Published online 7 October 2004

Introduction and synthesis: plant phylogeny

and the origin of major biomes

R. Toby Pennington

1



, Quentin C. B. Cronk

2

and James A. Richardson

ence of: (i) the pattern of accumulation of lineages through time; (ii) the time of origin of

monophyletic groups; (iii) when lineages arrived in different geographical areas; (iv) the time of origin of

biome-specific morphologies. This gives a powerful new view of the history of biomes that in many cases is

not provided by the incomplete plant fossil record. Dated plant phylogenies for angiosperm families such

as Leguminosae (Fabaceae), Melastomataceae sensu stricto, Annonaceae and Rhamnaceae indicate that

long-distance, transoceanic dispersal has played an important role in shaping their distributions, and that

this can obscure any effect of tectonic history, previously assumed to have been the major cause of their

biogeographic patterns. Dispersal from other continents has also been i mportant in the assembly of the

Amazonian rainforest flora and the Australian flora. Comparison of dated biogeographic patterns of plants

and animals suggests that recent long-distance dispersal might be more prevalent in plants, which has

multiple fossil calibrations. This allows cross-validation both of age estimates from different loci, and from

different fossil calibrations. For a more complete view of biome history, future studies should emphasize

full taxon sampling in ecologically important groups, and should focus on geographical areas for which few

species-level phylogenies are available, such as tropical Africa and Asia. These studies are urgent because

understanding the history of biomes can both inform conservation decisions, and help predict the effects of

future environmental changes at a time when biodiversity is being impacted on an unprecedented scale.

Keywords: molecular clocks; long-distance dispersal; plant biogeography; animal biogeography

The papers in this volume represent the proceedings of a

Royal Society Discussion Meeting held on 15–16 March

present-day maintenance of biomes is therefore an impor-

tant aspect of ecology. However, biomes are composed of

species, each of which has an individual evolutionary his-

tory. This is the point of connection between the ecology of

biomes and phylogenetics, but it is a connection that has

previously been little explored.

Understanding the history of biomes can inform conser-

vation decisions, and perhaps more critically, help predict

the effects of future environmental changes. Perhaps we

should value ancient biomes more highly than more recent

conserve ‘cradles’ of diversity where dated phylogenies tell

us the evolutionary process is active. Most importantly, a

deeper understanding of the Earth’s biodiversity is vital in

an age of projected species loss and global environmental

change. The nature and distribution of present-day biomes

determines, in large part, the functioning of the global

Author for correspondence ([email protected]).

One contribution of 16 to a Discussion Meeting Issue ‘Plant phylogeny

and the origin of major biomes’.

and one means of understanding how their component spe-

cies will react is to study the effects of past environmental

changes.

The aims of this introduction and overview are to guide

the reader through the following papers, to synthesize their

principal conclusions and to suggest some future directions

for research. The meeting generated lively discussion, and

because of the volume, range and complexity of the dis-

cussion points raised, we have taken the decision not to

include verbatim transcripts. Instead, we have used the dis-

‘biome’, a word with a long history and a multitude of uses.

We decline to be prescriptive or to give a single definition,

as the utility of the concept lies to a certain extent in its

flexibility. Instead, we give an overview of how the term has

been used.

The word biome was coined by Clements as early as

1916, but it was not taken up widely until Clements expan-

ded on the concept in a book (Clements & Shelford 1939);

it was also promoted by Carpenter (1939). Clements equa-

ted it to his concept of plant formation, but with the

to coin the term ‘ecosystem’, which included abiotic and

biotic components. It was ecosystem that came into cur-

rency as the basic unit of ecology, whereas biome was gen-

erally used for community classification at the macroscale

to indicate biotic commonality, particularly in physio-

gnomy, even over relatively diverse conditions of soil and

climate.

An example of a recent definition of the world’s major

biomes was provided by Olson et al. (2001), who modified

the definitions of Dinerstein et al. (1995) and Ricketts et al.

(vi) Mediterranean forests, woodlands, and shrub (tem-

(ix) Tropical and subtropical dry broadleaf forests.

(xi) Deserts and xeric shrublands.

sense, as it is used by most authors in this volume. Sub-

division of the biome may be obtained by incorporating

geographical and environmental data, or data on taxonomic

composition. An example of this process is the coining of

the term ‘ecoregion’, an ecologically relatively homo-

geneous region possessing a single biome. The ecoregion

was developed in Canada by Crowley (1967) and extended

by Bailey (1989). More recently, the ecoregion has been

adopted by the World Wide Fund for Nature (formerly

World Wildlife Fund, WWF) for global conservation plan-

lineages.

