Composition and species diversity of pine-wiregrass savannas of the Green Swamp, NorthCarolina*

Joan Walker & Robert K. Peet**Ikparlnwnr of’ Bioioxj> OlciA, Univcrsit.r~ of’ North Carolina, Chapel Hill, NC 27514, (/. ,‘$.A.

Keywords: Biomass, Coastal plain, Fire, Grassland, North Carolina, Phenology, Pinus palusrris, Primaryproduction, Savanna

Abstract

Fire-maintained, species-rich pines wiregrass savannas in the Green Swamp, North Carolina were sampledover their natural range of environmental conditions and fire frequencies. Species composition, speciesrichness, diversity (Exp H’, I/ C), and aboveground production were documented and fertilization experi-ments conducted to assess possible mechanisms for the maintenance of high species diversity in thesecommunities.

Although savanna composition varies continuously, DECORANA ordination and TWINSPAN classifi-cation of 2 I sites facilitated recognition of 3 community types: dry, mesic, and wet savannas. These savannasare remarkably species-rich with up to 42 species/O.25 m2 and 84 species/625 m*. Maximum richnessoccurred on mesic, annually burned sites. Aboveground production, reported as peak standing crop, wasonly 293 g . m 2 on a frequently burned mesic savanna but was significantly higher (375 g . m 2) on aninfrequently burned mesic site. Production values from fertilized high and low fire frequency sites wereequivalent. Monthly harvest samples showed that savanna biomass composition by species groups did notvary seasonally, but within groups the relative importance of species showed clear phenological progressions.

The variation in species richness with fire frequency is consistent with non-equilibrium theories of speciesdiversity, while phenological variation in production among similar species and the changing speciescomposition across the moisture gradient suggest the importance of equilibrium processes for maintenanceof savanna diversity.

Introduction

The savannas or grass sedge bogs of the coastalplain of the southeastern United States are charac-terized by scattered pines with an understory swardof mixed graminoids and forbs. These savannas are

well known for their floristic richness and especiallytheir numerous orchids, insectivorous plants andregional endemics (e.g. Wells, 1928; Wells &Shunk, 1928; Lemon, 1949). The descriptive re-ports of Wells and others prompted our study of thediversity of these communities in the Green Swampof North Carolina where the most extensive andbest preserved mesic savannas on the Atlantic coastare located.

Initial results showed that vascular plant speciesrichness was often near 40/ m2, a level of small-scalespecies diversity higher than any previously report-ed for North America and roughly equivalent to thehighest values reported in the world literature

Vegetatio 55, 163 179 (19X3).

0 Dr W. Junk Publishers, l’he Hague. Printed in the Nethcrt;inds.

164

(Whittaker, 1977b; Grime, 1979; Peet e/ al., 1983).The Green Swamp savannas, however, are uniqueamong communities reported to have high small-scale species diversity. Other equivalently species-rich communities have long histories of chronicgrazing or mowing. Although neither grazed normowed, these savannas have a long history of regu-lar burning. In addition, many of the species en-countered in the savannas are specific to this habi-tat and many are endemic to southeastern NorthAmerica. Few annuals and no adventives were en-countered. In contrast, previously described areasof high small-scale species diversity often containmany non-habitat-specific species (e.g. many spe-cies of the species-rich pastures in Britain and chalkgrasslands in Holland are described as also inhabit-ing roadsides and waste places; see Clapham et al.,1962; Bowen, 1968) and sometimes numerous an-nuals (Naveh & Whittaker, 1979; Shmida & Ellner,1984).

Many papers present theories to explain themaintenance of high species diversity(see reviews inWhittaker, 1977a; Tiiman, 1982; Peet PI al., 1983;Shmida & Ellner, 1984). These theories are oftendivided into two classes: equilibrium and non-equi-librium. Equilibrium theories derive from the com-petitive exclusion principle and require that specieshave significant differences in their ecologicalniches if they are to co-exist indefinitely (Schoener,1974, 1982; Grubb, 1977). Non-equilibrium theo-ries invoke external factors which interrupt com-petitive interactions and thereby allow co-existenceof species that might otherwise be unable to per-sist together (Paine, 1966; Connell, 1978, 1980;Grime, 1979; Huston, 1979; Peet PI al., 1983; Shmi-da & Ellner, 1984).

