insight review articles Global patterns in biodiversity Kevin J.Gaston Biodiversity and Macroecology Group,Department of Animal and Plant Sciences,University of Sheffield,Sheffield S10 2TN,UK (e-mail:k.i.gaston@sheffield.ac.uk) To a first approximation,the distribution of biodiversity across the Earth can be described in terms of a relatively small number of broad-scale spatial patterns.Although these patterns are increasingly well documented,understanding why they exist constitutes one of the most significant intellectual challenges to ecologists and biogeographers.Theory is,however,developing rapidly,improving in its internal consistency and more readily subjected to empirical challenge. iodiversity,the variety of life,is distributed ecosystems)of biological variation can be distinguished, heterogeneously across the Earth.Some areas most analyses of spatial variation concern biodiversity as teem with biological variation(for example, measured by the number ofspecies observed or estimated to some moist tropical forests and coral reefs), occur in an area(species richness).This results from wide- others are virtually devoid of life (for spread recognition ofthe significance of the species as a bio- example,some deserts and polar regions),and most fall logical unit,and from the practical issues of the ease and somewhere in between.Determining why these magnitude of data acquisition.Consideration of spatial differences occur has long been a core objective for variation in other measures of biodiversity,particularly ecologists and biogeographers.It constitutes a continuing, those concerning the difference between entities rather than an important,and to many an enthralling,challenge. simply their numbers,has been remarkably sparse(with the Indeed,the past decade has seen a veritable explosion of possible exception of patterns in body size and morpholo- studies documenting broad-scale (geographical)spatial gy).Thus,although much attention has been paid to latitu- patterns in biodiversity,seeking to explain them,and dinal variation in species richness,little is known about vari- exploring their implications.The reasons for this interest ation in the diversity of genes,individuals or populations are twofold.First,it reflects increased opportunity along latitudinal gradients. provided by improvements in available data and analytical The growth ofinterest in broad-scale spatial variation in tools,the former resulting mostly from extensive collation biodiversity has been particularly striking with regard to of existing specimen and species occurrence records,the four areas of enquiry:latitudinal gradients in species rich- establishment of dedicated distribution-mapping ness,species-energy relationships,relationships between schemes,and the use of remote-sensing technology (to local and regional richness,and taxonomic covariance in measure vegetation and other environmental variables). species richness.In this review,the progress being made in Second,it reflects concern over the future of biodiversity, each of these areas will be used to substantiate four broader and the resultant need to determine its current status,to cross-cutting observations about global patterns of biodi predict its likely response to global environmental change, versity:respectively,that no single mechanism adequately and to identify the most effective schemes for in situ explains all examples of a given pattern,that the patterns conservation and sustainable use.Many of these issues observed may vary with spatial scale,that processes operat- can be addressed satisfactorily only by resolving the ing at regional scales influence patterns observed at local historical mismatch between the fine resolution of study ones,and that the relative balance of causal mechanisms plots in ecological field work means that there will invariably be varia- (typically a few square metres)and, tions in and exceptions to any given pat- by comparison,the poor resolution tern. of land-use planning and models of environmental change. Latitudinal gradients in species A host of global patterns of spatial richness variation in biodiversity has been High proportions ofterrestrial and fresh- explored (Fig.1).This includes water species occur in the tropics patterns in hotspots and coldspots Moving from high to low latitudes the (highs and lows)of diversity (includ- average species richness within a sam- ing comparisons between biological pling area of a given size increases,as realms and between biogeographical has been documented for a wide regions),variation with spatial scale spectrum of taxonomic groups (for example,species-area relation- (including groups as different as ships and relationships between local protists,trees,ants,woodpeckers and regional richness)and along and primates)for data across a gradients across space or environmen- range of spatial resolutions.Such tal conditions(for example,latitude, gradients in species richness may be longitude,altitude,depth,peninsulas, steep (for a given area,tropical bays,isolation,productivity/energy assemblages are often several times and aridity).Although several differ- more speciose than temperate ones),and ent levels of organization (genes to have been a persistent feature of the 220 2000 Macmillan Magazines Ltd NATURE|VOL 40511 MAY 2000 www.nature.com
insight review articles 220 NATURE | VOL 405 | 11 MAY 2000 | www.nature.com Biodiversity, the variety of life, is distributed heterogeneously across the Earth. Some areas teem with biological variation (for example, some moist tropical forests and coral reefs), others are virtually devoid of life (for example, some deserts and polar regions), and most fall somewhere in between. Determining why these differences occur has long been a core objective for ecologists and biogeographers. It constitutes a continuing, an important, and to many an enthralling, challenge. Indeed, the past decade has seen a veritable explosion of studies documenting broad-scale (geographical) spatial patterns in biodiversity, seeking to explain them, and exploring their implications. The reasons for this interest are twofold. First, it reflects increased opportunity provided by improvements in available data and analytical tools, the former resulting mostly from extensive collation of existing specimen and species occurrence records, the establishment of dedicated distribution-mapping schemes, and the use of remote-sensing technology (to measure vegetation and other environmental variables). Second, it reflects concern over the future of biodiversity, and the resultant need to determine its current status, to predict its likely response to global environmental change, and to identify the most effective schemes for in situ conservation and sustainable use. Many of these issues can be addressed satisfactorily only by resolving the historical mismatch between the fine resolution of study plots in ecological field work (typically a few square metres) and, by comparison, the poor resolution of land-use planning and models of environmental change. A host of global patterns of spatial variation in biodiversity has been explored (Fig. 1). This includes patterns in hotspots and coldspots (highs and lows) of diversity (including comparisons between biological realms and between biogeographical regions), variation with spatial scale (for example, species–area relationships and relationships between local and regional richness) and along gradients across space or environmental conditions (for example, latitude, longitude, altitude, depth, peninsulas, bays, isolation, productivity/energy and aridity1,2). Although several different levels of organization (genes to ecosystems) of biological variation can be distinguished, most analyses of spatial variation concern biodiversity as measured by the number of species observed or estimated to occur in an area (species richness). This results from widespread recognition of the significance of the species as a biological unit, and from the