Embed Size (px)
Citation preview
Ecology, 80(6), 1999, pp. 1955-1969© 1999 by the Ecological Society of America
GENERALITY OF LEAF TRAIT RELATIONSHIPS:A TEST ACROSS SIX BIOMES
PETER B. REICH,' DAVID S. ELLSWORTH,2 MICHAEL B. WALTERS,3 JAMES M. VOSE," CHARLES GRESHAM,5
JOHN C. VouN,6 AND WILLIAM D. BOWMAN'
'Department of Forest Resources, University of Minnesota, Saint Paul, Minnesota 55108 USA^Department of Applied Science, Brookhaven National Laboratory, Upton, New York 11973 USA
^Department of Forestry, Michigan State University, East Lansing, Michigan 48824 USA4U.S. Forest Service, Coweeta Hydrological Laboratory, Otto, North Carolina 28763 USA
5Baruch Forest Institute, Clemson University, Georgetown, South Carolina 29442 USA^Division of Science, Florida Atlantic University, Davie, Florida 33314 USA
^Mountain Research Station, Institute of Arctic and Alpine Research, andDepartment of Evolutionary, Population, and Organismic Biology, University of Colorado,
Boulder, Colorado 80309 USA
Abstract. Convergence in interspecific leaf trait relationships across diverse taxonomicgroups and biomes would have important evolutionary and ecological implications. Suchconvergence has been hypothesized to result from trade-offs that limit the combination ofplant traits for any species. Here we address this issue by testing for biome differences inthe slope and intercept of interspecific relationships among leaf traits: longevity, net pho-tosynthetic capacity (Amax), leaf diffusive conductance (Gs), specific leaf area (SLA), andnitrogen (N) status, for more than 100 species in six distinct biomes of the Americas. Thesix biomes were: alpine tundra-subalpine forest ecotone, cold temperate forest—prairieecotone, montane cool temperate forest, desert shrubland, subtropical forest, and tropicalrain forest. Despite large differences in climate and evolutionary history, in all biomesmass-based leaf N (Afmass), SLA, Gs, and Amax were positively related to one another anddecreased with increasing leaf life span. The relationships between pairs of leaf traitsexhibited similar slopes among biomes, suggesting a predictable set of scaling relationshipsamong key leaf morphological, chemical, and metabolic traits that are replicated globallyamong terrestrial ecosystems regardless of biome or vegetation type. However, the intercept(i.e., the overall elevation of regression lines) of relationships between pairs of leaf traitsusually differed among biomes. With increasing aridity across sites, species had greaterAmax for a given level of Gs and lower SLA for any given leaf life span. Using principalcomponents analysis, most variation among species was explained by an axis related tomass-based leaf traits (Amax, N, and SLA) while a second axis reflected climate, G^-andother area-based leaf traits.
Key words: alpine; desert; leaf life span; leaf diffusive conductance; nitrogen; photosynthesis;scaling; specific leaf area; temperate forest; tropical forest.
INTRODUCTION Field and Mooney 1986, Korner 1989, Reich et al.
The idea that disparate plant species from different 1992)' but a comprehensive, quantitative evaluation ofbiomes may share similar leaf form and function has a these relationships among biomes is lacking. If corn-long history in ecology, but has rarely been tested in a binations of leaf traits recur m distantly related taxa
quantitative manner (Beadle 1966, Chabot and Hicks across a wide ^geographic range it would suggest that1982). It is clear that leaf life span, specific leaf area selection constrains the collective leaf traits possessed(SLA), nitrogen (N) concentration, leaf diffusive con- b? evefy sPedes (Reich et al- 1992> Reich 1993) and
ductance (Gs), and photosynthetic rate are fundamental provide evidence for convergent evolution,plant traits that vary greatly among species, often by Studies from a variety of species and ecosystems,orders of magnitude (Field and Mooney 1986, Reich et among plant groups taxonomically either broad (Fieldal. 1992). This variation may represent adaptation to and Mooney 1986, Evans 1989, Reich etal. 1991,1992,environmental heterogeneity that exists both locally and Reich 1993, Abrams et al. 1994, Mulkey et al. 1995)globally. It is well established that these leaf traits are or narrow (Chazdon and Field 1987, Williams et al.generally correlated, based on examination of variation 1989) have generally reported similar patterns of re-among and within species (e.g., Lugg and Sinclair 1981, lationships among leaf traits. Mass-based photosyn-
thetic capacity (Amass) and leaf nitrogen (Afmass) are usu-Manuscript received 26 January 1998; revised 28 August ally positively correlated, both are positively correlated
1998; accepted 7 August 1998. with SLA, and all three traits decline with increasing
1955
1956 PETER B. REICH ET AL. Ecology, Vol. 80, No. 6
TABLE 1. Description of the six study sites and related ecosystems used in the study of leaf trait relationships in the northern-hemisphere Americas.
Site Location Ecosystem typeElevation Temp. Precip. PET
(m) (°C) (mm) (mm)
Niwot Ridge, Colorado, 40°03' N, 105°36' Alpine tundra 3510 -3.0 900 350-400USA W
40°03' N, 105°36' Subalpine forest-meadow ecotone 3200W
Southern Wisconsin, 43°02' N, 89°28' W Cold temperate forest and prairie 275 8.0 820 700USA 43°21'N, 87°58'W Alkaline fen and bog 245
Coweeta, North Carolina, 35 000'N, 83°30' Montane humid temperate forest 700-850 12.5 1829 856USA W
Sevilleta, New Mexico, 34°37'N, 106°54' Desert grassland and shrubland 1400-1580 13.0 222 2428USA W Pinyon-juniper woodland 1890
34°36' N, 106°69'W
Hobcaw, South Carolina, 33°20'N, 79°13'W Warm temperate/subtropical forest 3-4 18.3 1295 1656USA
San Carlos, Amazonas, 1°56' N, 67°03' W Tropical rain forestVenezuela
120 26.0 3560 1971
Notes: Annual climate data (mean temperature, mean annual precipitation, and estimated potential evapotranspiration [PET])are shown for the meteorological station nearest to the main study area at each site. PET was calculated from open-panevaporation or the equivalent. Four of the sites (Colorado, North Carolina, New Mexico, and South Carolina) were part ofthe National Science Foundation Long-Term Ecological Research Program.
leaf life span. We have theorized that these relation-ships are universal at the broadest, continental to globalscales (Reich et al. 1992, Reich 1993) because of thefunctional interdependency among traits, plus the con-straints placed by biophysics and natural selection thatlead to ecological trade-offs (Coley et al. 1985, Coley1988, Field and Mooney 1986, Reich et al. 1992, Mul-key et al. 1995).
Thus, we hypothesize that interspecific proportionalscaling relationships (e.g., log Y = a + b log X) amongleaf traits will be quantitatively similar among diversebiomes, supporting the idea of convergent evolution.Alternatively, leaf trait correlations may occur but varyamong vegetation types or along climatic or edaphicgradients. To address these issues we quantified leafgas exchange rates, SLA, leaf N, and leaf life span,and their relationships for 10-43 species within eachof six sites representing different biomes in the Amer-icas. Our sites represent a range of biomes (Table 1)that vary in growing season length, mean air temper-ature, elevation, water availability, and soil fertility. Abrief summary paper (submitted after the original sub-mission of this manuscript) based in part on these samedata highlights the idea that the slopes of interspecifictrait relationships were similar among biomes andamong data sets (Reich et al. 1997). In this paper wetake the opportunity to (1) present the biome-specificdata, (2) explicitly contrast slopes among sites, (3)compare the elevations of these scaling equations, (4)contrast relationships among functional groups, and (5)make multivariate analyses of species leaf traits; col-lectively these represent the main objectives of thispaper and were beyond the scope of the previous sum-mary publication.
