Content uploaded by Roderic Page
Author content
All content in this area was uploaded by Roderic Page
Content may be subject to copyright.
88
The Auk 119(1):88–108, 2002
SEABIRD SUPERTREES: COMBINING PARTIAL ESTIMATES OF
PROCELLARIIFORM PHYLOGENY
M
ARTYN
K
ENNEDY
1
AND
R
ODERIC
D. M. P
AGE
Division of Environmentaland Evolutionary Biology, Institute of Biomedical and Life Sciences,
University of Glasgow, Glasgow G12 8QQ, United Kingdom
A
BSTRACT
.—The growing use of comparative methods to addressevolutionary questions
has generated an increased need for robust hypotheses of evolutionary relationships for a
wide range of organisms. Where a phylogeny exists for a group, often more than one phy-
logeny will exist for that group, and it is uncommon that the same taxa are in each of the
existing trees. The types of data used to generate evolutionary trees can also vary greatly,
and thus combining data sets is often difficult or impossible. Toaddress comparative ques-
tions for groups where multiple phylogenetic hypotheses already exist, we need to combine
different hypotheses in a way that provides the best estimate of the phylogeny for that group.
Here, we combine seven seabird phylogenies (based on behavioral, DNA–DNA hybridiza-
tion, isozyme, life history, morphological, and sequence data) to generate a comprehensive
supertree for the Procellariiformes using matrix representation with parsimony. This phy-
logeny contains 122 taxa and represents a conservative estimate of combined relationships
presented in the original seven source trees. We compared the supertree with results of a
combined sequence data supermatrix for 103 seabird taxa. Results of the twoapproaches are
broadly concordant, but matrix representation with parsimony provides a more compre-
hensive and more conservative estimate of the phylogeny of the group because it is lessin-
fluenced by the largest of the source studies (which uses a single, relativelyquickly evolving
gene). Genetic data sets that can be combined in a supermatrix approach are currently less
likely to be available than phylogenies that can be combined using some form of supertree
approach. Although there are limitations to both of those approaches, both would be simpler
if all phylogenetic studies made both their data sets andtreesthey generate availablethrough
databases such as TREEBASE. Received 8 December 2000, accepted 21 September 2001.
R
ESUMEN
.—El uso creciente de me´todos comparativos para abordar preguntasevolutivas
ha incrementado la necesidad de hipo´tesis robustas sobre relaciones evolutivasparaunaam-
plia variedad de organismos. Cuando existe una filogenia para un grupo, a menudo existira´
ma´s de una filogenia para ese grupo, y es poco frecuente que los mismostaxaeste´n presentes
en cada uno de los a´rboles. Los tipos de datos usados para generar a´rbolesevolutivos tam-
bie´n pueden variar enormemente, por lo que generalmente es difı´cil o imposible combinar
estos datos. Para responder a preguntas comparativas en grupos para los cuales existen mu´l-
tiples hipo´ tesis filogene´ticas, necesitamos combinar diferentes hipo´tesis de manera que ob-
tengamos la mejor estimacio´n de la filogenia de estos grupos. Aquı´ combinamos siete filo-
genias (basadas en comportamiento, hibridizacio´n de ADN–ADN, isoenzimas, historia de
vida, morfologı´a y datos de secuenciamiento) para generar un super-a´rbol integrador para
los Procellariiformes usando representacio´n de matrices combinada con parsimonia. Esta
filogenia contiene 122 taxa y representa una estimacio´n conservativa de las relacionespre-
sentadas en los siete a´rboles originales combinados. Comparamos el super-a´rbolcon losre-
sultados de una matriz de datos de secuencias combinada para 103 taxa de aves marinas. A
grandes rasgos, los resultados de las dos aproximaciones concuerdan, pero la representacio´n
de matrices combinada con parsimonia brinda una estimacio´n ma´s integral y ma´s conser-
vativa de la filogenia del grupo, debido a que esta´ menos influenciada por unodelos estudios
utilizados (la fuente con ma´s datos, que a su vez usa unu´nico gen de evolucio´nrelativamente
ra´ pida). Los juegos de datos gene´ticos que pueden ser combinadosen una super-matrizesta´n
por lo general menos disponibles que las filogenias que pueden ser combinadas usandoal-
guna aproximacio´n de tipo super-a´rbol. Aunque existen limitaciones para estas dos apro-
ximaciones, ambas serı´an ma´s simples si todos los estudios filogene´ticos pusieran a dispo-
1
E-mail: martyn.kennedy@bio.gla.ac.uk
January 2002] 89Seabird Supertrees
sicio´n tanto sus juegos de datos como los a´rboles que generan a trave´s de bases de datos
como TREEBASE.
R
ELATIONSHIPS WITHIN THE
tubenose sea-
birds (Procellariiformes) are of general interest
to many biologists, at least in part becausethey
are a diverse and wide-ranging group,and be-
cause they are commonly found in most oce-
anic regions of the world. Nunn and Stanley
(1998) generated a phylogeny for 85 species of
tubenose seabirds for their discussion ofeffects
of body size on rates of molecular evolution. By
being such a diverse group (e.g. showing such
a great range in body size), these birds allow
comparative questions like that of Nunn and
Stanley (1998) to be readily addressed when a
phylogeny is available. As well as ranging
greatly in size, the tubenose seabirds are par-
ticularly interesting because they are behavior-
ally and ecologically very diverse, and they
provide a model system for investigating cos-
peciation (e.g. Paterson et al. 2000).
In addition to Nunn and Stanley’s (1998)
phylogeny for tubenose seabirds, there aresev-
eral other phylogenies available for that group.
Some of those phylogenies include taxa not
present in Nunn and Stanley’s (1998) study,
whereas some also disagree with the relation-
ships found in their study. The storm petrels
(Hydrobatidae), for example, unexpectedly do
not form a monophyletic group in Nunn and
Stanley’s(1998)phylogeny.Giventhatthereare
a number of phylogenies available for that
group, precisely what the best estimate of the
phylogeny is remains uncertain.
Within the last few decades, there has been a
dramatic increase in number of studies using
phylogenies to address a wide range of issues.
Those issues include, for example, behavior
(e.g. Zyskowski and Prum 1999), biogeography
(e.g. Kennedy and Spencer 2000), coevolution
(e.g. Paterson et al. 2000), genetic systems (e.g.
Cruickshank and Thomas 1999), language (e.g.
Gray and Jordan 2000), rates of molecular evo-
lution (e.g. Johnson and Sorenson 1998), speci-
ation(e.g. Friesen and Anderson1997), and tax-
onomy (e.g. Kennedy et al. 1999). Because
biologists are becoming more convinced ofthe
utility of taking a phylogenetic approach to
questions they wish to address, robust hypoth-
eses about phylogenetic relationships for the
taxa of interest are required.
Even with ongoing advances in molecular
technology, phylogenetic hypotheses do notex-
ist for most of the world’s taxa. When phylog-
enies do exist for a group, they will often not
include all taxa of interest to the researcher. A
single phylogeny for the taxa of interest is not
always available, and studies have sometimes
had to combine two or more phylogenies to ob-
tain a tree that contains all the taxa (Sanderson
et al. 1998). Kennedy et al. (1996), for example,
had to combine four different source trees (two
generated from morphological data and two
from DNA–DNA hybridization data) to inves-
tigate homology of pelecaniform behaviors by
mapping them onto the best estimate of that
group’s phylogeny.