practical issues of the ease and magnitude of data acquisition. Consideration of spatial variation in other measures of biodiversity, particularly those concerning the difference between entities rather than simply their numbers, has been remarkably sparse (with the possible exception of patterns in body size and morphology). Thus, although much attention has been paid to latitudinal variation in species richness, little is known about variation in the diversity of genes, individuals or populations along latitudinal gradients. The growth of interest in broad-scale spatial variation in biodiversity has been particularly striking with regard to four areas of enquiry: latitudinal gradients in species richness, species–energy relationships, relationships between local and regional richness, and taxonomic covariance in species richness. In this review, the progress being made in each of these areas will be used to substantiate four broader cross-cutting observations about global patterns of biodiversity: respectively, that no single mechanism adequately explains all examples of a given pattern, that the patterns observed may vary with spatial scale, that processes operating at regional scales influence patterns observed at local ones, and that the relative balance of causal mechanisms means that there will invariably be variations in and exceptions to any given pattern. Latitudinal gradients in species richness High proportions of terrestrial and freshwater species occur in the tropics. Moving from high to low latitudes the average species richness within a sampling area of a given size increases, as has been documented for a wide spectrum of taxonomic groups (including groups as different as protists, trees, ants, woodpeckers and primates) for data across a range of spatial resolutions3,4. Such gradients in species richness may be steep (for a given area, tropical assemblages are often several times more speciose than temperate ones), and have been a persistent feature of the Global patterns in biodiversity Kevin J. Gaston Biodiversity and Macroecology Group, Department of Animal and Plant Sciences, University of Sheffield, Sheffield S10 2TN, UK (e-mail: k.j.gaston@sheffield.ac.uk) To a first approximation, the distribution of biodiversity across the Earth can be described in terms of a relatively small number of broad-scale spatial patterns. Although these patterns are increasingly well documented, understanding why they exist constitutes one of the most significant intellectual challenges to ecologists and biogeographers. Theory is, however, developing rapidly, improving in its internal consistency, and more readily subjected to empirical challenge. CONSERVATION INTERNATIONAL © 2000 Macmillan Magazines Ltd
insight review articles Figure 1 Spatial pattems in species richness. 1.4 a,Species-area relationship: 1.8 earthworms in areas 1.4 1.0 ranging from 100 m2 to 1.0 0. >500,000 km across 0.6 .6 Europe".b,Species-latitude 0.2 0 relationship:birds in grid cells 0.2 02 (-611,000 km2)across the New World4.c.Relationship 0.8 1.0121.41.61.82.0 22 between local and regional richness:lacustrine fish in Log area Log regional number of species North America (orange circles, 1,400 120 large lakes:blue circles,small ,200 100 lakes).d.Species-elevation relationship:bats in Manu 1,000 80 National Park Biosphere 800 60 Reserve,Peru 600 e,Species-precipitation 400 0 relationship:woody plants in 200 20 grid cells (20.000 km)in 0 0 southern Africa -60 40 -20 0 20 40 60 500 1.0001.5002.0002,5003.0003.500 N Latitude S Elevation(m) 600 500 400 ·。 300 200 100 0 0 2004006008001,0001,2001,400 Annual precipitation(mm) history of biodiversitys6.In the marine environment,open-ocean held an enduring fascination for biologists,particularly because of pelagic and deep-sea taxa also show broad latitudinal gradients in the obviously striking diversity of many tropical floras and faunas species richness,but some debate continues to surround evidence for when contrasted with their counterparts at high latitudes. shallow-water systems,particularly for non-calcareous taxa'. The latitudinal gradient in species richness,however complex it The growing number of increasingly refined analyses of latitudi- might be,is a consequence of systematic spatial variation in the nal gradients in species richness has begun to suggest some impor- balance of speciation and the immigration of species,which add tant nuances to this pattern,although the extent of their generality species to an area,and of the extinction and emigration of species, remains uncertain.Thus,it seems that declines in richness with which take them away.For very large areas,the effects of speciation latitude may be faster in the Northern than in the Southern and regional or globalextinction will predominate,and immigration Hemispheres,and that peaks in richness may not lie actually at the and emigration willbe less important.More than 25 different mecha- Equator itself but some distance away.Although poorly docu- nisms have been suggested for generating systematic latitudinal mented,such latitudinal asymmetries would be unsurprising given variation in these processes',commonly emphasizing reasons as to that these exist also in contemporary climate,in historical climatic why the tropics are highly speciose(although there is no a priori events,and in the latitudinal complexities ofthe geometry and area of expectation that either tropical or temperate zones in any sense land and ocean. represent an 'unusual'condition).These include explanations Indeed,the latitudinal gradient in species richness is a gross based on chance,historical perturbation,environmental stability, abstraction.Any underlying pattern is disrupted,sometimes habitat heterogeneity,productivity and interspecific interactions. markedly,by variation in species richness with other positional Many of these mechanisms are not mutually exclusive,and others variables (for example,longitude,elevation and depth),and merely offer different levels ofexplanation.Nonetheless,to some,en environmental ones(for example,topography and aridity).Thus, masse they have been perceived to constitute a gordian knot.Two the detailed pattern of change with latitude depends on where one recent attempts to cut it concern the importance of the physical struc- looks,reflecting the generally complex patterns ofspatial variation in ture of the Earth.First,null models that assume no environmental species richness.This indicates that consideration of latitudinal gradients,but merely a random latitudinal association between the gradients in richness in isolation from other gradients might not be size and placement(midpoint)of the geographical ranges of species, the most profitable way forward.In as much as latitude per se(and predict a peak of species richness at tropical latitudes This occurs likewise other positional variables)cannot be a determinant of because when the latitudinal extents of species in a given taxonomic species richness,but only a correlate of numbers of potentially causal group are bounded to north and south-perhaps by a physical con- environmental factors,this is doubtless correct.Nonetheless,more straint such as a continental edge or perhaps by a climatic constraint than any other pattern the latitudinal gradient in species richness has such as a critical temperature or precipitation threshold-then the NATURE|VOL 405|11 MAY 2000www.nature.com ☆©20o0 Macmillan Magazines Ltd 221
