� 2007 The Paleontological Society. All rights reserved. 0094-8373/07/3304-0005/$1.00
Paleobiology, 33(4), 2007, pp. 574–589
Fossil leaf economics quantified: calibration, Eocene case study,and implications
Dana L. Royer, Lawren Sack, Peter Wilf, Christopher H. Lusk,Gregory J. Jordan, Ulo Niinemets, Ian J. Wright, Mark Westoby,Barbara Cariglino, Phyllis D. Coley, Asher D. Cutter, Kirk R. Johnson,Conrad C. Labandeira, Angela T. Moles, Matthew B. Palmer,and Fernando Valladares
Abstract.—Leaf mass per area (MA) is a central ecological trait that is intercorrelated with leaf lifespan, photosynthetic rate, nutrient concentration, and palatability to herbivores. These coordinatedvariables form a globally convergent leaf economics spectrum, which represents a general continuumrunning from rapid resource acquisition to maximized resource retention. Leaf economics are littlestudied in ancient ecosystems because they cannot be directly measured from leaf fossils. Here weuse a large extant data set (65 sites; 667 species-site pairs) to develop a new, easily measured scalingrelationship between petiole width and leaf mass, normalized for leaf area; this enables MA estimationfor fossil leaves from petiole width and leaf area, two variables that are commonly measurable in leafcompression floras. The calibration data are restricted to woody angiosperms exclusive of monocots,but a preliminary data set (25 species) suggests that broad-leaved gymnosperms exhibit a similarscaling. Application to two well-studied, classic Eocene floras demonstrates that MA can be quantifiedin fossil assemblages. First, our results are consistent with predictions from paleobotanical and pa-leoclimatic studies of these floras. We found exclusively low-MA species from Republic (Washington,U.S.A., 49 Ma), a humid, warm-temperate flora with a strong deciduous component among the an-giosperms, and a wide MA range in a seasonally dry, warm-temperate flora from the Green RiverFormation at Bonanza (Utah, U.S.A, 47 Ma), presumed to comprise a mix of short and long leaf lifespans. Second, reconstructed MA in the fossil species is negatively correlated with levels of insectherbivory, whether measured as the proportion of leaves with insect damage, the proportion of leafarea removed by herbivores, or the diversity of insect-damage morphotypes. These correlations areconsistent with herbivory observations in extant floras and they reflect fundamental trade-offs inplant-herbivore associations. Our results indicate that several key aspects of plant and plant-animalecology can now be quantified in the fossil record and demonstrate that herbivory has helped shapethe evolution of leaf structure for millions of years.
Dana L. Royer. Department of Earth and Environmental Sciences, Wesleyan University, Middletown, Con-necticut 06459. E-mail: [email protected]
Lawren Sack.* Department of Botany, University of Hawai’i at Manoa, Honolulu, Hawai’i 96822Peter Wilf and Barbara Cariglino. Department of Geosciences, Pennsylvania State University, University
Park, Pennsylvania 16802Christopher H. Lusk, Ian J. Wright, and Mark Westoby. Department of Biological Sciences, Macquarie Uni-
versity, Sydney, New South Wales 2109, AustraliaGregory J. Jordan. School of Plant Science, University of Tasmania, Private Bag 55, Hobart 7001, AustraliaUlo Niinemets. Institute of Agricultural and Environmental Sciences, Estonian University of Life Sciences,
Tartu 51014, EstoniaPhyllis D. Coley. Department of Biology, University of Utah, Salt Lake City, Utah 84112Asher D. Cutter,† Conrad C. Labandeira, and Matthew B. Palmer. Department of Paleobiology, Smithsonian
Institution, Washington, D.C. 20013Kirk R. Johnson. Department of Earth Sciences, Denver Museum of Nature and Science, Denver, Colorado
80205Angela T. Moles.‡ Department of Biological Sciences, Macquarie University, Sydney, New South Wales
2109, AustraliaFernando Valladares. Centro de Ciencias Medioambientales, CSIC, E-28006 Madrid, Spain* Present address: Department of Ecology and Evolutionary Biology, University of California, Los Angeles†Present address: Department of Ecology and Evolutionary Biology, University of Toronto, Toronto, On-
tario M5S 3G5, Canada‡Present address: School of Biological, Earth, and Environmental Sciences, University of New South Wales,
New South Wales 2052, Australia
Accepted: 25 April 2007
IntroductionMany leaf traits strongly influence ecosys-
tem function (Dıaz et al. 2004; Wright et al.2004; Poorter and Bongers 2006; Shipley et al.
2006; Parton et al. 2007), but few have beenquantifiable from the fossil record. Amongthese traits, leaf dry mass per area (MA; alsocommonly abbreviated as LMA; MA is the in-
575FOSSIL LEAF ECONOMICS
FIGURE 1. Geographic and climatic distribution of cal-ibration sites. A, Geographic distribution of the 65 sitesused in the calibration data. Black symbols representsites where five or more species were sampled; graysymbols represent sites where four or fewer specieswere sampled (see ‘‘Materials and Methods’’). B, Cli-mate information and major biome type (Whittaker1975) for the calibration sites. SF � seasonal forest; WL� woodland; SL � shrubland. Biome boundaries areonly approximate and do not encompass all samples.Symbols follow panel A. See Appendices 1 and 2 for fur-ther details about sites.
verse of specific leaf area) is a key variable rep-resenting the dry mass cost of deploying pho-tosynthetic surface (Reich et al. 1997; Westobyet al. 2002). Species investing in a high MA
tend to have lower mass-based photosyntheticrates but longer leaf lifetimes (LL), such thattheir lower revenue (fixed carbon) per timemay be compensated by a longer-lasting rev-enue stream (Reich et al. 1997; Westoby et al.2002; Wright et al. 2004). Leaves with higherMA are more expensive to construct per unitarea, generally operate at lower nitrogen andphosphorus concentrations per unit mass,have slower rates of dark respiration, and arebetter defended against herbivory owing totheir greater thickness and/or toughness(Small 1972; Reich et al. 1997; Westoby et al.2002; Dıaz et al. 2004; Wright et al. 2004).These coordinated trade-offs form a ‘‘leaf eco-nomics spectrum’’ (Wright et al. 2004), whichrepresents one component of a general contin-uum running from specialization for rapid re-source acquisition (‘‘fast-return’’ species) to astrategy that maximizes resource retention(‘‘slow-return’’ species) (Grime 1974; Grubb1998). Leaf mass per area is also correlatedwith growth rates and the turnover of plantparts, and the influence of MA persiststhrough leaf ‘‘afterlife effects’’ into ecosystemprocesses including decomposition of litter(Kazakou et al. 2006) and mineralization of ni-trogen and phosphorus (Kobe et al. 2005).
Insect herbivory can be measured directlyfrom leaf fossils (Beck and Labandeira 1998;Labandeira 1998; Wilf and Labandeira 1999;Wilf et al. 2001, 2005), but the fundamentalleaf economic traits that influence herbivoryhave been difficult to quantify from fossils.Several methods for estimating LL for fossilspecies exist, but three of these, comparisonwith nearest living relatives (Chaloner andCreber 1990), leaf thickness (Chaloner andCreber 1990), and presence of leaf mats (Spicerand Parrish 1986), are qualitative; at best theycan distinguish deciduous from evergreen leafhabits (e.g., Wolfe 1987; Wolfe and Upchurch1987). A fourth method quantifies LL from thewood anatomy of conifers (Falcon-Lang2000a,b; Brentnall et al. 2005), but this methodis not yet applicable to angiosperms.
Here we analyze an extant data set drawn
from geographically widespread and climati-cally diverse sites (Fig. 1) to develop a newmethod for quantifying MA rapidly and ac-curately from the sizes and shapes of leaves.We assess several models for estimating MA
from petiole width (PW), leaf area (A), andleaf length, variables chosen because they canbe easily measured on most well-preservedleaf fossils. For example, although petiolelength has important biomechanical proper-ties (Niklas 1994), it is much less frequentlyavailable from fossils, because of incomplete-ness, than petiole width. We report results(Fig. 2) using the model
576 DANA L. ROYER ET AL.
FIGURE 2. Scaling relationship between petiole width(PW) and leaf dry mass per area (MA) for extant data. A,Calibration data for woody angiosperms. Solid andopen symbols represent species-site pairs that camefrom sites where five or more species and four or fewerspecies were sampled, respectively (see ‘‘Materials andMethods’’); triangles represent means for sites whereten or more species were sampled. Linear regression forspecies (solid black line) is log[MA] � 3.070 � 0.382 �log[PW2/A]; thin lines represent �95% prediction in-tervals (Sokal and Rohlf 1995). Linear regression forsites (gray line) is log[MA] � 3.214 � 0.429 � log[PW2/A]. B, Preliminary scaling relationship for broad-leavedspecies with distinct petioles from several gymnospermfamilies. Species-site pairs are plotted. The followinggenera are represented: Agathis, Gnetum, Podocarpus,Phyllocladus (cladodes), Saxegothaea, Torreya, and Taxus .The gray symbols correspond to the angiosperm data inpanel A. All relationships are significant at the familylevel using log-log linear regression except Taxaceae(Araucariaceae: r2 � 0.49, F1,7 � 5.85, p � 0.05; Gneta-ceae: r2 � 0.92, F1,3 � 22.2, p � 0.04; Podocarpaceae: r2 �0.44, F1,8 � 5.45, p � 0.05; Taxaceae: r2 � 0.09, F1,3 � 0.20,p � 0.70).
