multiple trait associations in relation to habitat … · 2019. 3. 29. · 635 ecological...

635

Ecological Monographs, 74(4), 2004, pp. 635–662q 2004 by the Ecological Society of America

MULTIPLE TRAIT ASSOCIATIONS IN RELATION TO HABITATDIFFERENTIATION AMONG 17 FLORIDIAN OAK SPECIES

JEANNINE CAVENDER-BARES,1,2,3,5 KAORU KITAJIMA,4 AND F. A. BAZZAZ3

1Department of Ecology, Evolution and Behavior, University of Minnesota, 1987 Upper Buford Drive,St. Paul, Minnesota 55108 USA

2Smithsonian Environmental Research Center, P.O. Box 28, Edgewater, Maryland 21037 USA3Department of Organismic and Evolutionary Biology, 16 Divinity Avenue, Harvard University,

Cambridge, Massachusetts 02138 USA4Department of Botany, 220 Bartram Hall, University of Florida, Gainesville, Florida 32611 USA

Abstract. Differentiation of species distributions along environmental gradients andphenotypic specialization help explain the co-presence of 17 oak species that might oth-erwise be expected to competitively exclude one another. In an effort to understand therole of niche differentiation in the co-presence of these congeneric species in north-centralFlorida, we examined the community structure of oak-dominated forests in this region inrelation to environmental variables and a suite of life history and physiological traits.Landscape distribution patterns of oaks and other woody species were determined fromrandomly established plots in three state parks. Soil moisture, nutrient availability, and fireregime were found to be critical factors influencing community structure, and the distri-bution of oak species was strongly correlated with these gradients. Detrended correspon-dence analysis of species’ distributions supported the grouping of oak species into threemajor community types: (1) hammock, (2) sandhill, and (3) scrub. Principal componentsof multiple traits also supported differentiation of oak species into these three groups.

There is clear evidence for ecological sorting among the oaks in these forests as speciespartition environmental gradients more than expected by chance, and most functional traitsof species correspond to species distributions as predicted based on a priori understandingof trait function. Across phylogenetic lineages, species showed evolutionary convergencein function and habitat preference. In contrast, leaf-level traits were conserved withinphylogenetic lineages and were not well correlated with local habitat factors, but ratherwith the broader geographic distributions and northern range limits of species. The cu-mulative results reported here provide strong evidence that oak species specialize for par-ticular niches via trade-offs in functional traits, and such niche partitioning contributes tothe high diversity of oak species at the landscape level. At smaller spatial scales, phylo-genetic diversity among the oaks is likely to be important in promoting their coexistencethrough other mechanisms.

Key words: detrended correspondence analysis; fire regime; functional traits; leaf life span;phylogenetic lineage; Quercus; soil moisture; soil nutrients.

INTRODUCTION

A critical issue in ecology is understanding the forcesthat maintain biological diversity, both taxonomic andfunctional. Central to this question has been a quest toexplain the patterns of species distributions and abun-dance across the landscape (Huston 1994) and to de-termine whether species specialize to partition the en-vironment or whether distributions are the result ofpredominantly stochastic processes (Diamond and Case1986, Brokaw and Busing 2000, Hubbell 2001). Al-though a number of mechanisms, including density de-pendent mortality, are important in regulating relativeabundance and diversity of species in both tropical(Wills et al. 1997, Wright 2002) and temperate systems

Manuscript received 10 January 2003; revised 30 October2003; accepted 15 November 2003; final version received 5 Jan-uary 2004. Correponding Editor: M. J. Lechowicz.

5 E-mail: [email protected]

(HilleRisLambers et al. 2002), such mechanisms do notexplain the association of functional traits with habitatand resource regimes as niche theory does (McKane etal. 2002). Indeed, a central goal of many ecologicalstudies has been to explain patterns of species distri-butions in relation to life history and functional traits(Monk 1966, Chabot and Hicks 1982, Hollinger 1992,Mulkey and Kitajima 1995, Reich et al. 1999, Ackerlyet al. 2002).

North Central Florida is at the confluence of thenorthern temperate flora and the southern subtropicalflora, providing a rich regional pool of species. Sev-enteen species of oaks are dominant or codominantwith pines in forest communities across the landscape.Due to the potential for close relatives to competitivelyexclude one another, the coexistence of multiple con-geners in this region is challenging to explain, partic-ularly given the lack of major elevational gradients thatoccur in other regions of high oak diversity (Whittaker

636 JEANNINE CAVENDER-BARES ET AL. Ecological MonographsVol. 74, No. 4

TABLE 1. A list of oak species in the study.

Species Code Clade

Generalhabitat

type

Firefrequencyranking‡

Canopyhabit

Q. falcata MichauxQ. hemispherica W. Bartram

ex Willd.Q. incana BartramQ. laevis WalterQ. laurifolia MichauxQ. myrtifolia Willd.

FAHE

INLVLAMY

redred

redredredred

sandhill/hammockhammock

sandhillsandhillhammockscrub

23

115

deciduousevergreen

brevideciduousdeciduousbrevideciduousevergreen

Q. nigra L.Q. pumila AsheQ. shumardii BuckleyQ. austrina SmallQ. chapmanii SargentQ. margaretta Ashe

NIPMSHAUCHMA

redredredwhitewhitewhite

hammockscrubhammock†hammock§scrubsandhill

5

43

1

brevideciduousbrevideciduousdeciduousdeciduousbrevideciduousdeciduous

Q. michauxii NuttallQ. stellata WangenheimQ. geminata SmallQ. minima (Sargent) SmallQ. virginiana Miller

MXSTGEMNVI

whitewhitelivelivelive

hammockhammock\scrubscrub¶hammock

52

5

deciduousdeciduousevergreenevergreenevergreen

Notes: Authorities are shown to the right of the species, along with a two-letter code and the major lineage to which itbelongs. General habitat type, fire-frequency ranking (Jackson et al. 1999), and canopy habit are given. Definitions of generalhabitat type are as follows: sandhill 5 high pineland, open woodland, deep sands, high fire frequency (1–15 yr) with lowseverity; scrub 5 dense thicket, sandy soil, infrequent fire (15–50 yr) with high severity; hammock 5 mesic or hydrictemperate hardwood forests with low or unpredictable fire frequency. Habitat descriptions are from Myers (1990), Platt andSchwartz (1990), Ewel (1990), and Kurz and Godfrey (1962).

† Soils underlain by limestone.‡ Ranking is from a model based on habitat types, excluding scrub habitats, as follows: (1) park-like forest/savanna; (2)

dry upland forest; (3) intermediate dry upland forest; (4) moist upland forest; (5) infrequently flooded bottomland forest(Jackson et al. 1999).

§ Often found on limestone bluffs.\ Upland soils frequently underlain by clay or marl.¶ Seepage areas of sandy-clay upland; restricted to sites subject to burning.

1956, Platt and Schwartz 1990). One possibility is thatoaks specialize into different niches, thus allowingthem to partition the landscape and maintain their di-versity. Here, we test the concept of the n-dimensionalniche, (as defined by Bazzaz [1987], based on earlierconceptions of the niche by Hutchinson [1957] andWhittaker [1975]) using the distribution of oak speciesacross critical environmental gradients. We refer to thephysical gradients of the environment as habitat factorsand the observed location of species across these gra-dients as habitats or realized niches.

While these oak species occur in close proximity,floristic studies have described them to be associatedwith distinct communities (hammock, sandhill, andscrub; Table 1), which vary markedly in soil wateravailability, fertility, and fire regime (Kurz and God-frey 1962, Ewel 1990, Myers 1990, Long and Jones1996). Oak species in this region also show significantvariation in phenology and traits related to water use(Cavender-Bares et al. 1999, Cavender-Bares and Hol-brook 2001), as found in other systems of high oakdiversity (Knops and Koenig 1994, Villar-Salvador etal. 1997, Damesin et al. 1998).

The overarching impetus for the study was to un-derstand factors that may help explain the maintenanceof high oak diversity at the landscape level (1–100-hascale) in north-central Florida. We first ask whether

species are nonrandomly distributed across environ-mental gradients. This allows us to investigate whetherdistinct community affiliations of north Florida oaksdescribed in the literature are supported by quantitativespatial analysis of species distributions relative to theabiotic environment. Second, if species are partitioningenvironmental gradients, what are the mechanisms, orfunctional traits, that allow them to do so? For example,do suites of leaf-level traits known to be associatedwith contrasting nutrient and water use strategies acrossa wide range of plant groups (e.g., Reich et al. 1997)allow oak species to specialize for particular resourceregimes? In an effort to answer these questions, wesought to (1) quantify distributions of 17 sympatric oakspecies with respect to each other and other woodyspecies; (2) analyze the distributions of species withrespect to critical environmental factors; (3) determinewhether species are partitioning their environmentmore than expected by chance using null models; and(4) determine whether and how functional traits of spe-cies correspond to their spatial distribution, their com-munity affiliation, and to abiotic features of their hab-itats.

As a result of the typical geological features in north-central Florida, namely sand deposition on a clay layeror limestone bedrock, even minor topographic changeswithin a few meters of elevation cause marked changes

November 2004 637HABITAT DIFFERENTIATION OF FLORIDIAN OAKS

FIG. 1. Schematic diagram showing environmental heterogeneity in woodlands of north-central Florida. The presence ofseveral important gradients, including water availability, soil fertility, and fire frequency and severity suggests the hypothesisthat oak species partition these gradients, allowing many species to coexist.

in soil hydrology, and consequently nutrient availabil-ity and fire regime (Fig. 1) (Myers and Ewel 1990). Ingeneral, the wetter the site, the higher the productivityand nutrient availability and the lower the fire fre-quency. As fire frequency decreases, the fuel load andthe fire intensity tend to increase (Jackson et al. 1999).However, in Florida, high temperatures combined withhigh humidity and soil moisture give rise to rapid de-composition; hence, fuel loading is not necessarily re-lated to time since last fire, as in other systems. As aresult, the more mesic and often lower elevation sites,dominated by hammocks of deciduous and evergreenbroadleaf trees, are considered to have high productiv-ity and high soil nutrient availability, as well as rareand unpredictable fire frequency (fire interval .100 yrwith variable fire severity [Ewel 1990]). The higherelevation sites on thick sand deposits are xeric andinfertile and, in the absence of fire suppression, arelikely to harbor either sandhill or scrub communities.Although soil water and nutrient availability appear tovary little between sandhills dominated by longleafpine (Pinus palustris) and scrub communities domi-nated by Quercus spp. or sand pine (P. clausa), theydiffer markedly in fire regimes (Myers 1985, 1990).The fire regime in scrub communities has been char-acterized as having a relatively frequent and predictablefrequency (on the order of every 50 yr), but also asbeing quite severe with crown fires that destroy mostof the aboveground biomass (Myers 1985, Myers1990). The sandhill fire regime has been described asboth frequent and predictable (1–15 yr) but consistslargely of low-intensity surface fires that seldom extendinto the crown (Myers 1990, Glitzenstein et al. 1995).

These contrasting fire regimes are thought to be facil-itated, at least in part, by the vegetation itself (Webber1935, Menges and Hawkes 1998, Carrington 1999).Frequent fires in the sandhill communities are encour-aged by grasses in the open understory and relativelyflammable litter of longleaf pine in the sandhill com-munities while a dense layer of evergreen shrubs de-creases fire frequency in scrub communities (Abraham-son and Hartnett 1990, Menges and Hawkes 1998).Unfortunately, precise records of fire frequency arelacking in most sites, but woody species compositionof the community itself is considered as a good indi-cator of fire frequency and intensity (Jackson et al.1999).

Based on theoretical considerations and prior em-pirical studies on the role of functional traits, we ex-pected species occurring in communities prone to fireto have mechanisms of fire resistance (e.g., thick bark[Vines 1968, Jackson et al. 1999]) or post-fire recovery(e.g., clonal spreading and rhizome resprouting [Men-ges and Kohfeldt 1995, Carrington 1999]), and species’traits related to growth and competitive ability to bepositively correlated with soil resource availability(Tilman 1984, 1988, Goldberg 1997). Differences instature, wood density, and seed size were also analyzedas important aspects of life history strategies that in-fluence species distributions in other systems (Westobyet al. 1992, Baker et al. 2003, Meinzer 2003, Reich etal. 2003). Based on previous studies linking leaf lon-gevity and specific leaf area to nutrient availability(Monk et al. 1989, Kikuzawa 1995, Reich et al. 1997,1999, Givnish 2002) or water availability (Wright etal. 2001, Westoby et al. 2002), we expected leaf level


traits, including leaf life span and period of canopyfoliation, to be negatively correlated with soil fertility,and we expected specific leaf area to be positively cor-related with soil moisture and soil fertility. We alsoexamined interspecific variation in species traits withrespect to both community affiliation and phylogeneticlineage to determine whether traits tend to show sim-ilarity within communities, within lineages or both.

The combination of quantitative data on spatial dis-tribution and community structure, species’ distribu-tions with respect to environmental factors, and spe-cies’ functional traits both in situ and in a commongarden provide us with a rare opportunity to examinewhether niche differentiation is an important factormaintaining diversity among oaks of North CentralFlorida.

METHODS

Quantification of species spatial associationsand environmental distributions

Study sites and field sampling.—Quantitative data onvegetation and soil variables were collected from ran-domly established 0.10-ha plots in three state parks innorth-central Florida. The parks included San FelascoHammock State Preserve, a 2803-ha park in AlachuaCounty with major sinkhole formations including San-chez Prairie and Church Swamp (31 plots); IchetuckneeSpring State Park, a 921-ha park bridging Columbiaand Suwanee counties and bisected by the Ichetuckneeriver (20 plots); and Manatee Springs State Park, a 960-ha park abutting the Suwanee River in Levy County(21 plots). Parks were chosen that have had minimaldisturbance in the last century, including large tracts ofold growth, and in which attempts have been made torestore the fire regimes. Random plots were establishedin each park by dropping pine needles on a topographicmap and marking where the tips fell. We then locatedthe approximate coordinates of the marks on the groundand recorded UTM (Universal Transverse Mercator)coordinates using a hand-held GPS instrument (GPS12, Garmin International, Inc., Olathe, Kansas, USA).Roads and park boundaries were excluded, as was thenorthern recent addition of San Felasco, which has ex-perienced recent logging and agricultural disturbance.Two additional plots were established at Payne’s PrairieState Preserve and Morning Side Nature Center in Ala-chua County to increase the sampling of less commonscrub species. Seventy-four plots were established intotal. Based on visual inspection of their physiogno-mies, 48 plots were in hammock, 18 plots were in san-dhill, and 8 plots were in scrub communities.

