RESEARCH REPORT 129

Copyright Q 2004, SEPM (Society for Sedimentary Geology) 0883-1351/04/0019-0129/$3.00

Paleotemperature Estimation Using Leaf-Margin Analysis:Is Australia Different?

DAVID R. GREENWOODSustainability Group, Victoria University, St Albans Campus, PO Box 14428, Melbourne City MC, VIC 8001, Australia,

E-mail: David.Greenwood@vu.edu.au

PETER WILFDepartment of Geosciences, Pennsylvania State University, University Park, PA 16802

SCOTT L. WINGDepartment of Paleobiology, National Museum of Natural History, Smithsonian Institution, Washington, DC 20560

DAVID C. CHRISTOPHELDepartment of Biological Sciences, Room 102 Olin Building, 2190 E. Iliff Ave., University of Denver, Denver, CO 80208

PALAIOS, 2004, V. 19, p. 129–142

Australia has numerous fossil floras suitable for paleocli-matic analysis, with potential to improve understanding ofSouthern Hemisphere climatic evolution. Leaf-marginanalysis (LMA) is a widely used method that applies pre-sent-day correlations between the proportion of woody dicotspecies with untoothed leaves and mean annual tempera-ture to estimate paleotemperatures from fossil megafloras.Australia’s unique history and vegetation imply that its leaf-margin correlation might differ from other regions; thesepossible differences are investigated here to improve paleo-climatic interpretations.

Australian rainforest vegetation shows nearly the sameregression slope as recorded in East Asia and the Americas,indicating a globally convergent evolutionary response ofleaf form to temperature. However, Australian sites tend tohave fewer toothed species at localities with the same tem-perature as Asian and American sites. The following fac-tors, singly or in combination, may account for this differ-ence: (1) Australia’s Cenozoic movement into lower latitudes,insulation from global cooling, and isolation from high-lat-itude cold-tolerant vegetation sources; (2) lack of highmountains as sources and refuges for cold-adapted taxa; (3)Pleistocene extinctions of cold-adapted taxa; and (4) thenear absence of a cold-climate forest ecospace in Australiatoday.

Application of Australian LMA to Australian Cenozoicfloras resulted in cooler temperature estimates than otherLMA regressions. However, Australian paleotemperatureestimates should account for the relative importance of cold-deciduous taxa. The timing and magnitudes of the extinc-tions of cold-adapted lineages are not known, and the mostconservative approach is to use Australian LMA as a mini-mum and non-Australian LMA as a maximum tempera-ture estimate.

INTRODUCTION

There is considerable interest in quantitative estimatesof terrestrial paleoclimates, including paleotemperatures.

Apart from the value of placing evolutionary events andgeological processes into a climatic framework (e.g., Up-church and Wolfe, 1987; Gregory and Chase, 1992; Huntand Poole, 2003), paleoclimate estimates offer tests of sen-sitivity studies using computer climate modeling (e.g.,Sewall et al., 2000; Shellito et al., 2003). Australia, the fo-cus of this study, lay at high Southern latitudes during theearly Paleogene (Veevers et al., 1991) and is a valuabletesting ground for climate modeling studies, which aremost sensitive to forcing factors at high latitudes (Petersand Sloan, 2000; Shellito et al., 2003). Paleontologicaldatalong have indicated that the middle Eocene and late earlyMiocene of southeastern Australia were much warmerthan today (e.g., Christophel, 1981; Greenwood, 1994;Kershaw et al., 1994; Macphail et al., 1994). Quantitativeestimates derived from proxies for Australian paleocli-mate are needed to improve understanding of these un-usual time intervals, which are better known on other con-tinents.

Terrestrial paleoclimates can be estimated from fossilfloras using correlations between leaf physiognomic (sizeand shape) attributes and climate variables in living for-ests. A number of variations on the leaf-physiognomic ap-proach have been demonstrated (e.g., Wolfe, 1993; Wingand Greenwood, 1993; Greenwood, 1994; Gregory and Mc-Intosh, 1996; Jacobs, 2002), but a method known as leaf-margin analysis (LMA) was the first to be quantified as anestimate of mean annual temperature (MAT) and remainsthe most widely used (e.g., Wolfe, 1971, 1978, 1979; Green-wood and Wing, 1995; Wing et al., 2000; Greenwood et al.,2003; Hunt and Poole, 2003). Leaf-margin analysis isbased on the strong positive relationship between MATand the proportion of woody dicot species in a floral sam-ple that has entire (untoothed) leaf margins (leaf-marginproportion, LMP). This trend was observed first amongfloras worldwide from areas without severe cold or mois-ture limitations (Bailey and Sinnott, 1915, 1916); it wasquantified later, as a linear regression, for East Asian me-sic vegetation (Wolfe, 1971, 1979; Wing and Greenwood,1993). A leaf-margin analysis consists of inverting the re-gression for application to fossil floras, where the propor-

130 GREENWOOD ET AL.

tion of untoothed species, the independent variable, isknown, and MAT, the dependent variable, is unknown.

The adaptive value of leaf teeth, or their absence, is notwell understood. Givnish (1979) argued that because ev-ergreen leaves tend to be thicker than deciduous leaves,and flow resistance decreases as leaves become thicker,more growth will occur in the intercostal area, smoothingleaf margins. Other explanations, to date, include differ-ential hydrodynamic expansion near the major veins ofrapidly expanding, deciduous leaves with low vein density(Mosbrugger and Roth, 1996); teeth as sites of spring pho-tosynthesis in young deciduous leaves (Baker-Brosh andPeet, 1997); and enhanced transpiration at leaf margins(Canny, 1990), which may compensate for reduced sapflow in cool environments (Wilf, 1997).

Margin type is not the only leaf character that is climat-ically informative, so Wolfe (1993) developed a multivari-ate database and analytical procedure known as CLAMP(climate leaf analysis multivariate program). This tech-nique incorporated a broader geographical coverage thanthe original LMA, used additional leaf characters, andconsidered a number of climate variables in a correspon-dence analysis. In addition, a number of derivative meth-ods have been proposed based on the CLAMP dataset (e.g.,Gregory and Chase, 1992; Wing and Greenwood, 1993;Gregory-Wodzicki, 2000). However, leaf-margin type isthe most significant character in every multivariate modelused to estimate temperature (Wolfe, 1993; Wing andGreenwood, 1993; Wilf, 1997; Gregory-Wodzicki, 2000).Thus, understanding regional or hemispherical variationin the LMA regression is fundamental to paleotempera-ture analysis, whether multivariate or univariate meth-ods are used. In addition, because the reproducibility ofleaf-margin analysis, based on a single unambiguouscharacter, mitigates observational errors in scoring mul-tiple characters (Wilf, 1997), tests on living floras havegenerated MAT estimates from LMA that are generally asgood as or better than multivariate approaches (Burnham,1997; Wilf, 1997; Wiemann et al., 1998; Gregory-Wodzicki,2000).

Accurate paleoclimate estimates may be derived bestusing modern calibration data that match, as closely aspossible, the climatic response of leaf physiognomy in thefossil floras studied (Wolfe, 1993; Wing and Greenwood,1993; Jordan, 1997; Greenwood and Wing, 1995; Stranksand England, 1997; Gregory-Wodzicki, 2000; Jacobs,2002). Regional climate history, past and present edaphicconditions, and regional peculiarities of plant phylogenetichistory may all shift leaf-climate relationships. For exam-ple, extreme cold and dryness, as found in subalpine sites,or poor soils may force relatively high leaf-margin propor-tions (Bailey and Sinnott, 1915, 1916; Webb, 1968), andthe foliage of some plant lineages is obligately toothed oruntoothed (Wolfe, 1993; Wing and Greenwood, 1993;Gregory and McIntosh, 1996). Examples of possible histor-ical and phylogenetic effects include the predominance offamilies in the Australian flora that characteristicallybear untoothed leaves, such as Myrtaceae, and the strongrepresentation in North American floras of deciduous treespecies with toothed leaves.

With regard to regional effects, of particular interesthere is that separate LMA regressions initially were pre-sented for the Northern and Southern hemispheres

(Wolfe, 1979; Upchurch and Wolfe, 1987). The SouthernHemisphere scale was described as a 4% increase in en-tire-margined species for an increase of 18C of MAT, com-pared to 3% for the Northern Hemisphere (Wolfe, 1979;Upchurch and Wolfe, 1987). This difference was attribut-ed to ‘‘. . . the almost total absence of deciduous plants inthe southern hemisphere . . . ’’ and the anecdotally recog-nized, but unquantified correlation of toothed marginsand deciduousness (Upchurch and Wolfe, 1987, p. 35). Un-fortunately, no data were ever presented to support theSouthern Hemisphere scale; consequently, the relativerepresentations of deciduous taxa in various regions of theSouthern Hemisphere remain unquantified.