FIGURE 3. Representative examples of fossil specimensused in study. The specimen in panel A (Alnus parvifolia,Republic) has a narrower petiole (5.3 mm; see whiteline) and larger leaf area (442.8 mm2) than the petioluleof the specimen in panel B (Caesalpinia pecorae, Bonanza;petiole width � 9.7 mm; leaf area � 191.2 mm2); this re-sults in a lower estimate of leaf dry mass per area for theA. parvifolia specimen (70.8 g m�2) than the C. pecoraespecimen (154.3 g m�2). Scale bars, 1 cm. The black linein the A. parvifolia specimen represents a conservativereconstruction of the leaf-margin segment that was notpreserved. See ‘‘Materials and Methods’’ for proceduraldetails on how petioles were measured.
2PWlog(M ) � a � b log . (1)A � �A
This model corresponds to the proposition,based on biomechanical and developmentalprinciples, that the cross-sectional area of thepetiole scales with the mass of the leaf. Thisrelationship is expected because the petiole is
important in the mechanical support of theleaf (Salisbury 1913; Niklas 1994; see ‘‘ModelFitting and Justification’’).
We apply our method to 187 fossil leavesfrom two Eocene fossil floras (Republic, Wash-ington, Klondike Mountain Formation; andBonanza, Utah, Green River Formation) (Figs.3, 4) where well-understood systematics(MacGinitie 1969; Wolfe and Wehr 1987), pa-leoclimate (MacGinitie 1969; Wolfe and Wehr1987; Wing and Greenwood 1993; Wilf et al.1998; Greenwood et al. 2005), and herbivory(Wilf et al. 2001, 2005; Labandeira 2002) allowtestable hypotheses. Republic is considered tobe dominated by deciduous species (Wolfeand Wehr 1987); thus, our hypothesis is thatthese species have low reconstructed MA. Bo-
577FOSSIL LEAF ECONOMICS
FIGURE 4. Correlation between insect herbivory and es-timated leaf dry mass per area (MA) for two Eocene fos-sil floras. Each data point represents a species mean, anderrors in MA represent �95% prediction intervals (Sokaland Rohlf 1995). Only species where �23 specimenscould be scored for insect herbivory are plotted. A, In-sect damage morphotypes (Wilf et al. 2001, 2005) versusMA; errors in herbivory represent �1�. Statistics of log-log linear regression for combined data: n � 18; r2 �0.64; F1,16 � 28.0; p � 0.0001. B, Percentage of leaves withinsect damage (Wilf et al. 2001, 2005); errors in herbiv-ory represent �1� of the binomial sampling error. Sta-tistics of log-log linear regression for combined data: n� 18; r2 � 0.67; F1,16 � 32.6; p � 0.0001. C, Percentage ofleaf area removed by insect damage; errors in herbivoryrepresent standard errors. Statistics of log-log linear re-gression for combined data: n � 15; r2 � 0.68; F1,13 � 27.4;p � 0.0001.
nanza putatively contains a mix of specieswith both short and long LL (MacGinitie 1969;Wilf et al. 2001); these interpretations werebased on qualitative methods discussed above
but correctly predicted the bimodal distribu-tion of herbivory observed at the site (Wilf etal. 2001). Thus, we hypothesized a broadrange of estimated MA values at Bonanza. Wecompare our reconstructions of MA with qual-itative observations of the floras and with di-rect measurements of insect herbivory, anduse them to refine understanding of plant andsite ecology as well as forest nutrient cyclingrates for these classic fossil floras.
Materials and Methods
Calibration Sites. To reconstruct MA fromleaf fossils, we first collected leaves to createan extant calibration from 667 species-sitepairs representing 468 species of woody an-giosperms from 65 geographically and cli-matically diverse sites (Fig. 1). We sampled1–20 mature, representative leaves or equiva-lent photosynthetic organs (phyllodes) (me-dian � 3; 88% of species-site pairs are basedon two or more leaves) from each of 5–86 spe-cies (median � 21) at 26 sites (Fig. 1A). Tobroaden our geographical coverage, we alsosampled 4–12 leaves (median � 10) from eachof one to four species at 39 additional sites(Fig. 1A; Appendices 1–2; Appendix A onlineat http://dx.doi.org.10.1666/pbio07001.s1).In aggregate, the sites represent most of themajor biomes where the foliage of woody an-giosperms is likely to be fossilized (Fig. 1B).We collected native, woody angiosperm spe-cies exclusive of monocots. We generally re-stricted our sampling to outer, exposed can-opy leaves (Appendices 1, 2) because they con-stitute the majority of leaf fossils (Spicer 1981).Leaves without obvious, distinct petioles wereexcluded. Herbaceous species were also ex-cluded because they rarely fossilize (Spicer1981).
Fossil Sites. We reconstructed MA and mea-sured insect herbivory for woody dicot spe-cies from two fossil lake floras. The first flora,Republic (Wolfe and Wehr 1987; Radtke et al.2005), is from the Klondike Mountain Forma-tion in northeastern Washington, U.S.A., andis late early Eocene in age (ca. 49 Ma [reportedin Radtke et al. 2005]). The climate at Republicis interpreted as humid and warm temperate(mean annual temperature [MAT] � 13C;mean annual precipitation [MAP] 1000
578 DANA L. ROYER ET AL.
mm) (Wolfe and Wehr 1987; Greenwood et al.2005). The second flora, Bonanza (MacGinitie1969), is from the uppermost Green River For-mation in northeastern Utah, U.S.A., and isearly middle Eocene in age (47.3 Ma [Smith etal. 2007]). In contrast to Republic, the climateat Bonanza has been interpreted as warmerand more seasonally dry (MAT � 15C; MAP� 840 mm) (MacGinitie 1969; Wing andGreenwood 1993; Wilf et al. 1998). The 187specimens with measurable leaf area and pet-iole width were selected from recent unbiasedcensus collections made by K.R.J. of 1019 dicotleaves at Republic and 894 at Bonanza, re-ported by Wilf et al. (2001, 2005). Both floraswere collected from single stratigraphic hori-zons (thickness of sampled horizons � 1.6 mand 0.1 m for Republic and Bonanza, respec-tively [Wilf et al. 2001, 2005]).
Leaf Measurements for Quantifying MA. Wemeasured petiole width (PW) perpendicularto the midvein in the plane of the leaf blade,at the basal-most insertion of the lamina intothe petiole. If the position of this measurementcorresponded to a locally thickened or wingedregion of the petiole, PW was measured justbasal to the feature. Leaflets and petioluleswere the units measured for compoundleaves, and phyllodes and basal attachmentsfor phyllodes; for our data set, simple leavesand leaflets did not differ in their scaling re-lationship between MA and PW2/A (slope: p �0.44; y-intercept: p � 0.45; likelihood ratiomethod of Falster et al. 2003). For a subset ofleaves from our calibration data set, we alsomeasured PW at the thinnest point and themidpoint of the petiole, but these alternativemeasurements did not yield improved corre-lations and tended to be highly correlatedacross species. Because complete petioles withbases are only rarely preserved, our protocolallows measurement of a greater number offossils than these alternatives. It is possiblethat a combination of petiole width and depthcorrelates more strongly with MA than doesPW alone, but original petiole depth is rarelypreserved in compressed fossils (Niklas 1978;Rex 1986).
We measured PW and leaf length with cal-ipers, often using clear acetate sheets for fos-sils to protect surfaces, or from high-resolu-
tion digital images (600 dpi minimum); leafarea was determined from digital images. Forthe extant calibration data, we calculated MA
from the dry mass and area of the leaf bladeand petiole (Cornelissen et al. 2003). Leaf massper area varies 30-fold and leaf area 3.5 ordersof magnitude in the calibration data; world-wide, MA varies about 50-fold and leaf areafive orders of magnitude (Wright et al. 2004).Our calibration data thus capture the majorityof the known variation in these variables. Giv-en the biomechanical basis for the scaling re-lationship (see ‘‘Model Fitting and Justifica-tion’’), PW could be better optimized for freshthan dry leaf mass. However, in a subset of 98species-site combinations, there was a strongcorrelation across species between dry massand fresh mass (r2 � 0.96; F1,96 � 2068; p �0.0001). Thus, for this subset, there was littledifference in the strength of correlation be-tween PW2/A and MA calculated on a fresh ordry mass basis (r2 � 0.81 and 0.77 for freshand dry mass, respectively); moreover, theslopes of the correlations were not significant-ly different (p � 0.72; likelihood ratio methodof Falster et al. 2003).
For PW measurements of fossils, only spec-imens where the petiole was clearly and com-pletely preserved at the point of measurementfor PW, described above, were used (Fig. 3;Appendix B online at http://dx.doe.org.10.1666/pbio07001/s2). One hazard with fossilpetioles is longitudinal splitting, creating thefalse appearance of a thin petiole; thus, fossilpetioles were inspected under binocular mi-croscopes to ensure that both petiole marginswere preserved before measurement at mag-nification. For specimens with partially pre-served leaf blades, only those specimenswhose full leaf areas could be reconstructedwith reasonable confidence were considered.Species represented by only one specimenwere excluded.