METHODS
Sites were selected to provide a wide range of en-vironmental conditions and terrestrial ecosystem types(Table 1). Two sites (in Colorado and Wisconsin) werelocated at the ecotone between biomes and include spe-cies common to both biomes. At several sites, mea-surements were made in more than one ecosystem typeor study area. The study site in Colorado was locatedat Niwot Ridge in the Front Range of the Rocky Moun-tains. Study plots were located in both wet and drymeadow tundra communities at 3510 m elevation, andin open subalpine forest-alpine meadow transition at3200 m. The species studied include common conifers,hardwood shrubs, and herbs. Soils at the site werelargely coarse-textured Inceptisols. The main studyarea in Wisconsin was the University of Wisconsin Ar-boretum, in Madison, Wisconsin (275-m elevation).Study plots were located in a mosaic of natural andrestored ecosystems, including forest, hedgerow, sa-vanna and tall grass prairie. Soils were largely medium-textured silt loams (Alfisols). We also examined speciesin a cedar-tamarack swamp and adjacent bog at Ce-darburg Bog, at the University of Wisconsin-Milwau-kee Field Station in Saukville, Wisconsin, ~130 kmnortheast of Madison. The climate in southern Wis-consin is humid continental, with cold winters andwarm summers. Species studied included a number ofcommon prairie and forest understory forbs, deciduoushardwood and coniferous forest tree species and broad-leafed evergreen bog shrubs.
A montane cool temperate forest was studied at theCoweeta Hydrological Laboratory, Otto, North Caro-lina. We studied naturally growing vegetation in sec-
September 1999 GENERALITY OF LEAF TRAIT RELATIONSHIPS 1957
ondary forest communities located at 700-850 m el-evation. Precipitation is abundant year-round, and theratio of precipitation to potential evapotranspiration ishigh (Table 1). Soils at this site were principally clayloam Ultisols. Species studied included a number ofcommon forest understory forbs, broadleafed decidu-ous and evergreen hardwood and evergreen coniferousforest tree species.
The lower Carolina coastal plain contains uplandpine-dominated forests as well as forested wetlands.We studied selected species from these communitieson the Hobcaw Forest in Georgetown County, SouthCarolina. The climate of the area is maritime, warmtemperate/humid subtropical. Annually, potentialevapotranspiration exceeds precipitation (Table 1).This difference is greatest during the growing season.Soils are sandy throughout the Hobcaw Forest.
Desert shrubland and pinyon-juniper woodland veg-etation were studied (1400-1540 m elevation) in theSevilleta National Wildlife Refuge, New Mexico. Soilsare sandy and classified as Aridosols of Haplargids-Torripsamments construction. The climate is warm andarid. Annually, potential evapotranspiration far exceedsprecipitation (Table 1).
The tropical rain forest site was located near SanCarlos del Rio Negro, Venezuela in the northern Am-azon basin. Although much of the data for the Vene-zuela site have been published before (e.g., Reich etal. 1991), given the design of this study their inclusionin this report is necessary. A total of 24 species werestudied in mature stands of three adjacent primary rainforest communities (of Bana, tall Caatinga, and TierraFirme forest) and in secondary successional standsgrowing on Tierra Firme sites (Reich et al. 1991, 1994).The different primary forests occur on Spodosols, Ul-tisols and Oxisols (Reich et al. 1994). The San Carlosregion is characterized year-round by abundant rainfalland warm temperatures.
Woody plants and perennial herbaceous species wereselected at each site (Table 2) based on the followingcriteria: we selected species expected to provide a gra-dient of leaf traits (based on prior general knowledge)and that were relatively abundant at each site. Bothbroadleaf and needle-leaf species (deciduous or ever-green) were selected at each site when possible.
Since leaf traits vary with leaf age, contrasts of gasexchange rates, N concentrations and SLA among spe-cies were made using leaves of a similar "physiolog-ical" age (i.e., ontogenetic stage). We used fully ex-panded young to medium-aged leaves of all species,which corresponds to the period when many leaf traitsare relatively stable (Reich et al. 1991). Gas exchangemeasurements were made during the growing seasonbetween 1987 and 1993. To minimize the potentiallyconfounding influence of shade, we tried to select"sun" leaves growing in relatively open conditions forall species at all sites. Measurements were made onopen-grown plants in all herbaceous dominated com-
munities and usually were made for open-grown treesor shrubs, or for mature trees in the upper canopy.Although leaf light microenvironment has a large im-pact on leaf traits, especially SLA (e.g., Ellsworth andReich 1992), the interspecific differences in leaf traitsin this study were large enough (often 25-50 X differ-ences) that smaller intraspecific differences due to vari-ation in leaf microenvironment (usually by factors ofless than two) would not have been significant.
Measurements of photosynthetic CO2 assimilationand leaf water vapor conductance were made underambient conditions at all sites with a portable leafchamber and infrared gas analyzer operated in the dif-ferential mode (ADC model LCA-2, Hoddesdon, En-gland). Measurements were made at mid- to late morn-ing (0800-1100 local time) when the following con-ditions were met: near full sunlight, relatively nonlim-iting vapor pressure deficits or temperatures. Thus,sampling was designed so that measurements weretaken to closely reflect leaf photosynthetic capacity inthe field at ambient CO2 concentration (Reich et al.1991, Ellsworth and Reich 1992). We took at least 10(but usually more) measurements per species from sev-eral individuals at each site, then averaged these forsubsequent analyses.
After measuring gas exchange rates, foliage was har-vested. The projected surface area of either the leaftissue or its silhouette was assessed by a digital imageanalysis system (Decagon Instruments, Pullman, Wash-ington). Total surface area was also calculated basedon the shape of each leaf type. The results were similarif total rather than projected surface area was used,although the quantitative relations differ. Given thatprojected area was measured, while total surface areawas estimated indirectly, data are expressed on a pro-jected area basis. Since SLA is by definition related toleaf thickness and density (Abrams et al. 1994, Garnierand Laurent 1994) we will use the terms leaf thickness,density, and SLA to convey roughly the same infor-mation.
In order to determine leaf life spans of broad-leavedspecies, leaf birth and death were monitored (usingtagging and/or drawings) for numerous leaves of atleast several plants per species (see Reich et al. 1991).In some instances, observations of leaf phenology wereused instead of direct measurement. For coniferous spe-cies with long-lived foliage, the average needle lon-gevity was calculated by counting the number of annualcohorts with at least 50% of their needles retained onthe branch.
In statistical analyses, individual data points repre-sent average values for a single species at a single site.A linear regression model was generally inappropriateto describe the relations between pairs of leaf traits,because the data were not normally distributed (Sha-piro-Wilk W test, all leaf traits P < 0.0001), the re-lationships were often not linear, and there was pat-terned heteroscedasticity in the residuals with the ab-
1958 PETER B. REICH ET AL. Ecology, Vol. 80, No. 6
TABLE 2. List of species, sites, and physiological data (mean per species per site).
Location
Colorado
North Carolina
New Mexico
South Carolina
Venezuela
Species
Acomastylis rosiiBistorta bistortoidesPsychrophila leptosepalaSalix glaucaSalix planifoliaVaccinium myrtillusArctostaphylos uva-ursiPinus flexilisPicea engelmaniiAbies lasiocarpa
Podophyllum peltatumVeratrum parviflorumHelianthus microcephalusRobinia pseudoacaciaEupatorium rugesumAcer rubrumLiriodendron tulipiferaQuercus coccineaGalax aphyllaPinus strobusPinus rigidaKalmia latifoliaRhododendron maximumTsuga canadensis
Baccharis angustifoliaEleagnus angustifoliaPopulus fremontiiGutierrezia sarothraeAtriplex canescensProsopis glandulosaQuercus turbinellaLarrea tridentataPinus edulisJuniperus monospermaJuniperus monosperma
Pterocaulon pycnostachyumTaxodium distichumVaccinium corymbosumQuercus laevisQuercus virginiana var.