Subsequent research has demonstrated substantiallythe same relationship between leaf-margin proportion andMAT in East Asia and the Americas, including SouthAmerica (Wilf, 1997; Wiemann et al., 1998; Gregory-Wod-zicki, 2000; Burnham et al., 2001; Kowalski, 2002). Onedata set of 30 sites from tropical South America showed,for a given temperature, leaf-margin proportion some-what above (corresponding to about 138C), but in generalagreement with, other data sets (Kowalski, 2002). The ex-ception was a group of cold sites from high elevations, forwhich toothed species were comparatively rare. As statedby Kowalski (2000), this appears to result from the unusu-al selective environment of Neotropical cloud forests,which are associated typically with thick, small, untooth-ed leaves (e.g., Leigh, 1999; Velazquez-Rosas et al., 2002).In addition, Kowalski’s (2002) results must be regarded aspreliminary because 17% of the species in her data setwere not scored for margin state due to unavailability ofherbarium specimens.

Preliminary studies of Australian mesic vegetation pro-duced a poor correlation between LMP and MAT, with asignificantly different regression slope and intercept fromEast Asia (Upchurch and Wolfe, 1987; Greenwood andChristophel, 1990; Greenwood, 1992, 2001). However, theanalyses of Australian LMA included a majority of siteswhere the margin type either was not known for some spe-cies (Greenwood and Christophel, 1990; Greenwood,2001), or was based on forest-floor litter collections from asmall number of sites (Greenwood, 1992). Jordan (1997)found that both LMA and multiple-regression models con-sistently over-estimated MAT for modern southeasternAustralian and New Zealand vegetation, and Kennedy(1998) found no relationship between leaf-margin propor-tion and MAT for New Zealand vegetation. However, NewZealand is a small and isolated landmass with an uniquebiogeographic history, and its forest cover exists within alimited range of MAT (58–158C). The New Zealand florasuffered major extinctions of plant lineages during theNeogene and Pleistocene (Lee et al., 2001). It is perhapsnot surprising that the relationship of leaf physiognomywith climate in modern New Zealand might be differentfrom the continental floras of Australia and the Americas.

This paper addresses the issue of regional effects on leafphysiognomy by re-examining the relationship of leaf-margin proportion to temperature for Australian vegeta-tion. The possible influence of soil type also is consideredbecause of the prevalence of nutrient-poor soils in manyAustralian ecosystems (Beadle, 1966). Based on a revisedcorrelation for Australia, new temperature estimates for aseries of Australian Paleogene floras are presented and

AUSTRALIAN LEAF-MARGIN ANALYSIS 131

FIGURE 1—Locations of sites for which floral list data (Webb et al.,1984; Appendix 1) were compiled, superimposed on a map showingthe limits of tropical rainforest and deciduous tropical woodlands inAustralia, and Australian fossil sites discussed in the text. The $ 400mm isohyet is indicated, demarking the limit of tree-dominated vege-tation.

compared to results derived from the same floras usingother calibrations. Factors are highlighted that might con-tribute to observed deviations in the Australian correla-tion, and general recommendations are presented for pa-leoclimatic estimation from Australian fossil floras.

MATERIALS AND METHODS

Extant Vegetation Database

In marked contrast to North America and Asia, the Aus-tralian flora is depauperate in deciduous trees and shrubs.The primary Australian woody vegetation is sclerophyl-lous and evergreen, and the dominant tree genera are Eu-calyptus, sensu lato (Myrtaceae), and Acacia (Mimosa-ceae), all of which have entire leaf margins (Crisp et al.,1999). Along the east coast, there are also areas of mesicforest lacking (or at least not dominated by) Eucalyptusthat often are termed rainforest in the Australian litera-ture, but are termed vine forests under Webb’s structural-physiognomic classification (Webb, 1959, 1968; Webb etal., 1984; Greenwood, 1996; Appendix 1). The Australianusage of rainforest is much broader than elsewhere, as itencompasses all forests where; (1) markedly sclerophyl-lous trees and shrubs (e.g., Eucalyptus) are either absentor uncommon; (2) the canopy is closed at least during thewettest part of the year; (3) life forms, or synusiae, char-acteristic of true rain forest are present; and (4) the domi-nant species are either species characteristic of true Aus-tralian rain forest (Webb, 1959, 1968; Greenwood, 1996),or are closely related to them. The greatest diversity of de-ciduous tree species in Australia is not found in temperateforests, but rather is found in the seasonally dry tropicalvine forest and vine thickets (Webb, 1959, 1968; Appendix1). Webb et al. (1984) produced a comprehensive survey ofsites throughout the climatic range of rainforest (or vineforest) within Australia. These forests are dominated bytrees and shrubs with broad laminae (length:width , 4.0),whereas Eucalyptus species are markedly sclerophyllousand typically stenophyllous (length:width k 4.0). Struc-turally, the rainforests differ also from the sclerophyllousforests by having closed canopies (i.e., .70% projected fo-liage cover), a significant presence of woody vines, andother synusiae characteristic of true rain forests (Webb,1959, 1968; Greenwood, 1996; Richards, 1996). The Webbet al. (1984) forest surveys varied in size from about 0.25hectare to several hectares, but the authors reported anintention to provide a representative floristic inventory ofthe vegetation type in each locale. Some data were basedon the mapping of all woody plants with a stem diameterat breast height . 15 cm, whereas others were surveyedusing transects and included woody vines.

Most plant families are found both in the Australiansclerophyllous forests and rainforests (e.g., Myrtaceae),and many genera have either species or closely relatedgenera in both biomes (Webb et al., 1984; Crisp et al.,1999). Nonetheless, the physiognomic-structural transi-tion between sclerophyllous forests and rainforest ismarked, and serves to highlight discrete ecophysiologicalstrategies (Webb, 1959, 1968). The tropical and subtropi-cal rainforests are often highly species rich and diverse(Appendix 1), and typically lack dominants in their cano-py. The seasonally dry tropical deciduous vine-forest-

thicket sites in the Webb et al. (1984) database contain sig-nificant numbers of obligately or facultatively deciduousspecies (e.g., Brachychiton spp. and Ficus virens). A smallnumber of sites from southeastern Australia containemergent Eucalyptus spp. and/or Acacia spp. in the cano-py; these sites are likely to be successional in response toforest fires. The cool-temperate rainforest sites, fromsoutheastern Australia and Tasmania, typically have lowwoody-dicot diversity (,10 spp./ hectare) and are domi-nated by toothed broadleaved evergreen species, such asNothofagus cunninghamii, N. moorei (Nothofagaceae),and Atherosperma moschatum (Atherospermataceae).

Australia’s rainforests floristically and physiognomical-ly reflect the predominant vegetation types of much of theAustralian continent for much of the Cenozoic (Christo-phel and Greenwood, 1989; Kershaw et al., 1994; Green-wood et al., 2003). In contrast, the sclerophyllous forestswere probably occupied a minor part of the landscape untilperhaps the late Neogene, when markedly seasonal rain-fall regimes became dominant (e.g., Gallagher et al.,2003). The sclerophyllous forests are adapted to low soilphosphorous and frequent wildfires, as well as seasonaldrought or low annual rainfall (Beadle, 1966; Hill, 1998).For these reasons, the Webb et al. (1984) database of rain-forest sites was selected here to assess LMA for Australia.

Sampling and Measurement

Data presented here are culled from floral lists by Webbet al. (1984) for over 600 rainforest sites. These are pre-dominantly located towards the relatively wet east andnorthern coasts as well as the southeast of Australia; ap-proximately two thirds of the continent is arid and lacksforest cover (Fig. 1). The subset of 113 sites that is usedhere have complete locality data and for which the leaf-margin types of all species could be determined (Fig. 1).This approach improves upon earlier analyses that usedall sites (Greenwood and Christophel, 1990; Greenwood,2001). Here, only woody dicot species from these floral

132 GREENWOOD ET AL.

lists were scored for leaf-margin type, consistent withstandard practice for paleoclimatic calibration data (e.g.,Wolfe, 1993). Climate values for the sites were extrapolat-ed using the BIOCLIM module (Busby, 1991) within theANUCLIM version 5.0 software (Houlder et al., 1999),based on geographic data (Appendices 1 and 2). BIOCLIMuses a mathematical climate surface of present-day Aus-tralian climate, based on standard meteorological decadalmeans, and a digital elevation model.

