A RT I C L E S
Permian Phytogeographic Patterns and Climate
Data/Model Comparisons
P. McAllister Rees, Alfred M. Ziegler, Mark T. Gibbs,1 John E.
Kutzbach,1 Pat J. Behling,1 and David B. Rowley
Department of the Geophysical Sciences, University of Chicago, Chicago, Illinois 60637, U.S.A.
(e-mail: rees@geosci.uchicago.edu)
ABSTRACT
The most recent global “icehouse-hothouse” climate transition in earth history began during the Permian. Warmer
polar conditions, relative to today, then persisted through the Mesozoic and into the Cenozoic. We focus here on
two Permian stages, the Sakmarian (285–280 Ma) and the Wordian (267–264 Ma; also known as the Kazanian),
integrating floral with lithological data to determine their climates globally. These stages postdate the PermoCarboniferous glaciation but retain a moderately steep equator-to-pole gradient, judging by the level of floral and
faunal differentiation. Floral data provide a particularly useful means of interpreting terrestrial paleoclimates, often
revealing information about climate gradations between “dry” and “wet” end-member lithological indicators such
as evaporites and coals. We applied multivariate statistical analyses to the Permian floral data to calibrate the nature
of floral and geographical transitions as an aid to climate interpretation. We then classified Sakmarian and Wordian
terrestrial environments in a series of regional biomes (“climate zones”) by integrating information on leaf morphologies and phytogeography with patterns of eolian sand, evaporite, and coal distributions. The data-derived biomes
are compared here with modeled biomes resulting from new Sakmarian and Wordian climate model simulations for
a range of CO2 levels (one, four, and eight times the present levels), presented in our companion article. We provide
a detailed grid cell comparison of the biome data and model results by geographic region, introducing a more rigorous
approach to global paleoclimate studies. The simulations with four times the present CO2 levels (4#CO2) match the
observations better than the simulations with 1#CO2, and, at least in some areas, the simulations with 8#CO2
match slightly better than those for 4#CO2. Overall, the 4#CO2 and 8#CO2 biome simulations match the data
reasonably well in the equatorial and midlatitudes as well as the northern high latitudes. However, even these highest
CO2 levels fail to produce the temperate climates in high southern latitudes indicated by the data. The lack of
sufficient ocean heat transport into polar latitudes may be one of the factors responsible for this cold bias of the
climate model. Another factor could be the treatment of land surface processes and the lack of an interactive vegetation
module. We discuss strengths and limitations of the data and model approaches and indicate future research directions.
Introduction
been developed for modern day floras by Walter
(1985). The general conclusion in this review was
that Permian climate was well differentiated, with
equator-to-pole gradients that appear to have been
similar to earth’s modern interglacial climate. Similar conclusions had been reached by others (Chaloner and Meyen 1973; Vakhrameev et al. 1978;
Meyen 1987), and the biome approach to climate
classification has been acknowledged in reviews of
Late Paleozoic vegetation (Wagner 1993; Utting and
Piasecki 1995; Gastaldo et al. 1996; Wnuk 1996).
The thrust of these reviews was to improve bio-
Our primary goal is to document global geographic
patterns of Permian climate parameters using fossil
floras and climate-sensitive sediments. This work
builds on an earlier review of Permian floral provinces (Ziegler 1990) in which the phytogeographic
units were plotted on paleogeographic maps and
assigned to a climate-based biome scheme that had
Manuscript received November 2, 2000; accepted June 14,
2001.
1
Center for Climatic Research, University of Wisconsin—
Madison, Madison, Wisconsin 53706, U.S.A.
[The Journal of Geology, 2002, volume 110, p. 1–31] 䉷 2002 by The University of Chicago. All rights reserved. 0022-1376/2002/11001-0001$01.00
1
2
P. M . R E E S E T A L .
stratigraphic correlation and detect global climate
change by examining broad aspects of local or regional floral successional patterns through long intervals of geologic time. These are laudable goals,
but we point out that mistakes may be made by
confusing the locally observed changes, which may
have resulted from the drift of continents across
climate zones, with real global climate change. The
paleolatitudinal changes of Pangea were substantial
during the Permian, amounting to ∼15⬚ northward
motion, or about one climate zone, for most basins
(Ziegler et al. 1997).
We maintain that the geographic variations in
floras must be documented before temporal variations can be assessed. Our approach is to integrate
the individual floral lists from across the earth for
time intervals that are as fine scale as possible,
which is generally the stage level. In this article,
we present a statistical analysis of a large compilation of taxonomic lists for two Permian stages,
the Sakmarian (285–280 Ma) and Wordian (267–264
Ma, also known as the Kazanian; ages from Jin et
al. 1997), building on the Wordian results of Rees
et al. (1999). A similar approach has been applied
to all stages of the Permian of Gondwana (Cuneo
1996), but there, the floral lists were grouped into
five to seven units (i.e., major geographic regions)
before the analysis, depending on the stage. Although we also grouped lists, we limited this spatially to distances of ∼100 km and vertically
throughout individual formations. So, we had in
excess of 100 composite lists per time interval, and
this allowed us to detect the gradients that are the
natural consequence of regional variations in climate parameters, such as rainfall and temperature.
Any discussion of real global change must be built
on a foundation that uses the biomes as building
blocks and that then determines the variation
through time of the area and latitude spanned by
each biome. Major problems with this goal include
the fact that significant areas of the earth are devoid
of a Permian record and the difficulty that the biome “boundaries” are by nature gradational and
therefore difficult to define. There is also a propensity for plants to be preserved under climate regimes that have relatively high precipitation, promoting higher rates of sedimentation and
preservation potential. Hence, we supplement the
floral data with lithological indicators of climate
such as coals, evaporites, and eolian sands. Finally,
the climate parameters controlling the Permian biomes may not have been directly analogous to their
recent counterparts. In fact, “no-analog” climates
and communities are known from the Holocene,
so we use the biome scheme as a general framework
to understand past climates. Our effort, together
with our work on the Jurassic (Ziegler et al. 1993,
1996; Rees et al. 2000), is therefore just a beginning.
The advantage of the biome approach is that it
transcends time and provides a uniform terminology for comparing floras throughout the Permian
and with other geological periods. The classifications of earlier workers were tied to paleocontinents that could change latitude or to floral taxa
that could evolve to other taxa, and this has resulted in confusion. We do retain the terms Angaran, Cathaysian, and Gondwanan for the north
temperate, tropical, and south temperate “realms,”
respectively. We make the very important point
that the paleocontinents for which these realms are
named each span a wide paleolatitudinal range and
incorporate a number of biomes. Moreover, individual paleocontinents may be host to more than
one realm, so, for example, low-latitude sites in
Gondwana may be occupied by the Cathaysian
realm. This simply reinforces the conclusion that
most land masses in the Permian were in contact
and that climate, and not geography, was the controlling factor in floral distributions (Ziegler 1990).
Our Permian biome maps are compared here
with ones resulting from a climate model study in
our companion article (Gibbs et al. 2002). That article is a refinement of an earlier study of Wordian
(i.e., Kazanian) climates (Kutzbach and Ziegler
1993) and incorporates elements of a new database
on climate-sensitive sediments of the Permian
stage intervals (Ziegler et al. 1998). The various approaches we use to interpret Permian climates are
discussed later, but the floral localities and climate
sensitive sediments are shown in figure 1A and 1B
to give an idea of the coverage of the geological
data. Broad-scale features of climate can be determined by studying the distributional patterns of
key lithologic indicators, although these provide
only end-member information about extremes of
precipitation and evaporation (e.g., coals, P 1 E;
evaporites and eolian sands, P ! E). We therefore
included paleobotanical, particularly fossil leaf,
data in order to derive more refined interpretations
of the complete climate spectrum (for details, see
Rees and Ziegler 1999; Rees et al. 2000). The paleogeographic reconstructions used here are ones
that have been recently updated (Ziegler et al. 1997)
and show details of paleotopography as well as the
positions of shorelines.
Compilation of Floral Data
Permian floral remains are quite evenly distributed
globally, and this provides a framework for under-
Journal of Geology
P E R M I A N P H Y T O G E O G R A P H I C PAT T E R N S
3
Figure 1. Sakmarian (A) and Wordian (B) paleogeographic maps (Mollweide projection with 45⬚ longitude lines)
showing the distributions of floral localities and lithological climate indicators (eolian sands, evaporites, and coals)
in each stage. Major geographical regions are highlighted.
standing Permian global phytogeographic patterns.
Our sampling strategy was designed to maximize
the climate signal in the floral data (Rees and Ziegler 1999). Taxonomic lists were compiled from
stratigraphic papers and paleobotanical monographs to achieve the widest geographic coverage
possible, but only papers that provided complete
lists of fossil assemblages were used—we excluded
taxonomic papers dealing with only a selected plant
group or groups. We were interested in unraveling
the climate signals and required as complete documentation as possible of the fossil assemblages to
serve as a proxy for the original vegetation, ecology,
and prevailing climate conditions. Complete lists
from stratigraphic descriptions did meet our standards, and so, to some extent, quality was sacrificed
for quantity and broad geographic coverage. Some
merging of lists was done, both temporally and geographically, within a geological formation to factor
out local community-level variations and to compensate for the failure to collect all elements of the
flora at the local bedding plane level. Of course,
this merging has already been performed for many
of the lists available in the literature. There is considerable variance in the way paleobotanists classify fossil plants, and moreover, none of the current
schemes is comprehensive. To maximize consis-
tency, our lists were classified (and synonymized
where necessary) using Meyen (1987), simply because his book has the most inclusive taxonomic
coverage, but it was necessary to supplement this
with Taylor and Taylor (1993) in certain instances.
The correlation of the basic geological unit, the
formation, to the stage level is of course imperfectly known and is currently in a state of flux in
the Permian. Our international correlations are
based on Jin et al. (1997) and are supplemented by
Zhuravleva and Ilina (1988) for the Angaran region
of Russia, the COSUNA charts for North America
(e.g., Hills and Kottlowski 1983), Jin et al. (1994)
for the Chinese microcontinents, and Langford
(1992) for the Gondwanan areas. Since many of the
floral lists are associated with interior basins that
were remote from marine sections, the accuracy of
the correlations is estimated to be Ⳳ1 stage. So,
short-term climate fluctuations are beyond the
scope of this analysis, but the gradual crosslatitudinal trends in continental motion and attendant climate changes are detectable, even if the
temporal control is less certain.
A total of 721 Permian floral localities, comprising 6252 plant occurrences, was compiled from the
literature, providing worldwide coverage throughout the period. Of these, 991 occurrences are rep-
Figure 2. Sakmarian (A) and Wordian (B) floral localities expressed as scaled pie diagrams showing the morphological
categories and numbers of genera present in each flora.
4
Figure 2
(Continued)
5
6
P. M . R E E S E T A L .
resented by fructifications, seeds, and wood, with
a further 376 occurrences represented by plant remains of uncertain affinities. It is standard paleobotanical practice to assign separate generic names
to different parts of the plant because the parts are
generally found separately and, in many cases, the
relationship of the parts to the original (i.e., complete) plant is unknown. Reproductive organs were
excluded from our analyses, which were based primarily on leaf genera to minimize possible duplication of organs representing the same plant. Moreover, the relationship between leaf morphology and
climate is better understood than with reproductive
organs, for the obvious reason that foliage interacts
with the atmosphere and is produced by the plant
to maximize efficiency under prevailing light, precipitation, temperature, and O2 and CO2 conditions. Wood genera and stems were also excluded,
except for stems and roots (e.g., Calamites, Lepidodendron, Stigmaria, and Vertebraria) with more
securely known affinities to foliage or else other
members of the lycopsids and sphenopsids in which
the stems and roots are more typically preserved
than the foliage. Thus, a total of 193 genera from
721 localities (4885 occurrences, or 78% of all compiled Permian floral data) was available for the analyses. Floral data were then selected from our two
target stages, the Sakmarian and Wordian. The Sakmarian comprises 112 genera from 128 localities
(799 occurrences), and the Wordian comprises 104
genera from 147 localities (1001 occurrences).
These were analyzed further to determine floristic
patterns and climate signals for each stage.
Plant Morphological Categories
In order to understand broad phytogeographic patterns, each genus was assigned to a coarser morphological category (at the level of plant class or
order), again based on Meyen (1987) and supplemented by Taylor and Taylor (1993). (Details on our
system of classification are provided in table 2.) The
categories parallel the major taxonomic subdivisions, which, in turn, often reflect the individual
physiognomic strategies of their constituent plants.
These coarse subdivisions can be useful as paleoclimatic tools if one accepts that leaf morphologies
typically represent environmental adaptations.
Other paleobotanists might choose to emphasize
different categories, but our scheme at least helps
to reveal the broad global vegetational patterns in
the Permian. We further believe that these patterns
are sufficiently pronounced to remain largely unaltered by fine-scale adjustments. The use of
higher-level taxa enables the raw data to be shown
for the Sakmarian and Wordian, with each floral
locality being represented by a pie diagram (fig. 2A,
2B). The size of each segment corresponds to the
number of genera represented by a particular morphological category, expressed as a percentage of all
genera at that locality. Note that, to enable all of
the pies and segments to be shown, their plotted
positions do not always correspond exactly to the
locality positions; these are shown accurately in
figure 1, and details are available on request. Major
differences between the Sakmarian and Wordian
(fig. 2A, 2B) include the marked decrease in Euramerican floral localities and increase in Angaran
ones and changes in floristic composition in China
(where gigantopterids and peltasperms become
more common and lycopsids decline).
The diameter of each locality pie diagram is
scaled according to the total number of genera present, and this, together with the number of different
higher-level taxa preserved at the locality, can give
some indication of floral diversity. Some caution
should be exercised here since sample sizes may
reflect different depositional environments (e.g.,
crevasse splay, floodplain, or deltaic), creating taphonomic biases. Bias also may be introduced by differences in preservation potential of different kinds
of plants or even different parts of the same plant.