A tree that results from combination ofmul-
tiple source-tree topologies has been termed a
‘‘supertree’’ (Sanderson et al. 1998). Intuitively,
the ideal way of combining several source trees
wouldappear to be to combine all source phy-
logenies’ data matrices into a single ‘‘super-
matrix’’thatcouldthenbeanalyzedtoestimate
the phylogeny. Sanderson et al. (1998) note,
however, that the supermatrix approach will
often not be tenable because of the cost in-
volved with filling in gaps in the data as well
as difficulties associated with combining some
types of data. They point out, for example, that
DNA–DNA hybridization data would not be
able to be included in a supermatrix, and that
homologizing characters would become in-
creasingly difficult as the size ofthe matrix in-
creased as more distantly taxa were added
(Sandersonet al. 1998).
The supertree approach offers an alternative
to the supermatrix approach. One method for
constructing supertrees is matrix representa-
tion with parsimony (MRP; Baum 1992, Ragan
1992). Matrix representation with parsimony
converts topologies of individual source trees
into a data matrix (for a general explanation,
see Sanderson et al. 1998). Once matrices for
each of the source trees are combined, super-
trees can be found using parsimony analysis.
Because matrices are derived from the source
trees’ topologies, MRP allows different data
types (e.g. sequences, morphology, behavior,
allozymes, DNA–DNA hybridization) to be
combined (Bininda-Emonds and Bryant 1998).
90 [Auk, Vol. 119K
ENNEDY AND
P
AGE
→
F
IG
. 1. The topologies of the source trees: (A) Austin (1996)—307 basesof cytochromeb; (B) Bretagnolle
et al. (1998)— 496 bases of cytochrome b; (C) Heidrich et al. (1998)— 1,043 bases of cytochromeb; (D) Imber
(1985)—morphology and life history information; (E) Nunn and Stanley (1998)— 1,143 bases of cytochrome
b; (F) Paterson et al. (1995)— 12S rRNA, isozyme, and behavioral life history data; and (G) Sibley andAhlquist
(1990)—DNA–DNA hybridization data. We have followed the classificationof Sibleyand Munroe (1990)for
consistency, but have used four genera accepted in the source phylogenies: Pseudobulweria is Pterodroma in
Sibley and Munroe (1990); Phoebastria and Thalassarche are Diomedea in Sibley and Munroe (1990); and Ha-
locyptena is Oceanodroma in Sibley and Munroe (1990). The taxa highlighted by a shaded background are
unique to that source tree.
1
Not classified in Sibley and Munroe (1990).
2
Not used in the analysis.
3
No species
name was associated with the genus in Sibley and Ahlquist’s (1990) tree. Because for the MRP analysis the
taxa labels must all contain both generic and specific names, the first species of the genus named in Sibley
and Munroe’s (1990) classification (see Appendix) that occurred in another of the source trees was allocated
to the genus.
Whereas some studies have had to cobble mul-
tiple source trees together (e.g. Kennedy et al.
1996), MRP uses an explicit optimality criteri-
on. As well as providing an optimal solution
(or solutions) to combining multiple source
trees, MRP allows different coding schemes
and different weights to be applied to certain
‘‘characters’’ (see Sanderson et al. 1998). Nodes
from source trees (which are effectively char-
acters in the MRP matrix), for example, could
be differentially weighted depending on the
level of support that the nodes received in
source trees. Whereas supertree and consensus
approaches have limitations (Steel et al. 2000),
and there is some debate over how to best con-
struct phylogenetic supertrees (e.g. seeSander-
son et al. 1998, Bininda-Emonds et al. 1999),
MRP is currently the most commonly used
method in construction of large supertrees.
Constructing supertrees from multiple
source trees is not necessarily a trivial exercise.
Bininda-Emonds et al. (1999), for example,
used 177 literature sources to construct super-
trees for the 271 extant species of Carnivora.
Finding all source trees for a large number of
taxa is potentially an extremely time consum-
ing task, without considering thatsource trees
then need to be coded into an appropriatesin-
gle-matrix format, and then analyzed.Finding
source trees would be an easier task if all stud-
ies that generate phylogenies were submitted
to a phylogeny database such as TREEBASE
(see Acknowledgments). If all trees for a group
were available in NEXUSfileformat(Maddison
et al. 1997) from TREEBASE, source trees could
then simply be combined intoaMRPmatrixus-
ing software such as RADCON (Thorley and
Page 2000). RADCON will takeall source trees
and output them as a single MRP matrix writ-
ten in NEXUS format ready for parsimony
analysis. An alternative (and supplementary)
approach to searching phylogeny databases is
to search the Web of Science (see Acknowledg-
ments)orsomeotherliteraturedatabaseforrel-
evant journal references and thus search for
any articles that contain phylogenies for the
group of interest.
In this article, we compare results of a MRP
analysis of several phylogenies available for tu-
benose seabirds with a supermatrix approach
using available sequence data. Because several
phylogenies available for members of the pro-
cellariiforms were generated from the same
typeof data (i.e.mitochondrialDNAsequence),
that group offers an ideal opportunity to com-
pare supertree and supermatrix approaches. A
robust, comprehensive, and conservative esti-
mate of what is currently known about thephy-
logeny of that group of birds will also allow fu-
ture comparative analyses to be readily
performed.
M
ETHODS
Data. We combined topological information from
seven different studies (see Fig. 1) using MRP. In ad-
dition to our preexisting knowledge of theliterature
(e.g. Sibley and Ahlquist 1990), we searched the Web
of Science in order to comprehensively surveythe lit-
erature for procellariiform phylogenies. Our initial
search criteria were the terms phylogen* and sea-
bird* (the asterisk wildcard allows any word with
that root to be found). After obtaining the list of
studies that included derivatives of both of those
terms, we investigated appropriateness of each of the
individual studies using their abstracts. For each in-
dividual study, we also surveyed studies that cited it
January 2002] 91Seabird Supertrees
92 [Auk, Vol. 119K
ENNEDY AND
P
AGE
F
IG
. 1. Continued.
January 2002] 93Seabird Supertrees
F
IG
. 1. Continued.
94 [Auk, Vol. 119K
ENNEDY AND
P
AGE
F
IG
. 1. Continued.
January 2002] 95Seabird Supertrees
and all studies cited in it. Following that procedure
we found the journal reference for all but one of the
source trees. We continued with that procedureusing
the combination of each of phylogen*, clado*, taxo-
nom*, and cladistic* with the terms seabird*, Proce-
llariiformes, diomedeidae, procellariidae, hydrobati-
dae, pelecanoididae, albatross*, petrel*, shearwater*,
storm-petrel*, and diving-petrel*. The combination of
taxonom* and petrel* provided the reference to the
study (Imber 1985) that the other searches had not
found.
Some studies found were not used as source trees
because they were by the same authors and the taxa
and data used overlapped between studies (i.e. Pat-
erson et al. 1993, 1995; Heidrich et al. 1996, Nunn et
al. 1996, Heidrich et al. 1998, Nunn and Stanley
1998). In this situation, we used the more compre-
hensive of the studies as our source trees. So that
each study had a single taxon for each species, we
pruned the additional subspecies or representatives
of the same species from the source trees. The source
trees were generated using TREEVIEW (Page 1996)
to provide NEXUS formatted files that could be
translated into a MRP matrix in RADCON. Because
of the reduced number of terminal taxa in some of
our source trees, they are not necessarily identicalto
those published in theoriginal papers, but represent
abridged versions of the originals’ topologies.
When the authors presented more than one topol-
ogy, we attempted to use the topology that they in-
dicated was the best estimate of their phylogeny.For
example, from Austin (1996) we used what the au-
thor described as a conservative hypothesis for the
evolutionary relationships of the Puffinus shearwa-
ters (figure 5 in Austin 1996; Fig. 1A). Austin (1996)
used a 307 base-pair (bp) fragment of cytochrome b
to construct his phylogeny. We used the maximum
parsimony (MP) tree found by Bretagnolle et al.