At each location, 20 3 50 m plots were establishedin an arbitrary orientation following the North CarolinaVegetation Survey Protocol (Peet et al. 1996). Withineach plot, the dbh of each oak tree over 1.3 m heightwas measured for calculation of basal area and thespecies recorded. The presence/absence of other woody

species was also recorded for each of the plots (Ap-pendix A).

Detrended correspondence analysis of species dis-tributions.—We used detrended correspondence anal-ysis (DCA), an indirect method of gradient analysis,to compute ordinations of plots and species using (1)presence/absence data for all woody species, includingoaks; (2) presence/absence of oaks only; and (3) rel-ative basal area of oaks only. DCA was performed us-ing the method of Hill (1979) with the DECORANAoption in PC-ORD (McCune and Mefford 1997). Wespecifically used indirect gradient analysis because wewanted to describe community structure and examinespecies associations independently of environmentalvariables before we determined whether environmentalvariables were correlated with the community struc-ture. Despite known limitations to DCA (e.g., Minchin1987), it is generally the most widely used and citedordination technique in the ecological literature and hasadvantages for our application in that plots and speciesare ordinated simultaneously.

We then tested whether values for soil variables,measured within each plot (see Soil nutrient and mois-ture measurements), were correlated with the plotsscores from the vegetation data for the first two axesof ordination. We also tested whether species traits,including fire frequency classification (see Speciesniche overlap), were correlated with species scores onthe first two axes of ordination.

Soil nutrient and moisture measurements.—Soilcores were collected in December 1998 from the top10 cm of the soil, excluding undecomposed litter, infour subplots within each plot for a total of 296 indi-vidual samples. Subplot samples were used to calculatea plot mean for each soil variable after analysis. Soilsamples were dried at 458C for several days and passedthrough a 3-mm mesh sieve. Samples were analyzedfor content of organic matter, calcium, nitrate N, am-monium N, labile phosphorus, potassium, and mag-nesium, as well as pH, at the soil testing laboratory ofat the University of Florida. Dried soil samples wereextracted with Mehlich I extractant (0.05 moles/L HCland 0.05 moles/L H2SO4) and the filtrate analyzed byinductively coupled argon plasma (ICAP) spectrosco-py. Organic matter content, reported as a percentageof total weight, was analyzed by a modified Walkley-Black dichromate procedure for samples below 5% or-ganic matter or by loss on ignition for samples above5% organic matter. Nitrate N and ammonium N wereextracted with KCl and analyzed colorimetrically witha lower detection limit of 0.2 mg/g. Ca, P, NH4-N andNO3-N are reported in units of mg/kg of soil.

Resin bags were also used to determine relative con-centrations of exchangeable phosphorus. These weremade from nylon hose filled with 14 g of anion andcation exchange resins (Ionac R NM-60, Sybron Chem-icals, Inc., Birmingham, New Jersey, USA) and sealedwith hot glue. They were initially soaked in 2 moles/L


KCl solution, rinsed with distilled water, and air driedfollowing procedures adapted from Binkley and Mat-son (1983). Bags were inserted into the soil in foursubplots within each plot to a depth of 5 cm in June1999 within 2 to 3 d during which no rainfall eventsoccurred. They were tied to markers with fishing lineso that they could be easily retrieved. Bags were leftin the ground for three months and removed at the endof summer within a 2–3-d period during which therewas no precipitation. Resins were extracted with 2moles/L KCl and filtrate was measured for P colori-metrically, reported in units of mg/kg of resin beads.

Soil water content integrated over 1 m depth wasmeasured within four subplots of each plot using timedomain reflectometry (TDR) and calibrated for two soiltypes (sandy soil with ,5% organic matter and sandysoil with .5% organic matter). Five sets of measure-ments taken during all seasons over a 14-mo periodbetween June 1998 and August 1999 were averagedfor each plot. Calibration of the TDR measurements isdescribed in Cavender-Bares and Holbrook (2001). Av-erage plot values of soil moisture and nutrients wereanalyzed in relation to plot scores from the ordinationanalysis.

One limitation of our study is that, for practical rea-sons, we measured soil water content rather than soilwater potential, which is more relevant to the physi-ological function of plants than soil water content. Dueto the variation of soil moisture release curves not onlyacross soil types, but also with soil depth, we did nothave a reliable way to convert soil moisture to soilwater potential. However, when we did estimate soilwater potential using soil moisture release curves forArredondo Fine Sand (Carlisle et al. 1988), the mostappropriate match for most plots, it did not significantlychange the results. Using soil water potential changesthe shapes of the distributions across the water gradi-ent, with higher resolution at drier sites and lower res-olution at wetter sites. Nevertheless, the relative orderof species along the spectrum from xeric to hydric soilsdoes not change. The shapes of the relationships be-tween species traits and the soil moisture regime alsochange depending on whether soil moisture or waterpotential is used. However, the association of traits withsoil water regime is clear in either case. Hence, noneof the major conclusions reached in the study are al-tered by this limitation.

Species niche overlap.—Pianka’s species overlap in-dex (Pianka 1973) was computed for all pairwise com-binations of oak species across seven resource levelsfor each of six soil factors. Overlaps were calculatedbased on relative basal area of species within plotsusing the following equation:

n

E EO ij ikiO 5 O 5jk kj

n n2 2(E ) (E )O Oij ik! i i

where Ojk is the overlap of species j on species k, i is

the resource level, n is the number of resource levels(n 5 7), and Eij is the proportion of basal area of speciesj in resource level i divided by the number of plots thatfall within resource level i (Lawlor 1980). Bin widthand number can influence the outcome of such anal-yses. We chose the smallest number of bins that gavegood resolution of species niche structure while min-imizing the number of empty bins. A sensitivity anal-ysis showed that results were robust to bin width andthat whether first and last bins were open or closed didnot matter. Overlap values of species pairs were testedagainst a null model in which we randomized the rawdata matrix by shuffling plot level values of relativebasal area across plots. Each randomization procedurewas run 1000 times and the significance of observedmean overlaps was tested against the distribution ofthe 1000 means of the random communities, as de-scribed above. This procedure was carried out by aprogram written in Visual Basic following methodssimilar to those described by Gotelli and Graves(1996), but it allowed us to randomize the raw datamatrix rather than the derived values of proportion ofbasal area per resource level.

Testing for diversity–resource availability relation-ships.—The number of woody species and number ofoak species per 0.10-ha plot were compared to meansoil fertility (nitrate N plus ammonium N) within plotsto determine whether diversity was related to soil fer-tility. Plots were rank ordered by their soil fertility anddivided into seven consecutive groups of 10 plots. Thenumber of woody species, the number of oak speciesand the soil fertility was then averaged for each of thegroups of plots.

Species, habitat preferences, range limits,and phenotypic traits

Species classification.—One issue that arises whenstudying promiscuous species, such as the oaks, iswhether species are actually distinct entities. In thisstudy region, species are phenotypically distinct (Kurzand Godfrey 1962), despite the tendency within thegenus for introgression to occur. In a separate study,genetic data showed little relevant intraspecific varia-tion and species monophyly was not strongly contra-dicted (Cavender-Bares et al. 2004), in contrast to pre-vious studies in other regions (e.g., Whittemore andSchaal 1991). In southern latitudes where the growingseason is long, including north-central Florida, tem-poral separation in flowering is thought to be an im-portant prezygotic isolating mechanism preventing hy-bridization (Nixon 1985, 1997). Farther north, as thelength of the growing season is reduced, such isolatingmechanisms may no longer be possible and hybridiza-tion events may increase in frequency, reducing thedistinctiveness of species. In this study region, how-ever, oak species appear to adhere well to the ecologicalspecies concept (VanValen 1976, Mishler and Dono-ghue 1982, Daghlian and Crepet 1983) and are treated


as distinct entities, following the taxonomic treatmentof Kurz and Godfrey (1962).

Species habitat preferences.—We calculated specieshabitat preferences for several soil resources as themean weighted soil moisture or nutrient value for eachspecies according to the following equation:

y x 1 y x 1 · · · 1 y x1 1 2 2 n nm* 5y 1 y 1 · · · 1 y1 2 n

where m* is the habitat preference (weighted mean);y1, y2, . . . yn are the basal area of the species withinplots 1, 2, . . . n; and x1, x2, . . . xn are the mean valuesof the environmental variable within plots 1, 2, . . . n(terBraak and Looman 1995).

Fire frequency classification.—Species were clas-sified into categories of increasing fire frequency basedon predictive models developed by Jackson et al.(1999) that link the following habitat types to fire re-gimes (see Table 1): (1) frequently flooded bottomlandforest, (2) infrequently flooded bottomland forest, (3)moist upland forest, (4) intermediate dry, (5) dry up-land forest, and (6) park-like forest/savanna (sandhill).Species were placed in these categories and assignedthe corresponding rank. No rank was given to the fivescrub species: Quercus chapmanii, Q. pumila, Q. min-ima, Q. myrtifolia, and Q. geminata.

Geographic distributions of species.—Northernrange limits of species were determined from distri-bution maps from Nixon (1997) and Burns and Honkala(1990), as described in Cavender-Bares and Holbrook(2001). Temperature and season length covary with lat-itude and range limit can effectively serve as a proxyfor season length.

Physiological and morphological measurements ofleaves of mature trees.—Measurement of several leaf-level traits were carried out on six to 10 individuals ofeach species at San Felasco Hammock State Preserveor Payne’s Prairie State Preserve. Traits measured in-clude gas exchange rates (leaf area and mass basis),laminar leaf area, specific leaf area, leaf nitrogen, andleaf chlorophyll content (data shown in Appendix B).All measurements were carried out on leaves of sunbranches in order to control for light level, which hasbeen shown to affect leaf life span and related leaf traitssignificantly (Williams et al. 1989, Reich et al. 1991,Ackerly 1996). Canopies of trees too tall to reach wereaccessed with the use of aerial lifts or scaffolding tow-ers. Five such towers were erected in steep ravine areasnot accessible with an aerial lift, allowing canopies of15 trees to be reached in this manner. All measurementsof gas exchange, SLA, total leaf N percent, and chlo-rophyll content were made during June on leaves ofthe first flush of the season. Carbon assimilation, sto-matal conductance, and transpiration were measuredwith a portable gas exchange system (LI-6400, Li-Cor,Inc., Lincoln, Nebraska, USA) under saturating light(2000 mmol·m22·s21) using the red–blue LED attach-

ment of the Li-6400. Leaf temperature was controlledat 308C. All gas exchange measurements were con-ducted in the morning between 08:00 and 11:00 hourson clear days in an effort to avoid stomatal depressionwith declining leaf water potential following methodsof Cavender-Bares and Bazzaz (2000). Five leaves ofeach individual were measured. The leaves were re-moved for measurement of laminar leaf area (leaf size),determined by measuring fresh leaf area with the pet-iole removed using a LI-3000 portable leaf area ma-chine (Li-Cor, Inc., Lincoln, Nebraska, USA). Leaveswere oven dried for one week at 658C and weighed tocalculate specific leaf area (SLA; cm2/g) and photo-synthesis on a mass basis (Amass). Percent total leaf ni-trogen (%N) was determined for three leaves of eachtree measured. Dried leaf tissue was ground, and ;200mg was used to determine %N using a Europa Ele-mental Analyzer (model ANCA-sl) attached to a Eu-ropa model 20–20 Stable Isotope Analyzer via a cap-illary interface (Europa Instruments, Crewe, UK).

Relative chlorophyll content of leaves was measuredwith a SPAD-502 chlorophyll meter (Minolta, Tokyo,Japan) on leaves at the time of leaf removal from thetree. A previous study of chlorophyll in oak leavesfound a linear relationship between relative SPAD unitsand chlorophyll content on an area basis within therange of values found in the current study. SPAD unitswere converted to chlorophyll content following meth-ods described in Cavender-Bares et al. (2000) using thefollowing equation:

y 5 5.53x 2 62.9

where y 5 chlorophyll content (mg/m2) and x 5 SPADabsorbance (relative).

Leaf life span and period of canopy foliation.—Leaflife span—the number of days a leaf survives—wascalculated for six to 10 mature trees of each speciesfrom leaf birth and death dates monitored on threebranches per tree. Only four, three, and three individ-uals were measured for Q. shumardii, Q. laurifolia, andQ. stellata, respectively, because of the difficulty inaccessing their canopies. Prior to leaf emergence,branches were flagged. After flushing occurred, theleaves were numbered with permanent ink on the tipof the leaf, and the birth date of each leaf was recorded.Each branch and each flush on each branch was colorcoded so that leaves could be counted with a spottingscope from the ground. Monthly surveys were con-ducted for a 14-mo period to determine the timing ofnew flushing events, to record the number of new leavesproduced and record the numbers of leaves still presentfrom each flush. By subtracting the number of remain-ing leaves from the number of leaves originally marked,survival time for each cohort of leaves was determinedand recorded. The monthly surveys provided a 15-dresolution for each measurement. Every severalmonths, new flushes were flagged and new leaves num-bered. Flushing and leaf-fall dates were also monitored


over a second year, and were similar to the first year;hence, only data from the first year are reported. Weused two measures of leaf life span: (1) the mean leaflife span of the first cohort of leaves flushing in springand (2) leaf life span averaged across all leaf cohorts.Species means for these two methods were highly cor-related (r 5 0.89). However, the first method was moreconsistent between the two years because it was lessdependent on the timing of the onset of cold weather.Subsequent flushes tended to get truncated by chillingevents, skewing the average. The first method is alsobetter correlated with seedling leaf life span measuredin a common environment without a cold period (r2 50.92 vs. r2 5 0.85). All analyses were done with bothmeasures, and the results were very similar. For brevity,we report only the first method.

Period of canopy foliation was averaged for six to10 individuals per species over a 2-yr period based onmonthly surveys. The period of canopy foliation wascalculated as the number of days per year that a treemaintained 50% or greater of its foliage relative tomaximum summer foliage. Canopy habit was definedby period of canopy foliation as follows: deciduous,300 d; brevideciduous $300 d and ,365 d; evergreen5 365 d (Table 1).

Whole-plant traits of mature trees.—Whole-planttraits, including asymptotic height, sapling outer barkthickness, maximum hydraulic conductance normal-ized by sapwood area (maximum Ks), whole-shoot tran-spiration normalized by sapwood area, Huber value(ratio of sapwood area to supported leaf area), nativepercent loss of conductivity (PLC, native embolism),wood density, radial growth rate, rhizome resproutingpotential, and seed mass, unless otherwise indicated,were measured for six to 10 mature trees across therange of their distributions at San Felasco or Payne’sPrairie. Data are shown in Table 1 and Appendix C.Methods for determining maximum Ks, whole shoottranspiration, Huber value, and native PLC are de-scribed in Cavender-Bares and Holbrook (2001).