New theoretical developments (e.g. Sanderson 1997,

2002; Near & Sanderson 2004; Thorne et al. 1998; Thorne

& Kishino 2002) offer a means of placing a dimension of

absolute time on phylogenetic trees derived from DNA

sequence data. Potentially, this provides a powerful new

‘ecological telescope’ to view biome history that may be

able to resolve issues that the fossil record cannot. If an age

can be assigned to a single node in a phylogenetic tree using

external evidence, these new methods allow the estimation

and their origins dated. The same is true of geographical

information, which can be mapped to infer ancestral areas.

Ecology has been impacted already by the phylogenetic

revolution (e.g. Ackerly & Donoghue 1998; Webb et al.

2002) and these impacts are poised to spread to our under-

standing of biomes.

Methods used to add a dimension of absolute time to mole-

cular phylogenetic trees make considerable assumptions,

not least that the initial calibration often relies upon the fos-

(a) Calibrating nodes

Much attention has been paid in the literature to techni-

ques and algorithms that enable the calculation of relative

ages in a molecular phylogenetic tree, even in the absence

of rate constancy (e.g. Sanderson 1997, 2002; Thorne et al.

1998; Thorne & Kishino 2002). Less attention has been

paid to how we might calibrate these phylogenies by assign-

ing ages to individual nodes, but calibration is potentially

the largest source of error in the dating of phylogenetic

trees. Two principal methods of calibration have been

2001a), creation of volcanic oceanic islands (Richard-

son et al. 2001b; Plana 2004).

The high frequency of long-distance dispersal events indi-

cated for many plant groups by authors in this volume

(Richardson et al. 2004; Plana 2004; Renner 2004; Lavin et

al. 2004; Pennington & Dick 2004; see further discussion

in x 6 below) highlights the danger of the use of geological

events, especially old ones, as calibration criteria. For

example, Renner, Plana and Lavin et al. demonstrate that

endemic radiations of Melastomataceae sensu stricto,

graphical origin through time.

(b) Multiple calibration points

It would appear that well-identified fossils represent the

most objective means of assigning ages to nodes. It is

clearly desirable to be able to assign fossils to multiple

nodes, because this provides a means of cross-validating

age estimates. This approach is used in this volume by Near

(2004; who uses fossil and geological calibrations). Near &

Sanderson (2004) present a useful new algorithmic

which may be misleading the analysis. These could be

removed from the analysis entirely, but it may be that these

have been assigned to an incorrect node, and that careful

reconsideration of their morphology is necessary.

It is unfortunately true for most studies, especially those

at the species level, that at best a single fossil calibration

might be available. Near & Sanderson (2004) acknowledge

this, but suggest that the limits to accurate age estimation

result in the disappointing conclusion that accurate diver-

gence time estimates require multiple reliable calibrations.

(c) Data and algorithms

Plant phylogenies are now often generated using DNA

sequence data from multiple loci, which ideally come from

the different genomes of the chloroplast, nucleus and, more

rarely, mitochondrion. This has the theoretical benefit of

increased phylogenetic accuracy. However, most published

studies dating molecular phylogenies have used DNA

sequence data from only a single locus. This may present

improve age estimation (Yang & Yoder 2003).

The Bayesian algorithm of Thorne & Kishino (2002)

implemented by Renner (2004) in this volume offers con-

siderable flexibility in the analysis of multiple loci in esti-

mating divergence times. It also allows easy calculation of

confidence limits around estimates. Interestingly, Renner’s

thorough analysis of Melastomataceae sensu stricto using

multiple loci and multiple calibrations produced age esti-

mates that differed little from those calculated using a

single locus, single calibration, and also assuming a strict

The disparity of results presented by Renner (2004) and

Yang & Yoder (2003) suggests that any conclusion that

multiple loci are required for accurate dates may not be

general. The more accurate estimates of Yang & Yoder

(2003) based upon multiple loci might just reflect the loci

and taxa studied. There is a clear need for more empirical

studies on systems that offer both numerous calibrations

and loci to assess the desirability of basing age estimates on

multiple loci and calibrations. We suggest that it is sensible

to calculate ages based upon separate loci to cross-validate

lengths, and the algorithm used to calculate ages may be

somewhat trivial relative to the choice of loci. It is vital that

there is a good level of nucleotide substitution for the

terminal taxa studied. If terminal taxa are separated by few

nucleotide subsititutions, and branch lengths in a molecu-

lar phylogenetic tree are short, all algorithms will produce

age estimates for adjacent nodes with wide confidence lim-

its that overlap. This makes estimation of the sequence of

cladogenesis in absolute time impossible.

areas (Pennington & Dick 2004; Renner 2004; Crisp et al.