(1998, figure 3; Fig. 1B) using a 496 bp fragment of
cytochromebWheneach speciesiscollapsedtoasin-
gle representative, the topologies of MP and maxi-
mum likelihood (ML) trees of Heidrich et al. (1998,
figure 2B, C; Fig. 1C) are equivalent. Heidrich et al.
(1998) used a 1,043 bp of cytochrome bto generate
their phylogeny. Imber (1985, figure 8) presented a
single topology (Fig. 1D) for the Gadfly Petrels (Pte-
rodroma) based on morphological and life-history in-
formation (the characters are discussed in his paper,
but there is no explicit data set or criterion for con-
struction of the phylogeny). Nunn and Stanley (1998,
figure 2) presented a consensus of their four equally
weighted MP trees (Fig. 1E) which were generated
from whole (1,143 bp) cytochrome bsequences. Pat-
erson et al. (1995) used a combination of mitochon-
drial 12S ribosomal RNA gene-sequence data, iso-
zyme information, and behavioral life-history data
to construct their best estimate of phylogeny (figure
4 in Paterson et al. 1995; Fig. 1F). Finally, Sibleyand
Ahlquist (1990, figure 368) used UPGMA on DNA–
DNA hybridization data to generate a hypothesis for
the evolutionary history of some of the seabirds (Fig.
1G). The procellariiform taxa, outgroups, their com-
mon names, and which source tree(s) contain them
are shown in the Appendix.
Because several source studies used mtDNA to
construct their trees, it is possible to compareresults
of our MRP analysis with results from analysis of a
character supermatrix for a large proportion of the
taxa. To construct that character supermatrix, we
downloaded the mtDNA data available for this
group from GenBank. The majority ofsequences for
this combined data set come from the whole cyto-
chrome-b-based study of Nunn and Stanley (1998; 90
taxa, this data set is in TREEBASE, study accession
number S351 and matrix accession number M468).
Some additional taxa were added using the partial
cytochrome bsequences of Austin (1996; Puffinus au-
ricularis,Pu. gavia,Pu. mauretanicus,andP. yelkouan)
and Bretagnolle et al. (1998; Pseudobulweria aterrima,
Ps. rostrata,Pterodroma baraui). We used the largest
fragmentofcytochromebavailabletousforeachtax-
on. Any additional taxon for which sequences orig-
inated in Heidrich et al. (1998) could not be included
because their sequence does not appear to havebeen
submitted to GenBank/EMBL. Heidrich et al. (1998)
did, however, use an unpublished sequence from
GenBank (Pu. tenuirostris, Nunn and Zino,accession
number U74352) and a sequence from Nunn et al.
(1996; Procellaria cinerea, accession number U48940)
that we have includedin our mtDNA data set.
Sequence data from another gene was added by in-
cluding the 12S rRNA data of Paterson et al. (1995),
givingthemtDNAdata set a total of 1,509 characters.
The aligned 12S data was obtained from the author.
In addition to 12 taxa for which there are cytochrome
bdata, the Paterson et al. (1995) data set includes an
additional four penguin species for which no cyto-
chrome bdata is available.
Analysis. The source-tree topologies were com-
bined and converted into a NEXUS file with a matrix
suitable for parsimony analysis using the ‘‘compo-
nent coding’’ option of MRP Supertree Consensus in
RADCON (Thorley and Page 2000). The resulting
MRP data set had 122 taxa and 188 ‘‘characters.’’ In
a MRP data set, the characters represent topologies
of source trees, where each node from a sourcetree
providesonecharacterto the matrix(Sanderson et al.
1998). That is, for each of the source trees, the taxa
present in the group derived from any given node
are coded as 1 for that node, whereas thetaxa not in
that group are coded as 0 for that node. All other taxa
(those present in one or more of the other source
trees,butnottheonebeingcoded)arecodedasmiss-
ing for that node. RADCON thus codes nodes for
each of the source trees for all taxa and combines
those into a single matrix. This data matrix was an-
alyzed using P
AUP
* 4.0b4a (Swofford 2000). Char-
acters were equally weighted and the starting-tree
96 [Auk, Vol. 119K
ENNEDY AND
P
AGE
January 2002] 97Seabird Supertrees
←
F
IG
. 2. The consensus MRP trees: (A) Adams consensus, (B) strict consensus. Each of the most parsi-
monious MRP trees had a length of 214, a CI of 0.8785, and a RI of 0.9681. The italicized letters associated
with branches of the Adams consensustree refer to groups mentioned in the text.
for branch swapping (TBR) wasobtained for the heu-
ristic search by stepwise addition using a random-
addition sequence. We defined the outgroup as the
hypothetical taxon RADCON constructs for thisuse.
This taxon (called ‘‘MRP outgroup’’) is added when
the MRP matrix is constructed and all of its charac-
ters are zeros (i.e. it shares no nodes with any of the
other taxa). The MRP data set is available in
TREEBASE.
The combined mtDNA data set has 103 taxa. We
used the partition-homogeneity test (Farris et al.
1995) to investigate whether the cytochrome-band
12S rRNA sequence can be analyzed as a single data
set. Only those 12 taxa for which we had data for
both genes were able to be used in this analysis. We
defined penguins as the outgroup for all mtDNA
analyses. To find the appropriate value to relatively
weight transitions and transversions, we evaluated
different ML models for the data set using MODEL-
TEST (Posoda and Crandall 1998). The best-fitting
model that estimates the transition to transversion
ratio was the HKY85 (Hasegawa et al. 1985) with
both proportion of invariable sites and gamma shape
parameter being estimated. For this model, the tran-
sition to transversion ratio was estimated at 10:1. We
thus weighted transversions 10:1 over transitionsfor
all our subsequent analyses and the starting-tree for
branch swapping (TBR) was obtained for the heuris-
tic search by stepwise addition using 100 random-
addition sequences. For reasons discussed below,the
same analyses were also conducted with the four
penguin species from the 12S rRNA data set exclud-
ed from our mtDNA data set. For this versionof the
data set, we investigated the strength of the associ-
ations implied by the sequencedata using bootstrap
analysis (Felsenstein 1985). For the bootstrap analy-
sis 1,000 replicates were performed using a fast-heu-
ristic search. The combined mtDNA data set isavail-
able in TreeBASE.
R
ESULTS
Matrix representation with parsimony. After
exploratory searches suggested that there were
at least 100,000 equally parsimonious trees, we
set the maximum number of trees to 10,000. We
stopped 10 searches with different random-ad-
dition sequences at this number of MP trees in
an attempt to ensure that we had found the op-
timal trees. We also ran additional exploratory
searches with other random-additionsequenc-
es, but found no trees shorter than 214 steps.
We compared each sample of 10,000 MP trees
to ensure they each gave representative consen-
sus trees. Nine of our 10 different random-ad-
dition sequence searches were stopped at
10,000 trees of 214 steps each, whereas theoth-
er search found 10,000 suboptimal trees. Each
of the optimal sets of 10,000 MP trees were
summarized as both Adams and strict-consen-
sus trees. All of the strict-consensus treeswere
identical (see Fig. 2B), but there wassome var-
iation in the Adams consensustrees. Fiveof the
Adams consensus trees were identical and this
topology was the most conservative (i.e. re-
solved the fewest branches) of the three topol-
ogies and thus we present that topology here
(see Fig. 2A).
Mitochondrial DNA. The partition-homoge-
neity test showed that there was no significant
difference in phylogenetic signal between the
cytochrome-band 12S sequences (1,000 repli-
cates, P50.69). We ran a series of exploratory
searches with different random-addition se-
quences and found no trees shorter than 8,636
steps. A completed heuristicsearchfound1,458
trees of 8,636 steps and those are summarized
as a strict (Fig. 3) consensustree.