The relative accessibility of plant localities will affect intensity of sampling, and the overall research
effort may be determined by financial resources or
the degree of intellectual interest. Despite the preceding caveats, a consistent pattern can still be seen
(fig. 2A, 2B) whereby the number of taxa per locality
is lowest in high latitudes and in “desert belt”
regions with eolian sands and evaporites (shown in
fig. 1). We interpret this pattern to reflect original
vegetation, with colder or drier climate conditions
being less conducive to the development of diverse
plant communities, and discuss this in more detail
below (see “Sakmarian and Wordian Climates”).
General adaptations of the various Permian plant
groups have been reviewed previously (Ziegler
1990), so the following discussion will be devoted
to the regularities in distribution patterns of each
group (fig. 2A, 2B). These patterns can be used to
reinforce the inferences that have been made concerning the morphological adaptations to precipitation and temperature effects that are seen in the
plants. The arborescent lycopsids, typified by Lepidodendron, were a mostly low-latitude group generally associated with coal swamps, so a tropical
rain forest environment can be assumed for most
occurrences. In the Permian, this biome was best
developed in the Chinese microcontinents and has
been referred to as a “Lycopsid Refugium” to con-
Journal of Geology
P E R M I A N P H Y T O G E O G R A P H I C PAT T E R N S
trast this with their wider distribution in the Carboniferous (Gastaldo et al. 1996). Lycopsids do appear in a number of mid- to low-latitude basins in
Angara and Gondwana, where they are represented
by different genera (e.g., Viatscheslavia to the north
and Lycopodiopsis to the south), but many of these
were smaller and simpler (Meyen 1976; Wagner
1993). Even so, arborescent forms do occur and are
taken to indicate warming of the climate following
the Permo-Carboniferous glaciation to the point
where frost-free conditions were established
(Guerra-Sommer et al. 1995). Here we would apply
a “warm temperate” biome designation with some
caution because some forms may have been exclusively herbaceous.
The sphenopsids, like the lycopsids, include arborescent (e.g., Calamites) and herbaceous forms
(e.g., some representatives of Phyllotheca), and,
again, the distinction may not always be immediately clear from the taxonomic lists. Unlike most
of the low-latitude representatives in Euramerica
and China, many Angaran and Gondwanan forms
were smaller and apparently herbaceous (Plumstead 1973; Meyen 1982) and extended to the highest latitudes, where they typically formed the
ground cover. The three “fern” categories were
somewhat more limited in their distribution, ranging only from equatorial to midlatitude regions,
with the pteridosperms, a dominantly arborescent
group, appearing to be exclusively in the tropical
through warm temperate biomes. They were well
represented during the Early Permian in areas like
Europe, which are interpreted as being relatively
hot and arid (Glennie 1984), but most other forms
occurred in wetter environments in the rain forest
and warm temperate biomes.
The gigantopterids have been interpreted as being the main group of Permian plants adapted to
the climbing vine or liana habit (Yao 1983) and are
the hallmark of the Cathaysian Floral Realm (i.e.,
China and Euramerica). They seem to have been
restricted to the tropical rain forest biome that was
best developed in China but are also found in such
low-latitude areas as Venezuela, Mexico, Texas,
southern Spain, and Arabia. Their distribution pattern through all equatorial land masses could suggest that dispersal was not a problem for these
plants (Mapes and Gastaldo 1986). Indeed, the
ocean separating China from Pangea may have been
populated with island “stepping stones.” However,
there are considerable uncertainties in identifying
fossil species—even when well preserved, there is
no certainty that they were truly closely related
biologically (in the sense of reproductive viability).
We make an important distinction here between
7
whole-plant reconstructions of fossil species (based
on physical attachment of different plant parts,
each one having been assigned previously to a distinct species) and morphological comparisons between specimens of the same type of plant organ
(whether stems, roots, reproductive organs, or
leaves). The Permian gigantopterids provide just
one illustration of the different definitions of extant
and fossil plant species. It is not surprising that
plants growing in similar climate zones, regardless
of geographic separation, would have developed
similar strategies—including leaf morphologies—
to maximize their overall efficiency and competitiveness, even if they were biologically incompatible, a point to which we return in our discussion
of high-latitude vegetation.
The peltasperms, cycadophytes, and ginkgophytes seem to be limited in distribution to midand low-latitude sites, including rain forest and
more environmentally stressed settings in Europe.
The Wordian records of peltasperms and cycadophytes from India (∼50⬚S) may be erroneous; it is
probable that these flora are younger (see the discussion of statistical results in the section “Wordian Genus Plots” below). It is interesting to note
that the ginkgophytes progressed to dominate highlatitude sites in the Mesozoic, while the cycadophytes continued to diversify in warmer climates
(Rees et al. 2000). However, Permian records of
ginkgophytes are less certain than Mesozoic ones
because the leaves, although resembling those of
Mesozoic ginkgophytes, lack the diagnostic features that are used to identify Ginkgo-like leaves
with certainty.
The Pinales are typified by having needle-like or
scaly leaves. They were best developed in lowlatitude settings with evaporites and eolian sands
or in sequences that were transitional to these arid
climates. They are also represented in Argentina
during the Sakmarian, often associated with cordaite and glossopterid genera, in environments that
were probably cool temperate.
The cordaites and glossopterids tended to dominate the high-latitude regions of Angara and Gondwana, respectively, to the exclusion of other arborescent forms. Their symmetry about the
equator can be seen in figure 2A and 2B. Glossopterids and many cordaites have large, tongue- or
strap-shaped leaves and are usually interpreted as
being deciduous, commonly occurring as leaf mats
and indicating cool temperate conditions (e.g.,
Ziegler 1990). However, the cordaites exhibit a
wide range of morphological variability. As well as
dominating the high northern latitudes, some cordaites also occurred in low latitudes and had a va-
8
P. M . R E E S E T A L .
riety of growth habits, ranging from bushes and
mangroves to large trees (Stewart 1983). The highlatitude cordaites and glossopterids are commonly
associated with temperate-latitude coal swamps
and gradually became more dominant components
of lower-diversity floras toward the highest latitudes (fig. 2A, 2B). They are thought to represent
seasonally cold climates (Ziegler 1990; Durante
1995) and provide another example of similar morphological response to climate in biologically and
geographically separate groups.
Multivariate Statistical Analyses of Floral Data
We applied multivariate statistics to the floral data
at the genus level to determine the finer-scale floristic patterns. We chose correspondence analysis
(CA), a method used commonly in studies of modern ecology and vegetational succession (Gauch
1982; Ter Braak 1992). With CA, two-dimensional
plots are produced showing variance within data
sets on a series of axes. The advantages of CA are
that it provides the same scaling of sample (locality)
and character (taxa) plots, enabling direct comparison, and can accommodate incomplete data matrices where some information is missing (Hill
1979a; Gauch 1982), as often occurs with the fossil
record (e.g., Rees and Ziegler 1999; Rees et al. 2000).
The version used is one of the programs in the
CANOCO: Canonical Community Ordination
package, compiled by Ter Braak (1992), an extension of the Cornell ecology program DECORANA
(Hill 1979b). The general procedure has been described by Shi (1993, p. 218): “Geometrically, ordination involves rotation and transformation of
the original multidimensional co-ordinate system
and reduction of high dimensionality so that major
directions of variation within the data set can be
found and more readily comprehended than by
looking at the original data alone.” Thus, we use
CA as a means of arranging all of the elements
(whether taxa or localities) relative to axes in multidimensional space according to their similarity to
each other. The greatest variation is shown on the
first axis, with other axes accounting for progressively less. A series of two-dimensional plots (one
for taxa and the other for localities) is produced
showing variance within data sets on the first four
axes. Taxa that frequently co-occur plot closest together, while those that rarely co-occur are farthest
apart. The same applies to the localities plot; those
that share many taxa plot closest to one another,
while those with little in common plot farthest
apart.
To standardize identifications as far as possible,
analyses of our floral lists were conducted at the
genus rather than species level (although the number of species of each genus was also recorded in
our database). This maximizes the probability that
original identifications were accurate and minimizes taxonomic distortions caused by different approaches of “splitters” and “lumpers” (who may
assign the same collection of fossil leaves to many
or few taxa, depending on the point of view). We
should also point out that fossil leaf genera and
species are often delimited taxonomically on the
basis of relatively coarse characters such as size,
shape, and venation pattern and so are frequently
defined by morphological and not necessarily true
biological criteria, as discussed earlier. The uncertainties increase when working at the species
rather than genus level. For instance, although not
infallible (see comments by Chaloner and Creber
1988), it is highly probable that a lanceolate Permian leaf with pronounced midrib and reticulate venation will be identified correctly as belonging to
the genus Glossopteris. However, it is far less certain that it will be assigned to the “correct” species
of that genus. As well as other factors, morphological differences between leaves may simply be due
to growth in different positions on the same tree
(e.g., sun and shade leaves). These, if found as isolated fossil specimens (especially without consideration of local field associations; e.g., Rees 1993),
may then be assigned to distinct species. We agree
with Taylor and Taylor (1993, p. 560) that “for the
time being, the interpretation of Glossopteris in a
broad sense appears to be the best approach until
a sufficient number of leaf types with attached reproductive organs can be found to more accurately
define particular species.” So, we carried out analyses at the genus presence-absence level, and the
resultant patterns are relatively coarse. However,
we believe this approach to be a necessary compromise in order to enable more reliable and accurate interpretations of overall phytogeographic
patterns for intervals in the geologic past.
Separate statistical analyses were conducted for
the Sakmarian and Wordian. Any locality with an
assigned age, however long ranging, that encompassed the Sakmarian or Wordian was included in
the analyses. This enabled more floral localities to
be included in the analyses for each stage, but this
was at the expense of temporal resolution, which
suffered accordingly. The effects of this can be seen
on our CA plots and are explained later. For the
Sakmarian, 58 genera from 108 localities (comprising a total of 693 genus occurrences) were analyzed,
with 69 genera from 121 localities (908 occurrences) analyzed for the Wordian. Genera with only
Journal of Geology
P E R M I A N P H Y T O G E O G R A P H I C PAT T E R N S
one or two occurrences were excluded from the
analyses, as were localities with only one or two
genera, to enable the main global and regional patterns to be observed. This avoids a problem with
CA in which genera that are rare and occur in localities with low total abundances are overemphasized and are effectively outliers, occurring at the
extreme ends of the axes. This would have had the
effect of grouping together other more common
genera so closely that assessment of the patterns
between them would have become problematic.
The same problem occurs with the locality plots.
However, as Gauch (1982, p. 152) remarked, “this
difficulty should not be overstated, because it occurs only with data sets having such rare species,
and this problem is easily avoided by deleting rare
species [or genera in our case] from a data set (because this deletion removes very little information
from the data set).” A downweighting option was
then applied, such that full weighting is applied to
the most frequently occurring (abundant) genus
(Amax), as well as to those with abundances down
to Amax/5. For example, Pecopteris is the most abundant genus in our Wordian data set (with 67 occurrences, i.e., A max p 67) and has full weighting.
Genera with fewer occurrences than 67/5 are progressively downweighted to our chosen minimum
of three occurrences (see Hill 1979b for further details of the downweighting procedure). Pecopteris
and 21 other genera have full weighting, down to
Rufloria and Calamites (each with 14 occurrences).
The results, with eigenvalues and cumulative
percentage variance of genus data for the first four
axes, are shown in table 1. Eigenvalues measure the
importance of an axis, with values between 0 and
1; the higher the value, the more important the axis
(see Ter Braak 1992). The percentage variance for
each axis may seem low (∼13% on axis 1, with
cumulative percentage variance of the first four
axes being approximately 33% for both the Sakmarian and Wordian results). However, Gauch
(1982, p. 141) commented that “in some cases, particularly with large and noisy data sets, the first
couple of PCA [principal components analysis] axes
may account for as little as 5% of the total variance
and yet be quite informative ecologically. On the
other hand, in other cases, 90% of the variance may
be accounted for, yet the result may be ecologically
meaningless or severely distorted. In the end, the
assessment of PCA results must be in terms of ecological utility; mere percentage of variance accounted for has not been found to be a reliable indicator of the quality of results.” Although Gauch
was referring directly to PCA, he also described it
as being computationally similar to CA (see also
9
Table 1. Statistical Results for Correspondence Analyses of Genera from Sakmarian and Wordian Floral
Localities
Stage
Sakmarian:
CA axis:
1
2
3
4
Total inertia
Wordian:
CA axis:
1
2
3
4
Total inertia
Eigenvalue
Cumulative
percentage
variance
.802
.477
.437
.326
5.887
13.6
21.7
29.1
34.7
.747
.571
.317
.235
5.731
13.0
23.0
28.5
32.6
Ter Braak 1992). Our global Sakmarian and Wordian
compilations qualify as “large and noisy” data sets,
so the relatively low percentage variance is unsurprising. Of far greater importance is determining
whether the results are meaningful in terms of phytogeographic patterns and inferred climates.
Sakmarian and Wordian CA axis plots for localities and genera are shown in figures 3 and 4, indicating their relative distributions on each of the
four axes. The locality plots (figs. 3A–3C, 4A–4C)
are coded by symbol according to major geographic
region and, to assist further interpretation, are
numbered according to “subregions” or countries
(see figs. 1 and 2 for paleogeographic locations). The
relative position of each locality is defined by its
constituent leaf genera; localities with many genera
in common plot closest together, and those with
little in common plot farthest apart. To understand
more fully the patterns shown, it is necessary to
study the corresponding genus plots (figs. 3D–3F,
4D–4F). The relative position of each genus is defined by its degree of association with other genera.
