofextant species is begun asking similar questions about our environment. nearly 5 Myr(in primates and carnivores anyway,and species in most other groups probably tend to be older rather than younger).So the Biodiversity and the stability and functioning of ecosystems tree can 'afford'at most about three species extinctions per year How many species can we lose before we start to affect the way ecosys- without shrinking overall.There have been roughly this many tems function?Principal environmental factors such as climate,soil documented species extinctions per year since 1600,and most type and disturbance strongly influence ecosystem functioning, extinctions must have passed us by.The rate has been increasing too: but likewise organisms can affect their environment.Some of the the last century saw the end of 20 mammalian species alone,a first ideas on how biodiversity could affect the way ecosystems pruning ofthe mammalian tree that would take at least 200 centuries function are attributable to Darwin and Wallace2s,who stated that a to redress. diverse mixture of plants should be more productive than a mono- Estimates of current and future rates of loss make even more culture.They also suggested the underlying biological mechanism: sobering reading.The rate at which tropical forest-probably the because coexisting species differ ecologically,loss of a species could habitat for most species-is lost is about 0.8%to 2%per year(call result in vacant niche-space and potential impacts on ecosystem it 1%for the purpose ofthis example).We must expect about 1%of processes.Defining ecological niches is not straightforward,but the tropical forest populations to be lost with it,a figure that may be Darwin and Wallace's hypothesis,if correct,provides a general as high as 16 million populations per year,or one every two biological principle which predicts that intact,diverse communities seconds.Most species have multiple populations,so rates of are generally more stable and function better than versions that have species loss will obviously be much lower.They are most commonly lost species.Recent experimental evidence(reviewed by Chapin et estimated through species-area relationships,although other al,pages 234-242,and McCann,pages 228-233),although pointing approaches are used too.Wilson famously used the species-area out important exceptions,generally supports this idea.Compared relationship to estimate an annual extinction rate of 27,000 species with systems that have lost species,diverse plant communities often -one species every twenty minutes.This and similar estimates have have a greater variety of positive and complementary interactions attracted criticism but recent work7 has shown that levels of and so outperform any single species4,and have more chance of species endangerment are rising in line with species-area predic- having the right species in the right place at the right time.This last tions,provided the analysis is conducted at the appropriate scale. 'sampling effect'mechanism has prompted much debate on the What are the implications of such rapid pruning for the tree of life? design,analysis and interpretation of experiments that aim Simulations in which species are wiped outat random'indicate that to manipulate biodiversity Although the sampling effect is most of the phylogenetic diversity would survive even a major biological in part-it requires both differences between species and extinction:up to 80%ofthe branch length could survive even if95% an ecological mechanism making some species more abundant than of the species were lost.This result assumes extinction to befall others-the probabilistic component(more diverse communities species at random;scenarios of non-random extinction can have have a greater chance of containing a species with particular proper- very different outcomes2.The current crisis,like previous mass ties)has made it controversial.Nevertheless,loss of species with key 216 2000 Macmillan Magazines Ltd NATURE VOL 40511 MAY 2000 www.nature.com
more to biodiversity than numbers of taxa. At present, it is hard to tell under what circumstances disparity precedes, or perhaps drives, species richness, and when the reverse applies. Different models can give very similar patterns of diversity and disparity over time60, and detailed studies at smaller scale62,63 may provide the greatest chance of an answer. The shrinking biosphere What about human impacts on biodiversity? A simple calculation shows that recent rates of species losses are unsustainable. If there are 14 million species at present7 , then each year the tree of life grows by an extra 14 Myr of branch length. The average age of extant species is nearly 5 Myr (in primates and carnivores anyway, and species in most other groups probably tend to be older rather than younger). So the tree can ‘afford’ at most about three species extinctions per year without shrinking overall. There have been roughly this many documented species extinctions per year since 160064, and most extinctions must have passed us by. The rate has been increasing too: the last century saw the end of 20 mammalian species alone, a pruning of the mammalian tree that would take at least 200 centuries to redress. Estimates of current and future rates of loss make even more sobering reading. The rate at which tropical forest — probably the habitat for most species — is lost is about 0.8% to 2% per year65 (call it 1% for the purpose of this example). We must expect about 1% of the tropical forest populations to be lost with it, a figure that may be as high as 16 million populations per year, or one every two seconds66. Most species have multiple populations, so rates of species loss will obviously be much lower. They are most commonly estimated through species–area relationships65, although other approaches are used too67. Wilson68 famously used the species–area relationship to estimate an annual extinction rate of 27,000 species — one species every twenty minutes. This and similar estimates have attracted criticism but recent work67,69,70 has shown that levels of species endangerment are rising in line with species–area predictions, provided the analysis is conducted at the appropriate scale. What are the implications of such rapid pruning for the tree of life? Simulations in which species are wiped out at random71 indicate that most of the phylogenetic diversity would survive even a major extinction: up to 80% of the branch length could survive even if 95% of the species were lost. This result assumes extinction to befall species at random; scenarios of non-random extinction can have very different outcomes72. The current crisis, like previous mass extinctions, is highly non-random73–76, with related twigs on the tree tending to share the same fate. This selectivity greatly reduces the ability of the phylogenetic hierarchy to retain structure in the face of a given severity of species extinction77,78. But how much structure is needed? Imagine if the only function of this article was the transfer of information. Many of the words could be deleted and you would still get the message. It would (we hope) be less pleasant to read. Similarly, for many people we need biodiversity because we like it; it should be conserved just as we conserve Mozart concertos and Van Gogh paintings79. But how many words could you delete before the meaning starts to get lost? Recently, ecologists have begun asking similar questions about our environment. Biodiversity and the stability and functioning of ecosystems How many species can we lose before we start to affect the way ecosystems function? Principal environmental factors such as climate, soil type and disturbance80,81 strongly influence ecosystem functioning, but likewise organisms can affect their environment82. Some of the first ideas on how biodiversity could affect the way ecosystems function are attributable to Darwin and Wallace28,83, who stated that a diverse mixture of plants should be more productive than a monoculture. They also suggested the underlying biological mechanism: because coexisting species differ ecologically, loss of a species could result in vacant niche-space and potential impacts on ecosystem processes. Defining ecological niches is not straightforward, but Darwin and Wallace’s hypothesis, if correct, provides a general biological principle which predicts that intact, diverse communities are generally more stable and function better than versions that have lost species. Recent experimental evidence (reviewed by Chapin et al., pages 234–242, and McCann, pages 228–233), although pointing out important exceptions, generally supports this idea. Compared with systems that have lost species, diverse plant communities often have a greater variety of positive and complementary interactions and so outperform any single species84,85, and have more chance of having the right species in the right place at the right time. This last ‘sampling effect’ mechanism has prompted much debate on the design, analysis and interpretation of experiments that aim to manipulate biodiversity86. Although the sampling effect is biological in part — it requires both differences between species and an ecological mechanism making some species more abundant than others — the probabilistic component (more diverse communities have a greater chance of containing a species with particular properties) has made it controversial. Nevertheless, loss of species with key insight review articles 216 NATURE | VOL 405 | 11 MAY 2000 | www.nature.com Figure 4 Species richness in major groups of organisms. The main ‘pie’ shows the species estimated to exist in each group; the hatched area within each slice shows the proportion that have been formally described. Data from ref. 7. Chordates Plants Molluscs Crustaceans 'Protozoa' Insects 'Algae' Arachnids Nematodes Fungi Viruses Bacteria Others © 2000 Macmillan Magazines Ltd