A potential error with fossils is that theirmorphology can change postmortem. How-ever, previous experiments that mimicked thefossilization process indicated little to nochange in the two-dimensional shape of leafblades and at most a 10% inflation in thewidth of xylem-rich tissues, such as petioles,that are buried in fine-grained sediment (Wal-
579FOSSIL LEAF ECONOMICS
FIGURE 5. Rate of leaf area removed by insect damage(Coley 1983) versus leaf dry mass per area (MA) for pre-sent-day vegetation (saplings) at Barro Colorado Island,Panama. Errors represent standard errors. Statistics oflog-log linear regression: n � 44; r2 � 0.36; F1,42 � 24.1;p � 0.0001. The offset to lower MA in these data relativeto the fossil reconstructions (Fig. 4) is a consequence ofsaplings having leaves with a lower MA than matureplants (Thomas and Winner 2002), and mature plantsconstitute the bulk of fossil plant deposits (Spicer 1981).The offset in herbivory relative to the fossil measure-ments (Fig. 4) is a consequence of the fragmentary na-ture of fossil leaves (see ‘‘Materials and Methods’’ forfurther details). Importantly, the dividing line in thePanama data between leaves that are highly damaged byinsects and those that are not corresponds to an MA of�50 g m�2, or a leaf life span (LL) of �12 months (95%of species with an MA �51.5 g m�2 have a LL of �12months, whereas 87% of species with an MA 51.5 g m�2
have a LL of 12 months). This relationship betweenherbivory and LL is consistent with the fossil data (see‘‘Results and Discussion’’) and further emphasizes thatthe fossil and Panama data sets are compatible.
ton 1936; Niklas 1978; Rex and Chaloner 1983;Rex 1986) such as the two fossil localities stud-ied here (MacGinitie 1969; Wolfe and Wehr1987); a 10% inflation of PW would lead to a7.6% overestimation of MA.
Leaf Measurements for Quantifying Insect Her-bivory. High rates of insect herbivory gener-ally correlate with trait values towards the‘‘fast-return’’ end of the leaf economics spec-trum, including high foliar nitrogen concen-tration and short LL (Coley 1983; Westoby etal. 2002); herbivory is predicted to inverselycorrelate with MA, but this has rarely been di-rectly tested in extant vegetation (Moles andWestoby 2000; Poorter et al. 2004) and neverbefore tested in fossil vegetation. To test forthe hypothesized negative correlation be-tween herbivory and MA, we compared pub-lished data on insect herbivory (Coley 1983) toMA determined from saplings of the same spe-cies in a present-day tropical forest on BarroColorado Island, Panama (Fig. 5).
Insect damage morphotypes and percent-age of specimens with insect damage werepreviously tabulated for the Republic (Wilf etal. 2005) and Bonanza (Wilf et al. 2001) fossilfloras. Only species for which �23 specimenscould be scored for insect herbivory were in-cluded here; this sample size represents acompromise between an adequate samplinglevel for precise results and the inclusion ofenough species to establish reliable site-leveltrends. To account for uneven sampling acrossspecies, insect damage morphotype data wererandomly subsampled to 23 specimens 5000times without replacement (Wilf et al. 2001),and the means of these subsamples are re-ported here (Fig. 4A). Both floras were scoredfor percentage of leaf area lost to insect dam-age following the method of Beck and Laban-deira (1998) (n � 1019 and 894 leaves for Re-public and Bonanza, respectively; only thosespecies where �23 specimens could be scoredfor insect herbivory were included in the tal-ly); species means were based on the arcsinetransformation of individual leaves (Sokal andRohlf 1995). We consider these measurements(Fig. 4C) minima because areas of the leaf thatwere not preserved, and that therefore mayhave been damaged or entirely removed by in-sects, cannot be analyzed.
Model Fitting and Justification
We fit several models (Table 1) to the ob-served relationships (Appendix A online) be-tween MA, PW, and other leaf dimensions.Previous work in two species has shown thatpetiole cross-sectional area correlates withsupported mass and area within species (Nik-las 1991a; Yamada et al. 1999). Additionally,several studies have examined relationshipsamong petiole biomechanical properties with-in and across species (Niklas 1991a,b, 1994,1999). Our study is the first to our knowledgeto demonstrate general scaling between peti-ole and lamina dimensions across diverse spe-cies, and to develop from these interrelation-ships a prediction of MA. Scaling might be in-fluenced by hydraulic supply as well as by me-chanical support because the petiole deliversthe transpiration stream to the leaf, and peti-
580 DANA L. ROYER ET AL.
TABLE 1. Models fitted. All models are based on individual species-site samples (n � 667). Both leaf dry mass perarea (MA; units in g m�2) and the predictor variables are handled on log scales to allow the use of power law al-lometries, and because variance increases with the mean whereas after log transformation scatter is more normallydistributed (Fig. 2A). All models are fitted using linear regression in order to minimize the sum-of-squares in they-dimension (i.e., MA), to facilitate retrodiction with fossils. This contrasts with the standardized major axis (SMA)estimation (also known as Model II regression, geometric mean regression, or reduced major axis), where errors inboth the x- and y-dimensions are minimized simultaneously (Falster et al. 2003; Warton et al. 2006); we use SMAto investigate the slopes of allometric relationships (see Model Fitting and Justification). The logic of each model isexplained in Model Fitting and Justification; Model E corresponds to equation (1) in the text. PW � petiole width(mm); A � one-sided projected area of leaf (mm2); L � leaf length (mm); N/A � not applicable because SMA cannotbe calculated for multivariate models.
Model a b c r2 Slope* SMA
A log(MA) � a � log(PW2/A) 4.930 �0.89 0.13 1.00B log(MA) � a � log(PW8/3/A) 4.870 �0.92 0.001 1.00C log(MA) � a � blog(PW4/A) 2.740 0.289 0.47 1.13 0.42D log(MA) � a � blog(PW/A) 2.870 0.307 0.42 1.09 0.47E log(MA) � a � blog(PW2/A) 3.070 0.382 0.55 1.04 0.51F log(MA) � a � blog(PW) � clog(A) 3.064 0.983 �0.386 0.58 1.05 N/AG log(MA) � a � blog[(PW)/(L � A)] 2.876 0.194 0.39 1.10 0.32
* Slope of measured vs. estimated MA linear regression fixed through the origin.
ole cross-sectional area correlates with xylemvessel area and with petiole hydraulic con-ductance per leaf area for leaves of a givenspecies (Salisbury 1913; Sack et al. 2002, 2003).However, across distantly related species, pet-iole cross-sectional area per leaf area does notnecessarily correlate with petiole or leaf hy-draulic conductance per leaf area becauseboth the numbers and sizes of xylem conduitswithin petioles vary strongly (Nardini et al.2005; Sack and Frole 2006).
Here we discuss the possible underlying bi-ology of the models, the statistical strengths oftheir fitting to the data, and possible interpre-tations of the observed scaling coefficients. Wenote that although our models consider peti-ole length implicitly as described below, we donot explicitly include petiole length as a pre-dictive variable because complete fossil peti-oles are rare. Also, we recognize that windload may affect scaling relationships betweenMA and petiole dimensions because plants inwindy habitats can have higher MA and small-er petiole cross-sectional area to allow easierbending and twisting for reducing drag (Nik-las 1996, 1998). However, because this adap-tation would lead to the opposite trend doc-umented here (Fig. 2), wind load is likely ofonly minor importance. Lastly, the relation-ships in this study were determined across di-verse species, but they have yet to be testedwithin species.
Model A is based on a simple scaling rela-
tionship between the ratio of the square of pet-iole cross-sectional area to leaf area versusMA. This relationship, MA � PW2/A, wherePW � petiole width and A � leaf area, wouldbe expected if the leaf behaved as a mass ap-plied to a vertical petiole that was just suffi-cient to support it. Preservation of compres-sive strength to maintain resistance to buck-ling then leads to the expectation of PW2 � M,where M � leaf mass, and thus MA � PW2/A.This scaling treats petiole length as varyinglittle, or at least independently of leaf size. Al-ternatively, if petiole length and width are co-optimized, a slightly different scaling follow-ing ‘‘elastic similarity’’ as for animal legsmight be expected (McMahon and Bonner1983; Peters 1983; Schmidt-Nielsen 1984), withPW8/3 � M (and MA � PW8/3/A; Model B).
Leaf mass is only rarely incident on a ver-tical petiole; instead, leaves are usually bettermodeled as end-loaded cantilevered beams(Niklas 1991b, 1999). Under this scenario, ifthe petiole supports the leaf mass with a fixeddeflection distance, at a given wind-load, andwithout leaf shape and petiole compositionand mechanical properties being influentialvariables, the expected relation is M � 3 EI/PL3, where E is the petiole elastic modulus, Ithe second moment of area of the petiole, theproduct EI the petiole flexural rigidity, and PLthe petiole length (Niklas 1991b, 1994, 1999).Indeed, previous work has shown that PL3
scales with EI as expected from the cantilever
581FOSSIL LEAF ECONOMICS
model across a diverse range of leaves (Niklas1991b, 1994, 1999). Model C applies this sce-nario, assuming PW to be independent of PL,and petiole shape to be relatively invariant; inthis case, M would be proportional to EI andI would be related to PW4, and MA � PW4/A.Model D applies the same scenario, addition-ally assuming that PW � PL; in this case, thecantilevered beam model simplifies to M �3(EI/PL3) � 3(E � PW4/PL3) � PW, and MA �PW/A.
Models E–G represent additional scenarios,with greater flexibility. Model E preserves theexpectation of the scaling of PW2 with M, asin Model A (and as observed to hold withingiven species, as discussed above), but allowsan allometric scaling, MA � (PW2/A)b. ModelF modifies Models C–E by allowing the ex-ponents to vary independently. Model G mod-ifies Model D by including leaf length as anadditional factor, reflecting the extra leverageof a given mass that is farther from the attach-ment point of the leaf.