Genera are coded by symbol according to “key”
higher taxa, with individual ones numbered (see
table 2 for details). We present our results so that
they can be studied at two different levels, according to expertise or interest of the reader. The
symbol-coded level facilitates understanding of the
general patterns. The numbering of geographic subregions and individual genera enables a more detailed appraisal of the data and derived patterns.
Sakmarian Locality Plots. Figure 3A shows CA
axis 1/axis 2 results for Sakmarian localities, with
each locality coded according to geographic region.
Gondwanan (Southern Hemisphere) sites have high
axis 1 scores, while Angaran (Northern Hemisphere
Figure 3. A–C, CA results for Sakmarian localities showing axes 1–4 scores for each. Major geographic regions are
highlighted. Certain subregions and countries are numbered: South America (1), Antarctica (2), Australia (3), Africa
(4), and India (5). D–F, CA results for Sakmarian genera showing axes 1–4 scores for each. Genera belonging to selected
morphological categories are highlighted. Each genus is also indicated by a number to enable more detailed comparisons (see table 2).
10
Figure 4. A–C, CA results for Wordian localities showing axes 1–4 scores for each. Symbol and numbering schemes
as in figure 3A–3C, with addition of the Russian Platform (6) and Siberian Platform (7) subregions. In contrast to the
Sakmarian, the Euramerican region is represented by only one locality in the Wordian CA. D–F, CA results for
Wordian genera showing axes 1–4 scores for each. Symbol and numbering schemes as in figure 3D–3F, with the
addition of gigantopterids. These occur as minor components of Sakmarian floras (fig. 2A) but did not qualify for
inclusion in the Sakmarian CA. Conversely, the Pinales (common in the Sakmarian) are represented by only one
genus (Walchia) in the Wordian CA.
11
12
P. M . R E E S E T A L .
Table 2. Genera and Their Assigned Morphological Categories Corresponding to Numbers Shown on the Sakmarian
and Wordian CA Genus Axis Plots
Genus
No.
Morphological
category
Alethopteris
Angaropteridium
Annularia
Annulina
Asansolia
Asterophyllites
Asterotheca
Baiera
Botrychiopsis
Buriadia
Calamites
Callipteridium
Callipteris
Cathaysiopteris
Chiropteris
Cladophlebis
Comia
Compsopteris
Cordaites
Crassinervia
Culmitzschia
Danaeites
Dichotomopteris
Dicksonites
Dicranophyllum
Dicroidium
Dorycordaites
Emplectopteridium
Emplectopteris
Ernestiodendron
Fascipteris
Gangamopteris
Gigantonoclea
Gigantopteris
Ginkgoites
Ginkgophyllum
Glossopteris
Glottophyllum
Gomphostrobus
Hermitia
Lebachia
Lepeophyllum
Lepidodendron
Lepidopteris
Linopteris
Lobatannularia
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
Pteridosperm
Pteridosperm
Sphenopsid
Sphenopsid
Fern
Sphenopsid
Fern
Ginkgophyte
“fern3”
Pinales
Sphenopsid
Pteridosperm
Pteridosperm
Gigantopterid
Ginkgophyte
Fern
Peltasperm
Peltasperm
Cordaite
Cordaite
Pinales
Fern
Fern
Pteridosperm
Dicranophyll
Peltasperm
Cordaite
Gigantopterid
Gigantopterid
Pinales
Fern
Glossopterid
Gigantopterid
Gigantopterid
Ginkgophyte
Ginkgophyte
Glossopterid
Cordaite
Pinales
Conifer
Pinales
Cordaite
Lycopsid
Peltasperm
Pteridosperm
Sphenopsid
Note.
Genus
No.
Morphological
category
Lycopodiopsis
Mixoneura
Nemejcopteris
Neomariopteris
Nephropsis
Neuropteridium
Neuropteris
Nilssonia
Noeggerathiopsis
Odontopteris
Oligocarpia
Palaeovittaria
Paracalamites
Paranocladus
Pecopteris
Phylladoderma
Phyllotheca
Plagiozamites
Poacordaites
Protoblechnum
Prynadeopteris
Pseudoctenis
Psygmophyllum
Pterophyllum
Raniganjia
Rhabdotaenia
Rhachiphyllum
Rhipidopsis
Rubidgea
Rufloria
Schizoneura
Sigillaria
Sphenobaiera
Sphenophyllum
Sphenopteridium
Sphenopteris
Stigmaria
Taeniopteris
Tingia
Todites
Trizygia
Vertebraria
Viatscheslavia
Walchia
Zamiopteris
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
Lycopsid
Pteridosperm
Fern
Fern
Cordaite
Fern
Pteridosperm
Cycadophyte
Cordaite
Pteridosperm
Fern
Glossopterid
Sphenopsid
Pinales
Fern
Peltasperm
Sphenopsid
“fern3”
Cordaite
Pteridosperm
Fern
Cycadophyte
Ginkgophyte
Cycadophyte
Sphenopsid
Glossopterid
Pteridosperm
Ginkgophyte
Glossopterid
Cordaite
Sphenopsid
Lycopsid
Ginkgophyte
Sphenopsid
Pteridosperm
“fern3”
Lycopsid
Cycadophyte
“fern3”
Fern
Sphenopsid
Glossopterid
Lycopsid
Pinales
Cordaite
Numbers are derived from the Sakmarian and Wordian CA genus axis plots in figures 3D–3F and 4D–4F.
mid- to high-latitude) sites from Russia and Mongolia have high axis 2 scores. Chinese, Euramerican
(European and North American), and North African/northern South American (i.e., low-latitude)
sites have low scores on both axes. Differences between the Chinese floras and those from Euramerica and North Africa/northern South America are
expressed on axis 3 (fig. 3B). Differences between
the Euramerican and North African/northern
South American floras are expressed on this axis
but are subtle, although floras from southern Europe, North Africa, Venezuela and the southwestern United States typically have lower axis 3 scores
than their northern neighbors, being more similar
to those from China. Axis 4 (fig. 3C) shows variations within Gondwana. Broadly speaking, two
groups of localities, “South America–India” (numbered 1 and 5) and “South America–AntarcticaAustralia-Africa” (1–4), can be seen along this axis.
Over half of the floras in the first group, with low
Journal of Geology
P E R M I A N P H Y T O G E O G R A P H I C PAT T E R N S
axis 4 scores, are dated as lower Sakmarian or
Asselian-Sakmarian, whereas over half of those in
the second group, with high axis 4 scores, are
slightly younger (dated as Sakmarian through Artinskian). So, the pattern seen on axis 4 could be
due to temporal changes, whether evolutionary
and/or climate related, rather than spatial variations in biogeography or climate.
Sakmarian Genus Plots. The corresponding genus
plots for the Sakmarian (figs. 3D–3F) aid further
interpretation of the locality patterns seen in figures 3A–3C. Glossopterids and cordaites typically
plot high on axes 1 and 2, respectively, while pteridosperms, lycopsids, and Pinales typically have
low scores on both axes (fig. 3D). As with the locality plot (fig. 3A), axis 1 accounts for the highest
degree of variance in the data. Gondwanan localities and glossopterid genera (Gangamopteris, Glossopteris, Rubidgea, and Vertebraria; numbered 32,
37, 75, and 88) comprise a well-delimited group that
plots high on this axis (fig. 3A, 3D), being distinct
both in a geographical and taxonomic sense. Major
differences between Northern Hemisphere localities and genera are therefore expressed on axis 2,
which accounts for less variance in the data but
nevertheless shows strong vegetational patterns.
Some of the exceptions to these patterns or otherwise noteworthy features are discussed here. Of
the cordaites, Noeggerathiopsis (55) plots high on
axis 1 in the Sakmarian, occurring commonly in
Gondwana although there are also records from Angara and China, particularly later in the Permian
(the genus has a high axis 2 score in the Wordian,
being more common in Angaran floras; fig. 4D).
Although Meyen (1987) suggested that Noeggerathiopsis might be affiliated with glossopterids, he
acknowledged that it is usually regarded as a cordaite genus. Unlike most other members of the cordaites, Crassinervia (20) has scaly leaves (Meyen
1987) and has the highest score on axis 2, occurring
commonly in Angaran and Mongolian floras, where
reduced leaf size may indicate more adverse environmental conditions. Two other genera (Dorycordaites [27] and Poacordaites [65]) have low
scores on axes 1 and 2, occurring in Euramerica
(North Africa and Europe) as components of lowlatitude vegetation. The genus Cordaites (19) has
been recorded in the Sakmarian from all four major
regions, in Gondwana, Euramerica, China, and Angara, hence its central position on the plot in figure
3D. This geographic distribution pattern (and resultant CA score) is probably a taxonomic artifact,
with strap-shaped leaves having been assigned to a
common and well-known “bin” genus. Note, however, the previously discussed dominance of cor-
13
daites in mid- and high-latitude Angaran assemblages, including Mongolia (fig. 2A, 2B).
Of the pteridosperms, most are found in China
and Euramerica and plot low on axes 1 and 2 (e.g.,
Alethopteris [1]), which contrasts with the position
of Angaropteridium (2; this plots high on axis 2
and, as the name implies, is restricted to the Angaran region). Most lycopsids (e.g., Lepidodendron
[43] and Sigillaria [78]) have similarly low axis 1/
axis 2 scores and low-latitude distributions, mostly
in China. One exception is Lycopodiopsis (47),
which is a recently described Gondwanan form.
The Pinales, an order of conifers, are typical Euramerican elements in the Sakmarian (fig. 2A), and
genera typically have low axis 1/axis 2 scores. Two
genera, Buriadia (10) and Paranocladus (60), occur
in South America and India and have high axis 1
scores. Genera representing other morphological
categories are shown by crosses. Botrychiopsis (9)
and Schizoneura (77) plot high on axis 1, being
Gondwanan representatives of the “fern3” and
sphenopsid categories. Other forms that typically
occur in Gondwana during the Sakmarian, and
which therefore plot toward the higher end of axis
1, include the ginkgophytes Chiropteris (15) and
Rhipidopsis (74). This latter genus has also been
recorded from Spain in the Sakmarian, hence its
more intermediate position on axis 1. Sphenopsids
such as Paracalamites (59) and Phyllotheca (63)
have high axis 1 scores and occur in Gondwanan
floras, although some have also been documented
from Angara and, more rarely, Euramerica. The
common sphenopsid genera in Euramerica, China,
and Angara include Annularia (3), Asterophyllites
(6), Calamites (11), and Sphenophyllum (80), which
plot low on axis 1 (fig. 3A, 3B).
Figure 3E shows genus axis 1 versus axis 3 scores.
The main feature of interest here is the separation
of the low-latitude Pinales of Euramerica (Culmitzschia [21], Ernestiodendron [30], Gomphostrobus [39], Lebachia [41], and Walchia [90]) with high
axis 3 scores from the Chinese lycopsid genera (Lepidodendron [43] and Stigmaria [83]) with low
scores. It is not surprising that Lepidodendron and
Stigmaria plot so closely together, being commonly
associated in plant assemblages and representing,
respectively, the stem and root organs of what is
usually considered to be the Lepidodendron tree.
A similar close association is seen on axis 1 (fig.
3D) for Glossopteris (37) and Vertebraria (88),
which represent the foliage and root systems of the
Glossopteris plant in Gondwana. Unlike the Chinese lycopsid genera, Sigillaria (78) occurs in Euramerican floras in the Sakmarian.
The axis 1/axis 4 genus plot (fig. 3F) shows dif-
14
P. M . R E E S E T A L .
ferences mainly between Gondwanan genera, with
the cordaite genus Noeggerathiopsis (55) plotting
high on axis 4 and the glossopterid Rubidgea (75)
having a low score. Noeggerathiopsis has been recorded from South America, Africa, and Australia
in floras dated as Sakmarian or Sakmarian-Artinskian. Rubidgea-bearing floras are all dated as Sakmarian and are from Brazil, with one exception
from India. Also evident is the separation of the
two genera of Pinales, Buriadia (10) and Paranocladus (60). Buriadia occurs in South America and
India, while Paranocladus is restricted to South
America. These patterns are reminiscent of those
seen on the corresponding locality plot (fig. 3C). Half
of the floras containing Buriadia are dated as lower
Sakmarian (and have low axis 4 scores) and may be
slightly older than those with Paranocladus (which
plot higher on axis 4). A similar pattern exists for
Botrychiopsis (9), which has a low axis 4 score and
occurs mostly in South American and Indian floras
in the Sakmarian, over half of which are dated as
Asselian-Sakmarian or lower Sakmarian.
Wordian Locality Plots. Similar patterns to the
Sakmarian are seen on the axis 1/axis 2 plot for
Wordian localities (fig. 4A). Gondwanan sites have
high axis 1 scores, Angaran and Mongolian sites
high axis 2 scores, and Chinese sites have low
scores on both axes. In contrast to the Sakmarian,
Wordian floras are present in south and north China
(fig. 2B). However, they have similar ranges of
scores on all four axes, indicating their close similarity (fig. 4A–4C). In addition, low-latitude Euramerican and North African/northern South American floras are largely absent in the Wordian (fig.
2B). Instead, axis 3 for the Wordian mainly separates Russian Platform (“southern Angaran”) localities (numbered 6, with low scores) from those
of the Siberian Platform (7, “northern Angara”) and
Mongolia, which have high scores (fig. 4B). As with
the Sakmarian, axis 4 shows variations within
Gondwana (fig. 4C), although in the Wordian, these
are mainly just between the Indian (5) and other
Gondwanan floras (from South America, Antarctica, Australia, and Africa; 1–4). Temporal effects
are again apparent, just as with the Sakmarian axis
4 plot (fig. 3C). The predominantly non-Indian floras (with low axis 4 scores) are dated in the range
of Roadian through Wordian or Roadian through
Capitanian (i.e., restricted to the Middle Permian),
whereas the Indian ones with high axis 4 scores are
dated as Wordian extending into the Late Permian.
