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:
[email protected])
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
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
Note. 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 Parana´ 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 Carbonife`re Permien: approches stratigraphique, biologique, se´dimentologique. 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. Pala¨ozoische 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.