The fits of Models C–G indicate a strongscaling of MA with petiole and lamina dimen-sions (Table 1); however, the fitted parametersdo not fit simply with many of the expecta-tions discussed above. For example, ModelsC–E and G show slopes b substantially lowerthan the expectations for a slope of 1, as de-termined by a standardized major axis (SMA)estimation (Falster et al. 2003). Further, allmodels indicate that PW relative to leaf area isinordinately high for leaves of high MA rela-tive to what simple support requirementswould require, under any of the above scenar-ios. This could be one explanation for the pre-viously demonstrated result that petiole flex-ural rigidity (EI) increases more strongly withleaf mass (M) than is predicted from the can-tilever model (EI � M with an exponent of 1.6–2.3 for diverse species sets [Niklas 1991a]).The disproportionate PW relative to leaf areafor leaves of larger MA and the consequentlyhigher petiole flexural rigidity would contrib-ute greater support stability given that thelaminar center of mass could be displacedover larger petiolar second moment of area.Such investment in greater safety is consistentwith the investment in greater constructioncost for leaves of higher MA, and their gener-
ally longer life spans (Villar and Merino 2001;Wright et al. 2004).
Model E, corresponding to equation (1), wasused for estimation because of its relativelyhigh goodness of fit (r2 � 0.55 for speciesmeans) and its low bias (the slope of the plotfor measured versus estimated MA is 1.04; Ta-ble 1). This model also has the advantage ofbeing a simple expression of the allometricscaling of petiole and lamina dimensions asdiscussed above. The parameters of Model Eindicate that an approximation of petiolecross-sectional area relative to leaf area scalesstrongly with MA, with SMA slope of 0.51(�0.026 95% confidence intervals); this modelis most compatible with petioles with circularand square cross-sections (kX2, where k is aconstant and X is the length of the side of asquare or the radius of a circle), however amixture of cross-sectional shapes will de-crease somewhat the predictive power of themodel. Model F has a slightly higher r2-valuethan Model E (0.58; Table 1), but this extra ex-planatory power is largely due to Model F’shaving an additional parameter.
Calculating Errors for MA Estimates. The cal-culation of 95% prediction intervals (PI) fol-lows Sokal and Rohlf (1995):
log PI
⎧ ⎫⎡ ⎤⎪ ⎪2¯⎢ ⎥1 1 (X � X )i2⎨ ⎬⎢ ⎥� log M � s � �⎪ � ⎪A Y·X ⎢ ⎥2k n x�⎩ ⎭⎣ ⎦
� t , (2)0.05[n�2]
where sY·X2 � unexplained mean square, k �
size of unknown sample, n � sample size ofcalibration data, Xi � mean log(PW2/A) of un-known sample, X � mean log(PW2/A) of cal-ibration data, x2 � sum of squares of cali-bration data, and t0.05[n�2] � critical value ofStudent’s distribution for (n � 2) degrees offreedom. Table 2 provides the necessary in-formation for calculating 95% PIs (for speciesand sites) from the regressions presented inFigure 2A; errors are asymmetric with respectto means because the regressions are based onlogarithmic relationships.
Results and Discussion
Testing Extant Vegetation. Fitting Model Eto our calibration data shows that the MA of
582 DANA L. ROYER ET AL.
TABLE 2. Parameters used to calculate 95% predictions intervals for estimates of leaf dry mass per area (MA).� unexplained mean square, n � sample size of calibration data, X � mean log(PW2/A) of calibration data, x22SY·X
� sum of squares of calibration data, and t0.05[n � 2] � critical value of Student’s distribution for (n � 2) degrees offreedom. See ‘‘Materials and Methods’’ for details.
S2Y·X n X x2 t0.05[n�2]
Species means 0.032237 667 �3.011 182.1 1.964Site means 0.005285 25 �2.857 5.331 2.069
individual species is estimated to a significantdegree (Fig. 2A; n � 667, r2 � 0.55, F1,666 � 825,p � 0.0001): 95% prediction intervals (Sokaland Rohlf 1995) are �� of observed values,60%
38%
assuming a sample size of three leaves. Thiserror is small compared to the �30-fold rangeobserved across species.
Estimates of MA are unbiased: the regres-sion slope of the measured versus estimatedMA is close to unity (1.04 � 0.04 95% confi-dence intervals). Mean values for sets of spe-cies at sites can be estimated very precisely be-cause of the lack of bias and increased samplesize (Fig. 2A; n � 25, r2 � 0.89, F1,24 � 186, p� 0.0001; 95% prediction intervals for individ-ual sites are �� of their observed values,16%
14%
assuming a sample size of ten species). Pre-liminary data from broad-leaved gymno-sperms (n � 25 species) match the correlationas well (Fig. 2B), suggesting applications thatinclude the pre-angiosperm record.
We tested whether MAT or MAP modulatedthe relationship between MA and PW2/A inour calibration data, using partial correlation.The relationship between PW2/A and MA re-mains significant and largely unchanged (fullcorrelation: r � 0.74; correlation after account-ing for MAT and MAP: r � 0.75 and 0.74, re-spectively; n � 667 and p � 0.0001 for bothtests). This insensitivity to environmental con-ditions contrasts with many other paleoeco-logical and paleoclimatological proxies (Royeret al. 2002) and reinforces the notion thatPW2/A is a faithful recorder of MA. Moreover,interrelationships among leaf economic vari-ables such as MA and LL are not strongly mod-ulated by phylogeny (Ackerly and Reich 1999);this is important for paleobiological studies,where fossil taxa may be extinct, be only dis-tantly related to taxa in the calibration data, orhave unknown affinities.
Because leaf economic traits are strongly in-
tercorrelated (Reich et al. 1997; Westoby et al.2002; Wright et al. 2004, 2005), our method hasthe potential to predict other traits. For ex-ample, in a worldwide compilation of leaf eco-nomic information (Wright et al. 2004), a MA
of 129 g m�2 for woody angiosperms corre-sponded to a mean LL of 12 months. We de-termined this LL category (�12 or 12months) for a subset of our data (n � 496 spe-cies-site pairs). A PW2/A of 0.0011, corre-sponding to an estimated MA of 87 g m�2, cor-rectly predicts the LL category 85% of the timein our calibration data. We adopt these MA val-ues to broadly distinguish between the short-lived ‘‘fast-return’’ (��87 g m�2) and long-lived ‘‘slow-return’’ (�129 g m�2) ends ofthe leaf economic spectrum.
Application to Fossil Record. We hypothe-sized low MA at Republic and a broader rangeof MA at Bonanza on the basis of previouslypublished, qualitative interpretations (see ‘‘In-troduction’’). Consistent with hypotheses, re-sults from Republic show domination by low-MA species (57–87 g m�2; Table 3). Conse-quently, it is likely that most of these specieshad leaf life spans of �12 months, suggestingthe presence of a deciduous forest among theangiosperms. Also consistent with hypothe-ses, the Bonanza flora shows a broader mix ofMA values (70–157 g m�2; Table 3). The mostabundant species at Bonanza have high MA
values (Table 3), suggesting that the vegeta-tion was dominated by species with long-livedleaves. The site mean of MA among angio-sperms is significantly higher at Bonanza thanRepublic (113.2 � versus 76.8 � g m�2; er-14.5 9.2
12.9 8.2
rors represent 95% prediction intervals; t1,14 �3.54, p � 0.003), and Bonanza is associatedwith a higher coefficient of variation (23.8%versus 12.8%; t1,17 � 2.85, p � 0.01 after arcsinetransformation; Sokal and Rohlf 1995).
Thus, these floras had very different ecolog-
583FOSSIL LEAF ECONOMICS
TABLE 3. Reconstructions of leaf dry mass per area (MA) for measurable species in the Republic and Bonanza fossilfloras. MA estimates for the Fabaceae (legumes) may be somewhat too high because of their short, pulvinulate pet-iolules.
n Abundance* (%) MA** (g m�2)
REPUBLICAlnus parvifolia (Betulaceae) 37 44.2 82.5Betula leopoldae (Betulaceae) 5 3.1 72.0Cercidiphyllum obtritum (Cercidiphyllaceae) 17 12.6 81.7Cornus sp. (Cornaceae) 2 0.8 57.3aff. Crataegus sp. (Rosaceae) 3 0.5 79.8Crataegus sp. (Rosaceae) 2 3.2 84.8Ericaceae sp. 2 0.4 84.0Itea sp. (Saxifragaceae) 3 2.0 77.9Macginitiea gracilis (Platanaceae) 2 1.8 69.2Photinia pageae (Rosaceae) 2 2.0 63.2Prunus sp. (Rosaceae) 2 0.6 68.4Rhus malloryi (Anacardiaceae) 7 2.9 85.6Sassafras hesperia (Lauraceae) 4 4.8 80.6Spiraea sp. (Rosaceae) 8 2.5 86.7Ternstroemia sp. (Theaceae) 2 0.2 87.0Ulmus sp. (Ulmaceae) 8 8.1 67.0Zelkova sp. (Ulmaceae) 2 0.5 85.9Zizyphoides flabella (Trochodendraceae) 2 1.7 63.0
BONANZAAllophylus flexifolia (Sapindaceae) 4 4.9 73.6Caesalpinia pecorae† (Fabaceae) 5 4.4 133.2Cardiospermum coloradensis† (Sapindaceae) 2 3.8 96.7Cedrelospermum nervosum‡ (Ulmaceae) 18 26.7 118.3Leguminosites lesquereuxiana (Fabaceae) 6 2.0 103.3Macginitiea wyomingensis‡ (Platanaceae) 7 5.3 70.0Parvileguminophyllum coloradensis† (Fabaceae) 5 33.1 156.6Populus tidwellii‡ (Salicaceae) 2 2.6 98.9Populus wilmattae‡ (Salicaceae) 3 2.9 82.7Rhus nigricans (Anacardiaceae) 13 7.2 115.2Salix cockerelli‡ (Salicaceae) 6 2.7 96.1Styrax transversa (Styracaceae) 4 0.8 72.2Syzygioides americana (Myrtaceae) 2 1.0 137.2
* Based on unbiased field census collections (Wilf et al. 2001, 2005). Values within floras do not sum to 100% because not all species are representedhere (see ‘‘Materials and Methods’’).