Wordian Genus Plots. As with the Sakmarian,
glossopterids and cordaites typically plot high on
axes 1 and 2, respectively (fig. 4D), while pteridosperms typically have low scores on both axes. The
genus Callipteris (13), which is common in the Sakmarian floras of Euramerica (and has a low axis 2
score), is instead present in the floras of the Russian
Platform by the Wordian and plots high on axis 2
(fig. 4D). Although there is evidence that at least
some species of Callipteris are peltasperms (e.g.,
Kerp 1982), we followed traditional taxonomic classification and assigned this genus to the pteridosperms, pending further work on the other species.
Unlike the Sakmarian, lycopsids are present in the
Wordian of Angara as well as the other provinces.
The genus Viatscheslavia (89) has a high axis 2
score, occurring in floras of the Russian Platform,
although other lycopsid genera (Lepidodendron
[43], Stigmaria [83], and Lycopodiopsis [47]) show
similar patterns to those seen in the Sakmarian.
Gigantopterids such as Cathaysiopteris (14), Gigantonoclea (33), and Gigantopteris (34) are common in Chinese floras and have low axis scores. In
contrast to the Sakmarian, Pinales are represented
by only one genus (Walchia [90]) in the Wordian
CA plots, being present in Angara and Euramerica.
Of the other plant groups, two peltasperm genera,
Dicroidium (26) and Lepidopteris (44), have the
highest axis 1 scores. They occur in Indian floras,
which, as noted above (fig. 4C), are dated as Wordian extending into the Late Permian (indeed, these
genera are more typical components of Triassic floras). As such, they show the least similarity with
other genera on the Wordian CA plots (fig. 4A, 4D).
On axis 3 (fig. 4E), the end members are the cordaite
genus Rufloria (76), which occurs in floras of the
Siberian Platform and Mongolia, and the conifer
Walchia (90), present in Euramerica and the Russian Platform. As with the Sakmarian CA plot, axis
4 (fig. 4F) shows variations between genera more
typical of Gondwanan floras. As mentioned above,
Dicroidium (26) and Lepidopteris (44) have the
highest axis 4 scores. At the other extreme is
Lycopodiopsis (47), which occurs mainly in what
may be slightly older floras from Brazil and Australia. Further work should lead to improved age
determinations and stratigraphic correlations and
enable greater resolution of these patterns and
interpretations.
Throughout the preceding discussion, we have
remarked upon the close similarity of the floras
from Angara and Mongolia (figs. 2–4). This is surprising if one looks at the paleogeographic location
of Mongolia, which our maps (figs. 1, 2) show to be
nearer to north China than to Angara, and would
lead to an expectation of closer similarity between
the floras from Mongolia and China. In this case,
there are inadequate paleomagnetic data to constrain the paleogeographic reconstruction, and fu-
Journal of Geology
P E R M I A N P H Y T O G E O G R A P H I C PAT T E R N S
ture maps should show Mongolia positioned closer
to Angara, as the floral data suggest. We discuss the
implications of this for climate interpretations and
model results below.
Correspondence analysis provides an objective
means of determining any patterns that exist in the
floral data, expressed in terms of varying similarity
of genera and localities relative to one another. The
interpretation of these patterns requires a knowledge of fossil plant taxonomy and paleogeography
because, although CA identifies the degree of variance in the data, it cannot specify the sources of
the variance. An individual leaf genus is defined by
morphological characters, which can be interpreted
in terms of broadscale environmental conditions.
The relative position of each genus on the generic
plot is defined by its degree of association with
other leaf genera, whereas the relative position of
each flora on the corresponding locality plot is defined by its constituent leaf genera. As explained
earlier, the number of genera and localities represented in the CA is lower than that used for the
pie diagrams (fig. 2A, 2B), which is why we used
two different approaches to analyzing the floral
data. Morphological categories and pie diagrams enable the maximum information to be used from
each floral locality, although the patterns are broad
and comparisons between localities remain subjective. Correspondence analysis provides a more detailed and objective means of assessing patterns between genera and localities. We believe this
combined approach to phytogeographic mapping
and paleoclimate interpretation to be more complete and rigorous than ones that use only selected
plant taxa because we first study the overall patterns and then make more detailed analyses and
interpretations within this broader framework. By
using as much of the original data as possible, in a
global whole-flora approach, we derive the closest
possible approximation to original vegetation,
which in turn enables us to infer prevailing climate
conditions.
Sakmarian and Wordian Climates
The combined floral and lithological data (figs. 1–4;
methodology summarized in fig. 5) were used to
determine global Sakmarian and Wordian biomes,
or climate zones (fig. 6A, 6D). We used a classification scheme in which the macroclimate of the
present-day land surface is expressed in terms of 10
major biomes (Walter 1985, as modified by Ziegler
1990). The Walter scheme was developed using data
from some 8000 meteorological ground stations
worldwide and is based on temperature, precipi-
15
Figure 5. Geologic information and processing required
in our approach to interpreting paleoclimates. Only the
main paleobotanical and lithological steps are shown in
detail; the wide range of other geologic information necessary to produce the paleogeographic maps is merely
outlined, the emphasis here being on the use of paleobotanical and other climate data.
tation, and the manner in which these parameters
are distributed through the annual cycle. So, the
biomes were rigorously defined using the physical
aspects of climate that are most influential in biological differentiation over the globe today. The
scheme is simple and therefore readily applicable
in the geologic past (table 3), where our understanding of vegetation and climates is limited by incomplete preservation (Ziegler 1990; Rees and Ziegler
1999; Rees et al. 2000).
The boundaries of high-latitude biomes are controlled by changes in temperature, whereas variations in precipitation influence those in lower latitudes (Lottes and Ziegler 1994). In the case of the
tundra, cold and cool temperate biomes, the growing-season length defines each (see table 4). The
number of months with an average temperature of
110⬚C was chosen to define the growing season because temperatures in this range are necessary for
most higher (vascular) plant growth. The warm
temperate biome is defined to include regions in
the temperate zone that do not experience a hard
frost and have sufficient precipitation throughout
Figure 6. Data- and model-derived biomes for the Sakmarian (A–C) and Wordian (D–F). A, D, Data-derived biomes
for the Sakmarian and Wordian. B, C, Sakmarian-modeled biomes from the 4#CO2 (SAK-4#CO2) and 8#CO2 (SAK8#CO2) circular orbit experiments. E, F, Wordian-modeled biomes from the 4#CO2 (WORD-4#CO2) and 8#CO2
(WORD-8#CO2) circular orbit experiments.
Journal of Geology
P E R M I A N P H Y T O G E O G R A P H I C PAT T E R N S
17
Table 3. Walter Climates and Biomes and Permian Equivalents
Number
1
2
2a
3
4
5
6
7
7a
8
9
Climates
Tropical, humid
Tropical, humid summers
Tropical, semihumid
Subtropical, arid
Warm temperate, dry
summers
Warm temperate, humid
Cool temperate
Cool temperate, dry
summers
Cool temperate, arid
Cold temperate
Polar
Modern vegetation
Permian biomes
Tropical rain forest
Tropical deciduous forest
Savanna
Desert
Sclerophyllous woody plants
Tropical ever wet
Tropical summer wet
Tropical summer wet
Desert
Winter wet
Temperate evergreen forests
Nemoral broadleaf deciduous
forest
Steppe
Warm temperate
Cool temperate
Desert
Boreal coniferous forest
Tundra
Midlatitude desert
Cold temperate
Tundra
the annual cycle. In the tropical and subtropical
regions, the rain forest and desert biomes represent
extremes of precipitation, while the summer-wet
(savannah) and winter-wet (Mediterranean) biomes
are related to the sharp seasonality of rainfall. We
define “wet” as a minimum of 40 mm of precipitation per month because this is approximately the
amount necessary to stimulate biotic productivity.
The other defined biomes are the high-latitude deserts, which are isolated from precipitation by continentality or mountain chains, and the glacial (essentially abiotic) biome. Of course, the temperature
and precipitation parameters of the geologic record
are difficult to measure directly, and the assignment of Permian floras to biomes must be based
on the assumption that plant responses to the climate stringencies were similar to those observed
today. At a highly simplified level, low-latitude
forms today may have large or small leaves relating
to wet or dry conditions, whereas higher-latitude
forms may be deciduous or small-leafed evergreen.
Similarly, for the geologic past, knowledge of plant
physiognomy and phytogeography enables us to interpret the prevailing environmental conditions.
The climate-sensitive sediments are very useful
as well (Lottes and Ziegler 1994; Ziegler et al. 1998).
Coals indicate sufficient precipitation through the
warmer months to stimulate growth and ensure a
constant water table, so these are found in the tropical rain forest belt as well as in the temperate biomes. Evaporites form in areas where evaporation
exceeds precipitation; therefore, the desert as well
as seasonally dry biomes are indicated, and eolian
sands seem to be restricted to the main desert belts.
Tillites and reefs are also critical in establishing
temperature extremes. Paleolatitudinal position
can also be helpful in biome assignment, particularly as it relates to the main precipitation belts of
the tropical and temperate zones. These are sepa-
Midlatitude desert
rated by the great subtropical deserts, centered at
about 25⬚ north and south, which owe their existence to the descending limbs of the Hadley cells,
a dynamic feature of earth’s circulation that seems
to have been a constant feature throughout at least
the Phanerozoic (Parrish 1998). We emphasize that
many boundaries today between biomes are gradational rather than distinct and that this must
have been true in the past. The lines dividing biomes on the paleogeographic maps are therefore
simply a cartographic device and are not meant to
imply sharp boundaries.
In the geologic record (see Rees et al. 2000 for
Jurassic examples), glacial and tundra biomes can
be delimited on the presence or absence of tillites
and/or the absence or occurrence of higher vascular
plants. The tundra versus cold temperate distinction is based on the recognition of stunted bushes,
or “ground cover,” versus arborescent forms (trees),
where winter conditions become too cold and the
growing season too short in the tundra even for
evergreen trees. Frozen substrates inhibit tree
growth, enabling smaller plants to dominate this
region. Cold and cool temperate biomes are distinguished on the basis of the predominance of evergreen versus deciduous elements in the vegetation.
This is related to the relative advantages of each
strategy, whereby evergreen plants outcompete deciduous ones in a shorter growing season at higher
latitudes. Evergreen trees respire, using food reserves, during dark, high-latitude winters, but the
low temperatures minimize expenditure of resources. Evergreen leaves tend to be small and thick
cuticled, minimizing water loss during the winter,
when root and trunk systems are less efficient. Deciduous trees would seem to be at an advantage in
cold winters since the absence of leaves means
lower resource requirements. However, deciduous
trees need to grow leaves in the spring, and the
18
P. M . R E E S E T A L .
Table 4. Walter Climates and Biomes (by Number) as a Function of the Number of Months with Temperatures of
10⬚C or More and the Number of Months with 40 mm or More Precipitation
Number of months having 40 mm or more of precipitation
0
1
2
3
4
5
6
7
8
9
10
11
12
Number of months
temperatures 110⬚C
3
3
7a
7a
7a
7a
7a
7a
7
7
8
9
9
3
3
7a
7a
7a
7a
7a
7a
7
7
8
9
9
3
3
7a
7a
7a
7a
7a
7
7
8
8
9
9
3
3
7
7
7
7
7
7
7
8
8
9
9
2a
2a
7
7
7
7
7
7
7
8
8
9
9
2
4
7
5
5
5
6
6
8
8
8
9
9
2
4
4
5
5
5
6
6
8
8
8
9
9
2
4
4
5
5
5
6
6
8
8
8
9
9
2
4
4
5
5
5
6
6
8
8
8
9
9
2
4
5
5
5
5
6
6
8
8
8
9
9
2
4
5
5
5
5
6
6
8
8
8
9
9
1
5
5
5
5
5
6
6
8
8
8
9
9
1
5
5
5
5
5
6
6
8
8
8
9
9
12
11a
10
9
8
7
6
5
4
3
2
1
0
Note. Criteria used to calculate the model biome results. Two additional criteria were added to help distinguish between biomes
1 and 5 and biomes 2 and 4. If the number of months having temperatures greater than 10⬚C was 12 but the “growing season degree
months” (GSDM) was less than the adjustable parameter GSDM0, the second row of the translation table was used instead of the
first. GSDM was defined as the quantity (mean monthly temperature [⬚C] minus 10); this quantity is summed over all months in
the year. After experimentation, GSDM0 was set at 155 degree months. This additional criterion was needed because both tropical
and temperate climate biomes may have all months well above 10⬚C, yet temperate-climate biomes have “winters” that may drop
to, say, 15⬚C, whereas tropical-climate biomes stay evenly warm. For example, for a humid region that has 12 mo 1 10⬚C, if the
temperature is 25⬚C each month, then GSDM is (25 ⫺ 10) # 12 p 180 and it is classified biome 1; however, if the temperature drops
to, say, 15⬚C for 3 mo, then GSDM is (25 ⫺ 10) # 9 ⫹ (15 ⫺ 10) # 3 p 150 and it is classified climate biome 5. To further distinguish
between climate biome 4 (warm temperate, dry summers, or “Mediterranean”) and climate biome 5 (warm temperate, humid), we
used the Koeppen criterion, that for Mediterranean climates (Cs, summer drought), the rainfall of the wettest winter month is at
least three times that of the driest summer month. For example, if the two primary criteria identified a region as climate biome
4, then the above mentioned criterion on summer versus winter rain is applied either to confirm the classification or change it
from 4 to 5.
growing season must be sufficient to enable this as
well as the overall growth of the plant and reproductive organs. The warm temperate biome generally experiences a short resting period in the winter but is populated with broadleaf evergreen
plants, as well as some of the more typically tropical growth forms, such as tree ferns and cycads.