geminataVaccinium arboreumPersea borboniaLyonia lucidaPinus serotinaPinus palustris
Manihot esculentaSolanum straminifoliaCecropia ficifoliaVismia lauriformisClidemia sericeaVismia japurensisBellucia grossularioidesGoupia glabraNeea obovataMiconia disparRetiniphyllum truncatumRhodognaphalopsis humilisProtium sp.Aspidosperma albumProtium sp.Caraipa heterocarpaOcotea costulataLicania heteromorpha
Func-tionalgroup
forbforbforbshrubshrubshrubshrubtreetreetree
forbforbforbtreeforbtreetreetreeforbtreetreeshrubtreetree
shrubtreetreeforbshrubshrubshrubshrubtreetreetree
forbtreeshrubtreetree
shrubtreeshrubtreetree
shrubshrubtreetreetreetreetreetreetreetreetreetreetreetreetreetreetreetree
Leaftype
broadbroadbroadbroadbroadbroadbroadneedleneedleneedle
broadbroadbroadbroadbroadbroadbroadbroadbroadneedleneedlebroadbroadneedle
broadbroadbroadbroadbroadbroadbroadbroadneedleneedleneedle
broadneedlebroadbroadbroad
broadbroadbroadneedleneedle
broadbroadbroadbroadbroadbroadbroadbroadbroadbroadbroadbroadbroadbroadbroadbroadbroadbroad
Leaflife
span(mo)
2.02.02.53.03.03.5
18.036.090.096.0
3.03.64.14.64.85.25.35.7
18.021.033.036.048.060.02.53.03.06.07.07.08.0
12.066.078.098.05.16.57.57.6
11.0
11.018.019.527.032.0
1.52.02.74.55.16.98.4
11.412.012.016.524.033.033.035.038.539.040.0
SLA(cnWg)
119.0121.0126.0122.8123.0171.078.427.332.739.3
323.0297.0251.0264.0469.0160.0165.0118.0125.092.048.994.548.981.6
105.0124.083.833.939.751.266.039.624.512.812.1
123.678.097.997.662.9
80.190.042.135.939.2
275.1287.4144.6114.3126.4105.478.1
129.9103.190.382.875.286.682.682.067.876.867.3
Leaf N(mg/g)
25.039.029.026.326.723.412.311.210.310.341.5
26.243.339.218.422.417.38.9
13.511.611.58.69.9
19.539.816.616.821.024.915.419.410.611.7
19.711.912.311.515.4
12.116.49.28.28.2
39.839.224.717.916.516.416.915.118.415.96.87.88.6
10.717.79.3
15.013.0
^mass(nmol-g-'-s-')
120.4136.0110.0124.3151.4104.531.825.911.611.5
261.0145.0259.6243.1382.0125.0170.0111.078.050.655.743.333.943.9
197.8175.1120.059.079.092.0
85.012.711.6
101.364.860.665.880.4
52.562.027.215.816.7
467.9378.5252.2159.3115.5140.4111.599.065.091.846.547.234.752.267.238.844.953.8
^area
10.811.18.7
10.212.26.14.49.73.52.96.54.4
10.59.46.87.7
10.29.74.96.0
11.34.86.85.5
18.714.114.414.717.117.9
18.34.17.7
7.88.36.26.8
12.8
6.46.76.24.13.9
16.913.517.713.89.1
12.814.07.86.4
10.35.66.24.06.58.25.75.98.0
G,(mmol-
673512462481542249191257
9681
169131536297224237450338
87198317172136141
575404495641350367
33010315196
365281145181361
1662031859467
113215562272
889597975
1707300479
1009490693321411381241329378
September 1999 GENERALITY OF LEAF TRAIT RELATIONSHIPS 1959
TABLE 2. Continued.
Location Species
Eperua purpureaEperua leucanthaLeguminosae sp.Micrandra spruceiMicropholis maguireiProtium sp.
Wisconsin Caulophyllum thalictroidesArisaema triphyllumDentaria laciniataErythronium americanumSilphium terebinthinaceumPodophyllum peltatumBaptisia leucophaeaTrillium grandifloraEchinacea purpureaSilphium integrifoliumSanguinaria canadensisPopulus deltoidesPopulus tremuloidesJuglans nigraBetula nigraAcer rubrumAcer saccharumPrunus serotinaCeltis occidentalesCarya ovataCornus floridaFraxinus americanaUlmus americanaBetula pumilaIlex verticillataLarix deciduaQuercus ellipsoidalisQuercus macrocarpaLarix laricinaCornus racemosaLonicera X bellaRhamnus catharticaAndromeda glaucophyllaChamaedaphne calyculataPinus strobusPinus banksianaJuniper virginianaPinus sylvestrisSarracenia purpureaPinus resinosaThuja occidentalisPicea glaucaPicea mariana
Func-tionalgroup
treetreetreetreetreetree
forbforbforbforbforbforbforbforbforbforbforbtreetreetreetreetreetreetreetreetreetreetreetreetreeshrubtreetreetreetreeshrubshrubshrubshrubshrubtreetreetreetreeforbtreetreetreetree
Leaftype
broadbroadbroadbroadbroadbroad
broadbroadbroadbroadbroadbroadbroadbroadbroadbroadbroadbroadbroadbroadbroadbroadbroadbroadbroadbroadbroadbroadbroadbroadbroadneedlebroadbroadneedlebroadbroadbroadbroadbroadneedleneedleneedleneedlebroadneedleneedleneedleneedle
Leaflifespan(mo)
43.044.050.050.050.551.0
1.11.93.03.03.53.74.04.04.05.05.05.05.25.55.55.55.55.65.65.75.86.06.06.06.06.06.06.06.76.8
13.013.021.027.0
27.030.036.048.060.060.0
SLA(cnWg)
95.172.158.455.358.1
425.0380.0297.0222.0133.0309.0106.3357.0128.5116.3321.0110.0121.0305.0118.0166.0125.099.0
121.0112.0108.0138.084.093.0
110.4139.095.0
114.083.3
123.0120.0115.075.8
114.974.041.030.034.078.134.044.835.034.0
Leaf N(mg/g)
15.212.521.410.88.0
10.3
58.263.653.042.014.444.735.951.615.016.653.623.622.129.622.821.018.520.723.919.814.019.819.315.115.521.021.023.514.314.519.023.013.911.917.012.416.413.911.411.77.6
12.412.1
^mass(nmol-g-'-s-')
55.135.147.743.3
33.0254.0
432.0263.0175.0244.0159.0209.0122.9116.0255.0162.0142.0175.0118.0108.0100.0125.0117.091.085.0
121.0116.067.272.8
102.0125.0149.452.395.0
118.0135.070.370.145.031.033.035.022.824.032.023.037.0
A™,(|j,mol-
m-2-s-')
3.76.77.4
5.7
5.9
14.211.913.47.9
15.05.89.8
10.07.9
14.811.95.8
10.06.57.6
11.79.78.47.38.6
13.87.66.68.9
13.213.66.77.09.8
11.09.36.16.17.66.5
10.32.96.37.26.69.2
G,(mmol-m-2-s-')
209173235298
274
13490
227359615164481499480478208595431162296
403316
408454255215250
598445250320440309299205231
144
163190
Notes: Data shown are specific leaf area (SLA), mass-based leaf nitrogen (leaf Wmass), mass-based net photosyntheticcapacity (Amass), area-based net photosynthetic capacity (A^J, and leaf diffusive conductance (G,) at photosyntheticcapacity.
solute value of the residuals increasing as a functionof the character value. Therefore, we used logarithmic(base-10) transformations of the data, which generallynormalized the data distribution, linearized the regres-sion functions, and stabilized the error term variances.
When any given leaf trait was used in regression asthe independent variable, there was no random sam-pling variation associated with it because the specieswere selected based on a priori knowledge to provide
a roughly continuous range of leaf traits. Based on theseconsiderations, Type I (sometimes also called ModelI) regression is appropriate (Steel and Torrie 1980, So-kal and Rohlf 1995). Given that there is both biologicaland error variation associated with our measures of alltraits used as independent variables, and that not allrelations involve direct causality, we also analyzedthese relationships using a Type II regression approachinvolving bivariate analyses (Steel and Torrie 1980,
1960 PETER B. REICH ET AL. Ecology, Vol. 80, No. 6
TABLE 3. Summary of multiple regression analyses for leaf traits in relation to site and other leaf traits.