The 113 sites closely approximate the complete climaticrange of rainforest environments in Australia. The sites’range of MAT is 3.38C to 24.98C, mean annual range oftemperature (MART) was 3.58C to 12.58C, and their rangeof mean annual precipitation (MAP) is 691 to 3686 mm/yr(Appendix 1). The MART throughout Australia today islow compared to most of North America (Fig. 2). The ma-jority of sites have a summer peak of rainfall; however,some temperate sites have a winter precipitation peakthat includes occasional snowfalls.

Leaf-margin type (toothed, untoothed, or both) for allspecies was obtained from published taxonomic descrip-tions in the Flora of Australia series (see Crisp et al., 1999)or from Hyland et al. (1999). A tooth was defined as a vas-cularized extension of the leaf margin with a correspond-ing sinus incised less than one quarter the distance to themidvein (Wilf, 1997; Ash et al., 1999). Facultativelytoothed species were scored as half-toothed (score of 0.5;Wolfe, 1993). Because species richness influences the de-gree of error in estimates based on LMA (Upchurch andWolfe, 1987; Wilf, 1997; Burnham et al., 2001), the data-base was subdivided into 2 subsets for analysis: all 113sites, and 74 sites, each of which contains at least 20 spe-cies of woody dicots. In Australia, mesic forests at lowMAT (, 108C) are species poor, typically with fewer thanten species of woody dicots per site (Appendix 1). Conse-quently, the removal of depauperate sites also deleted thecoldest sites from analysis, changing the minimum MATfrom 3.3 to 10.88C. However, a broad range of temperatureremained, from 10.8 to 24.98C, and MAP was in a compa-rable range (717 to 3193 mm/yr) to the full data set.

Webb et al. (1984) included limited data on soil charac-teristics for the majority of sites (Appendix 1). The siteswere classified into one of nine soil categories, based pri-marily on the parent rock (e.g., basalt, acid volcanic, basicvolcanic, limestone, sandstone, granite, or metamorphic),if known, and the manner of pedogenesis (e.g., ex situ soils,such as alluvium, versus in situ soils). To analyze thesedata for possible correlations of soil type with leaf-marginproportion, soils were grouped into fertile and infertile cat-egories, according to the definitions of Webb (1968). Thefertile category includes eutrophic to mesotrophic soilswith high-to-medium mineral-nutrient status (68 sites),and the infertile category includes oligotrophic soils defi-cient in some minerals important for plant growth, partic-ularly phosphorus (34 sites).

RESULTS

Leaf Margins and MAT

Analysis using least squares linear regression showsthat leaf-margin proportion (LMP) is correlated signifi-

cantly with MAT using all sites. The correlation also is sig-nificant for the subset of sites with at least 20 species:

(1) All 113 sites

MAT 5 22.0·LMP 1 1.32, s 5 13.08C2 221(r 5 57.3%, F 5 147, p , 10 )

(2) 74 sites $20 species

MAT 5 27.0·LMP 2 2.12, s 5 62.28C2 216(r 5 63.0%, F 5 122, p , 10 )

The standard errors for equations (1) and (2) are similar tothose derived from equivalent databases from other geo-graphical areas; for example, the error from LMA based onthe CLAMP database is 6 3.48C, and with the coldest sitesremoved it is 6 2.18C (Wilf, 1997).

The regressions for the 74 Australian sites with $ 20species each (used in equation 2) versus the sites in theEast Asian data set (Wolfe, 1979) showed a statisticallysignificant difference in slope (p K 0.001, slope equalitytest of Sokal and Rohlf, 1995, p. 495). This was the mostsevere test because the difference in slope is greater forEast Asia than for the other datasets (Fig. 3). In practice,this is not critical because the difference in slope for EastAsia only amounts to 0.78 C of difference in temperatureincrease per 20% of LMP, less than half the amount origi-nally suggested (Wolfe, 1979; Upchurch and Wolfe, 1987),and this value is negative, and less in absolute value, forthe other data sets besides East Asia. Thus, the responseof leaf-margin proportion to temperature is similar in Aus-tralia to elsewhere, as seen in the slope of the regression,but Australian vegetation is depleted in toothed species atall temperatures, as seen in the intercept (Fig. 3). Regard-ing the observation that the relationship between leaf-margin proportion and MAT in the Southern Hemisphereis different from the Northern Hemisphere (Upchurch andWolfe, 1987), there is a major difference in the interceptbut a negligible difference in the slope (Fig. 3). For SouthAmerica, the only other large area of the Southern Hemi-sphere that has been examined, no significant differencehas emerged (e.g., Gregory-Wodzicki, 2000).

Leaf Margins and Other Climate Variables

Results from univariate regression analyses of selectedclimatic variables (Appendix 2) on leaf-margin proportionare shown in Figures 4 and 5. Matching the results fromother regions (e.g., Wing and Greenwood, 1993; Wilf,1997), the strongest influences on LMP were temperaturevariables (Fig. 4), and no variable had a higher correlationcoefficient than MAT. All of the temperature variableswere correlated strongly with MAT, and when MAT isheld constant, the highest coefficient of partial correlationwith LMP was 20.24, for maximum temperature of thewarmest quarter. The highest-scoring variables all relateto the thermal load experienced by plants, such as season-al temperature extremes. The majority of sites were mesic(MAP .800 mm/yr); however, seasonal variation (as ei-ther precipitation seasonality or totals for driest, wettest,coldest, and warmest quarters; Fig. 5) was significant be-tween sites. Nonetheless, correlations of LMP with precip-itation variables were not significant (p . 0.05).

AUSTRALIAN LEAF-MARGIN ANALYSIS 133

FIGURE 2—Bivariate plot of mean annual range of temperature (MART) versus mean annual temperature (MAT) for all sites in the AustralianLMA calibration data set and sites from the CLAMP database for comparison (data from Wolfe, 1993, 1995). MART5the range from coldestmonth mean temperature (CMM) and warmest month mean temperature (WMM). Several North American and Australian cities are plotted forreference (from data compilation of Greenwood and Wing, 1995, and sources cited therein).

FIGURE 3—Bivariate plot of mean annual temperature versus pro-portion of woody dicot species with untoothed leaf margins for theAustralian data set of this study (subset with $20 species per site)and other databases. Other data from Wolfe (1979, 1995) and Greg-ory-Wodzicki (2000). LMA East Asia (Wolfe, 1979; Wing and Green-wood, 1993) slope 5 30.6; LMA derived from CLAMP ‘coldest sitesremoved’ (Wilf, 1997, p. 380) slope 5 24.4; LMA from Bolivia (Greg-ory-Wodzicki, 2000) slope 5 24.9.

Influence of Soil Type

The regression of LMP against MAT, contrasting fertileand infertile soil sites with at least 20 species each (Fig.4A), showed that the significance of the correlation washigher for fertile sites (n 5 46) relative to that calculatedusing all sites (n 5 23; i.e., r2 fertile sites 5 73% [p ,0.001], r2 infertile sites 5 61% [p , 0.001] versus r2 allsites $ 20 spp. 5 63% [p , 10216]). On visual inspection,neither the slope nor the intercept of the regression linefor solely infertile or solely fertile sites are shifted mark-edly relative to those for all sites with at least 20 species

(Fig. 4). This result suggests that the primary effect of dif-ferences in soil fertility is to increase the variance in LMPbetween sites that otherwise have similar environmentalcharacteristics. However, based on ANCOVA, the infertilesoils had significantly higher LMP than fertile soils (F1,66

5 5.55, p 5 0.021, MAT covariate), indicating that infertilesoils generally support vegetation with a higher propor-tion of untoothed species than do fertile soils with thesame MAT.

AUSTRALIAN PALEOGENE TEMPERATURES

A series of paleotemperature estimates is presented inTable 1 for ten Paleocene through Miocene floras, eachcontaining at least 20 species, from southeastern andsouthwestern Australia. Hunt and Poole’s (2003) estimatefor the middle Eocene Dragon Glacier flora from KingGeorge Island, off the Antarctic Peninsula, also is includ-ed; this flora is dominated reportedly by deciduous dicotsand contains many of the taxa typical of Australian Paleo-gene megafloras. Results using Australian LMA are com-pared to previous multiple regression analyses (Green-wood and Wing, 1995), and to results from East Asian(Wing and Greenwood, 1993) and Bolivian (Gregory-Wod-zicki, 2000) calibrations.