Seed size.—Size of acorns was determined both byvolume and mass. Volume was calculated from radius(r) and height (h) of the fresh acorn assuming an el-lipsoid shape using the equation: 4/3pr2(h/2). Acornswere dried at 708C for 3 d and weighed with and with-out the seed coat. Weights were only recorded for un-damaged acorns. A minimum of 30 acorns wereweighed per species, although volume was measuredfor 30–100 acorns per species. For 480 acorns from 16species, an empirical relationship was determined be-tween acorn volume (V) and acorn mass with seed coatremoved (M): M 5 0.6363V 2 0.1213, R2 5 0.97. Usingthis equation, seed mass for Q. stellata was determinedfrom data collected from the U.S. Forest Service (Burnsand Honkala 1990). We report seed mass without theseed coat as we assume that it is the most critical mea-sure of seed size (Bazzaz 1997).

Asymptotic height and rhizome resprouting poten-tial.—Asymptotic tree height was estimated by mea-suring height and dbh of 50–100 individuals of all sizesfor each species and fitting an exponential function tothe data as described by Thomas and Bazzaz (1999).We found our data to be in accord with data from theU.S. Forest Service and other published sources (Burnsand Honkala 1990, Nixon 1997, Greenberg and Simons1999). We used height information from Nixon (1997)for Q. minima and Q. pumila. For rhizome resproutingpotential and clonal spreading (RhResprPot), a rankwas assigned to each species based on seedling exper-iments, observations in the field, and published sources(Webber 1935, Kurz and Godfrey 1962, Burns andHonkala 1990, Guerin 1993, Menges and Kohfeldt1995, Nixon 1997, Greenberg and Simons 1999) ac-cording to the following criteria: (1) no resproutingfrom rhizomes or roots, (2) resprouting from rhizomespossible but rare, (3) occasional vegetative reproduc-tion from rhizomes and clonal spreading in small patch-es, (4) considerable sprouting from roots and under-ground runners almost always present, including clonalspreading and dome formation.

Outer bark thickness.—We adapted methods de-scribed by Jackson et al. (1999) and Adams and Jackson(1995) to measure outer bark (rhytidome) thickness ofsaplings. Log-log regression equations of outer barkthickness vs. stem diameter were used to estimate outerbark thickness at stem diameters representing approx-imately one-quarter of the asymptotic tree height, basedon relationships between stem diameter and height.Outer bark thickness of the short-statured runner oaks,Q. pumila and Q. minima, was estimated at 2 cm stemdiameter, which corresponds to a stem height that isclose to their maximum. We collected data for ;20individuals per species for Q. myrtifolia, Q. hemispher-ica, Q. pumila, Q. nigra, Q. minima, Q. chapmanii, Q.austrina, Q. falcata, Q. laevis, Q. margaretta, Q. in-cana, and Q. geminata. Data for remaining species wastaken from Jackson et al. (1999).

Radial growth increment.—Radial growth increment(mm/yr), a measure of absolute growth rate, was cal-culated from tree ring widths from tree cores of maturetrees for 17 species. Cores were taken only from treeswith straight trunks and fully exposed canopies. Coreswere sanded until polished and viewed under a dis-secting scope (Wild M3Z, Heerbrugg, Switzerland).Tree rings were measured to the nearest 0.01 mm usinga Velmex Unislide tree ring measuring system (VelmexCompany, New York, New York, USA) and softwaredesigned specifically for collecting tree ring data (Mea-sure J2X version 2.2.6, Voortech Consulting, Holder-ness, New Jersey, USA; tree ring data bank).6 Annualgrowth increments vary with stem diameter, age, lifehistory stage and growth conditions (Clark and Clark

6 ^http://www.voortech.com/projectj2x/measureJ2XInfo.html&


1999). Moreover, stem diameters and age at maturityvary greatly among these oak species, making com-parison of growth rates difficult. In an effort to accountfor these factors, we averaged 10 consecutive growthincrements for each individual during its maximumgrowth phase after reaching reproductive age. Repro-ductive ages for each species were based on data fromBurns and Honkala (1990) and Kurz and Godfrey(1962). For the runner oaks (Q. pumila and Q. minima),which are reproductive after only a few years, the lastfive annual growth increments were used.

Wood density.—Density of sapwood was determinedfrom tree cores or small segments for six to 10 indi-viduals of the same species as explained for radialgrowth. Volume was determined for fresh samples fromradius and length measurements of the cores or by vol-ume displacement in distilled water. Dry wood masswas determined by weighing oven-dried samples.

Seedling traits from a glasshouse experiment

Because species differences in phenotypes expressedin their native habitats reflect both genetic and envi-ronmental differences, in situ field measurements ofindividuals alone do not allow clear assessment of evo-lutionary divergence. Therefore, it is important to ex-amine the association of species traits expressed undercommon garden conditions with habitat preferences de-termined in the field. Hence, in addition to leaf andwhole-plant traits of adult trees in situ, seedling traitswere determined in a glasshouse for a subset of speciesincluded in the analysis.

Growth conditions.—Plants were grown in glass-house facilities at Harvard University. Acorns werecollected in the fall of 1996 in north-central Floridafor the nine species: Q. falcata, Q. geminata, Q. hem-ispherica, Q. laevis, Q. margaretta, Q. michauxii, Q.myrtifolia, Q. nigra, and Q. myrtifolia. Q. alba, con-sidered a hammock species, whose range begins justnorth of the field study area, was also included in thecommon garden study although it was not representedin the field study. Acorns were stored in a cold roomat 58C after collection and planted in 60 3 35 3 20cm tubs with Promix potting soil (Premier Horticulture,Dorval, Quebec, Canada) in the spring of 1997. Ger-mination took ;1–2 mo. Germinants were transplantedinto individual 12-L tree pots within 1–2 wk after emer-gence. The soil in these pots included a 2-cm layer ofgravel and 2 cm of Turfus (Profile Products, BuffaloGrove, Illinois, USA) at the bottom of the pot. Potswere then filled to the surface with sand and a smallamount of peat and field soil. At the time of trans-planting, the date was recorded, and initial measure-ments were made on stem height, root and stem di-ameter, number of leaves, and length of the longestleaf. Approximately 40 plants were transplanted perspecies, depending on germination rates and viabilityof acorns. At least 30 additional plants per species weremeasured, dried, and weighed to determine allometric

relationships between measured variables and biomass.Temperature was maintained at circa 288C during theday and 208C at night.

Seedling growth rates.—Absolute and relativegrowth rates were measured on all species in the ex-periment. Absolute growth rate (AGR, g/yr) was de-termined from dry biomass. Relative growth rate(RGR, g·g21·d21) was calculated as follows:

Log (M ) 2 Log (M )e final e initialRGR 5 (1)T 2 Tfinal initial

where Mfinal and Minitial are biomass at the time of harvest(Tfinal) and the time of transplanting (Tinitial) (Hunt 1982).Allometric equations were developed relating plant pa-rameters—including stem height, stem diameter, leafnumber, and length of longest leaf—to dry biomassbased on separate harvests of at least 30 seedlings perspecies (Appendix D). These equations were used toestimate Minitial from nondestructive measurements onthe common garden plants

Leaf-level traits.—Leaf life span was determined byrecording birth and death dates of leaves from apicalshoots on ;15 individuals per species. Leaves weresurveyed weekly for a 16-mo period, providing highresolution for these data. Gas exchange was measuredon 10–15 individuals per species in late July, control-ling light level (1500 mmol·m22·s21) using a LI-6400portable photosynthesis system (Li-Cor, Lincoln, Ne-braska, USA). After photosynthesis was measured, aleaf disk of known area was removed in the same areaon the leaf where gas exchange was measured. Largeprimary and secondary veins were avoided. The diskwas weighed with a microbalance for calculation ofAmass. Leaf laminar area (leaf size), and specific leafarea (SLA) were measured for multiple leaves per in-dividual at the time of harvest in the same manner asdescribed for mature trees. Whole plant transpirationrates were calculated by multiplying mean maximumtranspiration rates per species by the average leaf sur-face area per species.

Linking functional traits, habitat variables,and community affiliations

Independent contrast correlations.—Using hypoth-eses about phylogenetic relationships between species,determined in a separate study (Cavender-Bares et al.2004), we examined the direction and magnitude ofevolutionary change in particular traits in relation toother traits or habitat preferences, following the methodoriginally proposed by Felsenstein (1985) and stan-dardized by branch lengths (Garland et al. 1992). Byreconstructing ancestral states of traits, this method al-lowed us to examine evolutionary trade-offs and pat-terns of correlated evolution among traits and habitatpreferences. Analyses were carried out using fourequally likely tree topologies in the program CACTUS(Schwilk and Ackerly 2001).7

7 ^http://www.pricklysoft.net/software/cactus.html&


FIG. 2. Scatter diagrams showing species scores in rela-tion to the first and second axes after DCA using presence/absence for (A) all woody species within plots and (B) relativebasal area of oak species. Phylogenetic lineages of each spe-cies are shown with symbols. Three distinct community typescan be identified and are indicated with different shading(shown for oaks only): scrub, black; sandhill, gray; hammock,open symbols.

Principal components analysis.—PCA for traits ofmature trees and for traits of seedlings in the commongarden were carried out using Data Desk v. 6.0 (DataDescription, Inc., Ithaca, New York, USA) to determinewhether functional traits of species group by the threecommunity types—hammock, sandhill, and scrub.Traits included in the mature tree PCA were rhizomeresprouting potential, wood density, Huber value, max-imum leaf life span, radial growth rate, asymptoticheight, seed mass, native PLC, maximum Ks, Amass, lam-inar leaf area, whole-shoot transpiration, and SLA.Traits included in the seedling PCA were SLA, seedmass, laminar size, absolute growth rate, relativegrowth rate, whole plant transpiration, Amass, and leaflife span.

Testing for species similarity within communities andlineages.—Multiple analysis of variance (MANOVA)was used to determine whether variation among speciesin (1) spatial distribution (axis scores from the DCAanalyses) and (2) habitat preferences (weighted meanvalues of each soil factors for each species) could beexplained either by the preassigned community affili-ations, treated as a fixed effect with three levels, or byphylogenetic lineage, treated as a fixed effect with threelevels. For the ordination data, first- and second-axesscores were included as dependent variables in theMANOVA, but separate tests were run for the two dif-ferent DCA analyses. For habitat preferences, speciesmean, weighted values for soil moisture, soil pH, soilnitrate N plus ammonium N, soil calcium, and organicmatter content were treated as dependent variables. Wealso used univariate ANOVAs to determine whethercommunity affiliation or lineage explained interspecificvariation in functional traits. All analyses were carriedout in Data Desk v. 6.0.

RESULTS

Quantification of species distributions

Indirect gradient analyses.—The results of the or-dination analyses based on presence/absence of all spe-cies (Fig. 2A), on presence/absence of oaks only, andon relative basal area of oaks (Fig. 2B) indicate thatthe community structure we observed is well describedby the first and second axis of ordination. For brevity,the analysis of presence/absence of oaks only is notshown, but the results closely resemble the first anal-ysis. The eigenvalues for the first axis ranged from 0.66to 0.88 and from 0.42 to 0.53 for the second axis (Table2). Both ordinations show clustering of species withineach of the three community types. In the analysis usingrelative basal area of oaks, however, there is greaterresolution in the spatial distribution of hammock spe-cies and less resolution in the distribution of scrubspecies relative to the presence/absence ordination.

Average plot values of soil moisture and all soilfertility factors, except soil pH, were significantly neg-atively correlated with the first axis of ordination in all

three analyses (Table 2). Soil moisture and soil fertilityparameters across plots covaried to a large extent (Ta-ble 3) and cannot be statistically separated. The highestcovariance was found among the cations, Ca, Mg, andK. Fire frequency classification of tree species by Jack-son et al. (1999) (Table 1) was also highly correlatedwith this axis (see also Table 6) indicating that firefrequency covaries inversely with soil moisture andresource availability. Soil moisture was the only en-vironmental parameter significantly correlated with thesecond axis in the presence/absence analyses (Fig. 2A).In the analysis using relative basal area (Fig. 2B), thespecies are distributed somewhat differently across thetwo axes and the second axis is weakly correlated with


TABLE 2. Pearson and Kendall correlation coefficients (r) of soil parameters measured in 74plots in relation to the first and second ordination axes based on presence or absence of allspecies or relative basal area of oaks within the 74 plots.

Parameters

Presence/absence:all species

Axis 1(r)

Axis 2(r)

Relativebasal area:

oaks

Axis 1(r)

Axis 2(r)

Soil moistureNH4-NNO3-NOrganic matterP (soil samples)P (resin bags)

20.5520.5820.3320.4720.3820.38

0.310.210.090.270.050.16

20.5820.4720.4120.4320.1520.27

0.190.350.040.100.260.39

CaMgKpHEigenvalues

20.5920.5720.5820.19

0.66

0.200.180.26

20.280.42

20.5620.3620.5020.12

0.88

0.320.360.480.380.53

Notes: Eigenvalues are shown below the columns for each axis. Correlation coefficients .0.3or ,20.3 are shown in boldface (df 5 73).

TABLE 3. Pearson correlation coefficients for relationships among soil parameters within plots(N 5 74).

ParameterSoil

moisture OM NO3-N NH4-N P K Mg Ca

OMNO3-NNH4-NP

0.500.290.510.36

0.640.720.41

0.510.19 0.42

KMgCapH

0.510.540.500.00

0.530.580.670.14

0.290.450.770.53

0.740.590.670.31

0.630.820.480.20

0.770.670.29

0.780.41 0.66

Notes: Coefficients .0.30 are shown in boldface. The table indicates a high degree ofcovariance among soil fertility parameters and soil moisture. Organic matter is abbreviated asOM.

pH, K, Mg, Ca, NH4 and P, due to the low fertility inscrub plots (low axis scores) and higher fertility inhammock plots, particularly those in which Q. mi-chauxii and Q. shumardii occur (high axis scores).

The presence/absence ordination analysis shows aseparation of species along the second axis at the dryend of the gradient (right hand side in Fig. 2A). Inparticular, scrub species such as Q. chapmanii, Q. myr-tifolia, and Q. geminata clearly form a distinct grouprelative to sandhill species, such as Q. laevis, and Q.margaretta, and Q. falcata. For simplicity, the non-oakspecies are not labeled, however, sandhill oak specieswere found in association with Pinus palustris and thescrub oak species, particularly Q. chapmanii and Q.myrtifolia, were found in association with Lyonia fer-ruginea and L. lucida. Hammock oak species werefound in association with a high diversity of woodyspecies, including Liquidambar styraciflua, Pinus gla-bra, Carya spp., and Magnolia spp.