_ , Whole modelTlpppnnpnt
variable P
Wmass <0.0001
N^. <0.0001
SLA <0.0001
Amass <0.0001
Ama <0.0001
Gs <0.0001
Nm,ss <0.0001AU, <0.0001Amass <0.0001A«. <0.0001Gs <0.0001Amass <0.0001Aml 0.007Aarea <0.0001
r2
0.63
0.52
0.81
0.89
0.51
0.69
0.670.640.830.340.500.740.170.75
Site (intercept)
P
NS
<0.0001
0.0001
0.0001
0.0001
<0.0001
0.00010.00010.030.050.030.0002
NS<0.0001
F
18.9
6.4
20.5
5.8
8.8
10.610.62.62.32.65.4
14.9
Independentvariable
Variable
leaf lifespan
leaf lifespan
leaf lifespan
leaf lifespan
leaf lifespan
leaf lifespan
SLASLASLASLASLA
leaf Nmassleaf NmaGs
P
<0.0001
<0.0001
<0.0001
<0.0001
<0.0001
<0.0001
<0.0001<0.0001<0.0001<0.0001<0.0001<0.0001
NS<0.0001
F
137.0
20.7
189.1
723.7
68.3
74.4
169.269.5
300.518.030.6
250.2
205.7
Interaction(slope)
P
NS
NS
0.03
NS
NS
0.02
NSNS
0.030.030.007NSNSNS
F
2.7
3.0
2.72.63.4
Notes: The interactionit was dropped from theof the line.
term serves as a test for slope differences amongmodel, and in those cases the site term serves as a
sites. If the interaction term was not significanttest for differences among sites in the elevation
Sokal and Rohlf 1995). The significance levels and thefits of these Type II regression analyses were almostidentical to those obtained using Type I regression.Moreover, the bivariate normal density ellipses (P =0.95) were very similar to the 95% confidence intervalsfor individual observations (predictions of individu-als). Thus, despite some uncertainty about whether theType II or Type I model is more appropriate, the resultsare similar if analyzed either way. Data were furtheranalyzed using ANCOVA, separate slopes analyses,multiple regression, and principal components analyses(PCA) (IMP Statistical Software, SAS Institute). Anal-yses were made (using linear contrasts of the trans-formed variables) to test whether the slopes of the linesvaried among sites (i.e., the interaction between siteand independent traits, Table 3). If they did not differsignificantly, the interaction term was removed fromthe model and so-called "same slopes" analyses wereused to test for intersite differences among regressionlines (interpreted in this paper as a difference in theelevation of the total line). If the slopes did differ sig-nificantly ANCOVA was used to test for differencesamong sites in the dependent variable (Table 4) at thegrand mean of the covariate (the independent variable).PCA was also conducted to simultaneously explorespecies variation in multiple leaf traits.
RESULTS
Leaf Nmasy SLA, and gas exchange vs. leaf life span
Leaf 7Vmass, SLA, and life span varied nearly 10X,40 X, and 90 X, respectively, among species (Fig. 1,
Table 2). At every site (Fig. 1) leaf Afmass declined sig-nificantly (for all results reported, P < 0.001 unlessstated otherwise) in relation to increasing leaf life span(mean r2 = 0.60) and this relationship was similar forall data pooled. Sites did not differ significantly in theslope or elevation of the Nmass: leaf life span relation-ship (Table 3). Leaf N^ increased significantly (P <0.05) with increasing leaf life span at three of the sixsites (Fig. 1), and was very weakly related to leaf lifespan using all data pooled.
At every site SLA declined significantly with in-creasing leaf life span (Fig. 1) and the relationshipvaried among sites (Fig. 1, Table 3). At any given leaflife span, SLA was highest in humid temperate andtropical forests (North Carolina and Venezuela) andlowest in desert shrubland (New Mexico) (Fig. 1, Table4).
Amass varied 40-fold (from 12 to 468 nmol-g-'-s-')among all species (Fig. 1, Table 2). Amass decreasedmarkedly with increasing leaf life span at each of thesix sites (average r2 = 0.88) and for all data pooled (r2
= 0.78). The slopes of these relationships did not differamong sites, but the elevations of the lines did (Table3). Overall, species from humid temperate (North Car-olina) and tropical forests (Venezuela) had higher Amass
on average at any given leaf life span than those fromsubtropical (South Carolina), desert shrubland (NewMexico) or alpine/subalpine sites (Colorado) (Fig. 1,Tables 3 and 4).
Area-based net photosynthetic rate (Aarca) varied
September 1999 GENERALITY OF LEAF TRAIT RELATIONSHIPS 1961
TABLE 4. Site means of physiological measures of leaf traits.
Site
Leaf trait
Specific leaf area (cm2/g)
Leaf Wmass (mg/g)
Leaf N^, (g/m2)
Amiss (nmol-g~'-s~')
A,,^ (u,mol-mr2-s~')
Leaf life span (mo)
Value
absoluteadjusted
absoluteadjusted
absoluteadjusted
absoluteadjusted
absoluteadjusted
absolute
Colorado
81.374.1
19.118.2
2.412.36
57.451.3
7.16.9
8.4
Wisconsin
107.293.3
19.517.8
1.801.87
95.577.6
8.78.2
7.4
NorthCarolina
144.5147.9
17.818.6
1.291.26
107.2112.2
7.17.3
10.5
NewMexico
41.744.7
18.217.8
3.833.88
66.161.7
12.912.7
11.1
SouthCarolina
69.275.9
12.012.9
1.741.69
46.853.7
6.66.9
12.1
Venezuela
95.5117.4
14.517.0
1.531.44
77.6107.2
8.19.0
16.6
Notes: Absolute values were back-transformed from means of transformed values. Adjusted values were back-transformedfrom least-squares means adjusted for leaf life span using analysis of covariance.
6-fold among species (from 3 to 19There was a significant negative relationship betweenA^j, and leaf life span in five of six sites (Fig. 1) andfor all data pooled (r2 = 0.36). The slopes did not differsignificantly but the elevations of the lines did (Table3). At any given leaf life span species from desertshrubland (New Mexico) tended to have higher Aml.athan those from any other biome. Their lower SLAmore than compensates for slightly lower Amass in termsof effects on A^, (given that Axm = Amass/SLA).
The relationship between leaf diffusive conductance(Gs) and leaf life span was not significant in humidtemperate forest (North Carolina), but was significantat all other sites and for pooled data (Fig. 2). Tropicalrain forest species tended to have the highest Gs forany given leaf life span (Fig. 2). At all sites and forpooled data, A^., increased with Gs (average r2 = 0.75,Fig. 2). The slopes did not differ among sites, but theelevations of the lines did (Table 3). At any given Gs,AK^ tended to be highest in desert shrubland (the mostarid study area) and lowest in tropical rain forest (ahumid area).
Leaf N and gas exchange rates vs. SLA
Leaf Afmass was positively correlated to SLA at eachof the six sites (mean r2 = 0.66) and for all data pooled(r2 = 0.50, Fig. 3). The slopes did not differ amongsites, but the elevations of the lines did (Table 3). Forany given SLA, Afmass tended to be highest in the desertshrubland site. Leaf Amass increased with SLA for alldata pooled (r2 = 0.74) and at each site (r2 ranged from0.78 to 0.86) (Fig. 3). The slopes were different amongsites (P = 0.03), but this interaction explained a smallfraction of the total variation in Amass, compared to SLA(Table 3). At any given SLA, species in desert shrub-land (New Mexico) had the highest Amass, with speciesfrom other sites having lower values (Fig. 3, Table 4).
Leaf A/ana decreased with increasing SLA in five of
six sites and for all data pooled (r2 = 0.45, Fig. 3). Atany given SLA, species from the desert shrubland (NewMexico) site had higher N^ than other species (Table3, Fig. 3). Ajje., was not significantly correlated withSLA at 5 of 6 sites or for all data pooled (data notshown).