** We interpret leaves with MA values of �87 g m�2 to have leaf life spans of �1 year, and leaves with MA values 129 g m�2 to have leaf life spansof 1 year (see ‘‘Results’’ and ‘‘Discussion’’).
† Inferred by Wilf et al. (2001) to have long leaf life spans.‡ Inferred by Wilf et al. (2001) to have short leaf life spans.
ical structuring among woody angiosperms.Republic was dominated by species associatedwith relatively rapid mass-based rates of gasexchange and more rapid litter decomposi-tion, whereas Bonanza was dominated by‘‘slow-return’’ species, but with an importantsecondary component of ‘‘fast-return’’ spe-cies. This difference may have been driven bythe seasonally drier climate at Bonanza, a pat-tern consistent with observations in present-day vegetation of greater variance in MA inseasonally dry forests relative to moister for-ests (Niinemets 2001; Wright et al. 2005). Be-cause litter decomposition rates influence nu-trient turnover rates (Kobe et al. 2005) and re-gional biogeochemical cycling (Chapin 2003),
we infer that forest-wide nutrient cyclingamong woody angiosperms was probablymore rapid at Republic than Bonanza.
By quantifying the frequency, amount, anddiversity of insect damage on leaves from Re-public and Bonanza, we directly correlatedMA (and LL by extension) to insect herbivory(Fig. 4). We hypothesized a negative relation-ship because leaves with high MA are typicallyassociated with greater thickness and/ortoughness, higher amounts of chemical toxinsand/or other chemical deterrents, and lowerfoliar nitrogen concentrations (Coley 1983;Westoby et al. 2002); all of these characteristicshelp to minimize insect damage. At Republic,where all species have an estimated MA of �87
584 DANA L. ROYER ET AL.
g m�2, insect damage ranges from moderatelyhigh to very high (Fig. 4). In contrast, at Bo-nanza there is a greater range in MA and her-bivory levels, and these properties negativelycorrelate with one another. For both floras, thedividing line between species that are highlydamaged and those that are not correspondsto an MA of 90–100 g m�2 (Fig. 4), or an in-ferred LL of �12 months. These Eocene resultsare consistent with our predictions and withpatterns observed in an extant forest (Fig. 5),suggesting that strong insect selection of leaffunctional traits is of great antiquity.
Conclusions
An extant calibration indicates that leafmass per area can be easily reconstructed forfossils from the measurement of petiole widthand leaf area. This represents, to our knowl-edge, the first proxy for fossil MA. Some keyadvantages of the method include the follow-ing: the required measurements can be madequickly and accurately on most well-pre-served leaf fossils; the statistical errors for pre-dicting MA are at most � if three or more60%
38%
leaves are measured per species; and the bio-mechanical scaling relationship is not stronglymodulated by factors that can be difficult toevaluate in the fossil record (e.g., temperature,rainfall). Importantly, a preliminary analysissuggests that the method may also be appli-cable for some gymnosperm groups.
We quantified MA for 31 species in two well-understood Eocene floras, and we comparedthese estimates with measurements of insectherbivory and qualitative inferences of otherleaf economic variables for the same species.At Republic, where most species have an in-ferred short LL, we reconstructed low MA; atBonanza, where there is a broader range in in-ferred LL, we reconstructed a broader range inMA (Table 3). At both sites, there is a statisti-cally significant, inverse correlation betweenMA and insect damage (Fig. 4). Together, theseresults demonstrate a consistent, emergentpattern: Republic was dominated by ‘‘fast-re-turn’’ species, whereas Bonanza was domi-nated by ‘‘slow-return’’ species but with animportant secondary component of ‘‘fast-re-turn’’ species. More broadly, our results high-light the potential for quantifying leaf eco-
nomic information, including important as-pects of plant-animal interactions and con-straints on nutrient cycling rates, fromlesser-known floras.
Acknowledgments
This work arose from a working group of theARC-NZ Research Network for VegetationFunction, supported by the Australian ResearchCouncil. This work was also supported by thePetroleum Research Fund of the AmericanChemical Society grant 40546-AC8, the PennState Institutes of Energy and the Environment,a Macquarie University New Staff GrantA007220, the Estonian Academy of Sciences,and National Science Foundation grants DEB-0345750, EAR-0236489, and IOB-0546787. Wethank R. Burnham, N. Cellinese, D. Danehy, D.Dilcher, B. Ellis, P. Huff, C. Jander, E. Kowalski,T. Lott, E. Manzane, F. Marsh, D. Meade-Hunter,S. Passmore, M. Reynolds, C. Streeter, S. Trom-bulak, M. Wiemann, K. Wilson, S. Wing, G.Zotz, and the 45 members of the World Herbiv-ory Project for help in collecting and processingspecimens. We thank R. Burnham, K. Niklas,and J. Parrish for constructive reviews. This iscontribution 171 of the Evolutionary and Ter-restrial Ecosystems consortium at the NationalMuseum of Natural History.
Literature CitedAckerly, D. D., and P. B. Reich. 1999. Convergence and correla-
tions among leaf size and function in seed plants: a compar-ative test using independent contrasts. American Journal ofBotany 86:1272–1281.
Beck, A. L., and C. C. Labandeira. 1998. Early Permian insectfolivory on a gigantopterid-dominated riparian flora fromnorth-central Texas. Palaeogeography, Palaeoclimatology, Pa-laeoecology 142:139–173.
Brentnall, S. J., D. J. Beerling, C. P. Osborne, M. Harland, J. E.Francis, P. J. Valdes, and V. E. Wittig. 2005. Climatic and eco-logical determinants of leaf lifespan in polar forests of thehigh CO2 Cretaceous ‘greenhouse’ world. Global Change Bi-ology 11:2177–2195.
Chaloner, W. G., and G. T. Creber. 1990. Do fossil plants give aclimatic signal? Journal of the Geological Society, London 147:343–350.
Chapin, F. S. 2003. Effects of plant traits on ecosystem and re-gional processes: a conceptual framework for predicting theconsequences of global change. Annals of Botany 91:455–463.
Coley, P. D. 1983. Herbivory and defensive characteristics of treespecies in a lowland tropical forest. Ecological Monographs53:209–233.
Cornelissen, J. H. C., S. Lavorel, E. Garnier, S. Dıaz, N. Buch-mann, D. E. Gurvich, P. B. Reich, H. ter Steege, H. D. Morgan,M. G. A. van der Heijden, J. G. Pausas, and H. Poorter. 2003.A handbook of protocols for standardised and easy measure-
585FOSSIL LEAF ECONOMICS
ment of plant functional traits worldwide. Australian Journalof Botany 51:335–380.
Dıaz, S., J. G. Hodgson, K. Thompson, M. Cabido, J. H. C. Cor-nelissen, A. Jalili, G. Montserrat-Martı, J. P. Grime, F. Zarrin-kamar, Y. Asri, S. R. Band, S. Basconcelo, P. Castro-Dıez, G.Funes, B. Hamzehee, M. Khoshnevi, N. Perez-Harguindeguy,M. C. Perez-Rontome, F. A. Shirvany, F. Vendramini, S. Yaz-dani, R. Abbas-Azimi, A. Bogaard, S. Boustani, M. Charles,M. Dehghan, L. de Torres-Espuny, V. Falczuk, J. Guerrero-Campo, A. Hynd, G. Jones, E. Kowsary, F. Kazemi-Saeed, M.Maestro-Martinez, A. Romo-Dıez, S. Shaw, B. Siavash, P. Vil-lar-Salvador, and M. R. Zak. 2004. The plant traits that driveecosystems: evidence from three continents. Journal of Veg-etation Science 15:295–304.
Falcon-Lang, H. J. 2000a. A method to distinguish betweenwoods produced by evergreen and deciduous coniferopsidson the basis of growth ring anatomy: a new palaeoecologicaltool. Palaeontology 43:785–793.
———. 2000b. The relationship between leaf longevity andgrowth ring markedness in modern conifer woods and its im-plications for palaeoclimatic studies. Palaeogeography, Pa-laeoclimatology, Palaeoecology 160:317–328.
Falster, D. S., D. I. Warton, and I. J. Wright. 2003. (S)MATR: stan-dardised major axis tests and routines. Version 1.0. http://www.bio.mq.edu.au/ecology/SMATR.
Greenwood, D. R., S. B. Archibald, R. W. Mathewes, and P. T.Moss. 2005. Fossil biotas from the Okanagan Highlands,southern British Columbia and northeastern WashingtonState: climates and ecosystems across an Eocene landscape.Canadian Journal of Earth Sciences 42:167–185.
Grime, J. P. 1974. Vegetation classification by reference to strat-egies. Nature 250:26–31.
Grubb, P. J. 1998. A reassessment of the strategies of plantswhich cope with shortages of resources. Perspectives in PlantEcology, Evolution and Systematics 1:3–31.