Species distributions across resource axes and nicheoverlap.—Proportion of total basal area per soil mois-ture level are shown in Fig. 3 for seven resource bins.

The upper row shows sandhill species that are restrictedto dry sites, particularly Q. laevis and Q. margaretta,which only occur in the driest resource level. Q. incanaand Q. falcata occur in the first two driest resourcelevels, and Q. falcata also has a small presence in wet-ter sites. The second row shows scrub species, whichalso largely occur in dry sites, but are less restrictedthan sandhill species. Q. minima and Q. pumila (similardistribution as Q. minima, but not shown) actually oc-cur in intermediate soil moisture regimes. Q. geminatashows its highest abundance at low soil moisture butstill maintains a low presence even in the wettest soilmoisture level. Hammock species, shown in the thirdand fourth rows, all occur in the wettest soils but manyof them also occur in dry soils. Hammock species, ingeneral, tend to have broader distributions across thesoil moisture gradient than sandhill or scrub species.

Mean niche overlap of all species, based on Pianka’sniche overlap test, was significantly less than expectedacross gradients of soil moisture and pH (Table 4).Species pairwise overlap values for soil moisture aregiven in Appendix E. The same result was obtained if


FIG. 3. Plots show the proportional basal area of species across seven levels of soil moisture for 16 species. The proportionis calculated as the ratio of basal area in each level over the total basal area along the gradient; soil moisture levels shownare the midpoints of seven ranges of values. Q. pumila is not shown but has a distribution very similar to Q. minima.

soil moisture data were converted to soil water poten-tial based on soil moisture release curves for ArredondoFine Sand (Carlisle et al. 1988) and divided into sevenbins (P , 0.001; 1000 simulations . observed). Asensitivity analysis for soil moisture showed that theanalyses were robust to changes in bin width and thata two-fold increase in the number of bins still gave riseto lower niche overlap than expected. Forty-eight outof 136 total species pairs showed less overlap thanexpected across the soil moisture gradient. Only 10pairs showed more overlap than expected. Mean nicheoverlap of all species across all soil factors tested wassignificantly less than expected (P , 0.001, Table 4).Soil moisture is less important for niche partitioningamong hammock species, examined by themselves (P5 0.192, Table 4), although hammock species are par-titioning the pH gradient (P , 0.002). At the sametime, calcium is more important for niche partitioningamong hammock species (P , 0.056), than among allspecies (P , 0.49).

MANOVA results show that spatial distributions ofspecies are highly significantly explained by preas-signed community affiliations, but not by phylogeneticlineage. This result is the same whether DCA axesscores from the presence/absence analysis (Table 5) or

from the relative basal area analysis are used (notshown). Species habitat preferences with respect to soilproperties are also significantly explained by com-munity affiliation but not by phylogenetic lineage (Ta-ble 5). Species habitat preferences for soil moisture,soil nitrate and ammonium, soil calcium, organic mat-ter content, and pH differed among the communities,but not among lineages. Scheffe’s post hoc tests showthe significant difference in species soil preferencesamong communities arises from differences betweenpreferences of hammock and scrub species and betweenhammock and sandhill species, but not between sand-hill and scrub species. The same result emerges if soilproperties of scrub, hammock and sandhill plots arecompared, rather than species habitat preferences.

Functional traits of species in relation to communitystructure, habitat preferences,

and phylogenetic lineage

Species traits in relation to ordination axes.—Nativeembolism, radial growth rate, whole-shoot transpira-tion normalized by sapwood area, asymptotic height,seed mass, Amass, and maximum Ks of mature trees wereall significantly correlated with species scores for thefirst axis of ordination (Table 6). Seedling traits related


TABLE 4. Pianka’s pairwise niche overlap values averagedfor all oak species pairs and for hammock oak species onlyacross each of six soil factors (seven resource levels).

Soilfactor

Meanoverlap

Expectedoverlap P

All oak speciesSoil moisturepHCaP, exchangeableOrganic matterNO3-N 1 NH4-NAll soil factors

0.3120.3240.3940.4260.3570.4530.378

0.4290.4290.3950.4440.4130.4530.431

,0.001,0.001

0.490.330.150.56

,0.001

Hammock oak speciesSoil moisturepHCaP, exchangeableOrganic matterNO3-N 1 NH4-NAll soil factors

0.4960.3970.3690.4860.4930.6100.478

0.5330.5270.4540.5310.4760.5410.514

0.1920.0020.0560.1750.6190.9090.081

Notes: Lower observed mean overlap values than expectedare shown in boldface. Marginally significant values areshown in italics.

TABLE 5. MANOVA results testing whether communities or phylogenetic lineages, treated as main fixed effects, significantlyexplain the variance among groups of dependent variables: spatial distribution and habitat preferences.

Main effect

Spatial distribution(DCA axis scores)†

Wilks’lambda P df

Habitat preferences(soil factors)‡

Wilks’lambda P df

Phylogenetic lineage§Community\

0.8880.078

0.8070.0001

4, 266, 24

0.4770.081

0.7570.006

12, 1812, 18

† First and second axis scores for oak species from detrended correspondence analysis of occurrences of all woody specieswithin random plots.

‡ Species weighted mean values for soil moisture, organic matter content, exchangeable phosphorus, exchangeable nitrateand ammonium, pH, and calcium content.

§ Red, white, and live oaks.\ Sandhill, scrub, and hammock.

to gas exchange and growth, including absolute growthrate, whole plant transpiration, Amass, and RGR, weresignificantly correlated with the first axis of ordination.Fire-frequency classification of individual species (Ta-ble 1) was highly correlated with the first axis of or-dination for all ordination analyses (correlation ex-cludes scrub species shown in black). Both rhizomesprouting potential and sapling outer bark thicknesswere also correlated with the first axis of ordination,with hammock species having little rhizome resprout-ing potential relative to scrub species and sandhill spe-cies having greater bark protection than hammock andscrub species (Table 6, Fig. 4).

Period of canopy foliation for mature trees was cor-related with the second axis of Fig. 2A: scrub speciestend to be evergreen or brevideciduous while sandhillspecies are largely deciduous. Other leaf-level traits ofmature trees, including laminar leaf area, chlorophyllcontent, Amax, SLA, period of canopy foliation, andmean or maximum leaf life span were not correlated

with either of the DCA axes. Amax is defined as themaximum CO2 assimilation rate on a leaf-area basis,measured in the morning under saturating light at am-bient temperature and 350 mL/L CO2. Seedling leaf lifespan was also correlated with the second axis of or-dination: the two scrub species in the seedling exper-iment (Q. myrtifolia and Q. geminata) had longer-livedleaves (both evergreen) than the sandhill hill species(Q. margaretta, Q. laevis, and Q. falcata), all decid-uous. The brevideciduous scrub species, Q. pumila andQ. chapmanii, which have intermediate leaf life spans,and the sandhill species (Q. incana), which has a shortleaf life span (5.4 mo) were not included in the seedlingstudy, which accounts for the significant correlation forseedling leaf life span but not leaf life span of maturetrees.

Species traits in relation to environment.—Outerbark thickness of species was highly correlated withfire frequency ranking (r 5 0.92, P , 0.0001) for asubset of the species excluding scrub species (cf., thefire-frequency ranking by Jackson et al. [1999] doesnot include scrub species, Table 1). Sandhill specieshad significantly thicker bark than either scrub or ham-mock species, corresponding to high frequency/low se-verity ground fires in sandhill communities (Fig. 4Aand inset to 4B). The capacity of species to resproutfrom roots and rhizomes and to spread clonally wasmaximal in scrub communities, where severe canopyfires destroy the aboveground biomass, and least evi-dent in hammock communities (Fig. 4B), where canopyfires are rare. At the same time, asymptotic height wasgreatest in hammock communities, where competitionfor light is most intense, and fire frequency is very lowand unpredictable (Fig. 4C).

Species mean values for seed size and growth-relatedtraits increased with soil resource availability, shownas species habitat preferences for soil nitrate N plusammonium N and soil moisture (Fig. 5). Soil NO3-N1 NH4-N was chosen as the representative measure ofsoil fertility, as various soil measures are correlatedwith each other (Table 3). Seed mass, seedling absolutegrowth rate, mature tree growth rate (Fig. 4), were sig-nificantly correlated with both mean soil NO3-N 1


TABLE 6. Correlation coefficients for species means of trait values for mature trees and seedlings grown in a commongarden vs. axis scores for the two DCA ordinations.

Trait

Presence/absence:all species

Axis 1(r)

Axis 2(r)

Basal area:oaks

Axis 1(r)

Axis 2(r) N

Mature trees: fieldFire-frequency rankingSapling bark thicknessRhizome sprouting potentialPLCRadial growth

20.880.410.76

20.6620.60

0.8320.78

0.2520.1720.14

20.900.620.74

20.6120.68

0.120.30

20.710.400.51

1317171717

Whole-shoot transpirationAsymptotic heightSeed massAmass

Maximum Ks

Leaf size

20.5620.5320.5320.5120.4920.32

20.0820.4320.2320.4820.0720.39

20.4620.4120.4420.3920.5120.18

0.450.800.730.480.240.68

171717171717

Chlorophyll contentAmax

SLAPeriod of canopy foliationLeaf life spanWood density

20.2620.2320.18

0.050.090.12

0.2220.1920.19

0.490.440.47

20.0320.0820.2520.1120.06

0.02

20.2820.10

0.4820.5820.3520.61

171717171717

Huber value 0.18 0.16 0.21 20.40 17

Seedlings: common gardenAGRWhole-plant transpirationAmass

RGRSLA

20.8720.7720.6720.6520.31

0.140.020.430.39

20.28

20.6820.5720.7920.7220.11

0.510.61

20.0420.06

0.69

99999

LARAmax

Leaf life spanLaminar leaf area

0.5820.39

0.0520.54

20.600.500.73

20.03

0.7120.5720.2820.28

0.1120.3920.76

0.70

9999

Notes: Correlation coefficients $0.48 are shown in boldface. N is the number of species for which data were available.For mature trees, units are as follows: maximum Ks (maximum hydraulic conductance; kg H2O·s21·m21·MPa21); transpirationper sapwood area (mmol H2O·m22·s21); chlorophyll content (relative units); radial growth increment (cm/yr); fire frequencyranking (excluding scrub species; taken from Jackson et al. [1999]); asymptotic height (m); rhizome resprouting potential(index); sapling bark thickness (mm); period of canopy foliation (d); wood density (g/m3); leaf life span (d); maximumassimilation rate (Amax; mmol CO2·m22·s21), maximum assimilation rate-mass basis (Amass; mmol CO2·g21·s21); chlorophyllcontent (mg/m2); seed size (seed mass without the seed coat; g); leaf size (laminar leaf area; cm2); Huber value (ratio ofsapwood area to leaf area); native percentage loss of conductivity (PLC). For seedlings, units are: absolute growth rate (AGR;g/yr); relative growth rate (RGR; g·g21·d21); whole-plant transpiration (mmol H2O·m22·s21); laminar leaf area (cm2); SLA(cm2/g); leaf life span (d); Amax (mmol CO2·m22·s21); Amass (mmol CO2·g21·s21); leaf area ratio (LAR; m2/g).

NH4-N and mean soil moisture of species habitats. Bothmature tree radial growth and seedling RGR showed atrend of asymptotically higher values with higher re-source availability (Fig. 5E–H).

Traits known to be important in influencing watertransport and loss in other systems, including wooddensity, hydraulic conductance, native embolism(PLC), and whole-shoot transpiration, were examinedin relation to species soil moisture preferences usingstandard correlations and independent contrast corre-lations. Interestingly, using a standard correlation,wood density was not directly correlated with the soilmoisture gradient (Fig. 6A) because the trait is con-served within lineages, and live oaks have much denserwood than other species. Independent contrasts, how-ever, which allow trait conservatism within lineages tobe factored out, showed a significant decrease in wooddensity with increasing soil moisture (Fig. 6B), con-

sistent with findings in other systems (Meinzer 2003).Hydraulic conductance, native embolism (PLC), andwhole-shoot transpiration were all significantly corre-lated with soil moisture preferences and showed sig-nificant patterns of correlated evolution with habitatpreferences using independent contrasts (Fig. 6C–H).

These patterns give rise to very clear evolutionarytrade-offs, for example, between maximum height andthe ability of plants to resprout from rhizomes afterfire (Fig. 7A) or the native embolism of the xylemand the hydraulic capacity of stems (Fig. 7C), pro-viding support for a safety–efficiency trade-off (Sper-ry 1995, Hacke and Sperry 2001). At the same time,growth and height show a strong correspondence, asdocumented in other systems (Thomas 1996, Baker etal. 2003).

At the leaf level, Amass and SLA were significantlycorrelated with foliar N concentration when examined


FIG. 4. Species trait values for bark thickness (for saplings at 5 cm dbh), rhizome sprouting potential, and asymptoticheight in relation to fire regime, which varies by community. Phylogenetic lineages of species are indicated by shading. Theinset in (B) shows means for each trait within communities: height, gray circles and dotted line; rhizome resprouting, opensquares and solid line; bark, solid triangles and dashed line.

on an individual leaf basis (Appendix F). These rela-tionships varied across phylogenetic lineages, however,and were strongest within the live oak group. SLA wasnot significantly correlated with %N within the red oaksor white oaks. An overall correlation arises becauselive oaks have relatively low SLA and low %N whilewhite oaks tend to have higher %N and higher SLA(Appendix F). There is high variation in the relation-ships among leaf traits within the red oaks. In partic-ular, Q. incana behaves as an outlier species with rel-atively high Amass, high SLA, and short leaf life spans

but low %N. Indeed, Amass and %N are correlated withinthe red oaks only when Q. incana is removed from theanalysis (Appendix F). At the species level, leaf traitsincluding Amass and leaf life span were only weaklycorrelated with foliar N concentration (Table 7). Othertraits, however, including period of canopy foliationand SLA were more strongly correlated with leaf N.Yet, contrary to expectation, leaf-level traits were notcorrelated with soil fertility or soil moisture (Table 7).These results are the same regardless of which soilfactors are examined and if independent contrast cor-


FIG. 5. Correlations between species meansof four whole-plant traits in relation to soil fer-tility, as measured by nitrate plus ammoniumconcentration, or soil moisture of species’ hab-itats: (A, B) seed mass with seed coat removed;(C, D) absolute growth rates of seedlings in thecommon garden experiment; (E, F) relativegrowth rates of seedlings; and (G, H) radialgrowth rates of mature trees. In panels (E)–(H),two-parameter sigmoidal curves are fit to thedata. In panel (H), the runner oaks, Quercuspumila and Q. minima, are not included in thecurve. Symbol shapes indicate the phylogeneticgroups. Community affiliations are distin-guished by shading: scrub, black; sandhill, gray;hammock, open.

relations are used instead of standard correlations (datanot shown). If sandhill species are removed, leaf lifespan and period of canopy foliation tend to decline withincreasing soil fertility as expected, although the sta-tistical power is limited with only 13 species, and thesetrends are not significant. In contrast, leaf-level traits,with the exception of SLA, correspond well to thenorthern range limit of the species geographic distri-butions (Table 7). Southern species have longer leaflife spans and shorter periods of canopy foliation thannorthern species. In addition, southern species havesmaller leaves, lower Amass, and lower N concentrationthan northern species.