Leaf gas exchange rates and leaf N
At all sites and for pooled data, there were highlysignificant relationships between Amass and Afmass (Fig.4, r2 ranged from 0.7 to 0.9). The Amass-Afmass slopeswere not different among sites, but the elevations ofthe lines were significantly different. At any given leaf7Vmass, Amass tended to be higher in humid temperate andtropical rain forest and lower in desert shrub and alpinetundra/subalpine forest. The regression between Am^and Wajea (Fig. 4) was not significant in five sites, wasweakly significant (r2 = 0.26, P = 0.09) in North Car-olina, and was significant (P < 0.01) but with a lowr2 (0.07) for all data pooled.
Leaf traits and climate
Multiple regression was used to evaluate whetherleaf traits were significantly related to combinations ofclimate variables and other leaf traits. Although therewere significant models that included either mean an-nual precipitation or mean annual temperature, the cli-mate measure that was generally best related to leaftraits was an index of site moisture balance (precipi-tation minus potential evapotranspiration). SLA, Amass,and Gs were all significantly related to the additivecombination of site moisture balance index (positively)and leaf life span (negatively) (Table 5, Fig. 5). Thus,SLA, Amass, and Gs tend to be greater in moister ratherthan drier environments once leaf life span was con-sidered. In contrast, N^m was negatively related to sitemoisture balance and positively related to leaf life span(Table 5).
1962 PETER B. REICH ET AL. Ecology, Vol. 80, No. 6
c0O>_
if" '
100
10-
100c0cfsr 10-1
CO0
oO
"••- Colorado-o- Wisconsin
- North Carolina• New Mexico
-»• South Carolina-o- Venezuela -All sites
CD o
co 100CO
8? 10-o o
0^^
^ ̂
10 100 1 10 100 1 10
Leaf life span (months)
100 1 10 100
FIG. 1. Mass-based leaf N concentration (Wmnss), area-based leaf N content (A^), specific leaf area (SLA), mass-basednet photosynthetic rate (Amass, nmol-g~'-s~'), and area-based net photosynthetic rate (A^, u,mol-m~2-s~') in relation to leaflife span for plant species in six diverse ecosystems. Ecosystems are arranged from left to right panels by ranking (low tohigh) their mean annual temperature (alpine tundra/subalpine forest, Colorado [CO]; cold temperate forest, bog, and prairie,Wisconsin [WI]; mesic temperate forest, North Carolina [NC]; desert and scrub, New Mexico [NM]; warm temperate forest,South Carolina [SC]; tropical rain forest, Venezuela [VEN]). All regressions for Figs. 1-4 were significant at P < 0.001(based on linear regressions of base-10 logarithmically transformed data) unless shown otherwise immediately following thecorrelation coefficient. Correlation coefficients (r2) for leaf Nmss vs. leaf life span: CO (r2 = 0.92), WI (r2 = 0.54), NC (r2
= 0.73), NM (r2 = 0.51), SC (r2 = 0.46), and VEN (r2 = 0.51); for leaf JVarea vs. leaf life span: CO (r2 = 0.15, P = 0.27),WI (r2 = 0.25), NC (r2 = 0.07, P = 0.38), NM (r2 = 0.49), SC (r2 = 0.48), and VEN (r2 = 0.02, P = 0.52); for SLA vs.leaf life span: CO (r2 = 0.84), WI (r2 = 0.72), NC (r2 = 0.76), NM (r2 = 0.81), SC (r2 = 0.74), and VEN (r2 = 0.79); forleaf Amass vs. leaf life span: CO (r2 = 0.98), WI (r2 = 0.85), NC (r2 = 0.84), NM (r2 = 0.94), SC (r2 = 0.77), and VEN (r2
= 0.90); and for A^ vs. leaf life span: CO (r2 = 0.64), WI (r2 = 0.21), NC (r2 = 0.07, P = 0.34), NM (r2 = 0.65), SC(r2 = 0.41), and VEN (r2 = 0.69). Regression relationships for pooled data: log(Wm.,ss) = 1.57 - 0.34 log(leaf life span), r2
= 0.59; log (A^J = 0.17 + 0.09 log(leaf life span), r2 = 0.05, P = 0.02; log(SLA) = 2.43 - 0.46 log(leaf life span), r2
= 0.57; log(Amass) = 2.56 - 0.66 log(leaf life span), r2 = 0.78; log(Aarea) = 1.15 - 0.23 log(leaf life span), r2 = 0.36. Allrelationships significant at P < 0.1 are shown on the figures.
Multiple regression against SLA and site moisturebalance showed that Amass, Afmass, A .̂.,, and N^^ were allnegatively related to site moisture balance (Table 5).Thus, these measures all tend to be greater in arid thanhumid environments once variation in SLA is consid-
ered. Multiple regression of A^ vs. the combinationof Gs and site moisture balance showed that AXK tendsto be higher in arid environments for a given Gs, whichcan be interpreted as a strategy to enhance water con-servation.
September 1999 GENERALITY OF LEAF TRAIT RELATIONSHIPS 1963
02 50
'•53 »£?'II 1(H2°o E
» 2.
North CarolinaNew Mexico
r.*-•»• South Carolina-a-Venezuela -All sites
100 1 10 1001 10
Leaf life span (months)
100 1 10 100
50 400 3200 50 400 3200 50 400 3200 50 400 3200Leaf diffusive conductance
(mmol'rrr2>s-1)
FIG. 2. Leaf diffusive conductance (G,) in relation to leaf life span (top) and area-based net photosynthetic rate in relationto leaf diffusive conductance (below) for species in six diverse ecosystems. All other details are as in Fig. 1. Correlationcoefficients (r2) for leaf diffusive conductance vs. leaf life span (P < 0.001 unless noted otherwise): CO (r2 = 0.86), WI(r2 = 0.20, P = 0.01), NC (r2 = 0.17, P = 0.14), NM (r2 = 0.89), SC (r2 = 0.54), and YEN (r2 = 0.72); for A^ vs. leafdiffusive conductance: CO (r2 = 0.90), WI (r2 = 0.44), NC (r2 = 0.77), NM (r2 = 0.76), SC (r2 = 0.85), and YEN (r2 =0.75). Regression relationships for pooled data: log(G,) = 2.82 - 0.33 log(leaf life span), r2 = 0.30; logUa™) = -0.28 +0.48 log(G,), r2 = 0.54.
Site differences in mean leaf traits
Site rankings do not differ in most cases between theaverage leaf trait data and data standardized by leaflife span. However, site means fail to consider anydifferences in mean leaf life span across sites (sincespecies were not selected randomly). Mean leaf Afmass
did not differ markedly between sites except for SouthCarolina, where leaf Wmass was lower than other sites(Table 4). Substantial differences in mean SLA wereapparent across sites: SLA in desert was far lower thanin other sites and was highest in mesic sites. Giventhat high SLA and leaf Afmass both positively affect Amass
in general (Field and Mooney 1986, Reich et al. 1992,1994) it is not surprising that N. Carolina (with highmean SLA and Afmass) had higher mean Amass than othersites and that New Mexico, Colorado and South Car-olina (low SLA and/or low JVmass) had low Amass (Table4). As a consequence of having the "thickest" leaves(low SLA), Wan-a and AaK!L were also highest in NewMexico desert—shrubland.
Multivariate analyses
Simple PCA were performed on all species usingfive leaf traits (Fig. 6). The first two axes included 90%of the variation. The positive direction of Axis 1 (they axis) pointed towards species which, when definedin terms of the leading (and approximately equally im-portant) vectors, were of high SLA, Afmass, Amass, andshort leaf life span. Axis 2 (the x axis) contained onlyone important vector, Gs. Coniferous species occupiedonly the "low SLA" end of Axis 1 and the "low G"
end of Axis 2, with little overlap with woody pioneers,nonpioneer deciduous trees and shrubs, or herbaceousspecies. Evergreen broad-leaved species with low leafturnover rates occupied positions at slightly higher lev-els on both axes. Although groups do occupy differentpositions, species within the major groupings were of-ten located at distant positions from one another. Add-ing the site moisture balance index alone or with theother area-based measures (Am^, N^^) to the PCAchanges the values but not the overall structure, andthe climate variable and the area-based leaf traits loadonto the second axis, along with Gs. Hence, the firstaxis represents variation within and among sites,whereas the second axis largely represents variationamong sites, reinforcing the message that climate altersthe relationships among leaf traits, largely by alteringSLA.