history of biodiversity5,6. In the marine environment, open-ocean pelagic and deep-sea taxa also show broad latitudinal gradients in species richness, but some debate continues to surround evidence for shallow-water systems, particularly for non-calcareous taxa7 . The growing number of increasingly refined analyses of latitudinal gradients in species richness has begun to suggest some important nuances to this pattern, although the extent of their generality remains uncertain. Thus, it seems that declines in richness with latitude may be faster in the Northern than in the Southern Hemisphere8,9, and that peaks in richness may not lie actually at the Equator itself but some distance away10,11. Although poorly documented, such latitudinal asymmetries would be unsurprising given that these exist also in contemporary climate, in historical climatic events, and in the latitudinal complexities of the geometry and area of land and ocean. Indeed, the latitudinal gradient in species richness is a gross abstraction. Any underlying pattern is disrupted, sometimes markedly, by variation in species richness with other positional variables (for example, longitude, elevation and depth), and environmental ones (for example, topography and aridity). Thus, the detailed pattern of change with latitude depends on where one looks, reflecting the generally complex patterns of spatial variation in species richness. This indicates that consideration of latitudinal gradients in richness in isolation from other gradients might not be the most profitable way forward. In as much as latitude per se (and likewise other positional variables) cannot be a determinant of species richness, but only a correlate of numbers of potentially causal environmental factors, this is doubtless correct. Nonetheless, more than any other pattern the latitudinal gradient in species richness has held an enduring fascination for biologists, particularly because of the obviously striking diversity of many tropical floras and faunas when contrasted with their counterparts at high latitudes. The latitudinal gradient in species richness, however complex it might be, is a consequence of systematic spatial variation in the balance of speciation and the immigration of species, which add species to an area, and of the extinction and emigration of species, which take them away. For very large areas, the effects of speciation and regional or global extinction will predominate, and immigration and emigration will be less important. More than 25 different mechanisms have been suggested for generating systematic latitudinal variation in these processes2 , commonly emphasizing reasons as to why the tropics are highly speciose (although there is no a priori expectation that either tropical or temperate zones in any sense represent an ‘unusual’ condition12). These include explanations based on chance, historical perturbation, environmental stability, habitat heterogeneity, productivity and interspecific interactions. Many of these mechanisms are not mutually exclusive, and others merely offer different levels of explanation. Nonetheless, to some, en masse they have been perceived to constitute a gordian knot. Two recent attempts to cut it concern the importance of the physical structure of the Earth. First, null models that assume no environmental gradients, but merely a random latitudinal association between the size and placement (midpoint) of the geographical ranges of species, predict a peak of species richness at tropical latitudes13. This occurs because when the latitudinal extents of species in a given taxonomic group are bounded to north and south — perhaps by a physical constraint such as a continental edge or perhaps by a climatic constraint such as a critical temperature or precipitation threshold — then the insight review articles NATURE | VOL 405 | 11 MAY 2000 | www.nature.com 221 Figure 1 Spatial patterns in species richness. a, Species–area relationship: earthworms in areas ranging from 100 m2 to >500,000 km2 across Europe76. b, Species–latitude relationship: birds in grid cells (~ 611,000 km2 ) across the New World44. c, Relationship between local and regional richness: lacustrine fish in North America (orange circles, large lakes; blue circles, small lakes)61. d, Species–elevation relationship: bats in Manu National Park & Biosphere Reserve, Peru77. e, Species–precipitation relationship: woody plants in grid cells (20,000 km2 ) in southern Africa78. Number of species Number of species Number of species Log (number of species +1) Log local number of species 1.8 1.4 1.0 0.6 0.2 –0.2 1,400 1,200 1,000 800 600 400 200 0 600 500 400 300 200 100 0 –5 –3 –1 1 3 5 0.8 1.0 1.2 1.4 1.6 1.8 2.0 2.2 –80 –60 –40 –20 0 20 40 60 0 0 200 400 600 800 1,000 1,200 1,400 500 1,000 1,500 2,000 2,500 3,000 3,500 Log area Log regional number of species Latitude Annual precipitation (mm) °N °S Elevation (m) 1.4 1.2 1.0 0.8 0.6 0.4 0.2 0.0 120 100 80 60 40 20 0 a c b d e © 2000 Macmillan Magazines Ltd
insight review articles number ofways in which ranges can be distributed changes systemat- local climatic variability at high latitudes effectively increases the area ically between the bounds.Thus,whereas species with latitudinal of ecoclimatic zones that species can actually occupy,because it midpoints midway between the bounds can extend a little or a long requires that individuals have broad environmental tolerances way before those bounds are encountered,those with midpoints The observation that area alone is insufficient as a determinant of close to the bounds can extend only a little way before this occurs.A latitudinal gradients in species richness could equally be made about null model has been wanting from discussions of latitudinal almost any other factor that has been proposed as being important gradients in species richness.The 'mid-domain'model is thus likely (although critical tests are typically lacking).This highlights an issue to stimulate much interest.It is also likely to be most applicable for that has been central to much ofthe debate about the cause ofthis and groups whose distributions are genuinely limited by a physical other global patterns in biodiversity,namely the assumption that boundary(for example,those of large islands such as Madagascar), where a pattern is common to many taxa it must result from the same although its extension to two spatial dimensions is problematic, single mechanism-"wherever there is a widespread pattern,there given the longitudinal variation in land and ocean area.The is likely to be a general explanation which applies to the whole application of the model to other kinds of constraints is more ques- pattern To argue for a single primary cause may be to expect from tionable,as the position of those constraints that are recognized will ecological interactions a simplicity for which there is little evidence. be dependent on the inclusiveness ofthe set ofspecies considered. There is no necessary reason why latitudinal gradients exhibited by The second attempt to explain latitudinal gradients in species taxa as distinct as protozoa and mammals,and in environments as richness based on the physical structure ofthe Earth concerns the role structurally different as the deep sea and tropical forests,need be gen- of area (its importance has long been entertained 4s and recently erated in the same way,whatever the attractions of Occam's razor. brought to prominence).The tropics have a larger climatically Increasingly it seems that patterns in biodiversity are likely to be similar total surface area than any other ecoclimatic zone.This is generated by several contributory mechanisms221.The strongest because:(1)the surface area of latitudinal bands decreases towards and most general may be those where all the different mechanisms the poles;(2)the temperature gradient between the Equator and the pull in the same direction2.It is instructive that although numerous poles is nonlinear (the mean being relatively constant between mechanisms for latitudinal gradients in species richness have been approximately 20Nand 20S);and(3)the regions ofsimilar climate identified,and rather few processes that would oppose such a trend, immediately north and south of the Equator abut.It has been no single mechanism has ofitself proven sufficient. contended that,for a given species richness,larger mean geographi- cal-range sizes of species in the tropics result from the large area Species-energy relationships (which is not to be confused with any observed pattern in