Adult traits measured in situ vs. seedling traitsin a glasshouse

The degree to which traits measured in the field re-flect intrinsic properties of species rather than respons-es to the environment is not known. However, a sub-stantial component of these traits appears to be intrin-sic, given strong correlations between seedling traitsmeasured in a common environment and mature traits

measured in the field (Appendix G). Several of the traitsmeasured on seedlings could be directly compared tothe same traits or to similar kinds of traits as thosemeasured on mature trees in the field (Appendix G).Radial growth rates of adult trees were correlated withrelative growth rates of seedlings (RGR) measured un-der standardized favorable conditions in a glasshouse(r 5 0.66; Appendix G). Seedling absolute growth rateswere highly correlated with max. hydraulic conduc-tance (r 5 0.823) and with transpiration rates on asapwood area basis of mature trees (r 5 0.86). Phys-iological correlates of plant growth rates, e.g., tran-spiration rates, SLA, and Amax, were also correlatedbetween adult trees and seedlings. Shoot-level tran-spiration rates (per unit sapwood area) of adult treeswere also positively correlated with whole plant tran-spiration of seedlings (r 5 0.90; Appendix G). SLA ofseedlings and mature trees showed a strong correlation(r 5 0.81; Appendix G), although seedlings had muchhigher SLA than mature trees. Laminar leaf area wasalso significantly correlated (r 5 0.76) but seedlingshad smaller leaves than mature trees. Light saturated


FIG. 6. Correlations between species trait means and soil moisture (left-hand panels), as well as independent contrastcorrelations for contrasts of the same traits in relation to contrasts in soil moisture preferences (right-hand panels): (A, B)wood density; (C, D) maximum hydraulic conductance normalized by sapwood area (maximum Ks); (E, F) native percentageloss of hydraulic conductivity; and (G, H) whole-shoot transpiration normalized by supporting sapwood area. Symbols forthe left-hand panels indicate phylogenetic lineages. Due to the fact that species often share traits by common descent ratherthan through independent evolution, standard correlations, which assume independence of species, may not be appropriatein making inferences about adaptation (Felsenstein 1985). Independent contrast correlations (right-hand panels) take advantageof information about the evolutionary relationships among species and determine whether evolutionary change (magnitudeand direction) is correlated between two traits (in this case, between traits and species habitat preferences). Symbols andshading represent contrasts within or between lineages. R values for the plotted contrasts are given along with a range of Pvalues for contrasts using four equally likely phylogenies.

photosynthetic rates measured on a leaf area basis(Amax) in the common garden and in the field werestrongly correlated (r 5 0.81) although mature treeshad higher photosynthetic rates than seedlings. Amass

was not correlated between seedlings and mature trees

(r 5 0.41). Thus, the significant correlation of Amax

between adult and seedlings was largely due to SLA.Seedling leaf life span was very highly correlated withleaf life span of mature trees (r 5 0.92). The relation-ship was very close to the 1:1 line, but seedling leaf


FIG. 7. Independent contrast correlations of species traits showing evolutionary trade-offs and correlations: (A) contrastsin rhizome resprouting potential in relation to contrasts in asymptotic height; (B) contrasts in radial growth in relation toheight; (C) contrasts in percentage loss of hydraulic conductivity in relation to maximum hydraulic conductivity. R valuesfor the plotted contrasts are given along with a range of P values for contrasts using four equally likely phylogenies. Symbolsand shading represent contrasts within or between lineages, as in Fig. 5.

TABLE 7. Correlation coefficients (R) for leaf traits in relation to habitat factors, including species soil preferences andnorthern range limits (N latitude) and other leaf traits.

Habitat factorsand leaf traits

Leaflife

span

Period ofcanopy

foliationLeafsize SLA Amax Amass

FoliarN (%)

Chlorophyllcontent

Habitat factorsSoil NO3-N 1 NH4-NSoil pHSoil calciumSoil PSoil moistureN range limit

20.1720.1920.2020.2520.2020.563

20.2820.4620.3220.3720.0520.723

0.570.570.590.600.350.706

0.390.410.380.060.270.405

20.020.030.050.100.120.048

0.450.540.530.200.440.568

0.700.580.690.490.350.646

20.0520.34

0.000.140.20

20.159

Leaf traitsChlorophyll contentFoliar NAmass

0.16820.45620.568

0.2220.73820.475

0.10.7330.288

20.5120.5490.33

0.59620.095

0.495

0.0640.441

0.019

Amax

SLALeaf sizePeriod of canopy foliation

0.07320.6620.38

0.829

0.17420.65120.63

0.0050.345

20.61

Note: R values .0.5 or ,20.5 are shown in boldface.

life spans were shorter than adults especially in specieswith greater leaf longevity (Appendix G), an ontoge-netic difference that follows theoretical predictions(Kikuzawa and Ackerly 1999).

Principal components analysis of species traits

Eigenvalues for axis 1 and axis 2 of the principalcomponents analysis of traits of mature trees (Fig. 8A)explained 42% and 18%, respectively, of the total var-

iation. For simplicity, only the first two axes are pre-sented. Species clustered well according to their com-munity affiliation (Fig. 8A), although Q. falcata, a talltree from the sandhill community, was more closelyassociated with hammock species than sandhill speciesin the PCA. Scrub and hammock species were separatedby differences in rhizome resprouting potential, as-ymptotic height, radial growth rate, native embolism(PLC), maximum hydraulic conductance, and Amass


FIG. 8. (A) Principal components analysis of species traits for mature trees measured in the field. The following traitswere used: leaf life span, laminar leaf area (leaf size), carbon assimilation rate on a mass basis (Amass), specific leaf area(SLA), maximum hydraulic conductance (Ks max), Huber value (Hub Val), whole-shoot transpiration divided by sapwood area(WhShTr), native percentage loss of conductivity (PLC), asymptotic tree height (height), rhizome resprouting potential(RhRespr), outer bark thickness of saplings (bark), wood density, radial growth increment, and seed mass. (B) Eigenvectorsindicating direction and magnitude of the weighting of individual traits in the PCA are shown for the same axes as in panel(A). (C) Principal components analysis of species traits for seedlings measured in a common garden. Quercus alba (AL), ahammock species, was included in the common-garden experiment, although it does not occur in the field study sites. Thefollowing seedling traits were used: leaf life span (LL), laminar leaf area (leaf size), carbon assimilation rate on a mass basis(Amass), specific leaf area (SLA), whole-plant transpiration (WhPlTr), seed mass, absolute growth rate (AGR), and relativegrowth rate (RGR). Eigenvectors of the traits are shown in panel (D). For (A) and (C), community affiliations, phylogeneticlineages, and habitat types are distinguished by the same symbols and shading as in Figs. 1 and 3. Species codes (from Table1) are shown next to symbols.

(Fig. 8B). Sandhill species were intermediate betweenhammock and scrub with respect to these traits. Speciesthat group together in the scrub community, includingQ. minima, Q. pumila, Q. myrtifolia, and Q. geminatahave higher rhizome resprouting potential, greaterwood density, shorter stature, lower SLA, higher Hubervalues, and thinner bark than species that group to-gether in the sandhill community, including Q. laevis,Q. margaretta, and Q. falcata.

The PCA for traits measured on seedlings in thecommon garden experiment also distinguished betweenthe three community types (Fig. 8C). Again, Q. falcataaligned closely to hammock species as well as sandhillspecies. Scrub and hammock species differed with re-spect to absolute growth rates, seed size and wholeplant transpiration, with hammock species having high-

er values for these traits (Fig. 8D). Sandhill and ham-mock species were separated by differences in Amass andrelative growth rates, with sandhill species having low-er values for these traits than hammock species. Scrubspecies had longer leaf life spans, lower SLA, smallerseeds, and smaller leaves than sandhill species (Fig.8D).

Taking into consideration both the mature tree andseedling analyses, hammock species are characterizedby their tall stature, fast growth rates, high hydraulicconductance, high Amass, thin bark, and low rhizomeresprouting potential. Scrub species are characterizedby high rhizome resprouting potential, low growthrates, short stature, relatively long periods of canopyfoliation (although not necessarily long leaf longevi-ties), and thin bark. Finally, sandhill species have thick


TABLE 8. Univariate ANOVAs for leaf and whole-plant traits with community (three levels) or clade (three levels) treatedas the main fixed effects (df 5 2, 14).

Trait

Community(scrub, sandhill, hammock)

MS F P

Clade(red, white, live oak)

MS F P

Amax

Chlorophyll contentVessel diametersKsmax

Huber valueSLA

4.560396.017.491.3261.309

286.9

0.2400.7810.9711.2481.3831.526

0.7900.4770.4030.3170.2830.251

71.931158

56.930.8645.638

896.4

7.6872.9064.5980.766

17.198.886

0.0060.0880.0290.484

,0.00010.003

Whole-shoot transpirationLeaf life spanLeaf sizeWood densityPLC†

1.67368251291

0.0210.058

1.7402.6632.7613.1893.268

0.2110.1050.0980.0720.068

1.87484101388

0.0490.011

2.0083.6003.058

20.950.453

0.1710.0550.079

,0.00010.645

Seed massPeriod of canopy foliationAmass

N (%)Bark thickness

1.2975762

0.0030.4120.042

3.4103.9405.2766.357

14.73

0.0620.0440.0200.011

,0.0001

0.9779484

0.00020.4750.014

2.29310.19

0.2368.5012.050

0.1380.0020.7930.0040.166

Asymptotic heightRhizome resproutingRadial growth

504.510.069.360

15.7435.2117.59

,0.0001,0.0001,0.0001

11.791.0140.528

0.1150.6430.295

0.8920.5410.749

† PLC is percentage loss of conductivity.

bark, short leaf life spans, and are intermediate in manyof the traits that distinguish hammock and scrub spe-cies.

Analysis of variance for correspondence of traitsto community or lineage affiliations

A number of whole plant traits were significantlydifferent among the three communities (Table 8), cor-roborating the clusters shown by the principal com-ponents analyses. According to Scheffe’s post hoc tests,sandhill species had significantly thicker outer bark andgreater asymptotic height than scrub species, whilescrub species had greater rhizome resprouting potentialthan the other two groups, and hammock species hadgreater radial growth rates than the other two groups.Despite low statistical power, two measures of growthrates of seedlings grown under standardized conditionsshowed differences among species in the different com-munities (highest for hammock species) when exam-ined using MANOVA (F4,12 5 2.87, P 5 0.07), butthere were no differences among phylogenetic lineages(F4,12 5 1.99, P 5 1.6). Variation in leaf-level traitswere explained somewhat better by lineage than bycommunity affiliation, with the exception of Amass (Ta-ble 8). Significant or marginally significant differencesamong lineages were found for Amax (P 5 0.006), periodof canopy foliation (P 5 0.002), foliar N concentration(P 5 0.004), SLA (P 5 0.003), leaf life span (P 50.055), and chlorophyll content (P 5 0.088). Differ-ences for period of canopy foliation and leaf N werealso found among communities, but were less signifi-cant. Amass showed significant differences only amongcommunities. For mature trees, Amass was significantlylower for scrub species than hammock species, while

period of canopy foliation was significantly longeramong scrub species relative to sandhill species andlonger in live oaks than in white oaks. In general, traitsthat were similar within communities were not similarwithin lineages, and vice versa, with the exception offoliar nitrogen concentration, which was differentamong communities and among lineages (white oakshad higher leaf N than live oaks) and period of canopyfoliation (longer for live oaks than white oaks). Vesseldiameters, Huber value (ratio of sapwood area to sup-ported leaf area), and wood density differed amonglineages, showing significant evolutionary conserva-tism, but not among communities (Table 8).

Species diversity in relation to soil resources

The number of woody species per plot increased withsoil fertility initially (from less than six to more thanten) and then remained constant with subsequent in-creases in soil fertility. In contrast, the average numberof oak species per plot remained constant (between 3and 4) across all soil fertility levels, even at very lowfertility sites.

DISCUSSION

Our results show that oak species are not randomlydistributed across the Florida landscape but show clearpatterns of ecological sorting and specialization alongthe major environmental gradients in this region, in-cluding soil moisture, soil fertility, and fire regime. Oakspecies in different habitats possess functional traitsthat are known a priori to provide advantages for sur-vival in those habitats, and there appear to be funda-mental trade-offs in life history and physiological traitsthat prevent species from performing well across all


environments. Thus, at the landscape scale, i.e., thescale at which major shifts in soil moisture, fertilityand fire regime occur (1–100 ha), maintenance of highoak diversity can be explained by niche partitioning ofthe important environmental gradients. At smaller spa-tial scales, however, other mechanisms of species co-existence must be invoked (Cavender-Bares et al.2004).

Species distributions, community structure,and niche partitioning

Soil moisture, soil fertility, and fire regime play animportant role in explaining the distributions of oakspecies in north-central Florida. All soil moisture andsoil fertility factors, with the exception of pH, werenegatively correlated with the first axis of ordinationin the detrended correspondence analyses (Table 2) in-dicating that species are differentially distributed alonga gradient of soil moisture and availability of variousnutrients, which covaried to a large extent (Table 3).These results corroborate previous studies in the regionindicating that soil resource availability, driven by dif-ferences in soil hydrology, is important in the struc-turing of forest communities (Monk 1965, 1968, Hook1984, Hodges 1995). Studies of oak species in otherregions have also shown that they are distributed dif-ferentially across gradients of soil moisture or rainfall(Monk et al. 1989, Knops and Koenig 1994, Villar-Salvador et al. 1997).