The unusual result from the mtDNA super-
matrix of the nonmonophyly of the penguins
(see Fig. 3) is caused by none of the penguin
species having both cytochrome band 12S
rRNA sequences. The four penguin species
from the Paterson et al. (1995) 12S rRNA-based
study grouped with one of the storm-petrel
groups (see Fig. 3) within the procellariiforms.
Initial exploratory equally weighted analyses
had also placed these penguins within proce-
llariiforms, and it appeared that their position
within the trees may have been constrained by
the unexpected result (Fig. 1F) of the 12S se-
quence grouping the White-faced Storm-Petrel
(Pelagodroma marina) with the albatrosses (Dio-
medea epomophora and Thalassarche bulleri). In
further equally weighted analyses, we re-
moved the 12S sequence for the White-faced
Storm-Petrel and found that the problematic
penguins were placed more basally in the tree
98 [Auk, Vol. 119K
ENNEDY AND
P
AGE
←
F
IG
. 3. The strict consensus of 1,458 MP trees for
the mtDNA supermatrix with transversion weighted
10:1 over transitions. Each of the MP trees had a tree
length of 8,636, a CI of 0.3654, and a RI of 0.8129.
(though not necessarily with the other pen-
guins). Thus, the 12S sequence of the White-
faced Storm-Petrel appears to have pulled the
penguins of Paterson et al. (1995) inside of the
albatrosses, and thus away from the other pen-
guins. Because none of the penguins have se-
quence for both cytochrome band 12S, the
mtDNA supermatrix is only able to place the
penguins of Paterson et al. (1995) in a general
area of the tree that is loosely constrained by
positions of the other taxa for which there is
12S data. Furthermore, this inability to resolve
the position of these penguins generatesa great
number of additional optimal tree topologies.
Hence, we excluded the penguins that have
only 12S sequence from the subsequent super-
matrix analyses.
With the penguins of Paterson et al. (1995)
excluded from the analysis, we again ran a se-
ries of exploratory searches with different ran-
dom-addition sequences and found no trees
shorter than 8,464 steps. A completed heuristic
search found 18 trees of 8,464 steps, and those
are summarized as a strict (Fig. 4) consensus
tree. The bootstrap values (above 50%) are
shown on the branches of the strict-consensus
tree. Five branches not shown in the consensus
tree were supported by the bootstrap analysis:
Pygoscelis antarctica and Py. papua (60% sup-
port), Oceanodroma tethys and Halocyptena mi-
crosoma (80% support), Oc. tethys1Halocyptena
microsoma and Oc. melania (66% support), Phoe-
bastria albatrus and Ph. irrorata (53% support),
and Pagodroma nivea and Thalassoica antarctica
(60% support).
D
ISCUSSION
As a combined summary of existing knowl-
edge of evolutionary relationships of procella-
riiforms, both types of consensus tree for the
MRP analysis are well resolved. Given the
large number of equally optimal MRP trees
generated from the combination of several
source trees, both Adams and strict-consensus
trees (Fig. 2) represent conservative estimates
January 2002] 99Seabird Supertrees
of what is currently known about the phylog-
enyforthisgroup.Nounexpected,orwhatmay
be perceived to be wrong, groups were recov-
ered for the procellariiforms in this analysis—
although there are examples of circumstances
when that may occur (e.g. see Bininda-Emonds
and Bryant 1998). The Adams and strict-con-
sensus trees (shown in Fig. 2) differ only in
their level of resolution. The strict-consensus
tree collapses any branch that does not exist in
all 10,000 of the optimal trees, whereas theAd-
ams consensus tree retains the general struc-
ture common to all of the MRP trees. Adams
consensus places a taxon that is difficult to re-
solve (i.e. occurs in different places in the dif-
ferent optimal trees) as a polytomy at themost
basal node from which it is derived in all op-
timal trees. Thus the polytomy indicates that
the taxon is a member of that group in all op-
timal trees, but that it cannot be placed more
precisely. Adams consensus thus retains more
informative structure than the strict-consensus
treebysummarizing similarity ofoptimaltrees
and indicating which groups the difficult-to-
place taxa fall within. Because the Adams con-
sensus provides a more resolved estimate of
what is known about evolutionary relation-
shipsof the procellariiformsthanthestrictcon-
sensus tree, it provides a more useful starting
point for any future comparative analyses that
may be conducted with that group. One pos-
sibility, for example, is that taxa identified as
difficult to place by Adams consensuscouldbe
removedfrom a subsequentanalysisinorderto
obtain a more resolved phylogeny for any fu-
ture comparative analyses.
Most of the expected major groups were re-
covered by the MRP analysis, but monophyly
of storm-petrels (labeled aand bFig. 2A) re-
mains unresolved. This result is not surprising
given that the source tree that providesalmost
all information about storm-petrels separated
them into two groups (Fig. 1E; Nunn andStan-
ley 1998). The only other information about
storm-petrels in the source treessupports their
monophyly, but is a single node from the Sibley
and Ahlquist (1990) tree (grouping Oceanites
oceanicus with three Oceanodroma species). As
expected groups such as the albatrosses (cFig.
2A), diving petrels (dFig. 2A), gadfly petrels (e
Fig. 2A), prions (fFig. 2A), and shearwaters(g
Fig. 2A) were all monophyletic. The position of
some taxa, however, is difficult to resolve. The
differences between Adams and strict-consen-
sus trees make it apparent which taxa are dif-
ficult to place. The position of the Grey Petrel
(Procellaria cineria), for example, is difficult to
resolve (Fig. 2). The source trees suggest that it
could either be sister taxon to Bulwer’s Petrel
(Bulweria bulweria, Fig. 1B; Bretagnolle et al.
1998) or sister taxon to shearwaters (Fig. 1C;
Heidrich et al. 1998). Contradictory evidence
provided by those two source trees and lackof
a source tree that includes the Grey Petrel and
other Procellaria species makes it impossible to
resolve its position. Relationships within the
other Procellaria species and with the Pseudo-
bulweria species, however, are stable and re-
solved (if the position of the Grey Petrel is not
considered) which is better represented in the
Adams consensus tree than the strict-consen-
sus tree.
We would not, however, unconditionallyrec-
ommend accepting all relationships shown in
the Adams consensus tree that are unresolved
in the strict-consensustree. Relationshipswith-
in the outgroup penguins in the Adams con-
sensus tree, for example, show that combining
multiple source trees can lead to incorrect re-
lationships. The nonmonophyly of Eudyptes
and Pygoscelis in the Adams consensus tree is
an artifact of the degree of overlap betweenthe
penguin species in the source trees,rather than
being a novel result. The strict-consensus tree
shows that the structure within the penguins is
difficult to resolve. Of the three source trees
that include penguins, two have five species
(Fig. 1E, G; Sibley and Ahlquist 1990, Nunn
and Stanley 1998), whereas the other has four
penguin species (our Fig. 1F; Paterson et al.
1995). Two of these source trees share no pen-
guin species (Fig. 1E and 1F), whereas the third
(Fig. 1G) contains two penguin species from
each of the other two studies (Fig. 5). The lack
of monophyly in Eudyptes, for example, is due
to the three species occurring in the two source
trees in which penguins do not overlap (and
thus they group with other penguin taxa found
in their source trees rather than with one an-
other). Thus, some caution needs to be applied
to the use of the Adams consensus tree rather
than strict-consensus tree, and areas of dis-
agreement between the two techniques should
be evaluated to ensure that the increased res-
olution afforded by the Adams consensus tree
isnot spurious.Our finding, likethatforstorm-
100 [Auk, Vol. 119K
ENNEDY AND
P
AGE
January 2002] 101Seabird Supertrees
F
IG
. 5. The overlap between penguin species in
source trees (labeled as in Fig. 1 as E, F, and G). Note
that the two studies using mitochondrial genes (E
andF)havenotaxaincommon.