All of the Cenozoic MAT estimates are lower using Aus-tralian LMA (equation 2) than for either East Asian or Bo-livian LMA calibrations (Table 1). The greatest differencesare for the warmest estimates (e.g., Brandy Creek, GoldenGrove, Hotham Heights, and Nerriga), which differ bymore than 38C between Australian LMA (equation 2) andthe East Asian LMA (Wolfe, 1978; Wing and Greenwood,1993). In no case do the standard errors of the estimatesfrom East Asian and Australian LMA overlap, althoughoverlap occurs for the coolest floras between the Austra-

134 GREENWOOD ET AL.

FIGURE 4—Bivariate plots of selected temperature variables (Appendix 2) versus proportion of woody dicot species with untoothed leaf marginsfor the Australian data set of this study (subset with $20 species per site). (A) MAT versus LMP, Australian data subdivided into fertile soilsites V, and infertile soil sites D. (B) Temperature seasonality. (C) Minimum temperature of the coldest period. (D) Maximum temperature ofthe warmest quarter. (E) Mean temperature of the coldest quarter. (F) Mean temperature of the warmest quarter.

lian LMA and the Bolivian LMA. This result is consistentwith the differences among the intercepts and slopes of therespective regression equations (Fig. 3). All of the esti-mates derived from the East Asian LMA for the early andmiddle Eocene sites also analyzed by Greenwood andWing (1995) are consistently cooler using the AustralianLMA, but they are close to and overlap within the errorsfor the estimates using the 1995 multiple linear regressionequation.

The four revised Eocene estimates, although lower thancalculated by Greenwood and Wing (1995), all indicatemuch warmer conditions at their paleolatitudes than arefound at similar elevations and latitudes today (Green-wood et al., 2003). In addition, the revised Eocene esti-mates remove the appearance of a warmer Southern thanNorthern Hemisphere during the Eocene that was report-ed earlier (Greenwood and Wing, 1995).

The overall temporal pattern of the revised MAT esti-

AUSTRALIAN LEAF-MARGIN ANALYSIS 135

FIGURE 5—Bivariate plots of selected precipitation variables (Appendix 2) versus proportion of woody dicot species with untoothed leaf marginsfor the Australia (this study; $20 species per site subset). (A) mean annual precipitation. (B) Precipitation seasonality. (C) Precipitation of thedriest quarter. (D) precipitation of the wettest quarter. (E) Precipitation of the coldest quarter. (F) Precipitation of the warmest quarter.

mates is consistent with other paleoclimatic proxy datathat indicate warm episodes during the early Eocene andearly Miocene, but cooler conditions during the early Oli-gocene. However, the cold MAT of 4.28C that is estimatedhere for the early Oligocene Cethana flora is problematicand underscores the importance of evaluating multipleproxies. Carpenter et al. (1994) reported a diverse florafrom Cethana, including cycads and araucarian conifers,that is consistent with an ecotone in the modern Austra-

lian flora between sclerophyllous wet forest and temper-ate rainforest and an MAT of 10–148C. These values lie atthe upper end of the estimate derived using the EastAsian LMA. The unexpectedly low estimated MAT for theOligocene Cethana flora possibly is due to the diversity oftoothed sclerophyllous taxa, such as Banksieaephyllumspp. (Proteaceae), a result that contradicts Jordan’s (1997)hypothesis that LMA for Australian sites will be confound-ed by the predominance of obligately entire-margined

136 GREENWOOD ET AL.

TABLE 1—Estimates of MAT for southern Australian and Antarctic Paleocene to Miocene macrofloras, contrasting Australian LMA, East AsianLMA, and Bolivian LMA. Details of all floras are given in Greenwood and Christophel (2003) and Hunt and Poole (2003). Notes: (1) multiplelinear regression estimates from Greenwood and Wing (1995); (2) using LMA equations: East Asia MAT 5 30.6·LMP 1 1.141 (Wolfe, 1979;Wing and Greenwood, 1993), Bolivia MAT 5 37.9·LMP 2 3.83 (Gregory-Wodzicki, 2000), Australia MAT 5 27.0·LMP 2 2.12 (equation 2); (3)a minimum error of the estimate calculated as p[LMA] 5 slope*Ï((LMP*(1 2 LMP))/r), from Wilf (1997); (4) LMP scored from curated collections(University of Adelaide, Victoria University and Melbourne University); (5) LMP scored from a publication (descriptions and/or illustrations ofnamed taxa); (6) estimates and data from Greenwood et al. (2003); (7) estimates and data from Hunt and Poole (2003).

r LMP MAT1MAT2

[E Asia]MAT2

[Bolivia]MAT2

[Aust]s [LMAAust]3

Early MioceneEarly MioceneEarly OligoceneMiddle Eocene–OligoceneMiddle EoceneMiddle Eocene

Yallourn Clays4

Kiandra4

Cethana5

West Dale5

Anglesea4,6

Golden Grove4,6

272030292821

0.710.720.230.600.650.71

17.917.118.7

22.923.28.3

19.621.022.9

23.123.55.0

19.020.823.1

17.117.34.2

14.215.517.1

2.362.712.092.462.442.68

Middle EoceneEarly middle EoceneEarly–middle EoceneEarly–middle EoceneLate Paleocene

Dragon Glacier7

Nerriga4,6

Brandy Ck4,6

Hotham Heights4,6

Cambalong Ck4,6

3724282621

0.290.790.750.740.58

19.410.125.324.123.818.9

7.326.124.624.218.1

5.819.218.217.913.5

2.032.252.212.332.91

taxa. However, the compositions of the other floras inTable 1 are consistent with the MAT estimates present-ed here using Australian LMA (Greenwood et al., 2003;Greenwood and Christophel, 2004).

DISCUSSION

This study shows that the relationship between leaf-margin proportion and mean annual temperature is sta-tistically significant for Australian rainforest vegetation.This result matches that shown for unstressed environ-ments in other major regions around the world (Wolfe,1979; Wilf, 1997; Gregory-Wodzicki, 2000), and the similarslopes support global evolutionary convergence of leafphysiognomy in response to the selection of temperaturein mesic habitats. Leaf-margin characteristics appear tobelong with a long list of other leaf traits, such as lifespan,nutrient content, thickness, and herbivore resistance, thatshow global convergence under similar selective pressures(e.g., Coley et al., 1985; Reich et al., 1997). However, theAustralian correlation is weaker than for other analyses(e.g., LMA Australia r2 5 63% versus LMA East Asia r2 598%, LMA CLAMP ‘coldest sites removed’ r2 5 84%, andLMA Bolivia r2 5 89%), and the intercept differs so thatfewer toothed species are present at a given temperature.The latter observation marks a true regional difference be-tween Australia and other continents.

Hypotheses to explain the lack of toothed species in-clude: (1) significant phylogenetic effects in Australia com-pared to other continents, such as a prevalence of taxa un-able to develop marginal teeth; for example, the Austra-lian flora is dominated by Myrtaceae, the majority ofwhich are untoothed (Jordan, 1997); (2) additional factorsin the physical environment confounding the relationshipbetween leaf-margin proportion and MAT (e.g., low phos-phorous, low annual temperature amplitude, high vari-ability in rainfall); and (3) absence of historical factors fa-voring the evolution, survival, or migration to Australia oftemperate deciduous lineages with high proportions oftoothed species, or the loss of such lineages at some timeduring the Cenozoic.

Point (1) seems unlikely because sites dominated by ob-ligately entire-margined genera, such as Eucalyptus in theMyrtaceae, were largely excluded in this study. Moreover,the closed forest formations studied here share many taxawith similar forests in Africa, South America, and EastAsia (Crisp et al., 1999), and have done so for much of theCenozoic (Vadala and Greenwood, 2001; Greenwood andChristophel, 2004).

Regarding point (2), the present analysis of the influ-ence of other environmental factors, such as seasonalityand rainfall, showed that they appear to have no signifi-cant effect in determining the Australian LMP versusMAT intercept, or at least, any effect is masked by thestrong autocorrelation of the temperature variables andMAT (Figs. 4, 5). Analyses of sites with markedly infertilesoils, and of solely fertile soil sites, did not shift the inter-cept or the line to approximate the other datasets (e.g.,East Asia, North America; Fig. 4A), but did show that in-fertile sites were enriched in untoothed species relative tofertile soils sites, matching Webb’s (1968) original obser-vations. The analysis of the influence of soil was limited bythe available data, and further analysis based on more de-tailed soil information may be warranted in Australia andelsewhere.