Huff, P. M., P. Wilf, and E. J. Azumah. 2003. Digital future forpaleoclimate estimation from fossil leaves? Preliminary re-sults. Palaios 18:266–274.
Kazakou, E., D. Vile, B. Shipley, C. Gallet, and E. Garnier. 2006.Co-variations in litter decomposition, leaf traits and plantgrowth in species from a Mediterranean old-field succession.Functional Ecology 20:21–30.
Kobe, R. K., C. A. Lepczyk, and M. Iyer. 2005. Resorption effi-ciency decreases with increasing green leaf nutrients in aglobal data set. Ecology 86:2780–2792.
Kowalski, E. A., and D. L. Dilcher. 2003. Warmer paleotemper-atures for terrestrial ecosystems. Proceedings of the NationalAcademy of Sciences USA 100:167–170.
Labandeira, C. C. 1998. Early history of arthropod and vascularplant associations. Annual Review of Earth and Planetary Sci-ences 26:329–377.
———. 2002. Paleobiology of middle Eocene plant-insect asso-ciations from the Pacific Northwest: a preliminary report.Rocky Mountain Geology 37:31–59.
MacGinitie, H. D. 1969. The Eocene Green River flora of north-western Colorado and northeastern Utah. University of Cal-ifornia Publications in Geological Sciences 83:1–202.
McMahon, T. A., and J. T. Bonner. 1983. On size and life. Sci-entific American Library, New York.
Moles, A. T., and M. Westoby. 2000. Do small leaves expand fast-er than large leaves, and do shorter expansion times reduceherbivore damage? Oikos 90:517–524.
Nardini, A., E. Gortan, and S. Salleo. 2005. Hydraulic efficiencyof the leaf venation system in sun- and shade-adapted spe-cies. Functional Plant Biology 32:953–961.
New, M., D. Lister, M. Hulme, and I. Makin. 2002. A high-res-olution data set of surface climate over global land areas. Cli-
mate Research 21:1–25 (data available at http://www.cru.uea.ac.uk/cru/data/tmc.htm).
Niinemets, U. 2001. Global-scale climatic controls of leaf drymass per area, density, and thickness in trees and shrubs.Ecology 82:453–469.
Niinemets, U., F. Valladares, and R. Ceulemans. 2003. Leaf-levelphenotypic variability and plasticity of invasive Rhododendronponticum and non-invasive Ilex aquifolium co-occurring at twocontrasting European sites. Plant, Cell and Environment 26:941–956.
Niinemets, U., A. Portsmuth, D. Tena, M. Tobias, S. Matesanz,and F. Valladares. 2007. Do we underestimate the importanceof leaf size in plant economics? Disproportionate scaling ofsupport costs within the spectrum of leaf physiognomy. An-nals of Botany 100:283–303.
Niklas, K. J. 1978. Morphometric relationships and rates of evo-lution among Paleozoic vascular plants. Evolutionary Biology11:509–543.
———. 1991a. The elastic-moduli and mechanics of Populus tre-muloides (Salicaceae) petioles in bending and torsion. Ameri-can Journal of Botany 78:989–996.
———. 1991b. Flexural stiffness allometries of angiosperm andfern petioles and rachises: evidence for biomechanical con-vergence. Evolution 45:734–750.
———. 1994. Plant allometry: the scaling of form and function.University of Chicago Press, Chicago.
———. 1996. Differences between Acer saccharum leaves fromopen and wind-protected sites. Annals of Botany 78:61–66.
———. 1998. The influence of gravity and wind on land plantevolution. Review of Palaeobotany and Palynology 102:1–14.
———. 1999. A mechanical perspective on foliage leaf form andfunction. New Phytologist 143:19–31.
Parton, W., W. L. Silver, I. C. Burke, L. Grassens, M. E. Harmon,W. S. Currie, J. Y. King, E. C. Adair, L. A. Brandt, S. C. Hart,and B. Fasth. 2007. Global-scale similarities in nitrogen re-lease patterns during long-term decomposition. Science 315:361–364.
Peters, R. H. 1983. The ecological implications of body size.Cambridge University Press, Cambridge.
Poorter, L., and F. Bongers. 2006. Leaf traits are good predictorsof plant performance across 53 rain forest species. Ecology 87:1733–1743.
Poorter, L., M. van de Plassche, S. Willems, and R. G. A. Boot.2004. Leaf traits and herbivory rates of tropical tree speciesdiffering in successional status. Plant Biology 6:746–754.
Radtke, M. G., K. B. Pigg, and W. C. Wehr. 2005. Fossil Corylopsisand Fothergilla leaves (Hamamelidaceae) from the lower Eo-cene flora of Republic, Washington, USA, and their evolution-ary and biogeographic significance. International Journal ofPlant Sciences 166:347–356.
Reich, P. B., M. B. Walters, and D. S. Ellsworth. 1997. From trop-ics to tundra: global convergence in plant function. Proceed-ings of the National Academy of Sciences USA 94:13730–13734.
Rex, G. M. 1986. Further experimental investigations on the for-mation of plant compression fossils. Palaios 19:143–159.
Rex, G. M., and W. G. Chaloner. 1983. The experimental for-mation of plant compression fossils. Palaeontology 26:231–252.
Royer, D. L., C. P. Osborne, and D. J. Beerling. 2002. High CO2
increases the freezing sensitivity of plants: implications forpaleoclimatic reconstructions from fossil floras. Geology 30:963–966.
Royer, D. L., P. Wilf, D. A. Janesko, E. A. Kowalski, and D. L.Dilcher. 2005. Correlating climate and plant ecology with leafsize and shape: potential proxies for the fossil record. Amer-ican Journal of Botany 92:1141–1151.
Sack, L., and K. Frole. 2006. Leaf structural diversity is related
586 DANA L. ROYER ET AL.
to hydraulic capacity in tropical rainforest trees. Ecology 87:483–491.
Sack, L., P. J. Melcher, M. A. Zwieniecki, and N. M. Holbrook.2002. The hydraulic conductance of the angiosperm leaf lam-ina: a comparison of three measurement methods. Journal ofExperimental Botany 53:2177–2184.
Sack, L., P. D. Cowan, N. Jaikumar, and N. M. Holbrook. 2003.The ‘hydrology’ of leaves: co-ordination of structure andfunction in temperate woody species. Plant, Cell and Envi-ronment 26:1343–1356.
Sack, L., M. T. Tyree, and N. M. Holbrook. 2005. Leaf hydraulicarchitecture correlates with regeneration irradiance in tropi-cal rainforest trees. New Phytologist 167:403–413.
Salisbury, E. J. 1913. The determining factors in petiolar struc-ture. New Phytologist 12:281–289.
Schmidt-Nielsen, K. 1984. Scaling. Cambridge University Press,Cambridge.
Shipley, B., D. Vile, and E. Garnier. 2006. From plant traits toplant communities: a statistical mechanistic approach to bio-diversity. Science 314:812–814.
Small, E. 1972. Photosynthetic rates in relation to nitrogen re-cycling as an adaptation to nutrient deficiency in peat bogplants. Canadian Journal of Botany 50:2227–2233.
Smith, M. E., A. R. Carroll, and B. S. Singer. 2007. Synoptic re-construction of a major ancient lake system: Eocene GreenRiver Formation, Western United States. Geological Society ofAmerica Bulletin (in press).
Sokal, R. R., and F. J. Rohlf. 1995. Biometry, 3d ed. W.H. Free-man, New York.
Spicer, R. A. 1981. The sorting and deposition of allochthonousplant material in a modern environment at Silwood Lake, Sil-wood Park, Berkshire, England. U.S. Geological Survey Pro-fessional Paper 1143:1–77.
Spicer, R. A., and J. T. Parrish. 1986. Paleobotanical evidence forcool north polar climates in middle Cretaceous (Albian-Cen-omanian) time. Geology 14:703–706.
Thomas, S. C., and W. E. Winner. 2002. Photosynthetic differ-ences between saplings and adult trees: an integration of fieldresults by meta-analysis. Tree Physiology 22:117–127.
Villar, R., and J. Merino. 2001. Comparison of leaf constructioncosts in woody species with differing leaf life-spans in con-trasting ecosystems. New Phytologist 151:213–226.
Walton, J. 1936. On the factors which influence the external formof fossil plants; with descriptions of the foliage of some spe-cies of the Palaeozoic equisetalean genus Annularia Sternberg.Philosophical Transactions of the Royal Society of London B226:219–237.
Warton, D. I., I. J. Wright, D. S. Falster, and M. Westoby. 2006.Bivariate line-fitting methods for allometry. Biological Re-views 81:259–291.
Westoby, M., D. S. Falster, A. T. Moles, P. A. Vesk, and I. J. Wright.2002. Plant ecological strategies: some leading dimensions ofvariation between species. Annual Review of Ecology andSystematics 33:125–159.
Whittaker, R. 1975. Communities and ecosystems. Macmillan,New York.
Wilf, P., and C. C. Labandeira. 1999. Response of plant-insectassociations to Paleocene-Eocene warming. Science 284:2153–2156.
Wilf, P., S. L. Wing, D. R. Greenwood, and C. L. Greenwood.1998. Using fossil leaves as paleoprecipitation indicators: anEocene example. Geology 26:203–206.
Wilf, P., C. C. Labandeira, K. R. Johnson, P. D. Coley, and A. D.Cutter. 2001. Insect herbivory, plant defense, and early Ce-nozoic climate change. Proceedings of the National Academyof Sciences USA 98:6221–6226.