Published evidence suggests that both types ofmechanisms might be important in maintaininghigh species diversity in coastal plain savannas.Numerous studies have contrasted low diversityunburned savannas with more diverse burned sites(e.g. Heyward & Barnette, 1934; Stoddard, 1931;Lemon, 1949, 1967; Eleutarius& Jones, 1969; Vogl,1973). The maintenance of species diversity by fire,especially on low production sites, is consistentwith non-equilibrium theories (e.g. Grime, 1979;Huston, 1979). Other papers describe a strikingphenological variation among savanna species(Wells & Shunk, 1928; Folkerts, 1982; Gaddy,1982). Such differentiation may represent a form of

niche-partitioning consistent with equilibrium the-ories .

Our discovery of the extraordinary small-scalediversity of the Green Swamp savannas motivated!the research reported here. In this paper we firstdocument the variation in species composition andhigh species diversity of the Green Swamp savan-nas. We then examine evidence for two commonlysuggested mechanisms for maintenance of high di-versity in savannas: high fire frequency coupledwith low primary production - ~ a non-equilibriummechanism, and phenological differentiation anequilibrium mechanism.

Study area

At the time of European settlement mesic pinewiregrass savanna was widely distributed on fine,poorly drained, oligotrophic sands of the coastalplain from roughly the James River in Virginiasouth to Florida and then west to the MississippiRiver (Harper, 1906, 1914, 1943; Gano, 1917;Wells, 1932; Fernald, 1937; Wharton, 1978; Fol-kerts, 1982). Outlying communities floristicallysimilar to savannas could be found north into theNew Jersey pine barrens (Olsson, 1979) and west toeasternTexas(Streng& Harcombe, 1982). Region-ally, savannas can be viewed as communities oc-cupying the center of a soil moisture gradient alongwhich vegetation ranges from shrub bog throughsavannas and pine flatwoods to dry sandhills (seeWells, 1932; Christensen, 1979). Small pockets ofsavanna-like vegetation can be found in depres-sions and sinks within the flatwoods, and in seepswhere clay layers surface in the sandhill regions ofthe uppermost coastal plain (Wells & Shunk, 193 1).All of these savanna habitats have a fire cycle typi-cally of less than eight years (Christensen, 1981).

The Green Swamp, originally an area of roughly80 000 ha, is primarily an ombrotrophic peatlandwith elevation ranging from 12 to 18 m above sealevel. Emerging from the matrix of peat are lowislands of mineral soil on which savanna vegetationfrequently occurs. The islands vary in convexity,height, and drainage, though most rise less than1.5 m above the surrounding peat surface. Islandsize ranges from 1 to 20 ha. The more isolatedislands generally have lower fire frequencies andconsequently are often overgrown with shrubs. The

165

more species-rich areas are flat and poorly drainedwith the water table within a few centimeters of thesurface during spring and after heavy rains.

The mineral soils of the savannas developed fromPleistocene deposits of fine to coarse sand and clay.The most frequently encountered savanna soils areof the Leon (Aeric Haplaquod), Rains (Typic Pa-leayuult), L y n c h b u r g (Aeric Paleayuult), a n dForeston and Wrightsboro (Aquic Paleudult) se-ries. All of these soils are notably acidic and low innutrients, particularly nitrogen and phosphorus(Met7 et al., 1961; Plummer, 1963; Buol, 1973).