←
F
IG
. 4. The strict consensus of 18 MP trees for the mtDNA supermatrixwith penguins from the Paterson
et al. (1995) study excluded (i.e. 99 taxa in total) and transversion weighted 10:1over transitions. Eachof the
MP trees had a tree length of 8,464, a CI of 0.3640, and a RI of 0.8146.
petrels, suggests that the phylogenetic relation-
ships of penguins require more study (though
we did not explicitly sample studies including
penguin taxa and thus have not comprehen-
sively surveyed the penguin literature).
It is possible to evaluate the fit between
source trees and MRP consensus trees and to
find whether any of the individual sourcetrees
is particularly at odds with relationships sug-
gested by combination of source trees. In Table
1, we show bothconsistency index (CI) and re-
tention index (RI) for the all MRP characters,
and for those characters derived from each of
the source trees individually for both the strict-
consensus tree and one of the 10,000 MRP trees
(chosen at random). The CI and RI of all data
combinedissimilar to thatfoundforeachofthe
source trees for both the strict-consensus tree
and the single MRP tree, with none of the stud-
ies appearing to be a substantially worse fit
than any of the others. Because the single MRP
tree is more resolved than the strict-consensus
tree, its values are consistently higher. All
source trees are generally consistent with the
supertree and thus, the MRP trees are a good
representation of all source trees.
Combining topologies of the source treesal-
lows us to construct a phylogeny containing
122 taxa from a wide range of data types. Al-
though 90 of those taxa and 86 of the 188MRP
characters come from a single study (Nunn and
Stanley 1998), that study does not dominate the
supertree construction as much as it does the
supermatrix analysis. Two-thirds of Nunnand
Stanley’s (1998) taxa occur in one or more other
source tree, thus relative positions of most of
their taxa are subject to the constraint of where
the other source trees place them. Nunn and
Stanley’s (1998) study is no more consistent
with the strict-consensus MRP tree than the
other source trees are, suggesting that rather
than dominating the MRP analysis, it is a uni-
fying large study that allows for greater over-
lap between available source trees. For the
character supermatrix approach to combining
the seabird source studies, the Nunnand Stan-
ley (1998) study provides over 90% of the bases
in the supermatrix (1,143 bases for 90 taxa).
Thus, for this study on seabirds, the superma-
trix approach is dominated more by a single
large study than the supertree approach. (We
do not wish to imply that being dominated by
a single study is necessarily a problem, rather
that any errors in the dominant study are more
likely to be maintained.)
Combining sequence data from different
genes for multiple studies can potentially gen-
erate misleading relationships within resultant
phylogenies. Supermatrix approaches may be
affected by introduction of large amounts of
missing data (e.g. we have 12S sequence for
only 16 of 103 taxa) as well as not having over-
lapping sequence for some groups of taxa. Our
initial supermatrix analysis included penguins
from both Nunn and Stanley (1998) and Pater-
son et al. (1995) source trees. Those studies,
however, did not contain any penguin species
in common (see Fig. 5), and thus our analysis
gave an extraordinary result by grouping pen-
guins present in the Paterson etal. (1995) study
as sister taxa to one of the storm-petrelsubfam-
ilies (the Hydrobatinae; see Fig. 3). As noted
102 [Auk, Vol. 119K
ENNEDY AND
P
AGE
T
ABLE
1. Measures of fit for the source tree data on the strict consensus of the 10,000 MRP trees and on one
of the most parsimonious MRP trees (i.e. a fully resolved tree).
Tree
Number of
MRP
characters
Strict consensus tree
CI RI
A single MRP tree
CI RI
Combined
Austin (1996)
Bretagnolle et al. (1998)
Heidrich et al. (1987)
Imber (1985)
Nunn and Stanley(1998)
Paterson et al. (1995)
Sibley and Ahlquist (1990)
188
8
12
18
28
86
14
22
0.66
0.89
0.71
0.82
0.76
0.61
0.64
0.63
0.88
0.96
0.81
0.93
0.92
0.88
0.79
0.86
0.88
1.00
0.86
1.00
0.80
0.94
0.78
0.76
0.97
1.00
0.92
1.00
0.94
0.99
0.90
0.92
earlier, including penguins from the smaller
study in the supermatrix leads to penguins
from that study being placed in a part of the
tree that is constrained by positions of other
taxa for which there is 12S data available. Ex-
cluding penguin 12S sequences (and thus four
penguin species) removes the major cause of
the lack of resolution from the DNA data set
(i.e. number of MP trees is reduced from 1,458
to 18). This result shows that, as with degree of
overlap in source trees for supertree construc-
tion, caution needs to be taken when deciding
whether taxa from different studiescan validly
be combined in a supermatrix approach.
With exclusion of the four penguin taxa from
the Paterson et al. (1995) study, the final com-
binedDNAdata set is further dominatedbythe
NunnandStanley(1998)study.As with oursu-
pertree analysis and Nunn and Stanley’s (1998)
study, results of the supermatrix analysis also
suggest that the storm-petrels are not mono-
phyletic. In this analysis, however, the alba-
trosses (see Fig. 4) separate the two subfamilies
of the Hydrobatidae. Equally weighted parsi-
mony analysis with only Nunn and Stanley’s
(1998) data placed both subfamilies of storm-
petrels outside of albatrosses (see Fig. 1E),
whereas other exploratory analyses using
equal weights and the additionalcytochrome-b
sequence data placed them both nonmonophy-
letically inside the albatrosses. Bootstrap anal-
ysis of the final combinedDNA dataset weakly
supports separating the two subfamilies by the
albatrosses (i.e. 58% bootstrap support; Fig. 4),
and none of the sequence-based analyses we
performed place storm-petrels as a monophy-
letic group. Although storm-petrels may notbe
monophyletic (as noted earlier, that group re-
quires more extensive study), it is unlikely that
they are separated by albatrosses assuggested
by our weighted analysis (but not by any of our
other exploratory sequence analyses). This re-
sult may again be due to the 12S sequence
grouping the White-faced Storm-Petrel with al-
batrosses, and suggests that combining genes
with relatively little taxonomic overlap needs
to be approached with caution. The MRP su-
pertree, by incorporating more lines of (albeit
equivocal) evidence, thus provides a more con-
servative estimate ofrelationships of storm-pe-
trels by indicating a lack of resolution for the
two subfamilies.
Apart from the node separating penguins
from the procellariiforms (89%) and the node
separating penguins (aFig. 4), storm-petrels
and albatrosses from the remaining procella-
riiforms (96%; bFig. 4) there is a general lack
of bootstrap support for the deeper branchesin
our weighted parsimony analysis. This finding
suggests that genes for which sequence data is
available are evolving too rapidly to robustly
resolve relationships within that group of
birds. Several groups, however, are well sup-
ported by bootstrapping. Relationships within
albatrosses are generally well supported, and
monophyly of several groups including storm-
petrelsubfamilies,thegroupincludingfulmars
andgiant petrels, Procellariaspecies,Gadflype-
trels, diving petrels, and prions, are well sup-
ported (see Fig. 4). Thus, in this study the su-
pertree approach provides a less-resolved
phylogeny than the sequence supermatrix be-
cause of its relative conservatism (i.e. the su-
permatrix approach provides resolution that is
not well supported).