mean One factor thought to be important in modulating any effect of the range sizes at differentlevels ofrichness),and that these translate into physical structure ofthe Earth in determining latitudinalgradients in higher speciation rates (presuming larger ranges have higher species richness is the relationship between the number of species in probabilities of speciation)and lower extinction probabilities an area and ambient available ('usable')environmental energy. (presuming larger ranges have lower probabilities ofextinction)67. (This energy is usually estimated from models or indirectly from As a consequence,tropical regions have greater numbers of species other variables,and often used interchangeably with net primary than extratropical ones. productivity.)The form and cause ofthis relationship are some ofthe Area is almost certainly an important contributor to latitudinal most hotly debated topics in the study of global patterns in biodiver- gradients in species richness(indeed,area effects have a pervasive sity,with many fundamental issues as yet unresolved.Much of the influence on patterns of biodiversity).However,tests of the 'area discussion centres on the influence of spatial scale on observed model'have been limited (and often tangential),and have seldom relationships. sought the signal of the influence of area on latitudinal gradients At a relatively local scale (spatial resolution and extent),there is a when other factors have been controlled for.Moreover,as a sole marked tendency for a general hump-shaped relationship between explanation the area model is insufficient.To account fully for a species richness and available energy,with species richness increas- latitudinal gradient in species richness(rather than simply for the ing from low to moderate levels of energy and then declining again greater richness of the tropics)the model requires that ecoclimatic towards high levels of energy when a sufficient range of energy zones decline systematically in area moving from the Equator values is sampled At least across temperate to polar areas,at towards the poles.However,they do not do so (ecoclimatic zones at geographical scales there is substantial evidence for a broadly posi- high latitudes tend to belarge).Three possible explanations have tive monotonic relationship between species richness and energy been advanced for why a latitudinal gradient in species richness availabilitytobecommon(Fig.2).The bestcorrelates for plants might nonetheless be expressed:(1)low productivity/energy tend to be measures ofboth heat and water(such as actual evapotran- availability at high latitudes reduces the species richness they would spiration and net primary productivity),whereas for terrestrial,and gain as a result of area alonel.7 perhaps marine,animals the best correlates are measures of heat (2)zonal bleeding of tropical (such as mean annual temperature and potential evapotranspira- species into extratropical tion)For example,whereas the species richness of trees in regions smoothes out temperate Europe,eastern North America and East Asia increases species-richness gra- with primary productivity2,the richness of butterflies and birds in dients19;and areas of Britain increases with the temperature during the appropri- (3) high ate season2,and the species richness ofamphibians,reptiles,birds and mammals in areas of North America increases with annual potential evapotranspiration (estimated as a measure of the net atmospheric energy balance,independent of water availability2). The form taken by species-energy relationships at geographical scales,when extended to include subtropical and tropical areas,or at least to include the fullest range of variation in available energy (which may not be the same thing),remains unclear.There is evidence to suggest that they remain broadly positive and monoton- omem the answer may depend critically on the measure ofenergy used and the taxon concerned. 222 2000 Macmillan Magazines Ltd NATURE VOL 405|11 MAY 2000 www.nature.com
number of ways in which ranges can be distributed changes systematically between the bounds. Thus, whereas species with latitudinal midpoints midway between the bounds can extend a little or a long way before those bounds are encountered, those with midpoints close to the bounds can extend only a little way before this occurs. A null model has been wanting from discussions of latitudinal gradients in species richness. The ‘mid-domain’ model is thus likely to stimulate much interest. It is also likely to be most applicable for groups whose distributions are genuinely limited by a physical boundary (for example, those of large islands such as Madagascar), although its extension to two spatial dimensions is problematic, given the longitudinal variation in land and ocean area. The application of the model to other kinds of constraints is more questionable, as the position of those constraints that are recognized will be dependent on the inclusiveness of the set of species considered. The second attempt to explain latitudinal gradients in species richness based on the physical structure of the Earth concerns the role of area (its importance has long been entertained14,15 and recently brought to prominence16,17). The tropics have a larger climatically similar total surface area than any other ecoclimatic zone. This is because: (1) the surface area of latitudinal bands decreases towards the poles; (2) the temperature gradient between the Equator and the poles is nonlinear (the mean being relatively constant between approximately 207N and 207 S); and (3) the regions of similar climate immediately north and south of the Equator abut. It has been contended that, for a given species richness, larger mean geographical-range sizes of species in the tropics result from the large area (which is not to be confused with any observed pattern in mean range sizes at different levels of richness), and that these translate into higher speciation rates (presuming larger ranges have higher probabilities of speciation) and lower extinction probabilities (presuming larger ranges have lower probabilities of extinction)16,17. As a consequence, tropical regions have greater numbers of species than extratropical ones. Area is almost certainly an important contributor to latitudinal gradients in species richness (indeed, area effects have a pervasive influence on patterns of biodiversity). However, tests of the ‘area model’ have been limited (and often tangential), and have seldom sought the signal of the influence of area on latitudinal gradients when other factors have been controlled for. Moreover, as a sole explanation the area model is insufficient. To account fully for a latitudinal gradient in species richness (rather than simply for the greater richness of the tropics) the model requires that ecoclimatic zones decline systematically in area moving from the Equator towards the poles. However, they do not do so (ecoclimatic zones at high latitudes tend to be large10,17). Three possible explanations have been advanced for why a latitudinal gradient in species richness might nonetheless be expressed: (1) low productivity/energy availability at high latitudes reduces the species richness they would gain as a result of area alone10,17; (2) zonal bleeding of tropical species into extratropical regions smoothes out species-richness gradients18,19; and (3) high insight review articles 222 NATURE | VOL 405 | 11 MAY 2000 | www.nature.com local climatic variability at high latitudes effectively increases the area of ecoclimatic zones that species can actually occupy, because it requires that individuals have broad environmental tolerances3 . The observation that area alone is insufficient as a determinant of latitudinal gradients in species richness could equally be made about almost any other factor that has been proposed as being important (although critical tests are typically lacking). This highlights an issue that has been central to much of the debate