Species cluster spatially within three major com-munity types (hammock, scrub, and sandhill) that differin fire regime (Fig. 2A and B). Fire plays a major rolein maintaining oak diversity because both sandhill andscrub communities, which collectively contain eightoak species, are fire dependent. Differences in the se-verity and frequency of fire maintain these two verydistinct fire-dependent communities. In contrast, mosthammock species are not fire tolerant and they occurlargely where fire is excluded for long periods of time.The strong correlation of the first axis of ordination(Fig. 2A) with fire frequency index indicates that thesoil resource gradient and fire frequency covary as afunction of soil moisture regimes. Fire frequency andwater availability are generally inversely correlated,such that fire frequency increases as water availabilityor flooding period diminishes (Bonnicksen and Chris-tensen 1981, Swetnam and Betancourt 1990). It is like-ly that fire patterns reinforce distributions of vegetationcorresponding to water availability. However, becausethe fire frequency index does not include scrub species,the first axis of ordination (Fig. 2A) should be inter-preted only as an approximation of fire frequency, withhammock species (low axis scores) having very lowfrequency fires (on the order of every 100 yr) and scruband sandhill species (high axis scores) having higherfrequency fires (every ;50 yr for scrub; every 1–15 yrfor sandhill).

The second axis may be interpreted in terms of fireseverity, with scrub species (high axis scores) havinghigh severity canopy fires and sandhill species (lowaxis scores) having more frequent, low severity groundfires. While soil moisture is weakly correlated with thesecond axis, soil fertility is not correlated as there areno detectable differences in soil fertility between scruband sandhill communities. In a series of studies com-paring successional patterns after a range of fire ex-clusion intervals, Myers (1985, 1990) concluded thatscrub and sandhill communities are separated by theirfire regime, facilitated by the vegetation itself. Thisinterpretation is supported in the present study by theactual fire histories in these communities, to the extentthat they are known, and by current fire managementpractices. Oak species in these two communities differin various functional traits that are best interpreted toreflect adaptation to contrasting fire regimes as dis-cussed in the section of functional traits below. Ham-mock species (intermediate axis scores) incur unpre-dictable fire severity, depending on topography andfluctuations in drought and temperature (Ewel 1990).In the absence of fire, low-fertility sites can becomehammock communities that support several of the op-portunistic hammock oak species that do not survivefire, including Q. hemispherica and Q. nigra. In ad-dition, oak species that may be fire tolerant in termsof bark thickness but are also competitive in hammockcommunities (due to tall stature and relatively highgrowth and photosynthetic rates), including Q. falcataand Q. stellata (intermediate first-axis scores, Fig. 2A),may also be found in low-fertility hammock commu-nities where fire has been excluded. In contrast, somehammock oak species appear to be found only in nu-trient/cation rich hammocks (Q. michauxii, Q. shu-mardii, Q. austrina) or hydric hammocks (e.g., Q. laur-ifolia). These species tend to be large seeded, to havehigh hydraulic conductance and gas exchange rates andto be poor resprouters.

Partitioning the abiotic environments

Our results show strong evidence for niche parti-tioning across major soil resource gradients. Averageoverlap of species within plots, calculated accordingto Pianka’s niche overlap (Pianka 1973), was signifi-cantly less than expected by chance across all soil fac-tors (Table 4). The lowest overlap values, consideringall species pairs, were along soil moisture and pH gra-dients. In the north-central Florida landscape, nichedifferentiation among the soil moisture gradient cor-responds to the differentiation among fire regimes, asthese variables are coupled. Within hammock com-munities, however, differentiation across the soil mois-ture gradient is less pronounced than it is for the entirelandscape (Fig. 3, Table 4). Hammock species are stillsignificantly differentiated across the pH gradient butalso appear to be differentiated with respect to soilcalcium (Table 4). The partitioning of the pH and cal-


cium gradients among hammock species correspondsto the separation among hammock species on the sec-ond axis of the third DCA ordination (Fig. 2B), show-ing that Q. shumardii and Q. michauxii, in particular,prefer higher pH sites and sites with higher cation avail-ability than other species.

Specialization for specific habitats

Partitioning of the abiotic environment is consistentwith the hypothesis that species have suites of func-tional traits that allow them to specialize for particularranges along resource gradients where they can be mostcompetitive. A high degree of variation in the abioticenvironment can thus maintain high numbers of spe-cialists for these environments. Here we discuss evi-dence for the correspondence of critical functionaltraits of species to their habitat preferences for traitsknown a priori to be adaptive to particular environ-ments based on theoretical understanding of their func-tion and on previous empirical evidence.

Traits related to fire tolerance.—Traits related to firetolerance and post-fire recovery, such as bark thickness,rhizome resprouting potential and asymptotic heightare well distinguished among species of the three com-munity types (Table 8) and consequently to the threemajor fire regimes they experience. Species that grouptogether in scrub communities, including Q. minima,Q. pumila, Q. myrtifolia, Q. chapmanii, and Q. gemi-nata, have higher rhizome resprouting and clonalspreading than species that occur in sandhill commu-nities, including Q. laevis, Q. margaretta, Q. incana,and Q. falcata. The former group can sprout from rootsand forms clonal domes or very large and long-livedrhizomatous runners (Kurz and Godfrey 1962, Guerin1993). Q. margaretta, Q. laevis, Q. incana, and Q.virginiana can also resprout from rhizomes to someextent (Kurz and Godfrey 1962), but do not form clonaldomes or long lived runners and are more likely toresprout from aboveground stem tissue (Burns andHonkala 1990). Hammock species, in general, have lowrhizome resprouting potential, although most oaks canresprout from aboveground stems. Only Q. shumardiiand Q. michauxii show very limited resprouting ca-pacity (Burns and Honkala 1990), which may explain,in part, why they are restricted to the wettest sites.

Sandhill species have thicker outer bark as saplingsthan hammock or scrub species, made clear by thestrong correlation between sapling bark thickness andthe second DCA axis (Table 6, Fig. 2A and B) thatseparates sandhill species from scrub and hammockspecies. Outer bark thickness insulates against hightemperatures (Martin 1963, Bond and Wilgen 1996)and has been shown to affect survival during surfacefire (e.g., Uhl and Kaufmann 1990). Thick bark pro-vides an adaptive advantage to sandhill species, whichexperience frequent ground fires, particularly duringthe vulnerable juvenile life stage. Scrub species, whichexperience severe fires, do not invest in protective bark,

as all of the above ground biomass is likely to burn.They also have short stature, which minimizes loss oftissue after fire. Instead, they show a high ability torecover from severe fire by clonal spreading and re-sprouting from roots or rhizomes. Other authors haveshown that species in those contrasting fire regimesactually facilitate the perpetuation of those regimesthrough fuel loading, differences in the flammabilityof leaves and biomass, or the formation of clonal domestructures that prevent the spread of surface fires untilthere is sufficient fuel loading to promote severe crownfires (Webber 1935, Myers 1985, Guerin 1993).

Growth-related traits.—Traits related to growth rate,including radial growth rate, maximum hydraulic con-ductance, whole-shoot transpiration on a sapwood areabasis, seed mass, and AGR, are negatively correlatedwith the first axis in all three ordination analyses, suchthat species with faster growth and higher resource up-take rates tend to be found in hammock communitieswhere soil moisture and nutrients are more available(Table 6). Previous theoretical and empirical workshows that in relatively fertile soils, sites that mightbe considered ‘‘choice habitats’’ (Bazzaz 1991), com-petition for light is very high, and traits that promoterapid achievement of larger size, such as large seedsize and high absolute growth rates, are likely to befavored by selection (Tilman 1984, 1988). Suchgrowth-related traits are well correlated with specieshabitat preferences for soil resources (Fig. 5). Traitsrelated to rapid resource acquisition, including hy-draulic conductivity and whole shoot gas exchange ona sapwood area basis, were also directly correlated withresource availability (Fig. 6). Relative growth rates,measured on seedlings, and radial growth rate mea-sured on mature trees (Fig. 5E–H), showed differen-tiation among hammock communities on the one handand scrub and sandhill communities on the other. Clearevolutionary trade-offs between traits are also appar-ent. For example, species have specialized for eithertall stature and rapid mature growth rates (competitiveadvantages in resource rich environments where lightis limiting beneath the canopy) or for rapid resproutingfrom belowground roots and rhizomes (critical for re-covering from severe fire; Fig. 7A and B). Species withhigh capacity for hydraulic conductance also tended tohave high native embolism during a drought summer(Fig. 7C), indicating that oak species specialize foreither safety or efficiency in their water transport. Sim-ilar trade-offs in hydraulic architecture have been foundfor a range of species in other systems (Hacke andSperry 2001). Taken together, these trait-environmentrelationships and trait-trait correlations build a strongcase for ecological sorting and habitat specializationamong the oaks.

Leaf-level traits

Contrary to expectation and to previously docu-mented relationships in other systems (e.g., Karlsson


1992, Reich et al. 1992, Wright et al. 2001), leaf-leveltraits, including leaf life span, period of canopy foli-ation, specific leaf area, and Amax, were not correlatedwith any of the soil factors measured, including soilnitrate and ammonium or soil moisture (Table 7), al-though Amass was correlated with species habitat pref-erences for soil pH and soil calcium. The lack of cor-relation results from the phenomenon that species withcontrasting leaf phenologies often occur together, par-ticularly in hammock communities, but also to someextent in scrub communities. In a separate study, leaflife spans of oaks co-occurring within 0.10-ha plotswere found to be more contrasting than expected bychance (Cavender-Bares et al. 2004).

Given that the ranges of leaf life spans and otherleaf traits are relatively narrow relative to other studieslinking leaf traits to nutrient availability (Schlesingerand Chabot 1977, Chabot and Hicks 1982, Reich et al.1992, 1997, Karlsson 1994, Eckstein et al. 1999,Wright et al. 2001), their lack of correlation with soilfertility is not necessarily surprising. However, leaf-level traits do show correlations with leaf N concen-tration at the level of individual leaves (Appendix F)and/or at the species level (Table 7). While leaf N can-not be considered a measure of soil fertility, given com-plexities of interactions between soil moisture and nu-trient supply (Wright et al. 2001) and the conservativenature of leaf N within species and within lineages(Table 8), these results leave open the possibility thatat small spatial scales (tens of meters) species may bepartitioning microenvironments by foraging at differentdepths. No differences in predawn water potential havebeen found among species occurring together in thissystem (Cavender-Bares and Holbrook 2001), althoughfine root placement could nevertheless be occurring atdifferent depths leading to differential access to nutri-ent supplies (Goulden 1996, Hooper and Vitousek1997). Such a mechanism is testable and could con-tribute to species coexistence at small spatial scales.

Leaf longevity may also be linked to successionalstatus of the species within a given community (Mulkeyand Kitajima 1995), as well as to fire regime. Inter-estingly, the lack of correlation between leaf traits andsoil nutrients results partially from the short leaf lon-gevities and deciduous leaf habits of the sandhill spe-cies that grow in habitats with low soil resource avail-ability, but with relatively high light availability. Thisgives rise to significant or marginally significant dif-ference among community types with respect to severalleaf traits, including period of canopy foliation (Table8). Cost–benefit theory of leaf longevity at the leaf andwhole-plant level (Kikuzawa and Ackerly 1999, Giv-nish 2002) suggests that leaf-longevity is a function oftime to recover initial investment, which should beshorter under high light and nutrient availability. Fre-quent fires in sandhill communities maintain relativelyopen canopies and provide frequent pulses of nutrients(although post-fire nutrient pulses may be rather small;

Webber 1935, Abrahamson 1984, Myers 1985). Highlitter accumulation during the period of deciduousnessmay encourage fire when the trees are dormant andmore tolerant of fire (Glitzenstein et al. 1995), leadingto a positive feedback between short leaf longevity andfire frequency. In contrast to sandhill oaks, scrub oakshave more densely arranged foliage, apparently a great-er degree of self-shading, and lower daily photosyn-thetic income per unit leaf mass (J. Cavender-Bares,unpublished data). These traits are apparently coupledwith greater leaf longevity and evergreen habit as pre-dicted by cost. Evergreen habits of many scrub andsome hammock oaks would be disadvantageous wherefire occurs very frequently as in sandhill communities.

In contrast to soil fertility, the northern range limitsof species were significantly correlated with leaf lon-gevity, period of canopy foliation, and Amass (Table 7).Hence, leaf longevities may be more adapted to broaderclimatic factors, such as the length of the growing sea-son, than to local edaphic factors. As a result, oakspecies may not be optimally distributed across localsoil fertility gradients with respect to nutrient use ef-ficiency. Rather, in the assembly of the current com-munities, species may have preferentially tracked en-vironmental factors for which they had narrower tol-erances (Ackerly 2003), such as fire regime and soilmoisture. This hypothesis is supported by the fact thatspecies showed narrower niche distributions across thesoil moisture gradient than expected, but not across thenitrate and ammonium or phosphorus gradient (Table4). Leaf-level traits also tend to be conserved withinthe major lineages (Table 8), and may not be as evo-lutionarily labile as other traits, limiting the extent towhich leaf traits could have adapted through naturalselection to local soil fertility gradients.

Association of multiple traits in relationto habitat preference

The DCA ordinations show good correspondencewith the grouping of oak species into three main com-munities—hammock, sandhill, and scrub—distin-guished on the first axis by fire frequency and resourceavailability, and on the second axis by fire intensity.When the first and second axis scores of the speciesare analyzed as the dependent variables in MANOVA,the distinctiveness of the three communities are sig-nificant (Table 5). The community affiliations are alsosignificantly distinguished in terms of species prefer-ence for edaphic properties, including soil moisture andnutrient availability (Table 5). Scheffe’s post hoc anal-yses reveal that these distinctions arise from differ-ences between hammock and scrub species, as well asbetween hammock and sandhill species, rather thanfrom differences in soil preferences between scrub andsandhill species. This is consistent with other studiesthat have failed to find consistent chemical differencesin soils between the two communities, making fire re-gime the distinguishing abiotic factor (Myers 1985).


ANOVAs of individual traits show clear differencesin species from different communities for a number oftraits, particularly life history traits related to fire tol-erance and growth (Table 8, Fig. 4). In terms of adulttraits measured in situ, scrub and hammock oaks wereseparated by differences in rhizome resprouting poten-tial, height, radial growth rate, native embolism (PLC),maximum hydraulic conductance, and Amass. Sandhillspecies were intermediate between hammock and scrubspecies with respect to these traits. Rhizome resprout-ing and clonal spreading are a defining characteristicof the scrub oaks, in addition to short stature, slowgrowth, low hydraulic conductance, low rates of gasexchange, and low native embolism. In contrast, thehammock oaks are characterized by tall stature, fastgrowth, high gas exchange rates, high native embolism,and high hydraulic conductance. Sandhill species tendto have relatively tall stature, and intermediate growthand gas exchange rates. Thick bark is a critical func-tional trait that distinguishes sandhill species fromscrub and hammock species (Fig. 4).