DISCUSSION
Generality and scaling issues
The field data from six sites generally support thehypothesis that the slopes of interspecific relationshipsamong leaf traits will be similar among diverse biomes(Reich et al. 1992, Reich 1993) and suggest that theproportional scaling functions between leaf life span,-/Vm SLA, Amass, (and to a lesser extent Gs andacross species are general. Thus, for any given pro-portional difference in one trait among species, the pro-portional difference among species in any other traitwill be similar among all biomes. As an example, for
1964 PETER B. REICH ET AL. Ecology, Vol. 80, No. 6
•|2 1000
100-
IIQ-c10- -*•• Colorado
• Wisconsin• North Carolina• New Mexico
-»-South Carolina-o- Venezuela -All sites
100
(D°> 10
roo
^ 10'
ro"3 0.6
10 80 640 10 80 64010 80Specific leaf area (cm2/g)
64010 80 640
FIG. 3. Mass-based net photosynthetic rates and mass- and area-based leaf N in relation to specific leaf area for speciesin six diverse ecosystems. All other details are as in Fig. 1. Correlation coefficients (r2) for leaf Amas,. vs. SLA (P < 0.001unless noted otherwise): CO (r2 = 0.82), WI (r2 = 0.78), NC (r2 = 0.86), NM (r2 = 0.84), SC (r2 = 0.82), and YEN (r2
= 0.79); for leaf JVmass (mg/g) vs. SLA: CO (r2 = 0.76), WI (r2 = 0.62), NC (r2 = 0.77), NM (r2 = 0.50), SC (r2 = 0.68),and YEN (r2 = 0.64); and for leaf Naea (g/m2) vs. SLA: CO (r2 = 0.46), WI (r2 = 0.49), NC (r2 = 0.30, P = 0.05), NM(r2 = 0.72), SC (r2 = 0.57, P = 0.01), and YEN, NS); Regression relationships for pooled data: log(Amass) = -0.22 + 1.08log(SLA), r2 = 0.74; log(Wmass) = -0.20 + 0.52 log(SLA), r2 = 0.50; Iog(A?area) = 1.18 - 0.47 log(SLA), r2 = 0.45.
a 10X decrease in leaf life span in any biome, Amass
will increase by a factor of approximately five.In contrast to the relatively constant slopes, the el-
evations of the regression lines often differed by site,with climate-related variation among sites in SLA oftendriving these differences. Species in an arid desert eco-system had the lowest SLA on average or at any givenleaf N (Fig. 3) or leaf life span (Table 4, Fig. 1), andthe most humid sites had the highest average SLA. Asa result, the SLA-leaf life span relationship shifts to ahigher range of SLA across moist to arid gradients(compare desert with humid montane forest in Fig. 1).
Taken together, the data and analyses from our sixsites support the contentions that species, regardless ofbiome, tend to have a "syndrome" or set of linked leaftraits, with SLA, leaf life span, leaf Nmass, and Amass ofeach species generally falling together somewherealong a multiple trait continuum (Reich et al. 1991,1992, 1997). There were no exceptions (outliers): forinstance, no species have high Amass and high SLA, buta long leaf life span and low NmKS.
Functional group and phylogenetic considerations
At all five North American sites, species with highSLA and short leaf life span tended to be herbaceous,with broad-leaved deciduous tree species intermediatein leaf traits and species with low SLA and long leaf
life span generally being needle-leaved evergreen co-nifers. Thus, one could ask whether the observed leaftrait relationships, albeit general across biomes, mightbe due to comparisons of different plant functionaltypes or evolutionary groupings (in a phylogeneticallybroad sense). Several contrasts refute that idea and sug-gest that the trait relationships are general, both withinand across broad plant types and taxonomic groups,although different plant groupings do differ in a roughsense (see Fig. 6). First, in the tropical rain forest site,all species were broad-leaved evergreens (i.e., woodyAngiosperms), yet the leaf trait relationships in thisbiome were similar to all other biomes. Second, broad-leaved evergreen species in tundra, bog, and forest hab-itats (e.g., Sarracenia, Kalmia, Lyonia, Quercus vir-giniana) in Colorado, Wisconsin, and the Carolinas hadleaf traits similar to those of tropical rain forest broad-leaved evergreen and coniferous needle-leaved specieswith comparable leaf life spans. Third, needle-leaveddeciduous species (Taxodium and Larix) displayed thesame collective syndrome of leaf traits as broad-leavedspecies with similar SLA or leaf life spans, and oc-cupied similar positions along regression lines. Theseexamples suggest that the observed relationships aregenerally common within or across communities orecosystems dominated by species of various leaf type(needles, broad) or other functional groups (trees,
September 1999 GENERALITY OF LEAF TRAIT RELATIONSHIPS 1965
« 1000)fl)~
SinS.-?
sl 10o o
o
-•••-Colorado-o- Wisconsin
.|5 1000
100-
10-o-1
- North CarolinaNew Mexico
- South Carolina- Venezuela
0.6 2.5 100.6 2.5 iO 0.6 2.5
Leaf nitrogen (g/m2)
10 0.6
21 63 7 21 63 7
Leaf nitrogen (mg/g)
21 63 7
2.5 10
21 63
FIG. 4. Area-based net photosynthetic rate dj,mol-m~2-s~') in relation to area-based leaf N (g/m2), and mass-based netphotosynthesis in relation to mass-based leaf N (mg/g), for species in six diverse ecosystems. All other details are as in Fig.1. Correlation coefficients (r2) for A^, vs. N^, (P < 0.001 unless noted otherwise): CO (r2 = 0.01, P = 0.77), WI (r2 =0.03, P = 0.25), NC (r2 = 0.26, P = 0.09), NM (r2 = 0.10, P = 0.40), SC (r2 = 0.00, P = 0.89), and YEN (r2 = 0.07, P= 0.22); for leaf Amass vs. A?mass: CO (r2 = 0.90), WI (r2 = 0.70), NC (r2 = 0.86), NM (r2 = 0.64), SC (r2 = 0.81) and YEN(r2 = 0.70). Regression relationships for pooled data: log(Aa0.13 + 1.42 log(JVmass), r2 = 0.68.
a) = 0.85 + 0.24 log(A'ma), r2 = 0.07, P = 0.01; log(An
shrubs, herbs), given the broad variation within eachgroup and the overlap in trait combinations among suchgroups (Fig. 6, also see Reich et al. 1997). The con-vergence of these relationships among taxa is also sup-ported using phylogenetically independent contrasts(D. Ackerly and P. Reich, in press).
Controls on photosynthesis
Similarity in the Amass to Afmass relationship amongspecies at six sites, as well as in other independent datasets (Field and Mooney 1986, Reich et al. 1992) sup-ports its universal application among all species (butnot within species [see Reich et al. 1994]). In contrast,
TABLE 5. Summary of multiple regression analyses for leaf traits in relation to the site moisture balance index (mean annualprecipitation minus estimated mean annual pan evaporation) and other leaf traits.