Although the warm temperate biome may contain
some elements more typical of each of these neighboring biomes, changes in abundance of different
plant types, together with coal occurrences, enable
subdivisions of this part of the fossil floral and climate spectrum. Also, floral diversity decreases toward the highest latitudes, providing further information on the positions of high-latitude biome
boundaries. The seasonally dry (winter wet or summer wet) and desert biomes are recognized by the
extent of microphyllous plants on one side and the
typical occurrence of extensive evaporites and eolian sands defining the desert biome. The tropical
ever-wet biome is characterized by high diversity,
large arborescent forms, vines, and swamp deposits.
We discuss here some of the features seen on our
Sakmarian and Wordian biome maps (fig. 6A, 6D).
In the equatorial regions of the Sakmarian, the axis
of the Cathaysian Realm is represented by highdiversity floras of the Chinese microcontinents and
equatorial Pangea (i.e., Euramerica), and we assign
a tropical ever-wet biome to the parts of these
regions containing such floras. Upland settings and
island arcs may have helped to bridge gaps between
these areas, as has been mentioned. European and
United States floras are often referred to the
Euramerican or Atlantic Province, and although we
recognize the Euramerican geographic province, we
prefer to interpret the vegetation as representing a
seasonally dry variant of the Cathaysian Realm as
represented in China and assign it to a tropical
summer-wet biome. This area of Euramerica experienced a transition from the rain forest biome
of the Late Carboniferous to more stressed, presumably summer-wet conditions by the Early Permian (Glennie 1984). Apparently the transition was
not a smooth one, as floras typical of the Carboniferous alternate with Permian ones, and this has
proved confusing in defining the systemic boundary (Broutin et al. 1990). Coals do occur in the Early
Permian of Europe, but most of these are cannel
(algal) coals in local basins rather than the rain forest accumulations of higher-plant debris typical of
earlier times. There are exceptions to this pattern;
some of the Euramerican floras from southern Europe, North Africa, Venezuela, and the southwestern United States contain elements such as gigan-
Journal of Geology
P E R M I A N P H Y T O G E O G R A P H I C PAT T E R N S
topterids (fig. 2A) and have higher diversities and
lower CA axis 3 scores (fig. 3B) than those to the
north. We therefore assign these to the tropical
ever-wet biome (fig. 6A), while noting that they do
not contain classic rain forest elements, such as the
abundant lycopsids recorded for the Sakmarian of
China or Carboniferous of Euramerica. The high
diversity patterns are maintained in China into the
Wordian (fig. 2A), but floras are largely absent from
Euramerica (fig. 2B). The Euramerican region has
only a few low-diversity assemblages near the eastern shoreline by this time, and we assign it to the
tropical summer-wet biome.
North and south China were moving northward
during the Permian and, by the Late Permian, north
China was entering the seasonally drier zone, as
indicated by thinner coals and lower floral diversity
(Liu 1990; Li and Wu 1996). However, truly arid
conditions were not experienced until after the
Wordian stage (Wang 1993). The Permian desert
belts are made evident by the extensive evaporite
and eolian sand deposits (fig. 1A, 1B). They conveniently separate the equatorial Cathaysian
Realm (China and Euramerica) from the temperate
Angaran Realm to the north and the temperate
Gondwanan Realm to the south (fig. 6A, 6D).
Our interpretation of the north temperate Angaran Realm is based on the macrofloral interpretations of Durante (1995) and the extension of the
traditional Russian floral subdivisions into Greenland, Canada, and Alaska, using microfloral assemblages as tracers (Utting and Piasecki 1995). We assign the predominantly cordaite-bearing floras in
the midlatitude areas of Angara to a cool temperate
biome because of their deciduous nature, which
precludes a warm temperate assignment. The highest latitudes of Angara (the Siberian Platform) contain lower-diversity floras and “small-leafed cordaites” (Meyen 1982, p. 69), including the genus
Crassinervia discussed earlier (figs. 3D, 4D), and
we assign these to the cold temperate biome. South
of these areas is a coal-free belt adjacent to the evaporite deposits of the Russian Platform, which is
referred to as the Subangaran Province. We assign
this to the winter-wet biome because of the “significant admixture of Euramerican elements” (Durante 1995, p. 130) and because we feel that a setting like this would receive precipitation from
winter storms that would develop over the ocean
to the north. The Russian Platform includes the
Pechora Basin at the northern end of the Urals and
has high floral diversity plus abundant coals
throughout the Middle and Late Permian, so this
is assigned to a warm temperate biome on our Wordian map (fig. 6D). M. V. Durante (pers. comm.)
19
believes that the climate was warm here but points
out that the trees have annual growth rings, indicating some degree of seasonality.
The southern temperate Gondwanan Realm has
an axis of moderate diversity centered along the
50⬚S line, which we assume represents the cool
temperate biome because of the dominance of the
broad-leafed deciduous Glossopteris (Gould and
Delevoryas 1977). During the Permian, Gondwana
rotated about a pole near Australia so that the Paraná Basin of South America moved northward into
the warm, semiarid biome (Guerra-Sommer et al.
1995) while the Sydney Basin of Australia remained
at the transition of the cold and cool temperate
zones during the stages of interest here. Basins in
between all show some degree of warming, including the Karoo basins of South Africa (Falcon 1986)
and Tanzania (Kreuser et al. 1990), and the Gondwanan basins of India (Tiwari and Tripathi 1987;
Chandra and Chandra 1987). All of these regions
experienced glaciation before the Sakmarian, so the
climate warming must have been due in part to
deglaciation because the ice sheet would have generated and exported much cold air through gravitational outflow. Areas in western Argentina contain Permian desert sands thought to have
developed in the wind shadow of the marginal Andean arc systems (Limarino and Spalletti 1986).
High-diversity floras occur on the seaward side of
the arc system in Patagonia, which have been
thought to be “at odds with the present location of
the Patagonian plate” (Cuneo 1996, p. 78). A more
parsimonious explanation would link this site with
the warming effects of a maritime setting and a
warm poleward current (Ziegler 1998; see “Discussion” below).
At the other end of the temperature spectrum,
the high-latitude floras of Antarctica have very low
diversities (Cuneo et al. 1993), and we assign these
to the cold temperate biome. As mentioned earlier,
at high latitudes with cold winters and short growing seasons today, evergreen trees typically have an
advantage over deciduous ones, being able to photosynthesize as soon as light and temperature conditions reach sufficiently high levels, without having to produce new leaves. However, here we
consider the possibility that CO2 levels were higher
in the Permian than today. Leaf photosynthetic activity as well as plant growth and productivity
might well have been higher in such a regime, even
in a short growing season. Deciduous plants would
have had the advantage of minimizing expenditure
of resources during the dark winter dormancy,
whereas evergreen ones would have needed to continue nutrient provision to the leaves, expending
20
P. M . R E E S E T A L .
more resources. So, even with low winter temperatures and a short growing season, fossil deciduous
plants such as Glossopteris may have fared better
than evergreen ones. Thus, the preserved deciduous
plants may not necessarily indicate typically cool
temperate conditions as defined today, instead representing cold temperate climates with short growing seasons but in an atmosphere with elevated
CO2 levels. This idea, combined with lower floral
diversities, leads us to assign a cold temperate biome to these regions. Finally, we map a small area
of tundra in the interior of Antarctica on our Sakmarian map because we feel that regions distant
from the ocean would have had this climate, but,
admittedly, there are no deposits to support this
interpretation. Our model biome results, using a
range of CO2 levels, are discussed in the following
section.
Climate Data and Model Comparisons
The Sakmarian and Wordian paleogeographic maps
of Ziegler et al. (1997) were used as boundary conditions for experiments using the GENESIS Version
2 Global Climate Model (Thompson and Pollard
1997). Other boundary conditions included appropriately reduced solar luminosity (2.4% and 2.1%
relative to present for the Sakmarian and Wordian,
respectively), varied levels of atmospheric CO2
(one, four, and eight times present levels), and a
range of different orbital configurations in the case
of the Wordian. Prescribed land surface parameters
included uniform vegetation consisting of mixed
tree and grassland or savanna. This prescribed uniformity is clearly unrealistic; for instance, large areas of central Pangea were probably desert. However, prescribing the estimates of Pangean biomes
would have limited the utility of data/model comparisons as a means of assessing the accuracy of the
simulation because vegetation can have a significant effect on climate (e.g., Dutton and Barron
1997; Otto-Bliesner and Upchurch 1997). An intermediate (loamy) soil texture (43% sand, 39% silt,
and 18% clay) was also prescribed at every land grid
point. Our choice of a uniform “average” land surface in both the Sakmarian and the Wordian, although introducing a bias, allowed us to isolate
changes due to paleogeography and atmospheric
CO2 alone (cf. Fawcett and Barron 1998). Full details of the results of these experiments are given
in our companion article (Gibbs et al. 2002). Here,
we summarize the important aspects pertaining to
model biome determinations and then compare
these with the data-derived biomes.
In our companion article (Gibbs et al. 2002), we
found features that are typical of many previous
Pangean climate model simulations, such as high
aridity in central Pangea, large monsoons along the
Tethyan margins, and precipitation focused around
tropical mountains (Kutzbach and Gallimore 1989;
Kutzbach and Ziegler 1993; Otto-Bliesner 1993;
Barron and Fawcett 1995; Crowley et al. 1996). We
also tested model predictions against new global
compilations of Sakmarian and Wordian lithological climate indicators (Ziegler et al. 1998) such as
coals, evaporites, eolian sands, carbonate buildups,
tillites, glacial dropstones, oil source rocks, and
phosphorites. Overall, model performance is generally good when its predictions of temperature,
precipitation/evaporation ratio, and wind directions are tested against these indicators. Most important, the model captures temporal trends in paleoclimate that are evident in many locations.
These trends result from the general northward motion of Pangea by ∼15⬚ latitude through the Permian, moving particular regions in and out of different climate zones (Ziegler et al. 1997). Coals,
evaporites, and eolian sands provide useful information about extremes of precipitation and evaporation ratios. Although the climate model performs well against these indicators, of greater
interest is its performance with respect to the fossil
plant data and their inherent signal across the
global climate spectrum.
The simulated Sakmarian and Wordian model results (Gibbs et al. 2002) are expressed here in terms
of biomes based on the criteria for monthly averages of temperature and precipitation developed by
Kutzbach and Ziegler (1993). However, those criteria were modified here to incorporate a more realistic monthly growing-temperature threshold for
vascular plants of 110⬚ rather than 15⬚C, compatible with the Walter biome scheme (see earlier comments; tables 3, 4). As well as providing a means
of quickly and easily visualizing regional climates,
it allows us to make a comprehensive global comparison between the observed biome distributions
(fig. 6A, 6D) and the modeled biome distributions
(fig. 6B, 6C, 6E, 6F).
The model reproduces the data-derived biome
patterns reasonably well in the Tropics and northern high latitudes. However, even with a high CO2
level (eight times the present; fig. 6C, 6F), model
summer temperatures rise only just above freezing
in the highest southern latitudes. This region contains the greatest data/model discrepancy; cold
temperate conditions (biome 8) are indicated by the
paleobotanical data, whereas tundra (biome 9) is
predicted by the model. We discuss some of the
model shortcomings later but point out here that
Journal of Geology
P E R M I A N P H Y T O G E O G R A P H I C PAT T E R N S
riparian vegetation (adapted to life on river or lake
margins) is probably overrepresented in many fossil
plant assemblages, which may produce a significant bias affecting data/model biome comparisons.
For instance, in areas today where the regional vegetation is tundra, evergreen or even some deciduous
trees often grow along river margins. Riparian
plants are also more likely to have been preserved
in fossil depositional environments and tundra vegetation, comprising small-stature plants, will be
less recognizable as such even if preserved occasionally in these fossil assemblages (often being described as “unidentifiable leaf fragments” or else
ignored). It is therefore possible that a cold (or even
in some cases cool) temperate biome could be assigned to what was regionally predominantly tundra. The consequences of taphonomic biases inherent in the fossil record should be borne in mind
when comparing the data directly with coarseresolution model results. For example, the spatial
resolution of the GENESIS 2 model used here is
3.75⬚ latitude # 3.75⬚ longitude, so more local effects, such as riparian vegetation and its climate
signal, will not be captured. High-latitude data/
model discrepancies are common to other warm
intervals (e.g., Huber et al. 2000), and fully resolving them remains a fundamental problem in paleoclimatology (Barron et al. 1995; Crowley and North
1996; Schmidt and Mysak 1996; Huber et al. 2000).
A further discrepancy exists between both the
Sakmarian and Wordian data and model results, in
this case for Mongolia (fig. 6). As discussed earlier,
Mongolian floras are more similar to those from
Angara than north China (fig. 2A, 2B; figs. 3A, 4A)
and can be assigned a cool temperate biome (fig.
6A, 6D). However, the model results indicate a
warm temperate or even tropical ever-wet biome
(fig. 6B, 6C, 6E, 6F). Here, the problem is probably
with the paleogeographic reconstruction (Ziegler et
al. 1997), which should be modified to position
Mongolia closer to Angara. Paleomagnetic data are
inadequate to constrain the position of Mongolia,
and it is known that the collision between Mongolia and north China did not occur until the very
latest Permian. The post-Wordian portion of the
Permian is now thought to represent a significant
length of time, perhaps 16 m.yr. (Menning 1995),
allowing time for the Mongolian arcs and the north
China microcontinent to converge. In this case, we
have an excellent example of how fossil plant data
can be used to constrain the position of continental
fragments and terranes when other information is
lacking (see Ziegler et al. 1996 for Mesozoic examples). Although coarse model resolution is a major factor, incorrect specification of boundary con-
21
ditions (in this case, the paleogeography) also
contributes to data/model discrepancies.