Regarding point (3), several preliminary observationsare presented. First, a significant difference between thehistories of the Australian and both the North and SouthAmerican continents is Australia’s lack since the Eoceneof a land connection with high-latitude landmasses (i.e., apotential source of cold-adapted woody taxa). Webb (1968)noted a trend for decreasing diversity of deciduous treespecies from the tropics to the subtropics, and almost alldeciduous dicot species in the modern Australian flora aredrought-deciduous tropical trees in families such as Meli-aceae (e.g., Toona) and Sterculiaceae (e.g., Brachychiton).The sole tree species that is cold-season deciduous, Notho-fagus gunnii, is restricted to Tasmania. Nonetheless, someTasmanian Paleogene floras, such as Cethana, have a rel-atively high diversity of toothed taxa, such as Nothofagus.These higher-latitude sites may have supported some de-ciduous taxa, such as deciduous Nothofagus species, per-

AUSTRALIAN LEAF-MARGIN ANALYSIS 137

haps in response either to short winter days or low tem-peratures (Greenwood et al., 2003; Greenwood and Chris-tophel, 2004). Second, both North and South Americahave substantive high mountain ranges (the Rocky Moun-tains and Andes) that have acted as source areas for cold-adapted biota, whereas Australia has low relief now andhad low relief for the whole of the Cenozoic (Taylor, 1994).

A further difference between the histories of the Austra-lian and both the North and South American continents isthat Australia was essentially insulated from Cenozoicglobal climate cooling by its northward drift into middleand low latitudes. Australia may have cooled as little as58C over the Paleogene and Neogene (Macphail et al.,1994; Greenwood et al., 2003). Additionally, late Cenozoiccold intervals appear to have acted to force regional ex-tinctions in southeastern Australia, rather than to pro-mote the in situ evolution of cold-adapted tree taxa (Ker-shaw et al., 1994). In significantly smaller New Zealand,extinction and floral turnover appear to have had dramat-ic effects on the composition of its flora (Lee et al., 2001).

A final factor for consideration regarding point (3) isthat the present Australian climate, even at the highestlatitudes and elevations, does not favor deciduous, cold-adapted trees. The cold-climate (i.e., winter-snow) tree flo-ra is poorly adapted to extreme cold and is dominated bybroad-leaved evergreen trees with untoothed leaves, suchas the snow gum, Eucalyptus pauciflora, which does nottolerate sustained temperatures lower than 2238C, andtypically experiences winter mean minima .258C (Sakaiand Larcher, 1987; Costin et al., 2000; Jobbagy and Jack-son, 2000). According to Jobbagy and Jackson (2000), thealtitudinal tree-line worldwide approximates a warmmonth mean temperature (WMMT) of 6.8–9.38C, and inthe Southern Hemisphere it corresponds to a cold monthmean (CMMT) of .21.98C. The climatic limit at altitudefor broadleaved evergreen trees is a CMMT approximately228C. Therefore, the climate space exclusively occupied bybroadleaved cold-deciduous trees elsewhere in the world(i.e., WMMT .108C and CMMT ,22.08C) is largely ab-sent from present-day Australia. Southern Australia alsoappears to have lost significant diversity in the mesic mi-crothermal to mesothermal forest types, with a markedlylower diversity today than in the Paleogene (Greenwoodand Christophel, 2004), particularly for toothed taxa suchas Nothofagus. Tasmania today has only two species ofNothofagus, one of which is deciduous (N. gunnii), the oth-er evergreen (N. cunninghamii). Possibly as a result of thefactors listed above, there is no pool of cold-adapted,toothed woody dicot species to fill the cool ecospace in to-day’s relatively cold climate; hence, the modern flora is de-ficient in these elements relative to other continents. It isnot known when the Australian flora lost its deciduous mi-crothermal to mesothermal woody dicots, although somePaleocene and early Eocene high latitude floras may haverepresented deciduous forest (Greenwood et al., 2003;Greenwood and Christophel, 2004). From a practical per-spective, the different intercept for Australian LMA (Fig.3) means that previously published MAT estimates forAustralian fossil leaf floras should be reevaluated (Table1). For example, a flora with 80% non-toothed species willproduce an estimate of ;258C using East Asian LMA, but;208C using Australian LMA (equation 2). Multiple re-gression analysis shows some potential to lessen the dis-

crepancy (Table 1). However, leaf-margin analysis basedon modern Australian floras may not be suitable for somePaleogene floras that grew before significant extinctionsof cold-adapted lineages occurred. Accordingly, paleotem-perature estimates for the Cenozoic of Australia shouldtake into account the relative importance of deciduoustaxa. Some paleotemperature estimates may best be madeusing non-Australian LMA, derived from samples with ahigh representation of deciduous taxa. In other cases, Aus-tralian LMA may be more suitable. Until more is knownabout the timing and magnitudes of the extinctions ofcold-adapted lineages, as well as the deciduousness or ev-ergreenness of fossil leaf species, the most conservativeapproach is to use LMA based on Australia as a minimumtemperature estimate and East Asian LMA as a maxi-mum. Hunt and Poole (2003) applied this approach in a re-cent analysis of middle Eocene floras from King George Is-land, West Antarctica, noting that a Southern Hemi-sphere LMA was potentially inappropriate because thefossil floras were dominated by deciduous taxa.

In summary, the response of leaf-margin proportion toclimate, as seen in the slope of the regression, has yet toshow a regional effect of any importance. Local conditions,especially environmental stresses, have accounted for allof the major deviations. Australia is different only in thatthere are fewer toothed species at a given temperaturethan elsewhere, but the amount of decrease in toothedspecies with temperature is not different from other re-gions. Investigations of leaf physiognomy in stressed en-vironments, such as deserts, subalpine zones, and cloudforests, as well as in isolated areas, such as New Zealand,are important with regard to some fossil floras (e.g.,Stranks and England, 1997; Gregory-Wodzicki, 2000).However, it is equally important not to conflate regionaleffects, which is a deviation that holds across environ-ments within a wide area, with environmental effects onparticular floras within a region that are probably conver-gent in many regions with the same environments (e.g.,convergent physiognomies of deserts, subalpine zones,and cloud forests). For this reason, the division of leaf-cli-mate space into three domains (Gregory-Wodzicki, 2000),which are a mixture of specialized environments, smalllandmasses, and continents, appears premature.

CONCLUSIONS

The statistically significant relationship between leaf-margin proportion and mean annual temperature is wide-spread globally in unstressed environments. The slope ofthe regression in Australia is virtually identical to that ofEast Asia, North America, and Bolivia. This indicateswidespread convergence in the response of vegetation totemperature. However, the vertical intercept of the re-gression for the Australian database is higher than that ofthe majority of the other regional databases. Hence,woody dicot species with toothed leaf margins in Australiaare a lower proportion of total species at a site than else-where at the same temperature.

Leaf-margin analysis, calibrated for Australian vegeta-tion, can be used to estimate mean annual temperature forAustralian Cenozoic leaf floras, with the error of the esti-mate similar to, but slightly greater than, that for LMAfrom other databases. The caveat for application is that

138 GREENWOOD ET AL.

the Northern Hemisphere LMA may be more appropriatethan the Australian LMA for Australian Cenozoic florasdominated by deciduous dicots. The most conservative ap-proach is to use LMA based on Australia as a minimumtemperature estimate and non-Australian LMA as a max-imum (e.g., Hunt and Poole, 2003).

Multivariate approaches may help to compensate for re-gional differences in univariate leaf-climate relationships,including those related to historical events. However, pre-vious applications of multivariate approaches, based on aglobal database lacking Australian sites, have not im-proved accuracy significantly for Australian vegetation(Greenwood and Wing, 1995; Jordan, 1997). An Austra-lian multivariate database may be required to assess thispoint.

ACKNOWLEDGEMENTS

Primary funding for this study was provided by Austra-lian Research Council grants to Greenwood and Wing(SGS 40/96 and A39802019). Technical support was pro-vided by the CSIRO Tropical Forest Research Centre inAtherton, Queensland. We thank also the following peoplefor technical assistance in the laboratory and field: HeidiWittesch, Mark Scarr, and Cathy Greenwood. Wilf re-ceived funding from the Petroleum Research Fund of theAmerican Chemical Society (Grant no. 35229-G2) and theMichigan Society of Fellows.