Wilf, P., C. C. Labandeira, K. R. Johnson, and N. R. Cuneo. 2005.Richness of plant-insect associations in Eocene Patagonia: alegacy for South American biodiversity. Proceedings of theNational Academy of Sciences USA 102:8944–8948.
Wing, S. L., and D. R. Greenwood. 1993. Fossils and fossil cli-mate: the case for equable continental interiors in the Eocene.Philosophical Transactions of the Royal Society London B 341:243–252.
Wolfe, J. A. 1987. Late Cretaceous-Cenozoic history of decidu-ousness and the terminal Cretaceous event. Paleobiology 13:215–226.
Wolfe, J. A., and G. R. Upchurch. 1987. Leaf assemblages acrossthe Cretaceous-Tertiary boundary in the Raton Basin, NewMexico and Colorado. Proceedings of the National Academyof Sciences USA 84:5096–5100.
Wolfe, J. A., and W. C. Wehr. 1987. Middle Eocene dicotyledon-ous plants from Republic, northeastern Washington. U.S.Geological Survey Bulletin 1597:1–25.
Wright, I. J., and M. Westoby. 2002. Leaves at low versus highrainfall: coordination of structure, lifespan and physiology.New Phytologist 155:403–416.
Wright, I. J., P. B. Reich, and M. Westoby. 2001. Strategy-shiftsin leaf physiology, structure and nutrient content betweenspecies of high and low rainfall, and high and low nutrienthabitats. Functional Ecology 15:423–434.
Wright, I. J., P. B. Reich, M. Westoby, D. D. Ackerly, Z. Baruch, F.Bongers, J. Cavender-Bares, T. Chapin, J. H. C. Cornelissen,M. Diemer, J. Flexas, E. Garnier, P. K. Groom, J. Gulias, K. Hi-kosaka, B. B. Lamont, T. Lee, W. Lee, C. Lusk, J. J. Midgley, M.-L. Navas, U. Niinemets, J. Oleksyn, N. Osada, H. Poorter, P.Poot, L. Prior, V. I. Pyankov, C. Roumet, S. C. Thomas, M. G.Tjoelker, E. J. Veneklaas, and R. Villar. 2004. The worldwideleaf economics spectrum. Nature 428:821–827.
Wright, I. J., P. B. Reich, J. H. C. Cornelissen, D. S. Falster, P. K.Groom, K. Hikosaka, W. Lee, C. H. Lusk, U. Niinemets, J.Oleksyn, N. Osada, H. Poorter, D. I. Warton, and M. Westoby.2005. Modulation of leaf economic traits and trait relation-ships by climate. Global Ecology and Biogeography 14:411–421.
Yamada, T., E. Suzuki, and T. Yamakura. 1999. Scaling of petioledimensions with respect to leaf size for a tropical tree, Sca-phium macropodum (Sterculiaceae), in Borneo. Journal of PlantResearch 112:61–66.
587FOSSIL LEAF ECONOMICS
App
endi
x1
Det
ails
ofca
libr
atio
nsi
tes
use
din
stu
dy
(five
orm
ore
spec
ies
per
site
).C
lim
ate
dat
aar
efr
omw
eath
erst
atio
ns
wit
hin
0.1
ofth
esi
tes.
Site
Lat
.L
ong
.Sp
ecie
su
sed
MA
T(
C)
MA
P(m
m)
Au
thor
resp
onsi
ble
for
mea
sure
men
ts
Furt
her
info
rmat
ion
onsi
tes
and
sam
pli
ng
pro
toco
ls
Arc
hb
old
Bio
log
ical
Stat
ion
,Flo
rid
a,U
SA27
.2N
81.4
W15
22.3
1321
Roy
er,w
ith
som
esp
e-H
uff
etal
.200
3;K
owal
ksi
and
Dil
cher
Bar
roC
olor
ado
Isla
nd
,Pan
ama
9.2
N79
.9W
8625
.826
40ci
esfr
omH
arva
rd20
03:R
oyer
etal
.200
5:im
ages
ofle
aves
Big
Ham
moc
kN
atu
ral
Are
aan
dW
ild
life
Man
agem
ent
Are
a,G
eorg
ia,U
SA31
.9N
82.1
W25
19.5
1189
Fore
stby
Sack
,an
dso
me
spec
ies
from
avai
labl
eat
droy
er.
web
.wes
leya
n.e
du/
Dig
ital
Lea
fPhy
siog
nom
y.ht
m;
Coc
kap
onse
tSt
ate
Fore
st,C
onn
ecti
cut,
USA
41.4
N72
.5W
2410
.212
13B
arro
Col
orad
oIs
lan
dSa
cket
al.2
005
for
Bar
roC
olor
ado
Dil
cher
’sW
ood
slo
wla
nd
,Flo
rid
a,U
SA29
.6N
82.2
W24
20.6
1257
bySa
ckan
dM
oles
Isla
nd
leav
esm
easu
red
bySa
ck,a
nd
Dil
cher
’sW
ood
su
pla
nd
,Flo
rid
a,U
SA29
.6N
82.2
W20
20.6
1257
Sack
etal
.200
3fo
rH
arva
rdFo
rest
Du
keFo
rest
,Nor
thC
arol
ina,
USA
36.0
N78
.9W
2715
.211
08le
aves
mea
sure
dby
Sack
E.
N.H
uyc
kP
rese
rve
and
Bio
log
ical
Re-
sear
chSt
atio
n,N
ewYo
rk,U
SA42
.7N
74.5
W24
7.6
881
Flo
rid
aP
anth
erN
atio
nal
Wil
dli
feR
efu
ge,
Flo
rid
a,U
SA26
.2N
81.3
W18
24.3
1500
Har
vard
Fore
st,M
assa
chu
sett
s,U
SA42
.5N
72.2
W34
7.2
1068
Haw
kM
oun
tain
San
ctu
ary,
Pen
nsy
lvan
ia,
USA
40.6
N75
.9W
2310
.811
56
Hu
bbar
dB
rook
Exp
erim
enta
lFo
rest
,New
Ham
psh
ire,
USA
43.9
N71
.8W
155.
613
15
Inst
itu
tefo
rE
cosy
stem
Stu
die
s,N
ewYo
rk,
USA
41.8
N73
.8W
299.
510
86
Lit
tle
Pee
Dee
Stat
eP
ark
,Sou
thC
arol
ina,
USA
34.2
N79
.4W
2616
.911
49
Smit
hso
nia
nE
nv
iron
men
tal
Res
earc
hC
ente
r,M
ary
lan
d,U
SA38
.5N
76.3
W23
12.8
1132
Alc
orn
ocal
esN
atu
ral
Par
k,M
alag
a,Sp
ain
36.2
N5.
3W
2116
.788
0N
iin
emet
s&
Val
lad
ares
Nii
nem
ets
etal
.200
3,20
07
Ku
rin
gai
Ch
ase
Nat
ion
alP
ark
(hig
hP
),N
ewSo
uth
Wal
es,A
ust
rali
a33
.6S
151.
3E
1017
.511
48L
usk
Wri
ght
etal
.200
1;W
righ
tan
dW
esto
by20
02K
uri
ng
aiC
has
eN
atio
nal
Par
k(l
owP
),N
ewSo
uth
Wal
es,A
ust
rali
a33
.7S
151.
1E
1417
.511
48
Ok
atai
na,
Nor
thIs
lan
d,N
ewZ
eala
nd
38.8
S17
6.4
E20
12.0
2157
Rou
nd
Hil
lN
atu
reR
eser
ve(h
igh
P),
New
Sou
thW
ales
,Au
stra
lia
33.0
S14
6.2
E12
17.1
412
Rou
nd
Hil
lN
atu
reR
eser
ve(l
owP
),N
ewSo
uth
Wal
es,A
ust
rali
a33
.0S
146.
1E
517
.141
2
Fro
dsh
ams
Pas
s,Ta
sman
ia,A
ust
rali
a42
.8S
146.
4E
148.
517
80Jo
rdan
Roy
eret
al.2
005
for
sam
pli
ng
pro
toco
lH
obar
tW
etSc
lero
ph
yll
,Tas
man
ia,A
ust
rali
a42
.9S
147.
3E
1112
.162
0M
arg
aret
Riv
er,W
este
rnA
ust
rali
a34
.1S
115.
1E
2315
.411
48M
t.R
ead
,Tas
man
ia,A
ust
rali
a41
.8S
145.
6E
206.
537
10
Kok
e‘e
Par
k,K
aua‘
i,H
awai
‘i,U
SA22
.1S
159.
7W
1015
.525
00Sa
ckR
oyer
etal
.200
5fo
rsa
mp
lin
gp
roto
col
588 DANA L. ROYER ET AL.
App
endi
x2
Det
ails
ofca
libr
atio
nsi
tes
use
din
stu
dy
(fou
ror
few
ersp
ecie
sp
ersi
te).
Cli
mat
ed
ata
are
from
the
mod
elof
New
etal
.(2
002)
and
are
typ
ical
lyw
ith
in0.
2 of
the
site
s.