We were unable to improve resolution of the
supertree by weighting nodes from source
trees according to their relative levels of sup-
January 2002] 103Seabird Supertrees
port. If all source studies had been able to pro-
vide a comparable measure of support for their
nodes, we could have weighted the support for
each node appropriately in our matrix.Weight-
ing nodes based on their bootstrapsupport, for
example, may improve resolution of a super-
tree (see Bininda-Emonds and Bryant 1998).
Because our source trees weregenerated in sev-
eral different ways, however, they do not have
comparable support values.
Given the constraints of the two approaches
that have been discussed above, we suggest
that the MRP supertree presented here repre-
sents the best current estimate of relationships
of procellariiforms. The supertree combines in-
formation provided by more studies than are
available for the supermatrix approach, and se-
quence data currently available appear to be
evolving too rapidly to robustly resolve the
more basal relationships of that group. The su-
pertree thus provides the best available starting
point for future studies addressing questions
about procellariiforms from a phylogenetic per-
spective. The Adams consensus tree provides a
better starting point for comparative studies
because it is more resolved than the strict-con-
sensus tree (but polytomies within an Adams
consensus tree need to be interpreted withcau-
tion). We recommend, however, that any areas
of disagreement between the Adams and strict-
consensus trees should be evaluated to ensure
that relationships presented in the Adams con-
sensus tree are not artifacts of a lack of overlap
between source trees.
Possible uses for such a supertree include co-
evolutionary studies. Paterson et al. (2000), for
example, recently used a pruned version of the
Paterson et al. (1995) tree (Fig. 1F) to investi-
gate seabird and louse coevolution. Although
several branches of their phylogeny are not
supported by bootstrap analysis (figure 3A in
Paterson et al. 2000), with the exception of the
placement of Pelagodroma marina as sister taxon
to Diomedea epomophora, branches in their tree
are consistent with our supertree. Although a
single difference, that unusual relationship po-
tentially affects the results of Paterson et al.
(2000) and could alter the conclusions they
draw about the coevolutionary history of sea-
birds and their feather lice.
As more phylogenies become available for
different groups, it will become more common
for researchers to combine thesephylogenies in
some way in order to either address questions
at a different level or ina more comprehensive
way. In the foreseeable future, because data
available from the individual studies are un-
likely to be readily compatible (e.g. combining
sequence data can be difficult when there is lit-
tle overlap between taxa sequenced for differ-
ent genes), supertree approaches may bemore
generally used than supermatrix approaches.
In order for source trees to be more easily
found and for supertrees to be more readily
constructed, it would aid researchers if all po-
tential source trees were submitted to a phy-
logeny database (such as TREEBASE). Search-
ing GenBank and the Web of Science and other
databases can lead to studies being missed as
possible source trees if they could not be found
(book chapters would be particularly suscep-
tibleto this). Once the literatureforsourcetrees
have been found, it is possible that errors may
be introduced to the supertree matrix ifsource
trees are not regenerated accurately. To have
the data set and resulting phylogenies in aphy-
logeny database (see Cohen et al. 1998) would
eliminate those possible sources of error. As
with sequence data, alignments, data sets, and
the trees that are generated from them should
consistently be made available through a phy-
logeny database.
A
CKNOWLEDGMENTS
We thank Jennifer Bell for assistance with this
study. Kevin Johnson, Adrian Paterson, Rick Prum,
and an anonymous reviewer provided comments
during preparation of the manuscript. Peter Harri-
son generously gave permission for us to reproduce
figures from his book (Harrison 1987), which Jane
Paterson kindly redrew. Adrian Paterson provided a
copy of their aligned 12S data set. This project was
supported by Natural Environment ResearchCoun-
cil grant GR3/11075 and the Wolfson Foundation.
M.K. was supported by a New Zealand Science and
Technology Postdoctoral Fellowship.
TREEBASE is available online at www.treebase.
org. Web of Science is available at www.
webofscience.com.
L
ITERATURE
C
ITED
A
USTIN
, J. J. 1996. Molecular phylogenetics of Puffi-
nus shearwaters: Preliminary evidence from mi-
tochondrial cytochrome bgene sequences. Mo-
lecular Phylogenetics and Evolution 6:77–88.
B
AUM
, B. R. 1992. Combining trees asa way of com-
bining datasets for phylogenetic inference, and
104 [Auk, Vol. 119K
ENNEDY AND
P
AGE
the desirability of combining gene trees. Taxon
41:3–10.
B
ININDA
-E
MONDS
,O.R.P.,
AND
H. N. B
RYANT
. 1998.
Properties of matrix representation with parsi-
mony analyses. Systematic Biology 47:497–508.
B
ININDA
-E
MONDS
,O.R.P.,J.L.G
ITTLEMAN
,
AND
A.
P
URVIS
. 1999. Building large trees by combining
phylogenetic information: A complete phyloge-
ny of the extant Carnivora (Mammalia). Biolog-
ical Reviews 74:143–175.
B
RETAGNOLLE
,V.,C.A
TTIE
,
AND
E. P
ASQUET
. 1998.
Cytochrome-bevidence for validity and phylo-
genetic relationships of Pseudobulweria and Bul-
weria (Procellariidae). Auk 115:188–195.
C
OHEN
,B.L.,J.A.S
HEPS
,
AND
M. W
ILKINSON
. 1998.
Archiving molecular phylogenetic alignments
as NEXUS files. Systematic Biology 47:495–496.
C
RUICKSHANK
,R.H.,
AND
R. H. T
HOMAS
. 1999. Evo-
lution of haplodiploidy in dermanyssine mites
(Acari: Mesostigmata). Evolution 53:1796–1803.
F
ARRIS
,J.S.,M.K
A
¨LLERSJO
¨
,A.G.K
LUGE
,
AND
C.
B
ULT
. 1995. Constructing a significance test for
incongruence. Systematic Biology 44:570–572.
F
ELSENSTEIN
, J. 1985. Confidence limits on phyloge-
nies: An approach using the bootstrap. Evolu-
tion 39:783–791.
F
RIESEN
,V.L.,
AND
D. J. A
NDERSON
. 1997. Phylogeny
and evolution of the Sulidae (Aves:Pelecanifor-
mes): A test of alternative modes of speciation.
Molecular Phylogenetics and Evolution 7:252–
260.
G
RAY
,R.D.,
AND
F. M. J
ORDAN
. 2000. Language trees
support the express-train sequence of Austro-
nesian expansion. Nature 405:1052–1055.
H
ARRISON
, P. 1987. Seabirds of the World: A Photo-
graphic Guide. Christopher Helm, London.
H
ASEGAWA
,M.,H.K
ISHINO
,
AND
T. Y
ANO
.1985. Dat-
ing of the human–ape splitting by the molecular
clock of mitochondrial DNA. Journal of Molec-
ular Evolution 22:160–174.
H
EIDRICH
,P.,J.A
MENGUAL
,
AND
M. W
INK
. 1998.
Phylogenetic relationships in Mediterranean
and North Atlantic shearwaters (Aves: Procella-
riidae) based on nucleotide sequences of
mtDNA. Biochemical Systematics and Ecology
26:145–170.
H
EIDRICH
,P.,D.R
ISTOW
,
AND
M. W
INK
. 1996. Molec-
ular differentiation of Cory’s and Manx shear-
waters (Calonectris diomedea,Puffinus puffinus,
Puffinus yelkouan) and the Herring Gullcomplex
(Larus argentatus,L. fuscus,L. cachinnans). Jour-
nal fu¨ r Ornithologie 137:281–294.
I
MBER
, M. J. 1985. Origins, phylogeny andtaxonomy
of the gadfly petrels Pterodroma spp. Ibis 127:
197–229.