REFERENCES

ASH, A.W., ELLIS, B., HICKEY, L.J., JOHNSON, K.R., WILF, P., andWING, S.L., 1999, Manual of Leaf Architecture: Morphological De-scription and Categorization of Dicotyledonous and Net-VeinedMonocotyledonous Angiosperms: Smithsonian Institution, Wash-ington, D.C., 65 p.

BAILEY, I.W., and SINNOTT, E.W., 1915, A botanical index of Creta-ceous and Tertiary climates: Science, v. 41, p. 831–834.

BAILEY, I.W., and SINNOTT, E.W., 1916, The climatic distribution ofcertain kinds of angiosperm leaves: American Journal of Botany,v. 3, p. 24–39.

BAKER-BROSH, K.F., and PEET, R.K., 1997, The ecological significanceof lobed and toothed leaves in temperate forest trees: Ecology, v.78, p. 1250–1255.

BEADLE, N.C.W., 1966, Soil phosphate and its role in moulding seg-ments of the Australian flora and vegetation, with special refer-ence to xeromorphy and sclerophylly: Ecology v. 47, p. 992–1007.

BURNHAM, R.J., 1997, Stand characteristics and leaf litter composi-tion of a dry forest hectare in Santa Rosa National Park, CostaRica: Biotropica, v. 29, p. 384–395.

BURNHAM, R.J., PITMAN, N.C.A., JOHNSON, K.R., and WILF, P., 2001,Habitat-related error in estimating temperatures from leaf mar-gins in a humid tropical forest: American Journal of Botany, v. 88,p. 1096–1102.

BUSBY, J.R., 1991, BIOCLIM—A bioclimatic analysis and predictionsystem: in Margules, C.R., and Austin, M.P., eds., Nature Conser-vation: Cost Effective Biological Surveys and Data Analysis: CSI-RO, Melbourne, p. 64–68.

CANNY, M.J., 1990, What becomes of the transpiration stream?: NewPhytologist, v. 114, p. 341–368.

CARPENTER, R.A., HILL, R.S., and JORDAN, G., 1994, Cenozoic vege-tation in Tasmania: macrofossil evidence: in Hill, R.S., ed., TheHistory of Australian Vegetation: Cretaceous to Recent, Cam-bridge University Press, Cambridge, p. 276–298.

CHRISTOPHEL, D.C., 1981, Tertiary megafossil floras as indicators offloristic associations and palaeoclimate: in Keast, A., ed., Ecologi-cal Biogeography of Australia: Junk, The Hague, p. 379–390.

CHRISTOPHEL, D.C., and GREENWOOD, D.R., 1989, Changes in climate

and vegetation in Australia during the Tertiary: Review of Palaeo-botany and Palynology, v. 58, p. 95–109.

COLEY, P.D., BRYANT, J.P., and CHAPIN, F.S., III, 1985, Resourceavailability and plant anti-herbivore defense: Science, v. 230, p.895–899.

COSTIN, A., GRAY, M., TOTTERDELL, C., and WIMBUSH, D., 2000, Kos-ciuszko Alpine Flora: CSIRO Publishing, Melbourne, 233 p.

CRISP, M.D., WEST, J.G., and LINDER, H.P., 1999, Biogeography of theterrestrial flora: in Orchard, A.E., ed., Flora of Australia v. 1, 2nd

ed.: ABRS, Canberra, p. 321–368.GALLAGHER, S.J., GREENWOOD, D.R., TAYLOR, D., SMITH, A.J., WAL-

LACE, M.W., and HOLDGATE, G.R., 2003, The Pliocene climatic andenvironmental evolution of southeastern Australia: evidence fromthe marine and terrestrial realm: Palaeogeography, Palaeoclima-tology, Palaeoecology, v. 193, p. 349–382.

GIVNISH, T., 1979, On the adaptive significance of leaf form: in Sol-brig, O.T., Jain, S., Johnson, G.B., and Raven, P.H., eds., Topics inPlant Population Biology: Columbia University Press, New York,p. 375–407

GREENWOOD, D.R., 1992, Taphonomic constraints on foliar physiog-nomic interpretations of Late Cretaceous and Tertiary palaeocli-mates: Review of Palaeobotany and Palynology, v. 71, p. 142–196.

GREENWOOD, D.R., 1994, Palaeobotanical evidence for Tertiary cli-mates: in Hill, R.S., ed., The History of Australian Vegetation:Cretaceous to Recent: Cambridge University Press, Cambridge, p.44–59.

GREENWOOD, D.R., 1996, Eocene monsoon forests in central Austra-lia?: Australian Systematic Botany, v. 9, p. 95–112.

GREENWOOD, D.R., 2001, Climate—wood and leaves: in Briggs, D.E.,and Crowther, P.R., eds., Palaeobiology II: Blackwell Scientific,London, p. 480–483.

GREENWOOD, D.R., and CHRISTOPHEL, D.C., 1990, Leaf margin anal-ysis of Cretaceous and Tertiary climates: modern Southern Hemi-sphere rainforests as analogues: Geological Society of America,Abstracts with Program, v. 22,7, p. A350.

GREENWOOD, D.R., and CHRISTOPHEL, D.C., 2004, The origins andTertiary history of Australian ‘‘Tropical’’ rainforests: in Bermingh-am, E., Dick, C., and Moritz, C., eds., Tropical Rainforests: Past,Present, and Future: Chicago University Press, Chicago.

GREENWOOD, D.R., MOSS, P.T., ROWETT, A.I., VADALA, A.J., and KEE-FE, R.L., 2003, Plant communities and climate change in south-eastern Australia during the early Paleogene: in Wing, S.L., Gin-gerich, P.D., Schmitz, B., and Thomas, E., eds., Causes and Con-sequences of Globally Warm Climates in the Early Paleogene:Geological Society of America Special Paper 369, p. 365–380.

GREENWOOD, D.R., and WING S.L., 1995, Eocene continental climatesand latitudinal gradients: Geology, v. 23, p. 1040–1048.

GREGORY, K., and CHASE, C.G., 1992, Tectonic significance of paleo-botanically estimated climate and altitude of the late Eocene ero-sion surface, Colorado: Geology, v. 20, p. 581–585.

GREGORY, K.M., and MCINTOSH, W.C., 1996, Paleoclimate and pa-leoelevation of the Oligocene Pitch-Pinnacle flora, SawatchRange, Colorado: Geological Society of America Bulletin, v. 108, p.545–561.

GREGORY-WODZICKI, K.M., 2000, Relationship between leaf morphol-ogy and climate, Bolivia: implications for estimating paleoclimatefrom fossil floras: Paleobiology, v. 26, p. 668–688.

HILL, R.S., 1998, Poor soils and dry climate: the evolution of the Aus-tralian scleromorphic and xeromorphic vegetation: Australian Bi-ologist, v. 11, p. 26–29.

HOULDER, D., HUTCHINSON, M., NIX, H., and MCMAHON, J., 1999,ANUCLIM version 5.0 User Guide: Centre for Resource and En-vironmental Studies, Australian National University, Canberra,73 p.

HUNT, R.J., and POOLE, I., 2003, Paleogene West Antarctic climateand vegetation history in light of new data from King George Is-land: in Wing, S.L., Gingerich, P.D., Schmitz, B., and Thomas, E.,eds.: Causes and Consequences of Globally Warm Climates in theEarly Paleogene: Geological Society of America Special Paper 369,p. 395–412.

HYLAND, B.P.M., WHIFFIN, T., CHRISTOPHEL, D.C., GRAY, B., ELICK,R.W., and FORD, A.J., 1999, Australian Tropical Rain Forest Treesand Shrubs. An interactive identification system for trees and

AUSTRALIAN LEAF-MARGIN ANALYSIS 139

shrubs. CD-ROM and user guide: CSIRO Publishing, Melbourne,95 p.

JACOBS, B.F., 2002, Estimation of low-latitude paleoclimates usingfossil angiosperm leaves: examples from the Miocene Tugen Hills,Kenya: Paleobiology, v. 28, p. 399–421.

JOBBAGY, E.G., and JACKSON, R.B., 2000, Global controls of forest lineelevation in the northern and southern hemispheres: Global Ecol-ogy and Biogeography, v. 9, p. 253–268.

JORDAN, G.J., 1997, Uncertainty in palaeoclimate reconstructionsbased on leaf physiognomy: Australian Journal of Botany, v. 45, p.527–547.

KENNEDY, E.M., 1998, Cretaceous and Tertiary megafloras from NewZealand and their climate signals: Unpublished Ph.D. disserta-tion, The Open University, Milton Keynes, 2 vol.