Site
Lat
.L
ong
.Sp
ecie
su
sed
MA
T(
C)
MA
P(m
m)
Au
thor
resp
onsi
ble
for
mea
sure
men
tsFu
rth
erin
form
atio
non
site
san
dsa
mp
lin
gp
roto
cols
Arg
enti
na,
Bar
iloc
he:
Not
hofa
gus
fore
st41
.2S
71.4
W1
5.6
1008
Arg
enti
na,
Pu
erto
Mad
ryn
:Du
ne
gra
ssla
nd
42.8
S64
.1W
113
.720
1
Arg
enti
na,
Pu
erto
Mad
ryn
:Ste
pp
e42
.8S
64.1
W1
13.7
201
Arg
enti
na,
Tu
cum
an:Y
un
gas
N(s
ea-
son
ally
dry
fore
st)
23.7
S64
.8W
313
.852
0
Arg
enti
na,
Tu
cum
an:Y
un
gas
S(s
ea-
son
ally
dry
fore
st)
26.8
S65
.3W
311
.839
8
Au
stra
lia,
Ad
elai
de:
Ch
enop
odsh
rub
-la
nd
34.3
S13
9.5
E2
16.8
243
Au
stra
lia,
Ad
elai
de:
Scle
rop
hy
llsh
rub
-la
nd
34.3
S13
9.5
E2
16.8
243
Au
stra
lia,
Ad
elai
de:
Mal
lee
35.2
S13
9.1
E3
14.0
363
Au
stra
lia,
Ali
ceSp
rin
gs:
Spin
ifex
(Tri
odia
)g
rass
lan
dw
ith
emer
gen
teu
caly
pts
23.7
S13
3.9
E3
20.2
324
Au
stra
lia,
Arm
idal
e:C
alli
tris
woo
dla
nd
29.8
S15
1.1
E3
15.5
779
Au
stra
lia,
Bri
sban
e:Su
btr
opic
alra
info
-re
st28
.2S
153.
1E
417
.215
86
Au
stra
lia,
Dai
ntr
ee:T
rop
ical
rain
fore
st16
.1S
145.
4E
425
.619
18A
ust
rali
a,D
arw
in:V
ine
thic
ket
12.4
S13
0.8
E4
27.7
1658
Au
stra
lia,
Dar
win
:Sav
ann
a12
.4S
131.
1E
327
.615
62A
ust
rali
a,M
elb
ourn
e:G
rass
lan
dw
ith
Eu
caly
pt
over
stor
ey38
.0S
145.
6E
114
.098
9
Au
stra
lia,
Mel
bou
rne:
Eu
caly
pt
woo
d-
lan
dw
ith
brac
ken
un
der
stor
ey38
.4S
144.
9E
214
.487
7
Au
stra
lia,
Per
th:O
pen
scle
rop
hy
llw
ood
lan
d32
.0S
116.
0E
117
.210
80
Au
stra
lia,
Per
th:S
cler
oph
yll
shru
blan
d31
.7S
115.
9E
118
.377
4A
ust
rali
a,Sy
dn
ey:S
cler
oph
yll
shru
b-
lan
d33
.6S
151.
3E
217
.313
55
Au
stra
lia,
Tasm
ania
:Eu
caly
pt
woo
d-
lan
d42
.9S
147.
9E
311
.695
9
Au
stra
lia,
Tasm
ania
:Not
hofa
gus
fore
st42
.7S
146.
7E
310
.775
9A
ust
rali
a,To
owom
ba:
Eu
caly
pt
woo
d-
lan
d28
.1S
151.
7E
317
.470
1
Au
stra
lia,
Tow
nsv
ille
:Sav
ann
a19
.3S
146.
7E
123
.196
8A
ust
rali
a,To
wn
svil
le:V
ine
thic
ket
19.3
S14
6.8
E3
23.1
968
Isra
el,A
du
lam
:Med
iter
ran
ean
shru
b-
lan
d31
.6N
34.9
E4
19.4
407
Mol
esan
dth
eW
orld
Her
bivo
ryP
roje
ct(N
igel
An
dre
w,A
leja
nd
roB
isig
ato,
Ped
roB
len
din
ger
,Wil
-li
amB
ond
,Eli
zab
eth
Bor
er,S
arah
Bou
lter
,Lu
crec
iaC
ella
Piz
arro
,C
onn
ieC
lark
,Ph
ilip
pe
Coh
en,
Mos
heC
oll,
Will
Cor
nwel
l,W
illE
d-
war
ds,
Ras
mu
sE
jrn
æs,
Jose
Face
lli,
Ale
jan
dro
Farj
i-B
ren
er,F
lore
nci
aFe
rnan
dez
Cam
pon
,Ben
teG
raae
,G
ilbe
rto
Jam
anga
pe,
En
riqu
eJu
ra-
do,
Tif
fan
yK
nig
ht,
Bil
lL
ow,F
ai-
nes
sL
um
bw
e,B
enja
min
Mag
ana,
Jon
ath
anM
ajer
,Mig
uel
Mar
tın
ez-
Ram
os,P
eter
McQ
uil
lan
,Han
sM
elto
fte,
Ben
Moo
re,C
hri
sta
Mu
ld-
er,P
ablo
Per
i,N
igel
Pit
man
,Joh
nP
ouls
en,L
yn
da
Pri
or,K
ate
Rea
r-d
on-S
mit
h,J
org
eR
odri
gu
ez,E
ric
Seab
loom
,Jam
esSt
egen
,Die
goV
azqu
ez,R
uan
Vel
dtm
an,P
eter
Ves
k,A
nvo
nC
oult
er,H
ugo
von
Zei
pel
,Mat
tW
ald
ram
,Ch
arli
eZ
amm
it,Z
hen
gZ
hen
g)
Sit
ese
lect
ion
:Mol
eset
al.s
elec
ted
area
sof
nat
ura
lve
get
atio
n,i
nve
get
atio
nty
pes
char
acte
rist
icof
each
reg
ion
.
Spe
cies
sele
ctio
n:A
tea
chsi
te,M
oles
etal
.stu
die
dth
efo
ur
mos
tab
un
dan
tsp
ecie
s(b
yle
afco
v-er
)*.O
nly
woo
dy
dic
oty
led
ons
wer
ein
clu
ded
inth
ep
rese
nt
stu
dy.
Lea
fse
lect
ion
:Can
opy
acce
ssw
ason
lyav
aila
ble
aton
esi
te(D
ain
-tr
ee,A
ust
rali
a).A
tal
lot
her
site
s,M
oles
etal
.sam
ple
dle
aves
that
cou
ldb
ere
ach
edfr
omth
eg
rou
nd
.
Sam
plin
gpr
otoc
ol:T
wo
mat
ure
,bu
tre
cen
tly
exp
and
ed,l
eave
sw
ere
take
nfr
omea
chof
5p
lan
tsfr
omea
chsp
ecie
s.L
eave
sw
ere
kep
tco
olan
dm
oist
un
til
they
cou
ldb
esc
ann
edon
afl
atb
edsc
ann
er(f
aili
ng
this
,dig
ital
ph
otog
rap
hs
wer
eta
ken
ona
stan
dar
dg
rid
).L
eaf
area
was
calc
ula
ted
usi
ng
Imag
e-J
(rsb
.in
fo.n
ih.g
ov/i
j/).
Lea
ves
wer
ed
ried
at55
Cfo
r2
day
s,an
dw
eigh
ed.
*In
case
sw
her
eon
eof
the
sele
cted
spec
ies
did
not
hav
een
ough
reac
hab
lele
aves
,we
stu
die
dth
en
ext
mos
tab
un
dan
tsp
ecie
sin
-st
ead
.
589FOSSIL LEAF ECONOMICS
App
endi
x2
Con
tin
ued
.
Site
Lat
.L
ong
.Sp
ecie
su
sed
MA
T(
C)
MA
P(m
m)
Au
thor
resp
onsi
ble
for
mea
sure
men
tsFu
rth
erin
form
atio
non
site
san
dsa
mp
lin
gp
roto
cols
Isra
el,H
anad
iv:M
edit
erra
nea
nsh
rub
-la
nd
32.6
N34
.9E
319
.165
6
Mex
ico,
Ch
amel
a:D
ecid
uou
str
opic
alfo
rest
19.5
N10
5.0
W3
25.5
1366
Mex
ico,
Lin
ares
:Oak
woo
dla
nd
24.7
N99
.8W
222
.279
5M
exic
o,L
inar
es:T
hor
nsc
rub
24.8
N99
.5W
222
.279
5P
eru
,Los
Am
igos
:Tro
pic
alra
info
rest
onfl
ood
pla
in12
.6S
70.1
W3
24.9
3433
Per
u,L
osA
mig
os,S
ucc
essi
onal
veg
eta-
tion
nea
rri
ver
12.6
S70
.1W
124
.934
33
Per
u,L
osA
mig
os:T
rop
ical
rain
fore
st12
.5S
70.1
W4
24.9
3433
Rep
ubl
icof
Con
go,K
abo:
Bai
2.2
N16
.1E
121
.616
72R
epu
blic
ofC
ongo
,Kab
o:G
ilbe
rtio
den
-dr
on-d
omin
ated
trop
ical
rain
fore
st2.
1N
16.2
E3
21.6
1672
Rep
ubl
icof
Con
go,K
abo:
Mix
edtr
opi-
cal
rain
fore
st2.
1N
16.2
E4
21.6
1672
Sou
thA
fric
a,St
elle
nb
osch
:Fy
nb
os34
.0S
19.0
E1
14.6
1006
Sou
thA
fric
a,St
elle
nb
osch
:Kar
oo33
.6S
19.5
E1
15.8
807
Sou
thA
fric
aZ
ulu
lan
d:D
ryfo
rest
28.1
S32
.0E
421
.194
3So
uth
Afr
ica,
Zu
lula
nd
:Sav
ann
a28
.2S
31.8
E2
21.5
885