Climatological records are available for South-port, NC, approximately 35 km southeast of thestudy area. There the average annual precipitationof 1 3 18 mm is distributed fairly evenly throughoutthe year with the highest precipitation occurringfrom June through September. Mean annualtemperature is 16.8 ‘C with the July mean 26 ‘Cand the January mean 8 “C. Precipitation datafrom a U.S. Forest Service station 2.5 km north ofthe study site (1 600 mm/yr) indicate that annualprecipitation in the Green Swamp might beslightlyhigher than in Southport.

Like virtually ail of the coastal plain of the south-eastern United States, the Green Swamp has beensignificantly modified by man. Largeareas oforigin-al Tuxodium, N~vssa, Chamueq*puris swamp forestwere converted to pocosin by cutting and burningin the late 19th century. Subsequently largeareas ofpocosins and original savannas have been convert-ed to commercial pine plantations.

The remaining undeveloped savannas have alsobeen influenced by man, who, as archeological evi-dence suggests, has inhabited the Green Swampsavannas for at least 3 000 yr (Rights, 1947; Ward,1979). Burning was widely used by aboriginal peo-ples throughout the southeastern United States toimprove hunting (Lawson, 1714; Bartram, 1791;Maxwell, 1910; Vogl, 1973), and it is almost certainthat the early inhabitants burned these savannasregularly. Burning was continued by the Europeansettlers who burned the savannas to keep the landopen and to improve grazing conditions. (Our lim-ited information about early land use by the Euro-pean settlers in the study area comes from historicalrecords and anecdotal sources compiled by H.Mclver and R. Kologiski.) Although European set-tlements were already established near the CapeFear River in the 1660’s, it was not until the 19th

century that the Green Swamp savanna lands werevariously exploited for grazing and for tar, turpen-tine and timber production. Today, probably as aresult of these operations, few pines over40 cm dbhremain in the Swamp. Grazing by domestic stockended by the late 1930’sand turpentining by the late1940’s. Fire has continued on a 1 to 4 year cycle as amanagement toot to control fuel accumulation andto minimize growth of broadleaved trees.

M e t h o d s

The multiplicity of the questions to be addressedrequired that a variety of methods be used. It wasnecessary to sample both species composition andaboveground production of the savannas over therange of environmental conditions. Sampling witha series of nested yuadrats provided data to de-scribe species diversity at different size scales. Har-vest data were collected to test the hypothesis thatthere is phenoiogical variation in abovegroundprimary production. Biomass data also provide amore effective measure of species importance inthese complex communities than does frequency,and they provide a basis for comparison with otherstudies which relate diversity to standing crop (e.g.Grime, 1973, 1979; Al-Mufti et al., 1977; Willems,1980). Fertilizationexperimentswereused toexam-ine the interaction between increased site produc-tion and diversity. Because fire history has beenshown to affect both diversity and site production(Parrot, 1967; Christensen, I977), harvesting andfertilization experiments were conducted in siteswith histories of both annual burning and infre-quent burning.

Twenty-one savanna sites were selected to rep-resent a range of environmental conditions and firefrequencies. At each site species presence was re-corded in nested quadrats including 10 contiguoussquare0.25 m2 quadrats(= 1 X 2.5 m quadrat), a25m2, and a 625 m* quadrat. In each large quadrat fivesoil samples were collected from the top 10 cm ofmineral soil. Soils were analyzed by the North Ca-rolina State Soils Laboratory for exchangeablephosphate, potassium, calcium and magnesium,organic matter content, cation exchange capacity,

166

percent base saturation and pH. Topographic posi-tion scores were assigned, ranging from I for dryridges to4 for depressions. Soil drainage was scoredwith 0.5 for well-drained sites and 2.5 for the poorlydrained extreme. These two subjective scores weresummed to produce a single site moisture scalar.The length of the fire cycle during the last twodecades for each savanna was estimated based onfire records ofthe Nature Conservancy and FederalPaper Board Company, Inc., thedegree of isolationof the site, and the age of woody growth since thelast fire.