2004; Hill 2004), so understanding their biotic history

requires a broad geographical outlook. This breadth of

geographical view is the strength of the studies of species-

rich, cosmopolitan, ecologically important families pre-

sented here by Renner (Melastomataceae sensu stricto),

Lavin et al. (Leguminosae) and Richardson et al. (Anno-

naceae and Rhamnaceae).

Melastomataceae sensu stricto (3000 species) are gener-

ally pioneer and understorey trees and are an important

despite the fact that it belongs to an ancient family (82–

91 Myr old), whose biogeography was certainly influenced

by continental drift. The most extreme examples are

Rhamnaceae and Leguminosae, which are shown to have

undergone multiple transoceanic dispersal events.

The paper by Lavin et al. (2004) on legumes estimates

the ages for 59 ‘transcontinental crown clades’. Each is a

monophyletic group that comprises sister clades separated

by an oceanic barrier or large expanse of continental area,

such as one clade in North America, and its sister in South

producing these distribution patterns, and long-distance

dispersal is likely to be the primary explanation. Lavin

et al. (2004) suggest that the unified neutral theory of

biodiversity and biogeography of Hubbell (2001) might be

a potential explanation of the geographical and ecological

structure in the legume phylogeny because it is a dis-

clear that long-distance dispersal must have had a substan-

tial influence on the historical construction of the tree flora

diaceae, which have dispersed into Australia from overseas,

again emphasizing an important role for long-distance dis-

persal.

speciation

Jacobs (2004) reviews the palaeontological history of

African rainforest, woodland and savannah biomes. She

concludes that biomes corresponding with contemporary

rainforest and woodland did not originate until the Eocene,

when there is clear evidence for the presence of families

biome did not begin to expand until the Middle Miocene

(ca. 16 Myr ago), and did not become widespread until the

Late Miocene (ca. 8 Myr ago). This review provides a

background and time-scale against which Plana (2004)

evaluates evidence for the time-scale of species diversifi-

cation in West African (Guineo-Congolian) rainforests,

though her review also touches upon savannahs and East

African rainforests.

The high species richness of tropical rainforests was

quickly noted by nineteenth-century naturalists. As early

as 1878, Alfred Russell Wallace suggested a central role

result of a comparatively continuous and unchecked

development of organic forms; while in the temperate

regions there have been a series of periodical checks and

extinctions’. This inspired later papers from Dobzhansky

(1950) and Fischer (1960) that led to an influential view-

point that the high diversity of tropical rainforests resulted

data from Africa and the Neotropics showed that during

the Ice Age climates became drier and cooler (Aubre

ville

Plana (2004) evaluates the evidence from the few avail-

able dated phylogenies of African rainforest plant species

for Pleistocene species diversification. She presents strong

during the Pleistocene, and 40% since the beginning of the

Pliocene. Crisp et al. (2004) review dated phylogenies that

show massive speciation over the past 5 Myr in taxa such as

Chenopodiaceae, Brassicaceae and Gossypium that are

characteristic of the arid communities (Eremean biome) of

Australia. Recent, rapid speciation has clearly played a role

in producing a substantial proportion of the plant species

forest and savannah species, that the rainforest species are

often phylogenetically basal members of Miocene lineages.

Clearly, the species richness of African rainforests is the

result of a combination of both relatively ancient and more

recent speciation, a conclusion echoed in a study of season-

ally dry forest in the Neotropics (Pennington et al. 2004),

and in this volume for the Fynbos (Linder & Hardy 2004)

and Australian sclerophyll biomes (Crisp et al. 2004).

(c) Amazonian rainforest: the importance of

immigrants

America until the early Tertiary. At this time, South Amer-

ica was an island continent, and had been isolated since it

split from Africa ca. 100 Myr ago. It had been widely

assumed that the contemporary flora of the Amazon rain-

forest was derived almost entirely from lineages isolated by

this vicariance event (Gentry 1982; Young et al. 2002).

However, Pennington & Dick (2004) show that the Ama-

zon tree flora contains a substantial proportion (ca. 20%) of

species that are members of groups shown by molecular

phylogenies to have arrived in South America long after the

(d) The Cape Floristic Region

The ‘Fynbos Biome’ of the Cape Floristic Region is

noted for its high levels of plant species diversity and

endemism that is limited to relatively few clades. Some

authors even consider its number of endemic genera and

families sufficient to warrant describing it as a floristic king-

dom (Good 1974). The origin of this diversity is thought to

be a response to progressive aridification in the Pliocene

and Late Miocene that eliminated the tropical flora that

once occupied this region (Levyns 1964; Linder et al. 1992;

did not provide precise details of their initiation. Now we

can use molecular data in combination with the fossil rec-

ord and geological events to get more precise information

concerning the timing of these radiations. Linder & Hardy

(2004) present a phylogenetic analysis of the largest clade

of Restionaceae that is found in the Cape and demonstrate

that radiation was initiated between 20 and 42 Myr ago.