KERSHAW, A.P., MARTIN, H.A., and MCEWEN MASON, J.R.C., 1994,The Neogene: a period of transition: in Hill, R.S., ed., The Historyof Australian Vegetation: Cretaceous to Recent: Cambridge Uni-versity Press, Cambridge, p. 299–327.

KOWALSKI, E.A., 2002, Mean annual temperature estimation basedon leaf morphology: a test from tropical South America: Palaeo-geography, Palaeoclimatology, Palaeoecology, v. 188, p. 141–165.

LEE, D.E., LEE, W.G., and MORTIMER, N., 2001, Where have all theflowers gone? Depletion and turnover in the New Zealand Ceno-zoic angiosperm flora in relation to palaeogeography and climate:Australian Journal of Botany, v. 49, p. 341–356.

LEIGH, E.G., JR., 1999, Tropical Forest Ecology: A View from BarroColorado Island: Oxford University Press, New York, 245 p.

MACPHAIL, M.K., ALLEY, N.F., TRUSWELL, E.M., and SLUITER, I.R.K.,1994, Early Tertiary vegetation: evidence from spores and pollen:in Hill, R.S., ed., The History of Australian Vegetation:Cretaceousto Recent: Cambridge University Press, Cambridge, p. 189–261.

MOSBRUGGER, V., and ROTH, A., 1996, Biomechanics in fossil plant bi-ology: Review of Palaeobotany and Palynology, v. 90, p. 195–207.

PETERS, R., and SLOAN, L.C., 2000, High greenhouse gas concentra-tions and polar stratospheric clouds: a possible solution to highlatitude migration at the Latest Paleocene Thermal Maximum:Geology, v. 28, p. 979–982.

REICH, P.B., WALTERS, M.B., and ELLSWORTH, D.S., 1997, From trop-ics to tundra: global convergence in plant functioning:Proceedingsof the National Academy of Sciences USA, v. 94, p. 13730–13734.

RICHARDS, P.W., 1996, The Tropical Rain Forest, 2nd ed.: CambridgeUniversity Press, Cambridge, 599 p.

SAKAI, A., and LARCHER, W., 1987, Frost Survival of Plants: Respons-es and Adaptation to Freezing Stress: Springer-Verlag, Berlin,321 p.

SEWALL, J.O., SLOAN, L.C., HUBER, M., and WING, S.L., 2000, Climatesensitivity to changes in land surface characteristics: Global andPlanetary Change, v. 26, p. 445–465.

SHELLITO, C.J., SLOAN, L.C., and HUBER, M., 2003, Climate modelsensitivity to atmospheric CO2 levels in the Early–Middle Paleo-gene: Palaeogeography, Palaeoclimatology, Palaeoecology, v. 193,p. 113–123.

SOKAL, R. R., and ROHLF, F.J., 1995, Biometry, 3rd ed.: New York,W.H. Freeman, 887 p.

STRANKS, L., and ENGLAND, P., 1997, The use of a resemblance func-tion in the measurement of climatic parameters from the physi-ognomy of woody dicotyledons: Palaeogeography Palaeoclimatolo-gy Palaeoecology, v. 131, p. 15–28.

TAYLOR, G., 1994, Landscapes of Australia: their nature and evolu-tion: in Hill, R.S., ed., The History of Australian Vegetation: Cre-taceous to Recent: Cambridge University Press, Cambridge, p.60–79.

TRACEY, J.G., 1982, The Vegetation of the Humid Tropics of NorthQueensland: CSIRO, Melbourne, 124 p.

UPCHURCH, G.R., and WOLFE, J.A., 1987, Mid-Cretaceous to EarlyTertiary vegetation and climate: evidence from fossil leaves andwoods: in Friis, E.M., Chaloner, W.G., and Crane, P.R., eds., TheOrigins of Angiosperms and Their Biological Consequences: Cam-bridge University Press, Cambridge, p. 75–106.

VADALA, A.J., and GREENWOOD, D.R., 2001, Australian Paleogenevegetation and environments: evidence for palaeo-Gondwanic el-ements in the fossil records of Lauraceae and Proteaceae: in Met-calfe, I., Smith, J.M.B., and Davidson, I., eds., Faunal and FloralMigrations and Evolution in SE Asia–Australasia: Swets and Zei-tlinger Publishers, Lisse, p. 196–221.

VEEVERS, J.J., POWELL, C. MCA., and ROOTS, S.R., 1991, Review ofseafloor spreading around Australia. 1. Synthesis of patterns ofspreading: Australian Journal of Earth Sciences, v. 38, p. 373–389.

VELAZQUEZ-ROSAS, N., MEAVE, J., and VAZQUEZ-SANTANA, S., 2002,Elevational variation of leaf traits in montane rain forest tree spe-cies at La Chinantla, Southern Mexico: Biotropica, v. 34, p. 534–546.

WEBB, L.J., 1959, A physiognomic classification of Australian rainfo-rests: Journal of Ecology, v. 47, p. 551–570.

WEBB, L.J., 1968, Environmental determinants of the structuraltypes of Australian rain forest vegetation: Ecology, v. 49, p. 296–311.

WEBB, L.J., TRACEY, J.G., and WILLIAMS, W.T., 1984, A floristicframework of Australian rainforests: Australian Journal of Ecol-ogy, v. 9, p. 169–198.

WIEMANN, M.C., MANCHESTER, S.R., DILCHER, D.L., HINOJOSA, L.F.,and WHEELER, E.A., 1998, Estimation of temperature and precip-itation from morphological characters of dicotyledonous leaves:American Journal of Botany, v. 85, p. 1796–1802.

WILF, P., 1997, When are leaves good thermometers? A new case forLeaf Margin Analysis: Paleobiology, v. 23, p. 373–390.

WING, S.L., BAO, H., and KOCH, P.L., 2000, An early Eocene cool pe-riod? Evidence for continental cooling during the warmest part ofthe Cenozoic: in Huber, B.T., MacCleod, K., and Wing, S.L., eds.,Warm Climates in Earth History: Cambridge University Press,Cambridge, p. 197–237

WING, S.L., and GREENWOOD, D.R., 1993, Fossils and fossil climates:the case for equable Eocene continental interiors: PhilosophicalTransactions of the Royal Society, London B. Biological Sciences,v. 341, p. 243–252.

WOLFE, J.A., 1971, Tertiary climatic fluctuations and methods ofanalysis of Tertiary floras: Palaeogeography, Palaeoclimatology,Palaeoecology, v. 9, p. 27–57.

WOLFE, J.A., 1978, A paleobotanical interpretation of Tertiary cli-mates in the Northern Hemisphere: American Scientist, v. 66, p.694–703.

WOLFE, J.A., 1979, Temperature parameters of humid to mesic for-ests of Eastern Asia and relation to forests of other regions of theNorthern Hemisphere and Australasia: United States GeologicalSurvey Professional Paper, 1106, 37 p.

WOLFE, J.A., 1993, A method of obtaining climatic parameters fromleaf assemblages: U.S. Geological Survey Bulletin, 2040, 73 p.

WOLFE, J. A., 1995, Paleoclimatic estimates from Tertiary leaf assem-blages: Annual Review of Earth and Planetary Sciences, v. 23, p.119–142.

ACCEPTED OCTOBER 15, 2003

140 GREENWOOD ET AL.

APPENDIX 1Site data for 113 Australian rainforest samples (from Webb et al., 1984), arranged in order of increasing site richness. The 20-species

cutoff is indicated with a horizontal line (equation 2). The forest types follow the usage in the data set provided to us, which approxi-mately follows Webb (1968) and Tracey (1982). A key is given at the end of this table. MAT 5 mean annual temperature; MAP 5 meanannual precipitation. Soil types (source regolith): fertile sites—A 5 alluvial, B 5 basalt, L 5 limestone, and V 5 basic volcanic; infertilesites—M 5 metamorphic, G 5 granite, R 5 acid volcanic, S 5 sand, and N 5 sandstone.