about the cause of this and other global patterns in biodiversity, namely the assumption that where a pattern is common to many taxa it must result from the same single mechanism — “wherever there is a widespread pattern, there is likely to be a general explanation which applies to the whole pattern”20. To argue for a single primary cause may be to expect from ecological interactions a simplicity for which there is little evidence. There is no necessary reason why latitudinal gradients exhibited by taxa as distinct as protozoa and mammals, and in environments as structurally different as the deep sea and tropical forests, need be generated in the same way, whatever the attractions of Occam’s razor. Increasingly it seems that patterns in biodiversity are likely to be generated by several contributory mechanisms12,21. The strongest and most general may be those where all the different mechanisms pull in the same direction22. It is instructive that although numerous mechanisms for latitudinal gradients in species richness have been identified, and rather few processes that would oppose such a trend, no single mechanism has of itself proven sufficient. Species–energy relationships One factor thought to be important in modulating any effect of the physical structure of the Earth in determining latitudinal gradients in species richness is the relationship between the number of species in an area and ambient available (‘usable’) environmental energy. (This energy is usually estimated from models or indirectly from other variables, and often used interchangeably with ‘net primary productivity’.) The form and cause of this relationship are some of the most hotly debated topics in the study of global patterns in biodiversity, with many fundamental issues as yet unresolved. Much of the discussion centres on the influence of spatial scale on observed relationships. At a relatively local scale (spatial resolution and extent), there is a marked tendency for a general hump-shaped relationship between species richness and available energy, with species richness increasing from low to moderate levels of energy and then declining again towards high levels of energy when a sufficient range of energy values is sampled16,17,23. At least across temperate to polar areas, at geographical scales there is substantial evidence for a broadly positive monotonic relationship between species richness and energy availability to be common10,24–33 (Fig. 2). The best correlates for plants tend to be measures of both heat and water (such as actual evapotranspiration and net primary productivity), whereas for terrestrial, and perhaps marine, animals the best correlates are measures of heat (such as mean annual temperature and potential evapotranspiration)28,29,34. For example, whereas the species richness of trees in temperate Europe, eastern North America and East Asia increases with primary productivity27, the richness of butterflies and birds in areas of Britain increases with the temperature during the appropriate season25,26, and the species richness of amphibians, reptiles, birds and mammals in areas of North America increases with annual potential evapotranspiration (estimated as a measure of the net atmospheric energy balance, independent of water availability28). The form taken by species–energy relationships at geographical scales, when extended to include subtropical and tropical areas, or at least to include the fullest range of variation in available energy (which may not be the same thing), remains unclear. There is evidence to suggest that they remain broadly positive and monotonic, that they become mildly or strongly hump-shaped, and that they begin to break down altogether10,32,35–37; the answer may depend critically on the measure of energy used and the taxon concerned. CONSERVATION INTERNATIONAL © 2000 Macmillan Magazines Ltd
insight review articles resulting in a broadly positive relationship between species richness and energy availability at geographical scales (and at low-to- moderate energy levels at more local scales)are believed to be reasonably straightforward.Greater energy availability is assumed to enable a greater biomass to be supported in an area.In turn,this enables more individual organisms to coexist,and thus more species at abundances that enable them to maintain viable populations.The result is an increase in species richness with energy availability.This assumes a basic equivalence between species in their energetic requirements at different levels of energy availability.Although 12 14 there is some evidence in animal systems that average densities and Summer temperature(C) body sizes ofspecies in an area decrease as energy availability increas- es(that is,energy is divided more finely4),this will tend to enhance 000 the species-energy relationship provided these trends are sufficiently 900 marked compared with the scaling of metabolic rate to body mass. 800 There are important similarities between this 'more-individuals 700 model'and the area model as explanations of variation in species 600 richness".First,both to some degree concern variation in solar energy(and water availability),with the level and availability of this 300 energy source being important in the former case,and the spatial extent ofa givenlevel(as reflected in an ecoclimatic zone)in thelatter. 200 100 If ecoclimatic zones vary in available energy,then observed 10 15 20 species-energy relationships(and those between richness and lati- Sea surface temperature (C) tude)may reflect the joint effects of their area and this availability Second,the area model assumes that area influences richness through its effect on geographical-range size,and the more- 35 individuals hypothesis that energy influences richness through its effect on population size.There is a general,positive,interspecific 25 relationship between total population size(or local density)and size of geographical range.Any factor that increases one of these 5 variables will also be likely to increase the other.Both mechanisms 05 therefore depend,in effect,on some factor that is posited to influence the biomass available to be worked on by the processes of speciation 0.5 400 8001.2001.6002.000 and extinction,which will be a product of both area and available Potential evapotranspiration(mm yr-1) energy per unit area Presumably,it is for this reason that small areas tend to be species poor however high their energy input,where- as large areas tend to be species poor if there is low energy input. Figure 2 Species-energy relationships.a,Mean monthly summer temperature(C) Assuming that species-energy relations are causal and that a and richness of breeding birds in Britain (grid cells of 10 km x 10 km).b,Mean more-individuals model is operating,then it is unlikely that the path annual sea surface temperature and richness of eastern Pacific marine gastropods of causality is simple.Levels of available energy may constrain the (bands of 1 latitude).c.Potential evapotranspiration (mm yr)and richness of amount of biomass that is achieved in an area,but characteristics of Epicautabeetles (Meloidae)in North America(grid cells of2.5°×2.5°south of50°N, the biosphere,and particularly those of vegetation,are themselves 2.5°×5°n0thof50°031 known to be key influences on climate,including temperature and precipitation.For example,the coupling of an atmospheric model and a simple land-surface scheme has indicated that coastal defor- Any contingency ofthe gross or more detailed form of patterns in estation in West Africa has been a significant contributor to the biodiversity on the spatial extent and dispersion of sampling units is observed drought in the region;this deforestation has resulted in a not restricted to species-energy relationships.Indeed,the almost number of species being threatened with extinction.Complex pat- ubiquitous positive relationship between the numbers of species in terns of causality suggest an important connection between an area and the size of