J
OHNSON
,K.P.,
AND
M. D. S
ORENSON
. 1998. Com-
paring molecular evolution in twomitochondri-
al protein coding genes (cytochrome bandND2)
in the dabbling ducks (Tribe: Anatini). Molecu-
lar Phylogenetics and Evolution 10:82–94.
K
ENNEDY
, M., A. M. P
ATERSON
,J.C.M
ORALES
,S.
P
ARSONS
,A.P.W
INNINGTON
,
AND
H. G. S
PEN
-
CER
. 1999. The long and short of it: Branch
lengthsandtheproblem of placing the New Zea-
land short-tailed bat, Mystacina. Molecular
Phylogenetics and Evolution 13:405–416.
K
ENNEDY
,M.,
AND
H. G. S
PENCER
. 2000. Phylogeny,
biogeography, and taxonomy of Australasian
teals. Auk 117:154–163.
K
ENNEDY
,M.,H.G.S
PENCER
,
AND
R. D. G
RAY
. 1996.
Hop, step and gape: Do the social displays ofthe
Pelecaniformes reflect phylogeny? Animal Be-
haviour 51:273–291. (An erratum to this paper
was published in Animal Behaviour, 1996, 51:
1197.)
M
ADDISON
,D.R.,D.L.S
WOFFORD
,
AND
W. P. M
AD
-
DISON
. 1997. NEXUS: An extensible file format
for systematic information. Systematic Biology
46:590–621.
M
ARCHANT
,S.,
AND
P. J. H
IGGINS
. 1990. Handbook of
Australian, New Zealand and Antarctic Birds,
vol. 1. Ratitesto Ducks. Oxford University Press,
Oxford.
N
UNN
,G.B.,J.C
OOPER
,P.J
OUVENTIN
,C.J.R.R
OB
-
ERTSON
,
AND
G. G. R
OBERTSON
. 1996. Evolution-
ary relationships among extant albatrosses (Pro-
cellariiformes: Diomedeidae) established from
complete cytochrome-bgene sequences. Auk
113:784–801.
N
UNN
,G.B.,
AND
S. E. S
TANLEY
. 1998. Body size ef-
fects and rates of cytochrome bevolution in
tube-nosed seabirds. Molecular Biology and
Evolution 15:1360–1371.
P
AGE
, R. D. M. 1996. TREEVIEW: An application to
display phylogenetic trees on personal comput-
ers. Computer Applications in the Biosciences
12:357–358.
P
ATERSON
,A.M.,R.D.G
RAY
,
AND
G. P. W
ALLIS
.
1993. Parasites, petrels and penguins: Does
louse phylogeny reflect seabird phylogeny? In-
ternational Journal of Parasitology 23:515–526.
P
ATERSON
,A.M.,G.P.W
ALLIS
,
AND
R. D. G
RAY
.
1995. Penguins, petrels and parsimony: Does
cladistic analysis of behavior reflect seabird
phylogeny? Evolution 49:974–989.
P
ATERSON
,A.M.,G.P.W
ALLIS
,L.J.W
ALLIS
,
AND
R.
D. G
RAY
. 2000. Seabird and louse coevolution:
Complex histories revealed by 12S rRNA se-
quences and reconciliation analyses.Systematic
Biology 49:383–399.
P
OSODA
,D.,
AND
K. A. C
RANDALL
. 1998. MODEL-
TEST: Testing the model of DNA substitution.
Bioinformatics 14:817–818.
R
AGAN
, M. A. 1992. Phylogenetic inference basedon
matrix representation of trees. MolecularPhylo-
genetics and Evolution 1:53–58.
January 2002] 105Seabird Supertrees
S
ANDERSON
,M.J.,A.P
URVIS
,
AND
C. H
ENZE
. 1998.
Phylogenetic supertrees: Assembling the trees
of life. Trends in Ecology and Evolution13:105–
109.
S
IBLEY
,C.G.,
AND
J. E. A
HLQUIST
. 1990. Phylogeny
andClassificationofBirds:AStudyinMolecular
Evolution. Yale University Press, New Haven,
Connecticut.
S
IBLEY
,C.G.,
AND
B. L. M
UNROE
. 1990. Distribution
and Taxonomy of the Birds of the World. Yale
University Press, New Haven, Connecticut.
S
TEEL
,M., A. W. M. D
RESS
,
AND
S.B
OCKER
.2000.Sim-
ple but fundamental limitations on supertree
and consensus tree methods. Systematic Biology
49:363–368.
S
WOFFORD
, D. L. 2000. PAUP*. Phylogenetic Analysis
Using Parsimony (*and Other Methods), version 4.
Sinauer Associates, Sunderland, Massachusetts.
T
HORLEY
,J.L.,
AND
R. D. M. P
AGE
. 2000. RADCON:
Phylogenetic tree comparison and consensus.
Bioinformatics 16:486–487.
Z
YSKOWSKI
,K.,
AND
R. O. P
RUM
. 1999. Phylogenetic
analysis of the nest architecture of neotropical
ovenbirds (Furnariidae). Auk 116:891–911.
Associate Editor: R. Prum
106 [Auk, Vol. 119K
ENNEDY AND
P
AGE
A
PPENDIX
. Classificationfor the taxa used in this study(adapted from Sibleyand Munroe 1990).
Procellariidae (petrels and allies) Common names Source study
Macronectes giganteus
Macronectes halli
Fulmarus glacialis
Fulmarus glacialoides
Thalassoica antarctica
Antarctic Giant-Petrel
Hall’s Giant-Petrel
Northern Fulmar
Southern Fulmar
Antarctic Petrel
B, H, N, S
1
N
A, B, H, N
H, N
N, S
1
Daption capense
Pagodroma nivea
Lugensa
2
brevirostris
Pseudobulweria
2
aterrima
Pseudobulweria
2
rostrata
Pterodroma axillaris
Cape Petrel
Snow Petrel
Kerguelen Petrel
Mascarene Petrel
Tahiti Petrel
Chatham Islands Petrel
N, P
B, N
N, S
B
B
I
Pterodroma nigripennis
Pterodroma cervicalis
Pterodroma inexpectata
Pterodroma hypoleuca
Pterodroma leucoptera
Pterodroma cookii
Black-winged Petrel
White-necked Petrel
Mottled Petrel
Bonin Petrel
Gould’s Petrel
Cook’s Petrel
B, I, N
I
I, N, P
I, N, S
I
I, N, P, S
Pterodroma pycrofti
Pterodroma brevipes
Pterodroma defilippiana
Pterodroma longirostris
Pterodroma alba
Pterodroma arminjoniana
Pterodroma heraldica
3,4
Pycroft’s Petrel
Collared Petrel
Defilippe’s Petrel
Stejneger’s Petrel
Phoenix Petrel
Herald Petrel
I
I
I
I, N
I
I, S
I
Pterodroma sandwichensis
Pterodroma phaeopygia
Pterodroma neglecta
Pterodroma externa
Pterodroma baraui
Pterodroma ultima
Hawaiian Petrel
Galapagos Petrel
Kermadec Petrel
Juan Fernandez Petrel
Barau’s Petrel
Murphy’s Petrel
I
I, N
I, N
I, N
B, I
I
Pterodroma solandri
Pterodroma macroptera
Pterodroma magentae
Pterodroma lessonii
Pterodroma madeira
Pterodroma feae
Providence Petrel
Great-winged Petrel
Magenta Petrel
White-headed Petrel
Madeira Petrel
Cape Verde Petrel
I
I, N
I, N
I, N
I
I, N
Pterodroma mollis
Pterodroma incerta
Pterodroma cahow
Pterodroma hasitata
Halobaena caerulea
Pachyptila vittata
Soft-plumaged Petrel
Atlantic Petrel
Bermuda Petrel
Black-capped Petrel
Blue Petrel
Broad-billed Prion
I, N
I, N
I, N
I, N, S
N
N, P, S
1
Pachyptila salvini
Pachyptila desolata
Pachyptila turtur
Bulweria bulwerii
Procellaria aequinoctialis
Procellaria parkinsoni
Medium-billed Prion
Antarctic Prion
Fairy Prion
Bulwer’s Petrel
White-chinned Petrel
Black Petrel
N
N
N, P
B, H, N
N
N
Procellaria westlandica
Procellaria cinerea
Calonectris diomedea
Calonectris leucomelas
Puffinus pacificus
Puffinus bulleri
Westland Petrel
Grey Petrel
Cory’s Shearwater
Streaked Shearwater
Wedge-tailed Shearwater
Buller’s Shearwater
N, P
B, H
B, H, N, S
1
N
A, B, N
A, N
Puffinus carneipes
Puffinus creatopus
Puffinus gravis
Puffinus griseus
Puffinus tenuirostris
Flesh-footed Shearwater
Pink-footed Shearwater
Great Shearwater
Sooty Shearwater
Short-tailed Shearwater
A, N
A, N
A, N
A, N, P, S
A
January 2002] 107Seabird Supertrees
A
PPENDIX
. Continued.