It can be seen from figure 6 that the model results
using 8#CO2 (fig. 6C, 6F) match the data slightly
better than those using 4#CO2 (fig. 6B, 6E), particularly at high northern latitudes in the Wordian.
There is also some reduction in the areal extent of
the cool temperate, cold temperate, and tundra biomes, as would be expected. Such broad visual
comparisons are useful in order to see the main
patterns but do not enable detailed comparisons at
the local and regional level. We therefore compared
the data and model results on a geographic gridcell-by-grid-cell basis (2⬚ latitude # 2⬚ longitude),
enabling direct comparison of the biome result derived from the floral localities present within a grid
cell with the modeled biome for the same grid cell.
Figure 7 illustrates our approach, using a Wordian
floral locality (at 47.3⬚N, 36.8⬚E) from the Russian
Platform, which we assigned to the warm temper-
Figure 7. Example of a modeled biome grid cell result
available for comparison with the biome result derived
from the proxy data. The site is a Wordian coastal one
(at 47.3⬚N, 36.8⬚E) on the Russian Platform and was assigned a warm temperate biome 5 on the basis of the
floral data. It is compared here with the model biome
result from our 8#CO2 experiment (WORD-8#CO2) for
the corresponding 2⬚ latitude # 2 ⬚ longitude grid cell
(centered on 47⬚N, 37⬚E, and showing the adjacent
2⬚ # 2⬚ cells). The center cell shows a modeled cool temperate biome 6, so there is a mismatch between the data
and model results at this direct level of comparison.
However, there is a modeled biome 5 cell immediately
to the southeast of the central cell, so a match is indicated at this level of comparison.
22
P. M . R E E S E T A L .
ate biome 5. It is compared here with the model
biome result from our 8#CO2 experiment (WORD8#CO2) for the corresponding 2⬚ latitude # 2⬚ longitude grid cell (centered on 47⬚N, 37⬚E and showing the adjacent 2⬚ # 2⬚ cells). Note that the cells
with values of 0 are not land grid cells and so are
not assignable to a terrestrial biome. We used three
main levels of comparison: (1) “direct cell match”
of the data and model biome, (2) “direct match plus
most common of the adjacent cells,” and (3) “direct
plus any adjacent cell match.” Obviously, a level
1 match would be the ideal result. However, because of data and model uncertainties, including
coarse spatial resolution and temporal averaging,
the adjacent grid cells were also considered (levels
2 and 3). The level 2 comparison allows us to use
the most common biome score of the adjacent grid
cells for localities that did not produce a direct
match. Level 3 allows any adjacent cell with the
same biome value as the data to be counted as a
match, even if the other adjacent cells produced a
different value. The results in our example (fig. 7)
show no direct cell match at level 1, since the data
indicate biome 5 but the model produces biome 6.
Likewise, there is no match at level 2, since the
most common adjacent model cell is biome 6.
However, level 3 does provide a match, since the
model produces a biome 5 value in the adjacent
southeast grid cell. So, the locality illustrated in
figure 7 is a coastal site, on the western margin of
Angara, and is a warm temperate biome 5 from the
data but borderline biome 5/biome 6 from the
model. A fourth level of data and model cell comparison, “direct or any adjacent cell match Ⳳ biomes,” was also calculated (table 5). This essentially shows the maximum possible match between
the data and model results, allowing for possible
misinterpretations of biomes from the floral lists
(e.g., cool temperate 6 vs. cold temperate 8).
The results at the four levels of comparison for
all Sakmarian and Wordian localities, and for the
different model experiments, can be seen in table
6 and figure 8. These show the number of floral
localities and model grid cells that match, expressed as a percentage of all floral localities used
in each comparison. For comparison levels 1–3,
Wordian results for 8#CO2 and 4#CO2, for a range
of summer orbits (warm, cold, and circular), show
that 8#CO2 consistently produces a slightly better
match with the data than 4#CO2 (fig. 8A–8C). The
Sakmarian results (fig. 8D), for a circular orbit and
with the addition of a 1#CO2 experiment, show a
similar pattern to the Wordian, with 8#CO2 producing the best match but with both 4#CO2 and
8#CO2 being considerably better matches than
Table 5. Possible Range of Biomes Allowing for Misinterpretation of Data
Data biome
number
1
2
3
4
5
6
7
8
9
Model biome number
1
1
2
3
1
5
3
6
8
2
2
3
4
2
6
6
7
9
5
3
4
5
4
7
7
8
10
5
7
5
8
8
9
6
Note. The possible range of biomes allowing for misinterpretations of the floral data (e.g., if cool temperate biome 6 could
be warm temperate 5, midlatitude desert 6, or cold temperate
8). In reality, the situation is not as extreme since the phytogeographic patterns, distributions of lithological climate indicators, floral diversity patterns, and foliar morphologies all contribute to reduce the uncertainties.
1#CO2. However, the overall matches are slightly
lower than for the Wordian, and there is also less
distinction between the 8#CO2 and 4#CO2 results. The improved overall match of data and
model results between comparison levels 1–3 is due
to increasing use of results from adjacent model
grid cells. Direct comparisons of the three Wordian
orbital parameters are shown for 8#CO2 (fig. 8E)
and 4#CO2 (fig. 8F). There is little difference between the results for each level of comparison. All
of these experiments incorporated the large Gondwanan lakes shown on our paleogeographic reconstructions (figs. 1, 2, 6). An additional 4#CO2 experiment, without lakes (WORD-4#CO2, NLK),
was also conducted for the Wordian. The results
(table 6) show that the presence or absence of these
lakes does not significantly alter the percentage
match between the data and model biomes. Clearly,
the main differences in data and model grid cell
matches are due to the different CO2 levels used in
the model experiments. At the level 4 comparison
(“Ⳳbiomes”), overall floral data and model results
are excellent (producing matches between 84% and
93%), but there is no clear distinction between different CO2 values. Suffice it to say that this level
of comparison indicates the potential “best fit” between the data and model. Future studies will hopefully produce similarly high matches for the more
direct levels of comparison as we improve our data
and model approaches, and we discuss some of
these later.
These results (table 6; fig. 8) provide a quantitative assessment of the data and model comparisons. As such, they represent an advance over the
biome maps (fig. 6) and their interpretations. However, each histogram in figure 8 shows only the
Journal of Geology
P E R M I A N P H Y T O G E O G R A P H I C PAT T E R N S
23
Table 6. Percentage Match between Sakmarian and Wordian Floral Data and Model Results for Different Levels of
CO2 and Orbital Parameters for the Four Comparison Levels
Model run
WORD-8#CO2,
WORD-8#CO2,
WORD-8#CO2
WORD-4#CO2,
WORD-4#CO2,
WORD-4#CO2
WORD-4#CO2,
SAK-8#CO2
SAK-4#CO2
SAK-1#CO2
WSO
CSO
WSO
CSO
NLK
(1) Direct cell
match
(2) Direct ⫹
most common
adjacent cell
(3) Direct ⫹
any adjacent
cell
(4) Direct or adjacent
cell Ⳳ biomes
32
33
34
21
23
21
22
30
20
7
36
38
39
25
27
24
30
28
20
10
51
56
57
38
39
35
40
41
37
14
93
91
91
92
91
89
92
84
89
85
Note. See text for details and figure 8 for histogram plots. The experiments are Wordian 8#CO2 with lakes (for warm, cold, and
circular orbits: WORD-8#CO2, WSO; WORD-8#CO2, CSO; WORD-8#CO2); Wordian 4#CO2 with lakes (for warm, cold, and
circular orbits: WORD-4#CO2, WSO; WORD-4#CO2, CSO; WORD-4#CO2), plus 4#CO2 circular orbit without lakes (WORD4#CO2, NLK); and Sakmarian 8#CO2, 4#CO2, and 1#CO2, circular orbit with lakes (SAK-8#CO2, SAK-4#CO2, SAK-1#CO2).
percentage biome match between the different
model CO2 levels and data summed for the whole
world and does not reveal anything about geographic variations in the data and model matches.
We therefore plotted our results from the level 3
comparisons for 4#CO2 and 8#CO2 on the Sakmarian and Wordian maps (fig. 9). The Sakmarian
experiments were conducted using a circular orbit,
and so we chose to plot the circular orbit results
for the Wordian to enable direct comparison of the
spatial patterns in data and model matches between
each stage. The floral localities are shown as
squares on each map; solid ones indicate a match
with the model results, and open ones indicate a
mismatch. Neither CO2 level produces biomes that
match the data in the highest southern latitudes,
but each performs well in the highest northern latitudes. As expected from the results shown in figure
8, differences between Sakmarian 4#CO2 and
8#CO2 levels (fig. 9A, 9B) are minor, with 8#CO2
producing a few additional matches in Gondwana
but 4#CO2 doing slightly better in north China.
Differences between Wordian 4#CO2 and 8#CO2
levels (fig. 9C, 9D) are more pronounced, with
8#CO2 producing significantly more matches in
central Angara and north China, as well as some
additional matches in Gondwana. Thus, for the
Wordian, the model biomes using 8#CO2 provide
a better match with the data across a greater range
of geographic regions and biomes than 4#CO2 and
therefore reproduce the global climate patterns
more comprehensively.
We were consistent in our approach to analyzing
Sakmarian and Wordian floras, phytogeographic
patterns, and biome interpretations. For the data/
model biome comparisons shown in figure 9, we
compared the modeled Sakmarian and Wordian
4#CO2 and 8#CO2 results directly (using the “circular orbit plus lakes” experiments). For the Wordian level 3 comparison, the overall percentage
data/model match with 8#CO2 (57%) is significantly higher than 4#CO2 (35%), whereas differences for the Sakmarian are relatively minor (41%
vs. 37%). Although we are still in the early stages
of our work, we feel this suggests that CO2 levels
were higher in the Wordian than Sakmarian, which
is consistent with overall warming trends observed
from the geologic data. Our results are perhaps less
significant than the approach described here, which
can be applied to climate studies for any geologic
interval. We have developed quantitative and objective methods of analyzing the proxy climate data
and a direct means of comparing them with the
corresponding model results. Our results also provide indicators as to how we can improve these
approaches, and these are discussed below.
Discussion
The only other Permian global data and model studies so far conducted (Kutzbach and Ziegler 1993;
Rees et al. 1999) compared results for the Wordian.
We chose the same stage here to incorporate refinements in the data and model approaches and to
enable direct comparison with previous results, but
we also included the Sakmarian to investigate temporal changes through the Permian. The previous
results also showed a major discrepancy in the
southern high latitudes. Using the NCAR CCM1
model, Kutzbach and Ziegler (1993) predicted cold
temperate conditions (biome 8), whereas their data
indicated a cool temperate biome 6. However, they
noted (citing Truswell 1991) that a cold temperate
biome may be assignable based on relatively low
24
P. M . R E E S E T A L .
Figure 8. Percentage matches between floral data and different model results (see text for details and note to table
6 for explanation of model experiment abbreviations).
floristic diversity in this region compared with elsewhere in Gondwana. Our new results (fig. 2A, 2B;
fig. 6A, 6D) support a cold temperate interpretation
from the data for the Sakmarian and Wordian. Intriguingly, then, the new Wordian data interpretations presented here match the previous CCM1
simulation better than our new GENESIS simulation for this region (which produces tundra; fig. 6E,
6F), even though the CCM1 model resolution is
coarser (4.5⬚ latitude # 7.5 ⬚ longitude) than that of
GENESIS 2. It should be noted however that CCM1
predicts conditions that are too warm in the north-
Journal of Geology
P E R M I A N P H Y T O G E O G R A P H I C PAT T E R N S
25
Figure 9. Floral locality and model percentage matches for (level 3 comparison; see text for details) 4#CO2 and
8#CO2 in the Sakmarian (A, B) and Wordian (C, D) using results of the “circular orbit plus lakes” experiments. Solid
squares indicate a match between data and model biome results; open squares indicate a mismatch.
ern high latitudes compared with the data, whereas
the model results shown here (particularly for
8#CO2; fig. 6F) provide a better match to the cold
temperate conditions in the northern hemisphere
indicated by the data. Although Kutzbach and Ziegler (1993) used different values of atmospheric CO2
level (five times the present) and solar luminosity
reduction (1%), the net radiative forcing is essentially the same. We attribute this model-model difference primarily to different representations of latitudinal averages of poleward ocean heat transport
within the simple mixed-layer ocean scheme used
by the models. Kutzbach and Ziegler (1993) pre-
scribed values of ocean heat transport based on estimates from an ocean GCM experiment with an
idealized Pangean paleogeography (Kutzbach et al.
1990). In contrast, the GENESIS 2 scheme (Thompson and Pollard 1997) predicts substantially lower
ocean heat transport values based on the latitudinal
sea surface temperature gradient (see Gibbs et al.
2002 for a more detailed discussion).
Our new model results (fig. 6C, 6F) indicate that
even CO2 at eight times the present-day atmospheric level (PAL) is still insufficient to produce
the high-latitude conditions in the Permian inferred from floral data (fig. 6A, 6D), even when al-
26
P. M . R E E S E T A L .
lowing for uncertainties in interpreting the fossil
record. This level of CO2 is probably the highest
reasonable value that can be inferred for this time,
based on Berner’s (1994) geochemical cycle modeling (see discussion in Gibbs et al. 2002); clearly,
then, other climate forcing factors must be involved. One solution may be a “warm polar current” (Ziegler 1998), which the GENESIS 2 model
would be unable to resolve. This idea is based on
the fact that the poles during much of the geologic
past were less obstructed by continents than today,
allowing currents like the Norwegian Current to
transport heat toward the pole and keep it ice free
through the annual cycle. General circulation models (GCMs) cannot realistically fully reproduce past
climates unless the atmosphere can interact with
the ocean and vice versa. Preliminary experiments
with equilibrium asynchronous coupling (Liu et al.