that area (the species-area relationship)may species-energy theory and debates over the ecosystem function of itself vary in form with the absolute sizes of areas,their spatial rela- biodiversity1.52 tionships(for example,isolation),and their latitudinal position Even accepting that paths of causality may be complex,there are this is often forgotten when attempting to control for differences in some potentially significant difficulties with a more-individuals area in analyses of global patterns of biodiversity.Reconciliation of model. the patterns in biodiversity that are observed at different scales may 1.The assumption that the number of individual organisms provide significant insights into their determinants.If this is to be increases with available energy and total biomass may not apply to achieved,it is important to ensure that the scale of sampling and the plants,for which there is evidence that as standing crop increases the scale of processes that are postulated to explain patterns in species numbers of adult individuals per unit area actually declines (and richness are closely matched.One criticism of some discussion of their size increases),which should tend to reduce species richness species-energy relationships at broad scales has been that this has not rather than increase it.However,this argument is based in large been done;curiously,this has been interpreted,by different parties, part on findings from monospecific stands of species differing as yielding species-energy relationships that may be misleadingly substantially in their architecture,and it is unclear to what extent it strong or misleadingly weak.Matching scales of sampling and generalizes to multispecies stands and systems that are structurally processes is more readily achievable at local scales,and constitutes more similar (for example,temperate compared with tropical one of the most significant obstacles to testing mechanisms over forests).Evidence as to how overall biomass and numbers of broader areas. individuals change with species richness in animal systems is scant, Although other explanations have been offered,the processes even for well-known groupssuchas birds,and is plagued by apaucity NATURE|VOL 40511 MAY 2000 www.nature.com ☆©20o0 Macmillan Magazines Ltd 223
Any contingency of the gross or more detailed form of patterns in biodiversity on the spatial extent and dispersion of sampling units is not restricted to species–energy relationships. Indeed, the almost ubiquitous positive relationship between the numbers of species in an area and the size of that area (the species–area relationship) may itself vary in form with the absolute sizes of areas, their spatial relationships (for example, isolation), and their latitudinal position38,39; this is often forgotten when attempting to control for differences in area in analyses of global patterns of biodiversity. Reconciliation of the patterns in biodiversity that are observed at different scales may provide significant insights into their determinants. If this is to be achieved, it is important to ensure that the scale of sampling and the scale of processes that are postulated to explain patterns in species richness are closely matched. One criticism of some discussion of species–energy relationships at broad scales has been that this has not been done; curiously, this has been interpreted, by different parties, as yielding species–energy relationships that may be misleadingly strong or misleadingly weak40–42. Matching scales of sampling and processes is more readily achievable at local scales, and constitutes one of the most significant obstacles to testing mechanisms over broader areas. Although other explanations have been offered, the processes resulting in a broadly positive relationship between species richness and energy availability at geographical scales (and at low-tomoderate energy levels at more local scales) are believed to be reasonably straightforward. Greater energy availability is assumed to enable a greater biomass to be supported in an area. In turn, this enables more individual organisms to coexist, and thus more species at abundances that enable them to maintain viable populations. The result is an increase in species richness with energy availability. This assumes a basic equivalence between species in their energetic requirements at different levels of energy availability43. Although there is some evidence in animal systems that average densities and body sizes of species in an area decrease as energy availability increases (that is, energy is divided more finely44), this will tend to enhance the species–energy relationship provided these trends are sufficiently marked compared with the scaling of metabolic rate to body mass. There are important similarities between this ‘more-individuals model’45 and the area model as explanations of variation in species richness44. First, both to some degree concern variation in solar energy (and water availability), with the level and availability of this energy source being important in the former case, and the spatial extent of a given level (as reflected in an ecoclimatic zone) in the latter. If ecoclimatic zones vary in available energy, then observed species–energy relationships (and those between richness and latitude) may reflect the joint effects of their area and this availability37. Second, the area model assumes that area influences richness through its effect on geographical-range size, and the moreindividuals hypothesis that energy influences richness through its effect on population size. There is a general, positive, interspecific relationship between total population size (or local density) and size of geographical range46. Any factor that increases one of these variables will also be likely to increase the other. Both mechanisms therefore depend, in effect, on some factor that is posited to influence the biomass available to be worked on by the processes of speciation and extinction, which will be a product of both area and available energy per unit area29,47. Presumably, it is for this reason that small areas tend to be species poor however high their energy input, whereas large areas tend to be species poor if there is low energy input. Assuming that species–energy relations are causal and that a more-individuals model is operating, then it is unlikely that the path of causality is simple. Levels of available energy may constrain the amount of biomass that is achieved in an area, but characteristics of the biosphere, and particularly those of vegetation, are themselves known to be key influences on climate, including temperature and precipitation48. For example, the coupling of an atmospheric model and a simple land-surface scheme has indicated that coastal deforestation in West Africa has been a significant contributor to the observed drought in the region49; this deforestation has resulted in a number of species being threatened with extinction50. Complex patterns of causality suggest an important connection between species–energy theory and debates over the ecosystem function of biodiversity51,52. Even accepting that paths of causality may be complex, there are some potentially significant difficulties with a more-individuals model. 1. The assumption that the number of individual organisms increases with available energy and total biomass may not apply to plants, for which there is evidence that as standing crop increases the numbers of adult individuals per unit area actually declines (and their size increases), which should tend to reduce species richness rather than increase it35. However, this argument is based in large part on findings from monospecific stands of species differing substantially in their architecture, and it is unclear to what extent it generalizes to multispecies stands and systems that are structurally more similar (for example, temperate compared with tropical forests). Evidence as to how overall biomass and numbers of individuals change with species richness in animal systems is scant, even for well-known groups such as birds, and is plagued by a paucity insight review articles NATURE | VOL 405 | 11 MAY 2000 | www.nature.com 223 Summer temperature (°C) Number of species Number of species 0 20 40 60 80 8 10 12 14 16 Sea surface temperature (°C) Potential evapotranspiration (mm yr –1) 100 200 300 400 500 600 700 800 900 1,000 5 10 15 20 25 30 Square root of number of species –0.5 0.5 1.5 2.5 3.5 4.5 0 400 800 1,200 1,600 2,000 a b c Figure 2 Species–energy relationships. a, Mean monthly summer temperature (7C) and richness of breeding birds in Britain (grid cells of 10 km 2 10 km)33. b, Mean annual sea surface temperature and richness of eastern Pacific marine gastropods (bands of 17 latitude)10. c, Potential evapotranspiration (mm yr–1) and richness of Epicauta beetles (Meloidae) in North America (grid cells of 2.57 2 2.57south of 507 N, 2.57 2 57 north of 507 N)31. © 2000 Macmillan Magazines Ltd