Procellariidae (petrels and allies) Common names Source study
Puffinus nativitatis
Puffinus puffinus
Puffinus yelkouan
Puffinus mauretanicus
1
Puffinus auricularis
Puffinus opisthomelas
Christmas Island Shearwater
Manx Shearwater
Mediterranean Shearwater
Townsend’s Shearwater
Black-vented Shearwater
A, N
A, H, N
A, H
A, H
A
N
Puffinus gavia
Puffinus huttoni
Puffinus lherminieri
Puffinus assimilis
Pelecanoides garnotii
Pelecanoides magellani
Fulttering Shearwater
Hutton’s Shearwater
Audubon’s Shearwater
Little Shearwater
Peruvian Diving-Petrel
Magellanic Diving-Petrel
A
A, N, P
A, N
A, H, N
N
N
Pelecanoides georgicus
Pelecanoides urinatrix
Diomedea exulans
Diomedea amsterdamensis
Diomedea epomophora
South Georgia Diving-Petrel
Common Diving-Petrel
Wandering Albatross
Amsterdam Island Albatross
Royal Albatross
N, P, S
B, N, S
B, N
H, N
H, N, P, S
Diomedea dabbenena
3
Diomedea gibsoni
3
Diomedea antipodensis
3
Diomedea sanfordi
3
N
N
N
N
Phoebastria
5
irrorata
Phoebastria
5
albatrus
Phoebastria
5
nigripes
Phoebastria
5
immutabilis
Thalassarche
5
melanophris
Thalassarche
5
cauta
Thalassarche
5
chrysostoma
Thalassarche
5
chlororhynchos
Thalassarche
5
bulleri
Waved Albatross
Short-tailed Albatross
Black-footed Albatross
Laysan Albatross
Black-browed Albatross
Shy Albatross
Grey-headed Albatross
Yellow-nosed Albatross
Buller’s Albatross
N, S
H, N
N, S
H, N, S
N
N, S
H, N, S
N
H, N, P, S
Thalassarche bassi
3
Thalassarche eremita
3
Thalassarche salvini
3
Thalassarche impavida
3
N
N
N
N
Phoebetria fusca
Phoebetria palpebrata
Oceanites oceanicus
Garrodia nereis
Pelagodroma marina
Fregetta tropica
Sooty Albatross
Light-mantled Albatross
Wilson’s Storm-Petrel
Grey-backed Storm-Petrel
White-faced Storm-Petrel
Black-bellied Storm-Petrel
H, N
H, N
N, S
1
N
N, P
N
Fregetta grallaria
Hydrobates pelagicus
Halocyptena
6
microsoma
Oceanodroma tethys
Oceanodroma castro
White-bellied Storm-Petrel
European Storm-Petrel
Least Storm-Petrel
Wedge-rumped Storm-Petrel
Band-rumped Storm-Petrel
N
H, N
N
N
H
Oceanodroma leucorhoa
Oceanodroma tristrami
Oceanodroma melania
Oceanodroma hornbyi
Oceanodroma furcata
Leach’s Storm-Petrel
Tristram’s Storm-Petrel
Black Storm-Petrel
Ringed Storm-Petrel
Fork-tailed Storm-Petrel
B, N, S
N
N
S
N, S
Gaviidae (Loons)
Gavia stellata
Gavia immer Red-throated Loon
Common Loon S
S
Spheniscidae (Penguins)
Aptenodytes patagonicus
Pygoscelis papua
Pygoscelis adeliae
Pygoscelis antarctica
Eudyptes chrysocome
King Penguin
Gentoo Penguin
Adelie Penguin
Chinstrap Penguin
Rockhopper Penguin
N, S
1
N, S
1
P
N
N
108 [Auk, Vol. 119K
ENNEDY AND
P
AGE
A
PPENDIX
. Continued.
Procellariidae (petrels and allies) Common names Source study
Eudyptes pachyrhynchus
Eudyptes chrysolophus
Megadyptes antipodes
Eudyptula minor
Spheniscus demersus
Fiordland Penguin
Macaroni Penguin
Yellow-eyed Penguin
Little Penguin
Jackass Penguin
P
N
P, S
1
P, S
S
Taxa in Sibley and Munroe’s (1990) classification not included in any of the source studies
Pterodroma becki
Pterodroma macgillivrayi
Pachyptila belcheri
Pachyptila crassirostris
Bulweria fallax
Puffinus persicus
Puffinus bannermani
Puffinus heinrothi
Oceanites gracilis
Oceanodroma monorhis
Oceanodroma macrodactyla
Nesofregetta fuliginosa
Oceanodroma markhami
Oceanodroma matsudairae
Oceanodroma homochroa
Beck’s Petrel
Fiji Petrel
Slender-billed Prion
Fulmar Prion
Jouanin’s Petrel
Persian Shearwater
Bannerman’s Shearwater
Heinroth’s Shearwater
White-vented Storm-Petrel
Swinhoe’s Storm-Petrel
Guadalupe Storm-Petrel
Polynesian Storm-Petrel
Markham’s Storm-Petrel
Matsudaira’s Storm-Petrel
Ashy Storm-Petrel
Gavia arctica
Gavia pacifica
Gavia adamsii
Arctic Loon
Pacific Loon
Yellow-billed Loon
Aptenodytes forsteri
Eudyptes robustus
Eudyptes sclateri
Eudyptes schlegeli
Spheniscus humboldti
Spheniscus magellanicus
Spheniscus mendiculus
Emperor Penguin
Snares Penguin
Erect-crested Penguin
Royal Penguin
Humboldt Penguin
Magellanic Penguin
Galapagos Penguin
1
This species name was given to the genus in Sibley and Ahlquist’s (1990) tree.
2
Pterodroma in Sibley and Munroe (1990).
3
Not classified in Sibley and Munroe (1990).
4
Pterodroma heraldica was not classified in Sibley and Munroe (1990). It is sometimestreated as a subspecies of Pterodromaarminjoniana, hence
Sibley and Munroe (1990) gave it the common name of the Herald Petrel, which can be substituted for the Trinidade Petrel when Pterodroma
heraldica is classified and giventhat common name (see Marchant and Higgins 1990).
5
Diomedea in Sibley and Munroe (1990).
6
Oceanodroma in Sibley and Munroe (1990).