Whole model Site moisture balance Independent variable
Note: If the interaction term was not significant it was dropped from the model.t Dependent variables
Interaction
Traitsf
A^ass
Ararea
SLA
"mass
"area
Gs
Wmass
A'or
^massAcTAmass
^arca
A^
P
<0.0001
<0.0001
<0.0001
<0.0001
<0.0001
<0.0001
<0.0001<0.0001<0.0001<0.0001<0.0001<0.0001
0.01<0.0001
r2
0.59
0.39
0.75
0.83
0.39
0.39
0.640.590.780.130.120.710.090.71
P
NS
<0.0001
<0.0001
<0.0001
0.03
<0.0001
0.0020.002
<0.00010.001
NS
0.004NS
<0.0001
F
58.0
79.4
31.2
5.1
12.8
10.610.615.612.3
8.7
55.5
Variable
leaf lifespan
leaf lifespan
leaf lifespan
leaf lifespan
leaf lifespan
leaf lifespan
SLASLASLASLASLAleaf Nmisi
leaf N^Gs
P
<0.0001
<0.0001
<0.0001
<0.0001
<0.0001
<0.0001
<0.0001<0.0001<0.0001
0.0060.02
<0.0001NS
<0.0001
F
150.6
17.2
267.3
511.5
54.5
51.9
165.469.2
347.88.16.7
247.0
206.3
P
NS
NS
NS
NS
NS
NS
0.020.02
NS
NS
NS
NS
NS
NS
F
5.95.9
1966 PETER B. REICH ET AL. Ecology, Vol. 80, No. 6
810
100 x?^
FIG. 5. SLA in relation to the combination of leaf lifespan and the site moisture balance index (mean annual pre-cipitation minus potential evapotranspiration) for speciespooled from six sites in diverse biomes and climates (r2 =0.75).
relationships between leaf life span, A.̂ , and N^a weremore variable. The lack of correlation of leaf Afarea andleaf life span is the result of offsetting influences onNaKa of SLA and Nmass as they vary with leaf life span:parallel decreases in SLA and Afmass result in minimalor no net change in Afarea on average across the leaf life-span gradient (Fig. 1).
Variation in A^ among species was significantlycorrelated with leaf N^^ in the Field and Mooney
(1986) data survey, but was not related to Nml,a in anysite in this study or in a prior literature survey (Reichet al. 1992). This suggests that there is not a funda-mental relationship of Aarea to N^ among all species(as discussed previously in Reich et al. 1992, 1994).This may be partially ascribed to offsetting relation-ships: Amass is a positive function of Nmass and although7Vmass decreases with decreasing SLA, decreasing SLAincreases N^ for any given Afmass. This results in thepotential for leaves to have similar JVarea but differentyVmass. At a given A^, leaves with higher Nmass realizea higher A^ than leaves with lower Nmass, due to thepositive relationship between Amass and Nmass (Reich andWalters 1994), giving rise to considerable scatteramong species in the relationship of A^ to Afarea.
Why is the close association of multiple leaf traitsgeneral among species and biomes?
To build the physically sturdy foliage common inspecies with long leaf life span requires proportionallygreater carbon than nutrient investment (cell walls,thick waxy cuticles, etc.) and a dense structure. A dense(high mass per volume) or a thick structure correspondswith higher mass per unit area (low SLA) (Abrams etal. 1994, Gamier and Laurent 1994). High C/N ratiosand physically tough foliage both likely contribute tothe achievement of long leaf life spans (Reich et al.1991), since they confer some protection from herbiv-ory (Coley 1988) and physical weathering. Lower 7Vmass
is related to lower Amass because of the central role ofN in photosynthetic enzymes and pigments (Field andMooney 1986). Thus, lower SLA is associated with
• forbO pioneer, woodyT deciduous, woody (nonpioneer)v evergreen, woody
(leaf lifespan > 1 year)• coniferous, woody
o
6
4
2-
-2-
-6-3 -2 -1 0
Principal axis 2
FIG. 6. Principal components analysis ofspecies from six biomes, ordinated according tofive leaf traits (mass-based leaf N, mass-basednet photosynthetic capacity, leaf life span, spe-cific leaf area, and leaf diffusive conductance,using all 96 species for which data exist for allfive traits). The main contributors to high valuesof the first principal axis were increasing N,photosynthetic capacity, and specific leaf area,and decreasing leaf life span. The main con-tributor to high values of the second principalaxis was leaf diffusive conductance. The encir-cled groups of points represent five recogniza-ble plant groups. The pioneer, broadleaved de-ciduous (nonpioneer), and broadleaved ever-green (leaf life span > 1 yr) groups all includebroadleaved woody species and are further sep-arated based on successional habit and phenol-ogy-
September 1999 GENERALITY OF LEAF TRAIT RELATIONSHIPS 1967
greater tissue density and greater allocation of biomassto structural rather than metabolic components, en-hancing leaf strength and durability, but also resultingin greater internal shading and potential diffusionallimitations (Lloyd et al. 1992, Terashima and Hikosaka1995, Parkhurst 1994). Thus, low Nmass and low SLAcombine to promote low maximum rates of CO2 ex-change.
In species with the opposite leaf trait syndrome, thecombination of a high N investment in photosyntheticenzymes and pigments and high SLA can provide highmetabolism (gas exchange rates) and light harvestingper unit tissue mass (Field and Mooney 1986, Reichet al. 1992). However, such tissues are also nutritionallydesirable to herbivores (Coley et al. 1985) and less welldefended physically against biotic (herbivory) and abi-otic (physical weathering and stress) agents (Coley etal. 1985, Coley 1988, Reich et al. 1991).
Given the different roles of C vs. N in leaves, andof C or N allocated to structural vs. metabolic func-tions, it may be physically impossible to construct aleaf with very high Wmass and net CO2 exchange char-acteristics that is also physically robust and durable.Such constraints likely place a limit on how far anyspecies can occur (Fig. 7a) above the regression linesshown in Fig. 1-4. In contrast, although biophysicallyfeasible to build, no species has flimsy, unproductiveleaves; there would be disadvantages to leaves that areneither persistent nor productive and such a leaf traitsyndrome would likely be maladaptive. Thus, selectionshould place a limit on how far any species can be (Fig.7a) below the regression lines shown in Fig. 1-4. Giventhese constraints, the "solution" for every species isa position on the continuum of the leaf trait syndrome,ranging from species with leaves weighted toward pro-ductivity, others with leaves weighted toward persis-tence, and others at points in between.
Biogeographic differences
Do species in cold climates have high leaf JVmass? LeafAfmass did not differ among sites except that it was loweron infertile sandy soils in subtropical forest than allother sites. Thus the data presented offer no evidenceof higher leaf Afmass in high elevation or cold-climatesites, superficially disagreeing with Korner (1989).However, site differences in soils could mask any tem-perature trend for Nmass, given large variation in soilfertility and a low sample size (n = 6 sites). Moreover,other evidence does suggest that within phylogeneti-cally similar groups, plants adapted to colder environ-ments may have higher Nmass (Korner 1989, Reich etal. 1996).
Our data provide mixed support for the idea thatdesert plants are rich in N. On average, we found noevidence of higher average Nmlss in desert plants, dis-agreeing with long-mentioned hypotheses, but agreeingwith the recent analysis by Killingbeck and Whitford(1996). However, due to low SLA, N^ was higher in
XI"
limited by biophysical constraints(and ecological trade-offs)
limited byecological selection
Leaf trait x
FIG. 7. (a) Hypothetical limitations to leaf trait relation-ships due to biophysical constraints, physiology, ecology, andnatural selection, (b) Conceptualization of implications of leaftrait relationships for intra- and interecosystem comparisons.A shows typical variation within a site, B illustrates a hy-pothetical exception ("outlier") species, C illustrates leaftrait y in two different ecosystems at a common leaf trait x,D illustrates differences in the range of leaf trait x in twodifferent ecosystems, and E illustrates two species in differentecosystems that occupy the same relative intra-ecosystemranking.
desert-shrub vegetation than elsewhere, both on aver-age and for any given leaf life span. At any given SLA,moreover, N^, Nmass, and Amass were higher in desertshrub vegetation than in other biomes (Table 3) andhigher as a function of increasing site moisture deficit(Table 5). Thus, leaves of desert plants are N-rich com-pared to all others when comparing leaves of compa-rable structure. The high N of desert plants at any givenSLA offsets the negative correlation between Afmass andSLA and the tendency for desert plants to have lowSLA, resulting in desert plants having similar averageNmass as in other biomes.
Variation in SLA was related to site microenviron-ment (see Fig. 5) and large differences in SLA occuracross sites for any given leaf life span. These patternsare consistent with long-held common knowledgeabout variation in SLA, but to our knowledge, we areunaware of any published approach to quantifyingthese differences.