Site Forest typeElevation

(m) Latitude Longitude MAT (8C)MAP

(mm/yr) Soil Taxa Non-entire LMP (%)

436435428431408

MFFMFFSNVFMFFMFF

1570104020025

730

3783037842398003884636829

145855145843146818143837150802

5.68.4

12.613.211.1

16551589113911571203

——A——

34555

12223

33.350404060

427483204429157

SNVFSNVFMFFSNVFMFF

2001000800200

1400

3885429818358273980131810

146825152806149852146820152820

12.71311.112.610.2

1104951

101711331635

AMMAV

55777

33223

606028.628.642.9

19544

443496440

MFFDVTNMTSNVFMFF/NMT

1400550

150050

1200

3185817817418473782737850

151830144842146834149850146817

10.222.13.3

14.37.3

1272841

1918923

1580

VGBM—

7888

10

51335

71.412.537.537.550

254471162458206

NVFMFFCNVFANVFSNVF*

1006108080040

400

3383028826318102881737835

150823153807152825153830149810

11.413.413.619.311.8

12132072173817851199

ABVAA

1011111313

63715

6027.363.67.7

38.5448561164437163

C/SNVFLMVFCNVFMFF/NMTSNVF

1200280800

1350950

3080625838318104284031810

152825149857152825146830152825

11.419.813.63.7

12.7

1330691

173818261764

GRV—V

1314141414

93778

69.221.4505057.1

58498161159558

MVFFPMFFSNVF*MFFAMVF

101150400

1400300

1783028818318123181025840

146800153809152825152820152803

23.413.115.910.219.1

3686204414241635938

ABVVM

1515151516

16893

6.74053.36018.8

182454

7456559

ANVFCNVFENVFMFFAMVF

350600400925300

2882528815158002881625840

152845153816145805153810152802

17.116.222.714.419.1

1140246220141915931

VRMBA

1616171719

591

104

31.356.35.9

58.821.1

16544548245

388

CNVFMF/SNVFSNVFDVTSNVF*

800100098555040

3082230819298181780525828

152842152852152818144825153805

13.412.412.922.520.8

198524131123914

1612

BMMLS

1919192020

6101222

31.652.663.21010

114410280205474

NVFMFFCMVFSNVFENVF

30082040

4005

2185532820188043584129832

149820151825145841150810153833

2113.423.813.319.2

13161178214412951598

V—AMV

2020212122

510381

255014.338.14.5

12816090

243193

NVFCNVFANVFCNVFCNVF

200150150800400

2185431812248092783828821

149820152825151847152820152844

21.517.320.915.316.9

13751287124010811104

VA—BB

2222232324

4866

11

18.236.426.126.145.8

21219445011389

MFFCNVFCNVFMVFMVF

120040040

250400

2881430829308231582816850

153810152825152847145816145838

12.915.818.123.322.2

20751216176119101984

BBBGM

2425252626

119934

45.8363611.515.4

41162214247

560

MFFENVFMVFFPSNVFANVF

7504030

950410

3282017814208552180225840

151826145848148843148835149859

13.823.822.51819.1

1190313615962943717

BSAGV

2627272728

100596

38.50

18.533.321.4

55615893

ANVFMF/SNVFCNVF

1901100900

258283181017831

152807152820145835

19.811.919.114.1

97116362411

AVB

282829

7127

2542.924.1

AUSTRALIAN LEAF-MARGIN ANALYSIS 141

APPENDIX 1Continued.

Site Forest typeElevation

(m) Latitude Longitude MAT (8C)MAP

(mm/yr) Soil Taxa Non-entire LMP (%)

263481

SNVFSNVF

10001000

2881329816

153817152807

14.113

2380953

RM

2930

811

27.636.7

22199

465447312

ENVFMVFCNVFCNV/MFFMVF

100300950

130040

2881717800288233082218816

153835145840153807152833145859

19.222.714.110.823.7

19242035198718322121

SABBA

3131313132

59

11165

16.12935.551.615.6

554203609115246

AMVFCNVFMVFFP*NVFCNVF

19020040

300600

2582734845168162183526842

152806150845145828149812153834

19.815.724.520.917.1

9671827319317282210

VBAVB

3232333434

810368

2531.39.1

17.623.5

249223244455301

CNVFCNVFCNVFCNVFMVF

600800800900200

2683228813278382881518815

152834153817152820152829146816

16.915.215.314.322.7

12413167108113382599

ABBBG

3435353637

99

10104

26.525.728.627.810.8

18396

102217464

SNVFCMVFCNVFCNVFCNVF

800150400200250

3081517851248582681628822

152845145842151828152855152849

13.423.118.719.117.7

17662876808

16081085

—AMMA

3738383839

12789

10

32.418.421.123.725.6

16726446332

331

SNVFCNVFANVFENVFCMVF

100080025065040

3081728813288221385217808

152847153817152849143820145846

12.415.217.722.224

19643167108518832069

MBBAA

4141444545

12141266

29.334.127.313.313.3

22421846917

210

SNVFCNVFSNVFENVFCNVF

1000400750150400

2881326845288241183228812

153817152845153816142847153811

14.117.815.324.917.2

23801688277518271742

RBRNB

4951525354

1412203

11

28.623.538.55.7

20.4185486303214295

SNVFSNVFMVFENVFCMVF

800910200200600

3182430821188182583016803

152810152846146808153808145812

13.712.822.819.920.8

13062064238816762232

—MGSA

5454575757

16177

1017

29.631.512.317.529.8

16892

211171466213

SNVFSNVFCNVFCNVFCNVFENVF

40300800600900200

308202881128814288342882326800

153807153818153810153820153807153808

1817.815.11614.419.3

171321192326258920161586

NRBBBS

595965666970

92218221810

15.337.327.733.326.114.3

453209607170563593

CNVFCNVFCMVFNVFSNVFCNVF

50040040

2501000600

288152782016816308181782517816

153816152845145828153805145825145838

16.717.724.516.918.720.8

221313513193194714381965

BBVARB

72748186

101172

20149

232528

27.818.911.126.724.816.3

Forest code Webb-Tracey structural type Approximate equivalents

NMTMFF

Nanophyll mossy thicketMicrophyll fern forest

cool to warm temperate rainforest 6 sclerophyll emergents and/orconifers

SNVF Simple notophyll vine forest montane tropical to lower montane subtropical rainforestCNVFNVF

Complex notophyll vine forestNotophyll vine forest

lower montane tropical to lowland subtropical rainforest

LMVF Low microphyll vine forest semi-evergreen seasonal subtropical low forestAMVF Araucarian microphyll vine forest semi-evergreen tropical low forest 1 araucarian emergentsENVF Evergreen notophyll vine forest evergreen seasonal tropical forest 6 sclerophyll emergentsANVF Araucarian notophyll vine forest seasonal tropical forest 1 araucarian emergentsMVFCMVFMVFFP

Mesophyll vine forestComplex mesophyll vine forestMesophyll vine forest 1 fan palms

tropical rainforest of Richards (1996), but including tropical ever-green seasonal forest by some definitions (Webb 1959; Green-wood 1996)

DVT Deciduous vine thicket low monsoon forestMFF/NMT, MF/

SNVF, C/SNVFand CNV/MFF

Transitionary types

142 GREENWOOD ET AL.

APPENDIX 2Climate variables assessed in this study (Figs. 4, 5), definitions from Houlder et al. (1999). Variables are listed in the order of r2

(highest to lowest) in regression analysis with LMP.

Variable UnitsDefinition (from Houlder

et al., 1999)

Mean annual temperature (MAT) 8C The arithmetic mean of the mean temperature (temp.). of each month, wherethe mean month temp. is the mean of each day (i.e., the mean temp.through each daily diurnal cycle)

Mean temperature of the coldest quar-ter (MeanTcoldQtr)

8C The mean of the 3 successive months with the lowest mean temp.

Min. temp. coldest period (MinTcoldQtr) 8C Mean minimum temp. for the coldest monthMean temperature of the warmest quar-

ter (MeanTwarmQtr)8C The mean of the 3 successive months with the highest mean temp.

Max. temp. warmest period (Max-TwarmQtr)

8C Mean maximum temp. for the warmest month

Temperature seasonality N/A (temperature coefficient of variation) is the standard deviation of the weeklymean temperatures expressed as a % of the mean of those temperatures(in 8K)

Precip. seasonality N/A (precipitation coefficient of variation) is the standard deviation of the weeklyprecip. estimates expressed as a % of the mean of those estimates (i.e.,MAP)

Precipitation of the driest quarter(PrecipDriestQtr)

mm The mean of the 3 successive months with the lowest precipitation

Precipitation of the wettest quarter(PrecipWetQtr)

mm The mean of the 3 successive months with the highest precipitation

Precipitation of the coldest quarter(PrecipColdQtr)

mm The mean of the 3 successive months with the lowest mean temp. (i.e., win-ter precip.)

Precipitation of the warmest quarter(PrecipWarmQtr)

mm The mean of the 3 successive months with the highest mean temp. (i.e., sum-mer precip.)

Mean annual precipitation (MAP) mm/yr The annual sum of each monthly mean precipitation