insight review articles of strictly comparable studies from areas differing markedly in species richness. Local richness 2.Many taxa use such a small proportion of the total energy regional richness available in an area,or at least of the energy that is being measured, that it seems unlikely that detectable relationships with species richness would arise (especially given the likely magnitude of measurement errors).Thus,although species richness of birds tends Type I to increase with available energy,avian assemblages may,directly and indirectly,commonly exploit only a small proportion of the primary production ina locality.(The avian community ofthe forested water- Type ll sheds of the Hubbard Brook Experimental Forest has an average ingestion rate which represents 0.17%of ecosystem net annual Regional species richness productivity) 3.In its simplest form,the more-individuals model ignores the likely effects oftemporal variance in energy levels on species richness. High average levels of energy may not result in large numbers of Figure 3 Relations between local and regional species richness,illustrating the form of species if they are accompanied by high temporal variability in those type Iand type ll relationships and the limiting condition where local richness equals levels.The relationship between levels of available energy and their regional richness. variance may be broadly different between some terrestrial and marine systems (negative in the former,positive in the latter), perhaps explaining why even at very broad spatial scales high one rich source ofdata for testing hypotheses about the significance of richness may not be associated with high productivity in marine history. systems37 4.At regional scales,levels of species richness have not been Relationships between local and regional richness produced directly by present environmental conditions,as processes In exploring global variation in biodiversity,we need to understand ofspeciation and extinction do not operate on these timescales.Ifthe not only the importance of differences in spatial scale for the patterns more-individuals model is to apply this must mean that present that are observed (for example,hump-shaped species-energy environmental conditionsare a good proxy for past ones,or atleast of relationships at local scales and positive relationships at regional relative differences in the conditions in different areas. ones),but also how diversity at one scale might relate to that at anoth- Alternatives to the more-individuals model have been advanced er.Indeed,it is increasingly apparent that knowledge of the roles of to explain positive species-energy relationships.These have been pattern and process at different scales is at the very heart of an based particularly on variation with energy in levels of constraints on understanding ofglobal variation in biodiversity. geographical ranges,specialization,population growth rates and Two theoretical types of relationship have been contrasted number oftrophiclevels5.Foremost is the idea that the relationships between thelocal richness an assemblage might attain and the species may reflect physiological constraints on the distribution of species, richness of the region in which that assemblage residess(Fig.3). with energy availability capturing factors that limit distributions as a Local richness may bedirectly proportional to,but less than,regional result of metabolic considerations. richness,followingaproportional-samplingmodel(typeI).Alterna- In the absence of strong support for any of these alternative tively,as regional richness increases,local richness might attain a explanations,difficulties with the more-individuals model fuel ceiling above which it does not rise despite continued increases in growing speculation that at least some species-energy relationships regional richness(type II). may not be causal,and that energy availability may often be only a Acknowledging a number of technical concerns35-57,most real covariate ofsome other factor that is actually driving species richness. systems seem to exhibit an underlying type I relationship ；not Bird richness may,for example,be responding to a second-order uncommonly,regional richness explains alarge proportion(>75%) effect of greater vegetational complexity with increased available of variance in local richness,and local richness constitutes a marked environmental energy.Likewise,recent workhas shown that whereas proportion (>50%)of regional richness.For example,type I rela- sea surface temperature explained nearly 90%of geographical varia- tionships have been documented for fig wasps and their parasitoids tion in planktonic foraminiferal diversity throughout the Atlantic in southern and central Africa",tiger beetles in North America and Ocean,this temperature was also correlated with temperatures at in India,lacustrine fish in North America(Fig.Ic),and primates different depths.This indicates that the diversity may be controlled in Africa and in South America.The predominance of type I by the physical structure ofthe near-surface ocean and not directly by relationships is supported by the observation that some spatial gradi- available energy2. ents in species richness are documented both for localities and Continuing with this theme,there has been debate as to the respec- regions across those gradients (with obvious implications for the tive roles of contemporary levels ofenergy and ofhistorical factors in interpretation of regional collations offossil records). generatingglobalpatternsoftreespeciesrichnessinmoist forests.The A recurrent problem in studies of spatial patterns in biodiversity debate has centred on the extent to which differences in richness has been the conflation of pattern with mechanism.Nonetheless,the between continents and between latitudes result from variation in preponderance of examples of type I relationships,particularly annual actual evapotranspiration (a good,but not universal, where habitat type has been kept constant,backed up with other evi- predictor of primary productivity)or from long-term evolutionary dence(forexample,thelimitedsupport for communityconvergence, and geographicalprocesses-.The practical constraintsonconduct- density compensation and invasion resistance),indicates that there ing experiments at relevant scales mean that differentiating between are not hard limits to levels of local richness.That is,local assem- hypotheses necessarily requires that they make divergent testable blages do not seem to be saturated,in the way one might have expect- predictions,and even then may not enable the relative roles of ed if ecological interactions (for example,competition,predation different factors to be quantified.Historical factors have doubtless had and parasitism)limited local richness.Three potential anomalies a substantial role in shaping contemporary spatial patterns of arise if this conclusion is correct.First,it suggests that although biodiversity,but deriving such a priori predictions and quantifying ecological interactions are known to be strong in some circum- the part played by history can often prove difficult.Molecular stances,they may typically not be sufficient to have a marked effect on phylogenies,with estimated dates of diversification events,provide species richness.Second,it may be at odds with the more-individuals 224 2000 Macmillan Magazines Ltd NATURE VOL 40511 MAY 2000 www.nature.com