Implications
These comparisons of leaf trait relationships in wide-ly disparate ecosystems and biomes have several im-
1968 PETER B. REICH ET AL. Ecology, Vol. 80, No. 6
plications. First, they provide a quantitative basis forevaluating intra-ecosystem species differences (A, Fig.7b) and for identifying exceptions that might occur dueto unique plant adaptations (B, Fig. 7b). Second, iden-tification of general leaf scaling relationships enablescontrasts among ecosystems and biomes, and serves asa means of quantifying differences among them. Forexample, if species in a given ecosystem have higheror lower leaf N at any given SLA or leaf life span thanin another ecosystem (C, Fig. 7b), or if the range ofleaf N, leaf life span, or SLA is different in one eco-system compared to another (D, Fig. 7b), then this islikely due to key and quantifiable differences in theseecosystems (e.g., climate, soils, biology). Third, thesescaling relationships allow quantitative comparisons ofspecies in widely differing ecosystems. For example,do species (from different ecosystems) that share thesame relative intra-ecosystem ranking in leaf traitsshare the same successional position (E, Fig. 7b)? Ex-isting data suggest that such patterns often hold. Forexample, tree species of high SLA, high leaf Nmms, andshort leaf life span are more likely to be early succes-sional, inhabit high light microsites and display a fastgrowth rate in both temperate and tropical ecosystems(Uhl 1987, Reich et al. 1991, 1992). Finally, the un-covering of the generality of leaf trait relationshipsshould allow the selection of useful foliage attributesfor modeling vegetation productivity, distribution, anddynamics at diverse levels of scale (Running and Hunt1993, Leuning et al. 1995, Aber et al. 1996).
ACKNOWLEDGMENTSThis research was partially supported by National Science
Foundation Grants BSR 8819718, BSR 8857129, andIBN9296005, and by the NSF Long-Term Ecological Re-search Program. D. S. Ellsworth was supported by the U.S.Department of Energy (DE-ACO2-76CH00016) during thepreparation of this manuscript. We thank Chris Uhl, BrianKloeppel, Dan Bahauddin, and many others who contributedto this work in a wide variety of ways. We thank A. Thomasand Barbara G. Hallowell for providing unpublished data onleaf longevity in Galax.
LITERATURE CITED
Aber, J. D., P. B. Reich, and M. L. Goulden. 1996. Extrap-olating leaf CO2 exchange to the canopy: a generalizedmodel of forest photosynthesis validated by eddy corre-lation. Oecologia 106:267-275.
Abrams, M. D., M. E. Kubiske, and S. A. Mostoller. 1994.Relating wet and dry year ecophysiology to leaf structurein contrasting temperate tree species. Ecology 75:123-133.
Ackerly, D., and P. B. Reich. 1999. Convergence and cor-relations among leaf physiological traits: a test using in-dependent contrasts. American Journal of Botany, in press.
Beadle, N. C. 1966. Soil phosphate and its role in moldingsegments of the Australian flora and vegetation, with spe-cial reference to xeromorphy and scleromorphy. Ecology47:992-1007.
Chabot, B. E, and D. J. Hicks. 1982. The ecology of leaflife spans. Annual Review of Ecology and Systematics 13:229-259.
Chazdon, R. L., and C. B. Field. 1987. Determinants ofphotosynthetic capacity in six rainforest Piper species.Oecologia 73:222-230.
Coley, P. D., J. P. Bryant, and F. S. Chapin III. 1985. Resourceavailability and plant anti-herbivore defense. Science 230:895-899.
Coley, P. D. 1988. Effects of plant growth rate and leaflifetime on the amount and type of anti-herbivore defense.Oecologia 74:531-536.
Ellsworth, D. S., and P. B. Reich. 1992. Leaf mass per area,nitrogen content and photosynthetic carbon gain in Acersaccharum seedlings in contrasting forest light environ-ments. Functional Ecology 6:423-435.
Field, C., and H. A. Mooney. 1986. The photosynthesis-nitrogen relationship in wild plants. Pages 25-55 in T. Giv-nish, editor. On the economy of plant form and function.Cambridge University Press, London, UK.
Gamier, E., and G. Laurent. 1994. Leaf anatomy, specificmass and water content in congeneric annual and perennialgrass species. New Phytologist 128:725-736.
Killingbeck, K. T, and W. G. Whitford. 1996. High foliarnitrogen in desert shrubs: an important ecosystem trait ordefective desert doctrine? Ecology 77:1728-1737.
Korner, Ch. 1989. The nutritional status of plants from highaltitudes. A worldwide comparison. Oecologia 81:379—391.
Leuning, R., F. M. Kelliher, D. G. G. de Prury, and E.-D.Schulze. 1995. Leaf nitrogen, photosynthesis, conductanceand transpiration: scaling from leaves to canopy. Plant, Celland Environment 18:1183-1200.
Lugg, D. G., and T. R. Sinclair. 1981. Seasonal changes inphotosynthesis of field-grown soybean leaflets. 2. Relationto nitrogen content. Photosynthetica 15:138-144.
Lloyd, J., J. P. Syvertsen, P. E. Kriedemann, and G. D. Far-quhar. 1992. Low conductances for CO2 diffusion fromstomata to the sites of carboxylation in leaves of woodyspecies. Plant, Cell and Environment 15:873-899.
Mulkey, S. S., K. Kitajima, and S. J. Wright. 1995. Photo-synthetic capacity and leaf longevity in the canopy of a drytropical forest. Selbyana 16:169-173.
Parkhurst, D. F. 1994. Diffusion of CO2 and other gasesinside leaves. New Phytologist 126:449^-79.
Reich, P. B. 1993. Reconciling apparent discrepancies amongstudies relating life span, structure and function of leavesin contrasting plant life forms and climates: "the blind menand the elephant retold." Functional Ecology 7:721-725.
Reich, P. B., J. Oleksyn, and M. G. Tjoelker. 1996. Needlerespiration and nitrogen concentration in Scots pine pop-ulations from a broad latitudinal range: a common gardentest with field grown trees. Functional Ecology 10:768-776.
Reich, P. B., C. Uhl, M. B. Walters, and D. S. Ellsworth.1991. Leaf lifespan as a determinant of leaf structure andfunction among 23 tree species in Amazonian forest com-munities. Oecologia 86:16—24.
Reich, P. B., and M. B. Walters. 1994. Photosynthesis-ni-trogen relations in Amazonian tree species. II. Variation innitrogen vis-a-vis specific leaf area influences mass andarea-based expressions. Oecologia 97:73-81.
Reich, P. B., M. B. Walters, and D. S. Ellsworth. 1992. Leaflifespan in relation to leaf, plant and stand characteristicsamong diverse ecosystems. Ecological Monographs 62:365-392.
Reich, P. B., M. B. Walters, and D. S. Ellsworth. 1997. Fromtropics to tundra: global convergence in plant functioning.Proceedings of the National Academy of Sciences, USA94:13730-13734.
Reich, P. B., M. B. Walters, D. S. Ellsworth, and C. Uhl.1994. Photosynthesis-nitrogen relations in Amazonian treespecies. I. Patterns among species and communities. Oec-ologia 97:62-72.
Running, S. W, and E. R. Hunt. 1993. Generalization of aforest ecosystem process model for other biomes, BIOME-
September 1999 GENERALITY OF LEAF TRAIT RELATIONSHIPS 1969
BGC, and an application for global-scale models. Pages141-158 in Ehleringer, J. R., and C. B. Field, editors. Scal-ing physiological processes: leaf to globe. Academic Press,San Diego, California, USA.
Sokal, R. R., and F. J. Rohlf. 1995. Biometry. Third edition.W. H. Freeman, New York, New York, USA.
Steel, R. G. D., and J. H. Torrie. 1980. Principles and pro-cedures of statistics. A biometrical approach. Second edi-tion. McGraw-Hill, New York, New York, USA.
Terashima, I., and K. Hikosaka. 1995. Comparative eco-physiology of leaf and canopy photosynthesis. Plant, Celland Environment 18:1111-1128.
Uhl, C. 1987. Factors controlling succession following slash-and-burn agriculture in Amazonia. Journal of Ecology 75:377-407.
Williams, K., C. B. Field, and H. A. Mooney. 1989. Rela-tionships among leaf construction cost, leaf longevity, andlight environment in rain forest plants of the genus Piper.American Naturalist 133:198-211.