1999) between an atmosphere and an ocean GCM
show that simple modifications of today’s geography and sill depths (e.g., a wider and deeper Bering
Strait) allow warm currents to extend poleward of
present limits (Ziegler 1998). A warm polar current
could have arisen under the paleogeographic regime
of the Permian, where one large supercontinent
moved off the South Pole. The northward shift of
Gondwana through the Permian (Ziegler et al.
1997), coupled with a rise in atmospheric CO2 (Berner 1994), may have initiated deglaciation and polar
warming by allowing warm ocean currents to reach
the south polar region more effectively.
Another major paleoclimate data/model mismatch is “equable” climates in continental interiors; that is, model winter cooling greatly exceeds
that inferred from the geologic record. In our results, this discrepancy is particularly evident for
central southern Gondwana. The model predicts an
extensive area of cold temperate and tundra conditions (biomes 8 and 9, reflecting a short growing
season) for the Sakmarian and Wordian, whereas
we infer cool and cold temperate conditions (biomes 6 and 8) from the floral data. In fact, the discrepancy is worse than just a slight difference in
the exact length of the growing season but rather
between the equable climate indicated by the geologic record and the extreme seasonality predicted
by the climate model for this region. As discussed
in our companion article (Gibbs et al. 2002),
organic-rich lacustrine shales are present in Africa.
This would imply that winter air temperatures
never fell much below 4⬚C, yet the model predicts
substantially lower average winter temperatures
(⫺10⬚C or less). The idea here is that water achieves
its maximum density at 4⬚C, leading to turnover
of the water column and oxygenation and destruc-
tion of organic matter on the lake bed. Of course
there is a phase lag between the air and water temperatures. Furthermore, as well as the floral evidence discussed above, there is faunal evidence
(e.g., Karoo vertebrates) for equability (Yemane
1993; Ziegler 1993; but cf. Crowley 1994).
Although elevated CO2 increases the summer
maximum and winter minimum temperatures
slightly in the interior of southern Gondwana, our
model results indicate that this area is still subject
to extreme seasonality, and cold winters in particular. The presence of large freshwater lakes in this
region has been proposed to account for the discrepancy (Yemane 1993). The Ziegler et al. (1997)
Wordian map (figs. 1B, 2B, 6D) depicts a large seaway (the Parana-Karoo Inland Sea) that interconnected the basins of southern South America,
southern Africa, and Antarctica (see also Ziegler et
al. 1998), which, if anything, may perhaps be an
underestimate of the total area covered by freshwater lakes proposed by Yemane (1993). However,
in our companion article (Gibbs et al. 2002), we
conducted an extreme sensitivity experiment
whereby we excluded all of this huge seaway from
the map, replacing it with the same uniform soil
and vegetation as for other land grid points. This
experiment demonstrated that the presence of lakes
does not substantially ameliorate the regional climate if the “base” global climate is already cold,
that is, if the total net forcing from solar insolation,
atmospheric CO2, and poleward ocean heat transport allows temperatures to fall low enough in winter that sea ice can form on the lakes. Once this
critical threshold is reached, then this area is covered by a high-albedo surface, similar to the surrounding snow-covered land surface, and the regional climate cools substantially. This result is in
contrast to that of Kutzbach and Ziegler (1993),
who, as we have discussed above, used a much
higher prescribed ocean heat transport (and consequently, a much warmer global climate) than we
have used in this work. In Kutzbach and Ziegler’s
(1993) experiments, the ameliorating effect of lakes
in southern Gondwana on the regional climate was
more pronounced, not least because ice did not
form on the lakes.
Indeed, other investigators have found that the
inclusion of lakes in climate modeling studies does
not have a major effect, being limited regionally,
as would be expected from present-day observations (witness the present-day climate of Chicago).
Sloan (1994) found that incorporation of the Green
River Lake (∼15,000 km2) in a GENESIS Version 1
Early Eocene climate modeling study was critical
to reproducing the regional climate of western
Journal of Geology
P E R M I A N P H Y T O G E O G R A P H I C PAT T E R N S
North America. However, in her experiments, the
winter freeze line was deflected poleward only in
the immediate region of the lake (which contains
the most paleoclimate data sites for this interval);
elsewhere, winter temperatures remained substantially below freezing. Of particular relevance to our
work is Crowley and Baum’s (1994) study of Late
Carboniferous interglacial climates. Motivated by
Yemane’s (1993) observations, they prescribed several very large (albeit hypothetical) lakes in southern Gondwana (in a similar location to the Karoo
Inland Sea) with a total area of 4.2 # 10 6 km2. As
with our results, although these lakes locally increased winter temperatures by ∼10⬚C (with relatively little difference further away), winter temperatures still fell to ∼⫺40⬚C in central southern
Gondwana.
An alternative solution that has often been proposed for the “equable continental interior problem” is increased poleward ocean heat transport,
along with increased advection of heat into continental interiors. The climate model experiments
reported here have only a simple representation of
ocean heat transport and no representation of specific, geographically located warm currents that
could have contributed to polar warmth. However,
even a polar warm current (as discussed above)
might not be sufficient to advect heat so far inland.
As Crowley has frequently observed (e.g., Crowley
and Baum 1991; Crowley 1994), today the Gulf
Stream affects climate only one tenth of the way
into the Eurasian landmass; the problem is further
magnified by the huge size of Pangea. Thus, a warm
polar current, while a potential solution for explaining warm conditions at high-latitude coastal
sites, may not fully explain such conditions. As
Crowley and Baum (1994, p. 19) point out, this fundamental problem “is a major impediment to further progress in understanding paleoclimate fluctuations on supercontinents.”
Vegetation changes in high latitudes may also
have significant effects on climate (Dutton and Barron 1997; Otto-Bliesner and Upchurch 1997), and
vegetation feedbacks should therefore be expected
to have played a major role in enhancing the magnitude of Permian warming. It is probable that incorporation of vegetation feedbacks could reduce
the amount of atmospheric CO2 required to ensure
warm, ice-free polar conditions. Here, we have used
a uniform, mixed tree and shrub vegetation for all
land grid points to allow us to evaluate the direct
response to changing atmospheric CO2 (cf. Fawcett
and Barron 1998; Rees et al. 2000). As well as conducting interactive vegetation experiments (e.g.,
Foley et al. 1996), one of the next data-constrained
27
modeling steps will clearly be to conduct prescribed
vegetation experiments based upon more accurate
ecological interpretations of the paleobotanical
data.
Conclusions
The data/model comparisons reported in this article and its companion (Gibbs et al. 2002) are the
most thorough yet attempted for a pre-Pleistocene
interval, but much work remains to be done at
every level. We have pointed out a number of mismatches, large and small, in these comparisons, and
since we ourselves have prepared the maps, assembled the floral and sediment data, and run the climate models, we are aware of the uncertainties
throughout the analysis. We could have made adjustments to the paleogeography or model input
that would have better satisfied the data, but this
would have introduced obvious circularity in the
work. Accordingly, we used our published base
maps and employed conservative poleward heat
transport profiles as a test of the system. Overall,
we are satisfied with the general match between
the model and geological data, and the map patterns
seem to be correct even where the absolute values
of the meteorological parameters vary from the
expected.
The paleogeographic base maps are always subject to refinement. The orientation of Pangea is
well understood, but the positions of the smaller
Tethyan elements, like Mongolia, are still uncertain. More important for climate modeling studies
is the elevation of mountain ranges, and this is particularly difficult to establish geologically. In the
case of midlatitude desert deposits, it may be
tempting to postulate an adjacent mountain range
high enough to form a rain shadow, but this comes
dangerously close to circular reasoning. Conversely, the overestimate of the height or the width
of a mountain range in high latitudes may lead to
unwarranted snow production in the model results.
We feel that more attention should be devoted to
reconstructing paleotopography and that modeling
studies of pre-Pangean time intervals are premature
because of large uncertainties in the continental
orientations.
The floral and sedimentary information we have
assembled for the Permian is quite comprehensive
geographically and provides meaningful constraints
on precipitation, temperature, and wind directions
that can be compared directly with the model results. Further refinements in interpretation can
doubtless be made, and future detailed taphonomic
and taxonomic studies can be placed within our
28
P. M . R E E S E T A L .
framework of global phytogeography and climate.
However, the chief problem is in dating many of
these Permian deposits. The Permian time scale is
currently under revision, and continental deposits
are difficult to correlate. At best, we can hope to
document the slow climate transitions that result
from the latitudinal shifts of continents through
long intervals of time.
The atmospheric model will perform appropriately only if the input parameters are correctly
specified. Here, we see two general problems in our
current work. The geographic grid on which the
model is based is coarse, and this has had the effect
of reducing the elevation of narrow mountain
ranges that might have influenced circulation patterns. More important, although our model does
include an upper ocean that interacts with the atmosphere, we have found that the ocean heat transport calculated by the model is very small, and this
feature may be responsible in large part for producing polar climates that are too cold. Moreover,
the model calculates ocean heat transport as a dif-
fusion term rather than explicitly allowing for
warm currents to move heat into particular regions.
Our future work will incorporate a coupled oceanatmosphere model to approach this problem more
directly. Indeed, we are in the process of incorporating the effects of vegetation feedbacks and coupled ocean-atmosphere circulation in the next generation of higher-resolution climate models. This,
combined with our more rigorous and standardized
approach to compilation and interpretation of the
proxy data, will ultimately lead to a more complete
and accurate understanding of climates and climate
change throughout geologic time.
ACKNOWLEDGMENTS
This work was supported by National Science
Foundation grants ATM 96-32160, EAR 96-32286,
and ATM 00-00545. We thank Bob Gastaldo, Bill
DiMichele, and two others for their helpful
reviews.
REFERENCES CITED
Barron, E. J., and Fawcett, P. J. 1995. The climate of Pangea: a review of climate model simulations of the
Permian. In Scholle, P. A.; Peryt, T. M.; and UlmerScholle, D. S., eds. The Permian of northern Pangea.
Vol. 1. Paleogeography, paleoclimates, stratigraphy.
New York, Springer, p. 37–52.
Barron, E. J.; Fawcett, P. J.; Peterson, W. H.; Pollard, D.;
and Thompson, S. L. 1995. A “simulation” of midCretaceous climate. Paleoceanography 10:953–962.
Berner, R. A. 1994. GEOCARB II: a revised model of atmospheric CO2 over Phanerozoic time. Am. J. Sci. 294:
56–91.
Broutin, J.; Doubinger, J.; Farjanel, G.; Freytet, P.; Kerp,
H.; Langiaux, J.; Lebreton, M. L.; Sebban, S.; and Satta,
S. 1990. Le renouvellement des flores au passage Carbonifère Permien: approches stratigraphique, biologique, sédimentologique. C. R. Acad. Sci. Paris 311:
1563–1569.
Chaloner, W. G., and Creber, G. T. 1988. Fossil plants as
indicators of Late Palaeozoic plate positions. In
Audley-Charles, M. G., and Hallam, A., eds. Gondwana and Tethys. Geol. Soc. Lond. Spec. Publ. 37:
201–210.
Chaloner, W. G., and Meyen, S. V. 1973. Carboniferous
and Permian floras of the northern continents. In Hallam, A., ed. Atlas of palaeobiogeography. Amsterdam,
Elsevier, p. 169–186.
Chandra, S., and Chandra, A. 1987. Vegetational changes
and their climatic implications in coal-bearing Gondwana. Palaeobotanist 36:74–86.
Crowley, T. J. 1994. Pangean climates. In Klein, G. D.,
ed. Pangea: paleoclimate, tectonics, and sedimentation during accretion, zenith, and breakup of a supercontinent. Geol. Soc. Am. Spec. Pap. 288:25–39.
Crowley, T. J., and Baum, S. K. 1991. Toward reconciliation of Late Ordovician (∼440 Ma) glaciation
with very high CO2 levels. J. Geophys. Res. 96:
22,597–22,610.
———. 1994. General circulation model study of Late
Carboniferous interglacial climates. Palaeoclimates 1:
3–21.
Crowley, T. J., and North, G. R. 1996. Paleoclimatology
(2d ed.). Oxford, Oxford University Press, 339 p.
Crowley, T. J.; Yip, K.-J.; Baum, S. K.; and Moore, S. B.
1996. Modelling carboniferous coal formation. Palaeoclimates 2:159–177.
Cuneo, N. R. 1996. Permian phytogeography in Gondwana. Palaeogeogr. Palaeoclimatol. Palaeoecol. 125:
75–104.
Cuneo, N. R.; Isbell, J.; Taylor, E. L.; and Taylor, T. N.
1993. The Glossopteris flora from Antarctica: taphonomy and paleoecology. In Proceedings of the 12th International Congress on Carboniferous/Permian Stratigraphy and Geology (Buenos Aires, 1991) (vol. 2).
Nanjing, Nanjing University Press, p. 13–40.
Durante, M. V. 1995. Reconstruction of Late Paleozoic
climatic changes in Angaraland according to phytogeographic data. Stratigr. Geol. Correlation 3:123–133.
Dutton, J. F., and Barron, E. J. 1997. Miocene to present
vegetation changes: a possible piece of the Cenozoic
cooling puzzle. Geology 25:39–41.
Falcon, R. M. S. 1986. A brief review of the origin, for-
Journal of Geology
P E R M I A N P H Y T O G E O G R A P H I C PAT T E R N S
mation, and distribution of coal in southern Africa.
In Anhaeusser, C. R., and Maske, S., eds. Mineral deposits of southern Africa. Johannesburg Geol. Soc. S.
Afr., p. 1879–1898.