The PCA for traits measured on seedlings in thecommon garden experiment also distinguished betweenthe three types of community affiliations. Q. falcataaligned closely to hammock species as in DCA anal-ysis. Scrub and hammock species differed with respectto absolute growth rates, and whole plant transpiration,with hammock species having higher values for thesetraits. Sandhill and hammock species were separatedby differences in Amass and relative growth rates, withsandhill species having lower values for these traitsthan hammock species. Although this analysis includesa smaller number of species, it corresponds well to PCAfor traits of mature trees and indicates that the threetypes of community affiliations are differentiated bygrowth-related traits.

Interpreting trait–environment correlations

One difficulty in interpreting trait–environment cor-relations in field studies is that the differences in traitsbetween species may be the result of both genetic dif-ferences and plasticity if species are measured in dif-ferent environments. Common garden experiments canhelp determine the extent to which trait differencesobserved among species in the field are genetic vs.environmentally induced (Givnish et al. 2004). Prac-tically, however, it is very difficult to adequately re-produce all of the representative field conditions in acommon garden. Moreover, many traits cannot be mea-sured on seedlings, making the study of long-lived spe-cies particularly challenging. In our study, the highdegree of correlation between traits measured in thefield and in the common garden indicates that trait dif-ferences among species measured on mature trees inthe field have a strong genetic component and are notmerely the result of plasticity in traits of species thathave contrasting distributions. There were importantdifferences in magnitude between seedlings and adults

for some of the traits, but the relative ranking of thespecies was similar (Appendix G). This result is im-portant given that ontogenetic shifts in traits can besignificant in oaks (Cavender-Bares and Bazzaz 2000),and it gives us confidence in the interpretation of nichedifferentiation as an explanation for co-occurrence be-cause trait differences among species appear to have agenetic basis.

The relative importance of maternal effects in thecommon garden experiment is unknown. However, theintraspecific variation in seed size, which is one mea-sure of maternal effects (Sultan 1996), was small rel-ative to the variation across species (Fig. 5). Given thatseeds were collected from a range of sites and that meanseed sizes compare well with previously publishedranges, it is likely that seed size, itself, reflects an in-trinsic property of the species. Nevertheless, the roleof maternal effects in influencing species distributionsis an important issue warranting further study.

Explaining oak diversity at smaller spatial scales

At large spatial scales (1–100 ha), environmentalheterogeneity supports a high number of oak speciesspecialized for different habitats. At smaller spatialscales, where soil moisture, fertility and fire regime arenot changing significantly (i.e., within 0.1-ha plots),other mechanisms must be invoked to explain the co-occurrence of multiple oak species. In many systems,increasing soil fertility permits increasing biologicaldiversity until a saturation point is reached (e.g., Hus-ton 1994, Cornwell and Grubb 2003). In this system,it also appears that increasing nutrient supply permitsincreases in diversity of woody plants, particularly atthe low end of the soil fertility gradient (Fig. 9). How-ever, the same pattern is not shared by the oaks, whichmaintain a constant number of oak species per 0.1-haplot across the entire soil fertility gradient (Fig. 9) aswell as across communities (not shown). This suggeststhat there is a limit to the maximum diversity of oakspecies at small spatial scales that cannot be explainedby resource availability, other abiotic factors, or com-munity composition.

We suggest that within the oak genus, phylogeneticdiversity may control the maximum number of oaksthat can occur together at small spatial scales. In arelated study, Cavender-Bares et al. (2004) providestrong evidence that these communities show a distinctpattern of phylogenetic overdispersion, whereby oakspecies from different lineages (more distantly related)are more likely to occur together than species from thesame lineage (more closely related). This can be viewedto some extent in Fig. 2A and B, which shows that inall communities, species from different phylogeneticlineages are represented. Among the regional speciespool of oaks, it appears that there is sufficient phylo-genetic diversity to allow the co-presence of three orfour oak species per plot, regardless of the abiotic en-vironment or local species composition. Within plots,


FIG. 9. Mean number of woody species (black circles) oroak species (open circles) in relation to mean soil fertility(nitrate N 1 ammonium N). All 0.1-ha plots were rank-ordered by their soil fertility and divided into seven consec-utive groups of 10 plots each. The number of woody species,the number of oak species, and the soil fertility were averagedfor each of the groups of plots to determine whether woodyspecies diversity or oak species diversity per 0.10-ha plotincreased with soil fertility.

more distantly related species may co-occur due to dif-ferences in seed maturation time (two years in red oaks,one year in white and live oaks) that may give rise toasynchrony in masting and temporal staggering of re-generation among red and white or live oak lineages.Climate variability can differentially affect seed sur-vival which may stagger regeneration among lineages.If suitable patches for seedling regeneration are lim-iting, this would reduce competition and increase thelikelihood of co-occurrence among red and white pluslive oaks (Chesson and Warner 1981, Chesson 1985).Differences in leaf longevity among live oaks and whiteoaks, or among red oaks of different lineages, may beanother mechanism that promotes co-occurrence by re-ducing competition for nutrient resources eitherthrough temporal differences in resource uptake (Sakai1992) or to differences in nutrient use efficiency cou-pled with differences in rooting depth, as discussedabove. Patterns of host specificity of pathogens or her-bivores that lead to higher density dependent mortality(Janzen 1970, Connell 1971) if too many species ofthe same phylogenetic lineage occur in close proximitycould also promote the coexistence of more distantlyrelated species. Understanding the mechanisms of co-existence at this scale requires further investigation andis the subject of ongoing research (J. Cavender-Bares,unpublished data).

Conclusions

This study provides strong evidence that at the land-scape scale niche partitioning and habitat specialization

contribute to the coexistence and maintenance of highdiversity within the oak genus in north-central Florida.We have shown that species exhibit less overlap in theirdistributions across environmental gradients than ex-pected and that species show significant differentiationin functional traits that allow them to specialize forcontrasting physical environments, particularly with re-spect to fire frequency and severity, soil moisture, andsoil fertility. Three previously described communitytypes—hammock, sandhill, and scrub—are clearly dis-tinguished by species composition, the abiotic envi-ronment, and the functional traits of the oak species inthem. Species within the three major phylogenetic lin-eages of oaks represented in Florida are distributedacross these communities and environmental gradients,exhibiting evolutionary convergence in habitat pref-erences. This pattern of phylogenetic overdispersionand evolutionary convergence is explored in detail ina related study (Cavender-Bares et al. 2004) and mayexplain coexistence of multiple oak species at smallspatial scales. While the results presented here do notpreclude other mechanisms of coexistence, they pro-vide a strong case for habitat differentiation at the land-scape level (1–100-ha scale) and an explanation forboth species diversity and functional diversity.

ACKNOWLEDGMENTS

Jimi Sadle, Jason Teisinger, and Rachel Seman are grate-fully acknowledged for their assistance in all aspects of thefield work. Eric Weis, in particular, but also Nicole Lancasterand Tiffany Fecht, were instrumental in the completion of thecommon garden experiment. We thank David Ackerly forcritical advice on the design of the study, statistical analysisof the data, and for comments on earlier versions of the man-uscript. We also thank Martin Lechowicz, Janneke Hill-RisLambers, and two anonymous reviewers for valuable com-ments which improved the manuscript. Christine Muth, Se-bastian Catovsky, and Catherine Lovelock provided valuablecommentary on earlier versions of the manuscript. AngusGholson and Terry Henkel are gratefully acknowledged forassistance with the local flora and locating field sites. TerryHenkel also helped collect acorns for the common gardenexperiment. We thank Francis Putz, Tom Sinclair, NicholasSturgeon, Mike Binford, Walter Judd, the Florida District IIPark Service, Randy Brown, and Sam Cole for valuable dis-cussions, equipment, and/or assistance in the field. Fundingwas provided by a Mellon Foundation Grant and by an NSFDissertation Improvement Grant (DEB-9801455).

LITERATURE CITED

Abrahamson, W. G., and D. C. Hartnett. 1990. Pine flatwoodsand dry prairies. Pages 103–149 in R. L. Myers and J. J.Ewel, editors. Ecosystems of Florida. University of CentralFlorida Press, Gainesville, Florida, USA.

Ackerly, D. D. 1996. Canopy structure and dynamics: inte-gration of growth processes in tropical pioneer trees. Pages619–658 in A. Smith, editor. Tropical forest plant eco-physiology. Chapman and Hall, New York, New York,USA.

Ackerly, D. D. 2003. Community assembly, niche conser-vatism and adaptive evolution in changing environments.International Journal of Plant Sciences 164:S165–S184.

Ackerly, D. D., C. A. Knight, S. B. Weiss, K. Barton, andK. P. Starmer. 2002. Leaf size, specific leaf area and mi-crohabitat distributions of chaparral woody plants: con-


trasting patterns in species level and community level anal-yses. Oecologia 130:449–457.

Adams, D. C., and J. F. Jackson. 1995. Estimating the al-lometry of tree bark. American Midland Naturalist 134:99–106.

Baker, T., M. Swaine, and D. Burslem. 2003. Variation intropical forest growth rates: combined effects of functionalgroup composition and resource availability. Perspectivesin Plant Ecology Evolution and Systematics 6:21–36.

Bazzaz, F. A. 1987. Experimental studies on the evolutionof niche in succussional plant populations: a synthesis. Pag-es 245–272 in P. J. Edwards, editor. Colonization, succes-sion and stability. Blackwell Scientific Publications, Ox-ford, UK.

Bazzaz, F. A. 1991. Habitat selection in plants. AmericanNaturalist 137:S116–S130.

Bazzaz, F. A. 1997. Allocation of resources in plants: stateof the science and critical questions. Pages 1–30 in J. Grace,editor. Plant resource allocation. Academic Press, San Di-ego, California, USA.

Binkley, D., and P. Matson. 1983. Ion exchange resin bagmethod for assessing forest soil nitrogen availability. SoilScience Society of America Journal 47:1050–1052.

Bond, W. J., and B. W. v. Wilgen. 1996. Fire and plants.Chapman and Hall, London, UK.

Bonnicksen, T. M., and N. L. Christensen. 1981. Integratingsummary. Pages 177–180 in W. A. Reiners, editor. Fireregimes and ecosystem properties. USDA Forest ServiceGeneral Technical Report WO-26.

Brokaw, N., and R. T. Busing. 2000. Niche versus chanceand tree diversity in forest gaps. Trends in Ecology andEvolution 15:183–188.

Burns, R. M., and H. H. Honkala. 1990. Silvics of NorthAmerica: hardwoods. Agriculture Handbook 654. U.S. De-partment of Agriculture, Forest Service, Washington, D.C.,USA.

Carlisle, V., F. Sodek, M. Collins, L. Hammond, and W. Har-ris. 1988. Characterization data for selected Florida soils.University of Florida, IFAS Soil Department, Soil Char-acterization Laboratory, Gainesville, Florida, USA.

Carrington, M. E. 1999. Post-fire seedling establishment inFlorida sand pine scrub. Journal of Vegetaion Science 10:403–412.

Cavender-Bares, J., D. D. Ackerly, D. A. Baum, and F. A.Bazzaz. 2004. Phylogenetic overdispersion in Floridianoak communities. American Naturalist 163:823–843.

Cavender-Bares, J., S. Apostol, I. Moya, J. M. Briantais, andF. A. Bazzaz. 1999. Chilling-induced photoinhibition intwo oak species: are evergreen leaves inherently better pro-tected than deciduous leaves? Photosynthetica 36:587–596.

Cavender-Bares, J., and F. A. Bazzaz. 2000. Changes indrought response strategies with ontogeny in Quercus rub-ra: implications for scaling from seedlings to mature trees.Oecologia 124:8–18.

Cavender-Bares, J., and N. M. Holbrook. 2001. Hydraulicproperties and freezing-induced xylem cavitation in ever-green and deciduous oaks with contrasting habitats. Plant,Cell and Environment 24:1243–1256.

Cavender-Bares, J., M. Potts, E. Zacharias, and F. A. Bazzaz.2000. Consequences of CO2 and light interactions for leafphenology, growth, and senescence in Quercus rubra.Global Change Biology 6:877–887.

Chabot, B. F., and D. Hicks. 1982. The ecology of leaf lifespans. Annual Review of Ecology and Systematics 13:329–359.

Chesson, P. L. 1985. Coexistence of competitors in spatiallyand temporally varying environments: a look at the com-bined effects of different sorts of variability. TheoreticalPopulation Biology 28:263–287.

Chesson, P. L., and R. R. Warner. 1981. Environmental var-iability promotes coexistence in lottery competitive sys-tems. American Naturalist 117:923–943.

Clark, D. A., and D. B. Clark. 1999. Assessing the growthof tropical rain forest trees: issues for forest modeling andmanagement. Ecological Applications 9:981–997.

Connell, J. H. 1971. On the role of natural enemies in pre-venting competitive exclusion in some marine animals andin rain forest trees. Pages 298–310 in P. J. d. Boer and G.R. Gradwell, editors. Dynamics of populations. Center forAgricultural Publishing and Documentation, Wageningen,The Netherlands.

Cornwell, W. K., and P. J. Grubb. 2003. Regional and localpatterns in plant species richness with respect to resourceavailability. Oikos 100:417–428.

Daghlian, C. P., and W. L. Crepet. 1983. Oak catkins, leavesand fruits from the Oligocene Catahoula formation and theirevolutionary significance. American Journal of Botany 70:639–649.

Damesin, C., S. Rambal, and R. Joffre. 1998. Co-occurrenceof trees with different leaf habit: a functional approach onMediterranean oaks. Acta Oecologica 19:195–204.

Diamond, J., and T. J. Case. 1986. Community ecology. Harp-er and Row, New York, New York, USA.

Eckstein, R. L., P. S. Karlsson, and M. Weih. 1999. Leaf lifespan and nutrient resorption as determinants of plant nu-trient conservation in temperate-arctic regions. New Phy-tologist 143:177–189.

Ewel, K. C. 1990. Swamps. Pages 281–323 in R. L. Myersand J. J. Ewel, editors. Ecosystems of Florida. Universityof Central Florida Press, Orlando, Florida, USA.

Felsenstein, J. 1985. Phylogenies and the comparative meth-od. American Naturalist 125:1–15.