of strictly comparable studies from areas differing markedly in species richness. 2. Many taxa use such a small proportion of the total energy available in an area, or at least of the energy that is being measured, that it seems unlikely that detectable relationships with species richness would arise (especially given the likely magnitude of measurement errors). Thus, although species richness of birds tends to increase with available energy, avian assemblages may, directly and indirectly, commonly exploit only a small proportion of the primary production in a locality. (The avian community of the forested watersheds of the Hubbard Brook Experimental Forest has an average ingestion rate which represents 0.17% of ecosystem net annual productivity53.) 3. In its simplest form, the more-individuals model ignores the likely effects of temporal variance in energy levels on species richness. High average levels of energy may not result in large numbers of species if they are accompanied by high temporal variability in those levels. The relationship between levels of available energy and their variance may be broadly different between some terrestrial and marine systems (negative in the former, positive in the latter), perhaps explaining why even at very broad spatial scales high richness may not be associated with high productivity in marine systems37. 4. At regional scales, levels of species richness have not been produced directly by present environmental conditions, as processes of speciation and extinction do not operate on these timescales. If the more-individuals model is to apply this must mean that present environmental conditions are a good proxy for past ones, or at least of relative differences in the conditions in different areas. Alternatives to the more-individuals model have been advanced to explain positive species–energy relationships. These have been based particularly on variation with energy in levels of constraints on geographical ranges, specialization, population growth rates and number of trophic levels45. Foremost is the idea that the relationships may reflect physiological constraints on the distribution of species, with energy availability capturing factors that limit distributions as a result of metabolic considerations30. In the absence of strong support for any of these alternative explanations, difficulties with the more-individuals model fuel growing speculation that at least some species–energy relationships may not be causal, and that energy availability may often be only a covariate of some other factor that is actually driving species richness. Bird richness may, for example, be responding to a second-order effect of greater vegetational complexity with increased available environmental energy. Likewise, recent work has shown that whereas sea surface temperature explained nearly 90% of geographical variation in planktonic foraminiferal diversity throughout the Atlantic Ocean, this temperature was also correlated with temperatures at different depths. This indicates that the diversity may be controlled by the physical structure of the near-surface ocean and not directly by available energy32. Continuing with this theme, there has been debate as to the respective roles of contemporary levels of energy and of historical factors in generating global patterns of tree species richness in moist forests. The debate has centred on the extent to which differences in richness between continents and between latitudes result from variation in annual actual evapotranspiration (a good, but not universal, predictor of primary productivity) or from long-term evolutionary and geographical processes40–42. The practical constraints on conducting experiments at relevant scales mean that differentiating between hypotheses necessarily requires that they make divergent testable predictions, and even then may not enable the relative roles of different factors to be quantified. Historical factors have doubtless had a substantial role in shaping contemporary spatial patterns of biodiversity, but deriving such a priori predictions and quantifying the part played by history can often prove difficult. Molecular phylogenies, with estimated dates of diversification events, provide one rich source of data for testing hypotheses about the significance of history. Relationships between local and regional richness In exploring global variation in biodiversity, we need to understand not only the importance of differences in spatial scale for the patterns that are observed (for example, hump-shaped species–energy relationships at local scales and positive relationships at regional ones), but also how diversity at one scale might relate to that at another. Indeed, it is increasingly apparent that knowledge of the roles of pattern and process at different scales is at the very heart of an understanding of global variation in biodiversity. Two theoretical types of relationship have been contrasted between the local richness an assemblage might attain and the species richness of the region in which that assemblage resides54 (Fig. 3). Local richness may be directly proportional to, but less than, regional richness, following a proportional-sampling model (type I). Alternatively, as regional richness increases, local richness might attain a ceiling above which it does not rise despite continued increases in regional richness (type II). Acknowledging a number of technical concerns55–57, most real systems seem to exhibit an underlying type I relationship54,56,58; not uncommonly, regional richness explains a large proportion (>75%) of variance in local richness, and local richness constitutes a marked proportion (>50%) of regional richness. For example, type I relationships have been documented for fig wasps and their parasitoids in southern and central Africa59, tiger beetles in North America and in India60, lacustrine fish in North America61 (Fig. 1c), and primates in Africa and in South America62. The predominance of type I relationships is supported by the observation that some spatial gradients in species richness are documented both for localities and regions across those gradients (with obvious implications for the interpretation of regional collations of fossil records). A recurrent problem in studies of spatial patterns in biodiversity has been the conflation of pattern with mechanism. Nonetheless, the preponderance of examples of type I relationships, particularly where habitat type has been kept constant, backed up with other evidence (for example, the limited support for community convergence, density compensation and invasion resistance), indicates that there are not hard limits to levels of local richness63. That is, local assemblages do not seem to be saturated, in the way one might have expected if ecological interactions (for example, competition, predation and parasitism) limited local richness. Three potential anomalies arise if this conclusion is correct. First, it suggests that although ecological interactions are known to be strong in some circumstances, they may typically not be sufficient to have a marked effect on species richness. Second, it may be at odds with the more-individuals insight review articles 224 NATURE | VOL 405 | 11 MAY 2000 | www.nature.com Regional species richness Local species richness Local richness = regional richness Type I Type II Figure 3 Relations between local and regional species richness, illustrating the form of type I and type II relationships and the limiting condition where local richness equals regional richness. © 2000 Macmillan Magazines Ltd