The starting dates for the radiation of seven other clades

were also evaluated with an estimated range of between 7

and 20 Myr ago. Linder and Hardy suggest that these

and Goldblatt (1997) hypotheses that state that the mod-

ern species richness of the Cape flora is the result of radi-

ation into the arid habitats of the western portion of the

Cape Floristic Region. The initiation of these radiations

appears to be lineage specific starting in the Late Oligo-

cene, and continuing through the Late Miocene into the

Hemisphere forests with those reviewed for animals in the

same areas by Sanmartı

n et al. (2001). There are striking

´

plant immigrant lineages enter a community.

The final paper (Davies et al. 2004) takes a different

approach and asks why some biomes contain more species.

There may be contemporary ecological explanations for

this phenomenon in that different biomes may allow

differing numbers of species to coexist. However, why did

mote higher species richness nearer the Equator. The

amount of energy input into a system may limit the number

of species that coexist in an area or influence evolutionary

rates. Davies et al. (2004) demonstrate that angiosperm

families that are exposed to a high-energy load tend to be

both more species rich and possess faster evolutionary

rates, although it is not known whether one drives the

among lineages. This may be due to interactions between

biological traits and the environmental conditions in which

Webb et al. 2002). In this final section we highlight some

research areas that we see as especially important.

Because dating phylogenies relies on assessing amounts

of change in DNA sequences, much attention has been

focused upon issues such as correcting for deviations from

a strict ‘molecular clock’ (Sanderson 1997, 2002; Thorne

et al. 1998; Thorne & Kishino 2002; Near & Sanderson

2004). However, much discussion at the meeting, includ-

ing points raised by Michael Donoghue (Yale University),

Susanne Renner (Ludwig Maximilians University Munich)

and Sir Peter Crane (Royal Botanic Gardens, Kew) high-

lighted the vital role to be played by both fossils and

morphological data.

Crane pointed out that though the method presented by

Near & Sanderson (2004) provides a means of identifying

possibly misleading fossil calibrations, this should not sub-

stitute for careful study of fossils before they are used in an

analysis. Different fossils will have different values as

cific phylogenetic node because they have more character

information than isolated organs; (ii) the age precision

offered by fossils varies greatly, representing an age range

rather than an absolute date, and the best fossils for cali-

bration will have smaller age ranges.

Crane also highlighted the importance of the analysis of

morphological data to accurately position fossils in a phylo-

genetic framework. The theoretical ideal might be to score

the fossil and extant taxa for morphological characters, and

infer the placement of the fossil taxa from a morphological

morphological characters (Scotland et al. 2003), and miss-

ing data for fossils in a simultaneous analysis because their

DNA cannot be sequenced. This may cause their position

in the resulting phylogenetic trees to be unstable (Platnick

et al. 1991). However, these approaches will be impossible

in most cases simply because the fossils are fragmentary

cular–morphological phylogenetic analysis, allowing indi-

vidual morphological character states to be assigned to

(b) The value of phylogenetic studies of ecologically

The purpose of this volume is to investigate the history of

biomes, but many of the phylogenetic studies presented

arose originally from unrelated studies focused on the sys-

tematics of different groups. The papers in this volume by

Lavin et al. (2004; Leguminosae), Renner (2004;

Melastomataceae sensu stricto), Plana (2004; Begonia), Lin-

der & Hardy (2004; Restoniaceae), Richardson et al.

(2004; Annonaceae and Rhamnaceae), Gravendeel et al.

are generally small trees that form an important component

of the rainforest understorey virtually worldwide. Begonia

species are ubiquitous terrestrial herbs or trunk epiphytes

in tropical rainforests worldwide. Phylogenies of these

groups are complementary because they permit inference

of the history of different aspects of the rainforest biome.

However, for all these taxa, the full potential of their phylo-

genies will not be realized until taxon sampling is more

complete. For example, sampling of Begonia species out-

side of Africa–Madagascar is sparse (Plana et al. 2004),

tropical and Asian Begonia species would be a massive

undertaking. The collection of tall, legume trees is even

tougher. However, the scientific value of better-sampled

phylogenies would repay the investment.

(c) Geographical areas with few phylogenies

phylograms allowing an assessment of radiation history.

Linder & Hardy (2004) review evidence from eight phylo-

genies for clades largely endemic to the Cape Region of

South Africa, and a recently established collaborative net-

work for the study of the evolution of the Cape flora will

ensure that more phylogenies will appear soon (J. A. Haw-

kins, personal communication). By contrast, the review of

the main expanse of African rainforest in the Guineo-

Congolian region by Plana (2004) deals with phylogenies

of two plant genera. Similarly, we are aware of very few