Fawcett, P. J., and Barron, E. J. 1998. The role of geography
and atmospheric CO2 in long term climate change:
results from model simulations for the Late Permian
to present. In Crowley, T. J., and Burke, K. C., eds.
Tectonic boundary conditions for climate reconstructions. Oxford, Oxford University Press, p. 21–36.
Foley, J. A.; Prentice, I. C.; Ramankutty, N.; Levis, S.;
Pollard, D.; Sitch, S.; and Haxeltine, A. 1996. An integrated biosphere model of land surface processes,
terrestrial carbon balance, and vegetation dynamics.
Global Biogeochem. Cycles 10:603–628.
Gastaldo, R. A.; DiMichele, W. A.; and Pfefferkorn, H.
W. 1996. Out of the icehouse into the greenhouse: a
Late Paleozoic analog for modern global vegetational
change. GSA (Geol. Soc. Am.) Today 20:1–7.
Gauch, H. G., Jr. 1982. Multivariate analysis in community ecology. Cambridge, Cambridge University
Press, 298 p.
Gibbs, M. T.; Rees, P. M.; Kutzbach, J. E.; Ziegler, A. M.;
Behling, P. J.; and Rowley, D. B. 2002. Simulations of
Permian climate and comparisons with climatesensitive sediments. J. Geol. 110:33–55.
Glennie, K. W. 1984. Early Permian-Rotliegend. In Glennie, K. W., ed. Introduction to the petroleum geology
of the North Sea. Oxford, Blackwell, p. 41–60.
Gould, R. E., and Delevoryas, T. 1977. The biology of
Glossopteris: evidence from petrified seed-bearing and
pollen-bearing organs. Alcheringa 1:387–399.
Guerra-Sommer, M.; Cazzulo-Klepzig, M.; and MarquesToigo, M. 1995. Palaeoclimatic implications of Lycophyta in the Gondwana of southern Brazil. IG,
UFRGS, Porto Alegre, Pesquisas 22:21–31.
Hill, M. O. 1979a. Correspondence analysis: a neglected
multivariate method. Appl. Stat. 23:340–354.
———. 1979b. DECORANA: a FORTRAN program for
detrended correspondence analysis and reciprocal averaging: ecology and systematics. Ithaca, N.Y., Cornell
University.
Hills, J. M., and Kottlowski, F. E. 1983. Correlation of
stratigraphic units of North America (COSUNA) project: southwest/southwest mid-continent region [and
other charts in the series]. Tulsa, Okla., American Association of Petroleum Geologists.
Huber, B. T.; MacLeod, K. G.; and Wing, S. L. 2000. Warm
climates in earth history. Cambridge, Cambridge University Press, 462 p.
Jin, Y.-G.; Glenister, B. F.; Kotlyar, G. V.; and Sheng, J.Z. 1994. An operational scheme of Permian chronostratigraphy. In Jin, Y.-G.; Utting, J.; and Wardlaw, B.
R., eds. Permian stratigraphy, environments and resources. Vol. 1. Palaeontology and stratigraphy. Palaeoworld Spec. Issue 4:1–13.
Jin, Y.-G.; Wardlaw, B. R.; Glenister, B. G.; and Kotlyar,
G. V. 1997. Permian chronostratigraphic subdivisions.
Episodes 20:10–15.
Kerp, J. H. F. 1982. Aspects of Permian palaeobotany and
29
palynology. II. On the presence of the ovuliferous organ Autunia milleryensis (Renault) Krasser (Peltaspermaceae) in the lower Permian of the Nahe area
(FRG) and its relationship to Callipteris conferta
(Sternberg) Brongniart. Acta Bot. Neerl. 31:417–427.
Kreuser, T.; Wopfner, H.; Kaaya, C. Z.; Markwort, S.;
Semkiwa, P. M.; and Aslandis, P. 1990. Depositional
evolution of Permo-Triassic Karoo basins in Tanzania
with reference to their economic potential. J. S. Afr.
Earth Sci. 10:151–167.
Kutzbach, J. E., and Gallimore, R. G. 1989. Pangaean climates: megamonsoons of the megacontinent. J. Geophys. Res. 94:3341–3357.
Kutzbach, J. E.; Guetter, P. J.; and Washington, W. M.
1990. Simulated circulation of an idealized ocean for
Pangean time. Paleoceanography 5:299–317.
Kutzbach, J. E., and Ziegler, A. M. 1993. Simulation of
Late Permian climate and biomes with an atmosphere/ocean model: comparisons with observations.
Philos. Trans. R. Soc. Lond. Ser. B 341:327–340.
Langford, R. P. 1992. Permian coal and palaeogeography
of Gondwana. Bureau of Mineral Resources (Australia), Record 1991/95, Palaeogeography 39, 136 p.
Li, X., and Wu, X. 1996. Late Paleozoic phytogeographic
provinces in China and its adjacent regions. Rev. Palaeobot. Palynol. 90:41–62.
Limarino, C. O., and Spalletti, L. A. 1986. Eolian Permian
deposits in west and northwest Argentina. Sediment.
Geol. 49:109–127.
Liu, G. 1990. Permo-Carboniferous paleogeography and
coal accumulation and their tectonic control in the
north and south China continental plates. Int. J. Coal
Geol. 16:73–117.
Liu, Z.; Gallimore, R. G.; Kutzbach, J. E.; Xu, W.; Golubev, Y.; Behling, P.; and Selin, R. 1999. Modeling
long-term climate changes with equilibrium asynchronous coupling. Clim. Dynamics 15:325–340.
Lottes, A. L., and Ziegler, A. M. 1994. World peat occurrence and the seasonality of climate and vegetation. Palaeogeogr. Palaeoclimatol. Palaeoecol. 106:
23–37.
Mapes, G., and Gastaldo, R. A., 1986. Late Paleozoic nonpeat accumulating floras. In Broadhead, T. W., ed.
Land plants: notes for a short course. University of
Tennessee, Department of Geological Sciences, Studies in Geology 15, p. 115–127.
Menning, M. 1995. A numerical time scale for the Permian and Triassic periods: an integrated time analysis.
In Scholle, P. A.; Peryt, T. M.; and Ulmer-Scholle, D.
S., eds. The Permian of northern Pangea. Vol. 1. Paleogeography, paleoclimates, stratigraphy. New York,
Springer, p. 77–97.
Meyen, S. V. 1976. Carboniferous and Permian lepidophytes of Angaraland. Palaeontogr. Abt. B Palaeophytol. 157:112–157.
———. 1982. The Carboniferous and Permian floras of
Angaraland (a synthesis). Biol. Memoirs 7:1–110.
———. 1987. Fundamentals of palaeobotany. London,
Chapman & Hall, 432 p.
Otto-Bliesner, B. L. 1993. Tropical mountains and coal
30
P. M . R E E S E T A L .
formation: a climate model study of the Westphalian
(306 Ma). Geophys. Res. Lett. 20:1947–1950.
Otto-Bliesner, B. L., and Upchurch, G. R., Jr. 1997. Vegetation-induced warming of high-latitude regions during the Late Cretaceous period. Nature 385:804–807.
Parrish, J. T. 1998. Interpreting pre-Quaternary climate
from the geologic record. New York, Columbia University Press, 338 p.
Plumstead, E. P. 1973. The Late Palaeozoic Glossopteris
flora. In Hallam, A., ed. Atlas of palaeobiogeography.
Amsterdam, Elsevier, p. 187–205.
Rees, P. M. 1993. Caytoniales in Early Jurassic floras from
Antarctica. Geobios 26:33–42.
Rees, P. M.; Gibbs, M. T.; Ziegler, A. M.; Kutzbach, J. E.;
and Behling, P. J. 1999. Permian climates: evaluating
model predictions using global paleobotanical data.
Geology 27:891–894.
Rees, P. M., and Ziegler, A. M. 1999. Palaeobotanical
databases and palaeoclimate signals. In Jones, T. P.,
and Rowe, N. P., eds. Fossil plants and spores: modern techniques. Geol. Soc. Lond. Spec. Publ., p.
240–244.
Rees, P. M.; Ziegler, A. M.; and Valdes, P. J. 2000. Jurassic
phytogeography and climates: new data and model
comparisons. In Huber, B. T.; MacLeod, K. G.; and
Wing, S. L., eds. Warm climates in earth history. Cambridge, Cambridge University Press, p. 297–318.
Schmidt, G. A., and Mysak, L. A. 1996. Can increased
northward ocean heat flux explain the warm Cretaceous climate? Paleoceanography 11:579–593.
Shi, G. R. 1993. Multivariate data analysis in paleoecology and paleobiogeography—a review. Palaeogeogr.
Palaeoclimatol. Palaeoecol. 105:199–234.
Sloan, L. C. 1994. Equable climates during the Early Eocene: significance of regional paleogeography for
North American climate. Geology 22:881–884.
Stewart, W. N. 1983. Paleobotany and the evolution of
plants. Cambridge, Cambridge University Press, 405
p.
Taylor, T. N., and Taylor, E. L. 1993. The biology and
evolution of fossil plants. Englewood Cliffs, N.J., Prentice-Hall, 982 p.
Ter Braak, C. J. F. 1992. CANOCO: a FORTRAN program
for canonical community ordination. Ithaca, N.Y., Microcomputer Power.
Thompson, S. L., and Pollard, D. 1997. Greenland and
Antarctic mass balances for present and doubled atmospheric CO2 from the GENESIS Version 2 Global
Climate Model. J. Clim. 10:871–900.
Tiwari, R. S., and Tripathi, A. 1987. Palynological zones
and their climatic inference in the coal-bearing Gondwana of peninsular India. Palaeobotanist 36:87–101.
Truswell, E. M. 1991. Antarctica: a history of terrestrial
vegetation. In Tingey, R. J., ed. The geology of Antarctica. Oxford, Clarendon Press, p. 499–537.
Utting, J., and Piasecki, S. 1995. Palynology of the Permian of northern continents: a review. In Scholle, P. A.;
Peryt, T. M.; and Ulmer-Scholle, D. S., eds. The Permian of northern Pangea. Vol. 1. Paleogeography, pa-
leoclimates, stratigraphy. New York, Springer, p.
236–261.
Vakhrameev, V. A.; Dobruskina, I. A.; Meyen, S. V.; and
Zaklinskaja, E. D. 1978. Paläozoische und mesozoische Floren Eurasiens und die Phytogeographie dieser
Zeit. Jena, VEB Gustav Fischer, 268 p.
Wagner, R. H. 1993. Climatic significance of the major
chronostratigraphic units of the upper Palaeozoic. In
Proceedings of the 12th International Congress on
Carboniferous/Permian Stratigraphy and Geology
(Buenos Aires, 1991) (vol. 1). Nanjing, Nanjing University Press, p. 83–108.
Walter, H. 1985. Vegetation of the earth and ecological
systems of the geo-biosphere (3d ed.). New York,
Springer, 318 p.
Wang, Z. 1993. Evolutionary ecosystem of PermianTriassic redbeds in north China: a historical record of
global desertification. In Lucas, S. G., and Morales,
M., eds. The nonmarine Triassic. New Mexico Museum of Natural History and Science Bulletin 3. Albuquerque, New Mexico Museum of Natural History
and Science, p. 471–476.
Wnuk, C. 1996. The development of floristic provinciality during the Middle and Late Paleozoic. Rev. Palaeobot. Palynol. 90:5–40.
Yao, Z.-Q. 1983. Ecology and taphonomy of gigantopterids. Bull. Nanjing Inst. Geol. Palaeontol. Acad. Sinica
6:63–84 (in Chinese with English summary).
Yemane, K. 1993. Contribution of Late Permian paleogeography in maintaining a temperate climate in
Gondwana. Nature 361:51–54.
Zhuravleva, I. T., and Ilina, V. I. 1988. The upper Paleozoic of the Angara Basin: fauna and flora. Novosibirsk,
Nauka, 265 p.
Ziegler, A. M. 1990. Phytogeographic patterns and continental configurations during the Permian period. In
McKerrow, W. S., and Scotese, C. R., eds. Palaeozoic
Palaeogeography and Biogeography. Geol. Soc. Lond.
Mem. 12:363–379.
———. 1993. Models come in from the cold. Nature 361:
16–17.
———. 1998. Warm polar currents. EOS: Trans. Am. Geophys. Union, Spring Meeting 98(suppl.):S50.
Ziegler, A. M.; Gibbs, M. T.; and Hulver, M. L. 1998. A
mini-atlas of oceanic water masses in the Permian
period. Proc. R. Soc. Vic. 110:323–343.
Ziegler, A. M.; Hulver, M. L.; and Rowley, D. B. 1997.
Permian world topography and climate. In Martini,
I. P., ed. Late glacial and postglacial environmental
changes: Quaternary, Carboniferous-Permian and
Proterozoic. Oxford, Oxford University Press, p.
111–146.
Ziegler, A. M.; Parrish, J. M.; Yao, J.; Gyllenhaal, E. D.;
Rowley, D. B.; Parrish, J. T.; Shangyou, N.; Bekker, A.;
and Hulver, M. L. 1993. Early Mesozoic phytogeography and climate. In Allen, J. R. L.; Hoskins, B. J.;
Sellwood, B. W.; Spicer, R. A.; and Valdes, P. J., eds.
Palaeoclimates and their modelling with special reference to the Mesozoic era. Philos. Trans. R. Soc.
Lond. Ser. B 341:203–243.
Journal of Geology
P E R M I A N P H Y T O G E O G R A P H I C PAT T E R N S
Ziegler, A. M.; Rees, P. M.; Rowley, D. B.; Bekker, A.; Li,
Q.; and Hulver, M. L. 1996. Mesozoic assembly of
Asia: constraints from fossil floras, tectonics and pa-
31
leomagnetism. In Yin, A., and Harrison, T. M., eds.
The tectonic evolution of Asia. Cambridge, Cambridge University Press, p. 371–400.