Garland, T., P. H. Harvey, and A. R. Ives. 1992. Proceduresfor the analysis of comparative data using phylogeneticallyindependent contrasts. Systematic Biology 41:18–32.

Givnish, T. J. 2002. Adaptive significance of evergreen vs.deciduous leaves: solving the triple paradox. Silva Fennica36:703–743.

Givnish, T. J., R. A. Montgomery, and G. Goldstein. 2004.Adaptive radiation of photosynthetic physiology in the Ha-waiian lobeliads: light regimes, static light responses, andwhole-plant compensation points. American Journal ofBotany 91:228–246.

Glitzenstein, J. S., W. J. Platt, and D. R. Streng. 1995. Effectsof fire regime and habitat on tree dynamics in north Floridalongleaf pine savannas. Ecological Monographs 65:441–476.

Goldberg, D. E. 1997. Competitive ability: definitions, con-tingency and correlated traits. Pages 286–306 in J. L. Harp-er, editor. Plant life histories. Cambridge University Press,Cambridge, UK.

Gotelli, N. J., and G. R. Graves. 1996. Null models in ecol-ogy. Smithsonian Institution Press, Washington, D.C.,USA.

Goulden, M. L. 1996. Carbon assimilation and water-use ef-ficiency by neighboring Mediterraean-climate oaks that dif-fer in water access. Tree Physiology 16:417–424.

Greenberg, C., and R. Simons. 1999. Age, composition, andstand structure of old-growth oak sites in the Florida highpine landscape: implications for ecosystem managementand restoration. Natural Areas Journal 19:30–40.

Guerin, D. N. 1993. Oak dome clonal structure and fire ecol-ogy in a Florida longleaf pine dominated community. Bul-letin of the Torrey Botanical Club 120:107–114.

Hacke, U. G., and J. S. Sperry. 2001. Functional and eco-logical xylem anatomy. Perspectives in Plant Ecology Evo-lution and Systematics 4:97–115.

Hill, M. O. 1979. DECORANA—A FORTRAN program fordetrended correspondence analysis and reciprocal averag-ing. Cornell University, Ithaca, New York, USA.


HilleRisLambers, J., J. S. Clark, and B. Beckage. 2002. Den-sity-dependent mortality and the latitudinal gradient in spe-cies diversity. Nature 417:732–735.

Hodges, J. D. 1995. The southern bottomland hardwood re-gion and brown loam bluffs subregion. Pages 227–269 inJ. W. Barrett, editor. Regional silviculture of the UnitedStates. John Wiley and Sons, New York, New York, USA.

Hollinger, D. Y. 1992. Leaf and simulated whole-canopy pho-tosynthesis in two co-occurring tree species. Ecology 73:1–14.

Hook, D. D. 1984. Waterlogging tolerance of lowland treespecies of the south. Southern Journal of Applied Forestry8:136–149.

Hooper, D., and P. Vitousek. 1997. The effects of plant com-position and diversity on ecosystem processes. Science277:1302–1305.

Hubbell, S. 2001. The unified neutral theory of biodiversityand biogeography. Princeton University Press, Princeton,New Jersey, USA.

Hunt, R. 1982. Plant growth curves: the functional approachto plant growth analysis. Edward Arnold Publishers, Lon-don, UK.

Huston, M. A. 1994. Biological diversity: the coexistence ofspecies on changing landscapes. Cambridge UniversityPress, Cambridge, UK.

Hutchinson, G. 1957. Concluding remarks. Cold Spring Har-bor Symposium on Quantitative Biology 22:415–427.

Jackson, J. F., D. C. Adams, and U. B. Jackson. 1999. Al-lometry of constitutive defense: a model and a comparativetest with tree bark and fire regime. American Naturalist153:614–632.

Janzen, D. 1970. Herbivores and the numbers of tree speciesin tropical forests. American Naturalist 104:501–528.

Karlsson, P. S. 1992. Leaf longevity in evergreen shrubs:variation within and among European species. Oecologia91:346–349.

Karlsson, P. S. 1994. Photosynthetic capacity and photosyn-thetic nutrient-use efficiency of Rhododendron lapponicumleaves as related to leaf nutrient status, leaf age and branchreproductive status. Functional Ecology 8:694–700.

Kikuzawa, K. 1995. The basis for variation in leaf longevityof plants. Vegetatio 121:89–100.

Kikuzawa, K., and D. D. Ackerly. 1999. Significance of leaflongevity in plants. Plant Species Biology 14:39–45.

Knops, J. H., and W. D. Koenig. 1994. Water use strategiesof five sympatric species of Quercus in central coastal Cal-ifornia. Madrono 41:290–301.

Kurz, H., and R. K. Godfrey. 1962. Trees of northern Florida.University of Florida, Gainesville, Florida, USA.

Lawlor, L. R. 1980. Structure and stability in natural andrandomly constructed competitive communities. AmericanNaturalist 116:394–408.

Long, T. J., and R. H. Jones. 1996. Seedling growth strategiesand seed size effects in fourteen oak species native to dif-ferent soil moisture habitats. Trees 11:1–8.

Martin, R. E. 1963. Thermal properties of bark. Forest Prod-ucts Journal 13:419–426.

McCune, B., and M. J. Mefford. 1997. PC-ORD, multivariateanalysis of ecological data. MjM Software, GlenedenBeach, Oregon, USA.

McKane, R. B., L. C. Johnson, G. R. Shaver, K. J. Nadel-hoffer, E. B. Rastetter, B. Fry, A. E. Giblin, K. Klelland,B. L. Kwiatkowski, J. A. Laundre, and G. Murray. 2002.Resource-based niches provide a basis for plant speciesdiversity and dominance in arctic tundra. Nature 415:68–71.

Meinzer, F. C. 2003. Functional convergence in plant re-sponses to the environment. Oecologia 134:1–11.

Menges, E., and C. V. Hawkes. 1998. Interactive effects offire and microhabitat on plants of Florida scrub. EcologicalApplications 8:935–946.

Menges, E. S., and N. Kohfeldt. 1995. Life history strategiesof Florida scrub plants in relation to fire. Bulletin of theTorrey Botanical Club 122:282–297.

Minchin, P. R. 1987. An evaluation of the relative robustnessof techniques for ecological ordination. Vegetatio 69:89–107.

Mishler, B. D., and M. J. Donoghue. 1982. Species concepts:a case for pluralism. Systematic Zoology 31:491–503.

Monk, C. D. 1965. Southern mixed hardwood forests of northcentral Florida. Ecological Monographs 35:335–354.

Monk, C. D. 1966. An ecological significance of evergreen-ness. Ecology 47:504–505.

Monk, C. D. 1968. Successional and environmental rela-tionships of the forest vegetation of north central Florida.American Midland Naturalist 79:441–457.

Monk, C. D., D. W. Imm, R. L. Potter, and G. Parker. 1989.A classification of the deciduous forest of eastern NorthAmerica. Vegetatio 80:167–181.

Mulkey, S., and K. Kitajima. 1995. Photosynthetic capacityand leaf longevity in the canopy of a dry tropical forest.Selbyana 16:169–173.

Myers, R. L. 1985. Fire and the dynamic relationship betweenFlorida sandhill and sand pine scrub vegetation. Bulletinof the Torrey Botanical Club 112:241–252.

Myers, R. L. 1990. Scrub and high pine. Pages 150–193 inR. L. Myers and J. J. Ewel, editors. Ecosystems of Florida.University of Central Florida Press, Orlando, Florida, USA.

Myers, R. L., and J. J. Ewel. 1990. Ecosystems of Florida.University of Central Florida Press, Orlando, Florida, USA.

Nixon, K. C. 1985. A biosystematic study of Quercus seriesVirentes (the live oaks) with phylogenetic analyses of Fa-gales, Fagaceae and Quercus. University of Texas, Austin,Texas, USA.

Nixon, K. C. 1997. Fagaceae. Pages 436–506 in Flora ofNorth America Editorial Committee, editor. Flora of NorthAmerica north of Mexico. Oxford University Press, NewYork, New York, USA.

Peet, R. K., T. R. Wentworth, and P. S. White. 1996. TheNorth Carolina vegetation survey protocol: a flexible, mul-tipurpose method for recording vegetation composition andstructure. Castanea 63:262–274.

Pianka, E. R. 1973. The structure of lizard communities.Annual Review of Ecology and Systematics 4:53–74.

Platt, W. J., and M. W. Schwartz. 1990. Temperate hardwoodforests. Pages 194–229 in R. L. Myers and J. J. Ewel,editors. Ecosystems of Florida. University of Central Flor-ida Press, Orlando, Florida, USA.

Reich, P. B., D. S. Ellsworth, M. B. Walters, J. M. Vose, C.Gresham, J. C. Volin, and W. D. Bowman. 1999. Generalityof leaf trait relationships: a test across six biomes. Ecology80:1955–1969.

Reich, P. B., C. Uhl, M. B. Walters, and D. S. Ellsworth.1991. Leaf life span as a determinant of leaf structure andfunction among 23 Amazonian tree species. Oecologia 86:16–24.

Reich, P. B., M. B. Walters, and D. S. Ellsworth. 1992. Leaflife-span 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 94:13730–13734.

Reich, P., I. Wright, J. Cavender-Bares, J. Craine, J. Oleksyn,M. Westoby, and M. Walters. 2003. The evolution of plantfunctional variation: traits, spectra and strategies. Inter-national Journal of Plant Sciences 164:S143–S164.

Sakai, S. 1992. Asynchronous leaf expansion and sheddingin a seasonal environment: result of a competitive game.Journal of Theoretical Biology 154:77–90.


Schlesinger, W. H., and B. F. Chabot. 1977. The use of waterand minerals by evergreen and deciduous shrubs in Oke-fenokee Swamp. Botanical Gazette 138:490–497.

Schwilk, D. W., and D. D. Ackerly. 2001. Flammability andserotiny as strategies: correlated evolution in pines. Oikos94:326–336.

Sperry, J. S. 1995. Limitations on stem water transport andtheir consequences. Pages 105–124 in Plant stems. Aca-demic Press, London, UK.

Sultan, S. E. 1996. Phenotypic plasticity for offspring traitsin Polygonum persicaria. Ecology 77:1791–1807.

Swetnam, T. W., and J. L. Betancourt. 1990. Fire–SouthernOscillation relations in the southwestern United States. Sci-ence 249:1017–1020.

ter Braak, C., and C. Looman. 1995. Regression. Pages 29–77 in O. vanTongeren, editor. Data analysis in communityand landscape ecology. Cambridge University Press, Cam-bridge, UK.

Thomas, S. C. 1996. Asymptotic height as a predictor ofgrowth and allometric characteristics in Malaysian rain for-est trees. American Journal of Botany 83:556–566.

Thomas, S. C., and F. A. Bazzaz. 1999. Asymptotic heightas a predictor of photosynthetic characteristics in Malay-sian rain forest trees. Ecology 80:1607–1622.

Tilman, D. 1984. Plant dominance along an experimentalnutrient gradient. Ecology 65:1445–1453.

Tilman, D. 1988. Plant strategies and the dynamics and struc-ture of plant communities. Princeton University Press,Princeton, New Jersey, USA.

Uhl, C., and J. B. Kaufmann. 1990. Deforestation, fire sus-ceptibility, and potential tree responses to fire in the easternAmazon. Ecology 71:437–449.

VanValen, L. 1976. Ecological species, multispecies, andoaks. Taxon 25:233–239.

Villar-Salvador, P., P. Castro-Dıez, C. Perez-Rontome, and G.Montserrat-Martı. 1997. Stem xylem features in three

Quercus (Fagaceae) species along a climatic gradient in NESpain. Trees 12:90–96.

Vines, R. G. 1968. Heat transfer through bark, and the re-sistance of trees to fire. Australian Journal of Botany 16:499–514.

Webber, H. J. 1935. The Florida scrub, a fire-fighting asso-ciation. American Journal of Botany 22:344–361.

Westoby, M., D. S. Falster, A. T. Moles, P. A. Vesk, and I.J. Wright. 2002. Plant ecological strategies: some leadingdimensions of variation between species. Annual Reviewof Ecology and Systematics 33:125–159.

Westoby, M., E. Jurado, and M. Leishman. 1992. Compar-ative evolutionary ecology of seed size. Trends in Ecologyand Evolution 7:368–372.

Whittaker, R. H. 1956. Vegetation of the Great Smoky Moun-tains. Ecological Monographs 26:1–80.

Whittaker, R. H. 1975. Communities and ecosystems. Secondedition. MacMillan, New York, New York, USA.

Whittemore, A. T., and B. A. Schaal. 1991. Interspecific geneflow in sympatric oaks. Proceedings of the National Acad-emy of Sciences (USA) 88:2540–2544.

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.

Wills, C., R. Condit, R. B. Foster, and S. P. Hubbell. 1997.Strong density- and diversity-related effects help to main-tain tree species diversity in a neotropical forest. Proceed-ings of the National Academy of Sciences (USA) 94:1252–1257.

Wright, I. J., P. B. Reich, and M. Westoby. 2001. Strategyshifts in leaf physiology, structure and nutrient content be-tween species of high- and low-rainfall and high- and low-nutrient habitats. Functional Ecology 15:423–434.

Wright, S. J. 2002. Plant diversity in tropical rain forests: areview of mechanisms of species coexistence. Oecologia130:1–14.

APPENDIX A

A list of non-oak woody species used in the gradient analyses is available in ESA’s Electronic Data Archive: EcologicalArchives M074-014-A1.

APPENDIX B

A table showing leaf-level traits for 17 oak species is available in ESA’s Electronic Data Archive: Ecological ArchivesM074-014-A2.

APPENDIX C

A table showing whole-plant traits for 17 oak species is available in ESA’s Electronic Data Archive: Ecological ArchivesM074-014-A3.

APPENDIX D

A table of seedling biomass equations as a function of other measured variables for calculation of relative growth rate(RGR) is available in ESA’s Electronic Data Archive: Ecological Archives M074-014-A4.

APPENDIX E

Species pairwise niche overlap values across a soil moisture gradient are presented in ESA’s Electronic Data Archive:Ecological Archives M074-014-A5.


APPENDIX F

Graphs showing leaf trait relationships for individual leaves for Amass, SLA, and %N are presented in ESA’s ElectronicData Archive: Ecological Archives M074-014-A6.

APPENDIX GGraphs showing the relationship between traits measured on seedlings grown in a common garden and traits measured on

mature trees measured in the field are presented in ESA’s Electronic Data Archive: Ecological Archives M074-014-A7.

multiple trait associations in relation to habitat … · 2019. 3. 29. · 635 ecological...

Documents