Six rectangular 12 X 24 m permanent plots wereestablished for measurement of net abovegroundproduction. Three plots were set up in an annuallyburned mesic savanna. One of these plots was fertil-ized two weeks after the burn with 5.4 g/ m2 N, I .3g/m2 P and 2.7 g/m2 K. Paired control and fertil-ized plots were established on a mesic savanna witha 3- 4 year fire cycle. A final control plot was estab-lished on a drier, annually burned site. All sites wereburned in February, 6 weeks before data collectionwas initiated.

Net aboveground production was estimated byharvesting biomass in 10 square 0.5 X 0.5 m quad-rats at 2 to 4 week intervals from April 7 throughSeptember 2. Five regularly spaced quadrats werecentered in alternating meters along each of twotransects. This design provided buffer strips be-tween sampled quadrats, minimized tramplingdamage, and ensured sampling across any within-plot variation. Aboveground biomass was cut towithin 0.5 cm of the soii, placed in plastic bags, andstored at 4 ‘C until sorting. Samples were sortedinto live and dead parts by species, dried at60 65 o C for 4% hr, and weighed. Production isreported as maximum standing crop biomass, ameasure which underestimates site production. Inthis report we do not account for asynchronyamong species biomass peaks, nor do we estimatebelowground production.

From mid-1979 through 1981 a major droughtoccurred on the North Carolina coastal plain (U.S.Dept. Commerce, 1977-83). During this periodmany of the savanna forbs did not flower, whichsuggested a decline in vigor and may also have ledto a decline in the available seed pool. In a season of

normal precipitation immediately following thedrought (1982), we resampled a control plot (Plot 3)to assess year to year consistency of our results andthe possible impact of the drought.

The 2 1 stand x 15 I species data matrix was ordi-nated using Detrended Correspondence Analysis(DCA: Hill, 1979a; Hill & Gauch, 1980). DCAordination of both presence/absence and frequencydata, both with and without woody species, pro-duced similar arrangements of stands and species.The effects of fire history on the ordination wereminimized by using only herbaceous species andpresence/absence data in the final analysis. In addi-tion to helping identify environmental factors high-ly correlated with major trends in species composi-tion, the ordination provided the species order for aphytosociologicai summary table (see Appendix).

The stands were divided into three communitytypes based on position on the first DCA axis.TWINSPAN (Hill, 1979b), a polythetic devisiveclustering program (see Cauch & Whittaker, 198 I;Gauch, 1982) was used to further examine possiblegroupings of stands.

Among methods for quantifying species diversi-ty, species richness, measured as species per unitarea, is probably the most computationally simpleand easily interpretable (Peet, 1974; Whittaker,1977a). The nested quadrat sampling scheme al-lowed determination of species richness at severalsize scales. Two dominance-weighted indices ofspecies diversity, Simpson’s index (C = V/>,2) andthe Shannon-Wiener index (H’= Q, logp,), werecalculated. Importance values (p,) were measuredas relative aboveground biomass for harvestedplots and relative frequency of occurrence in IO0.5 X 0.5 m quadrats for nested-quadrat samples.While C is strongly affected by the importance ofthe most abundant species. H’gives more weight tospecies of intermediate importance. We use the re-ciprocal form of C( Hill’s N2) and the exponentiatedform of tl’(Hill’s N,) which are interpretable as thenumber of equally common species that will pro-duce the observed heterogeneity (see Hill, 1973;Peet, 1974). Evenness was calculated as (N,-I)/(Nt-I) which is Alatalo’s (1981) revision of Hill’s(1973) ratio J:I,L.

Results

The DCA ordination of the twenty-one 625 mzplots is shown in Figure 1. Values of the moisturescalar strongly correlate with the first axis (Spear-man rank correlation = 0.80, ~3 < 0.00 I), which canbe interpreted as a moisture gradient from wet todry sites (see ordered species list in Appendix). Thesecond ordination axis varies with fire freyuencysuch that the annually burned sites are consistentlylocated above the infrequently burned sites.

Soil pH is negatively correlated with the first axisscore (Spearman rank correlation = 0.44). Amongenvironmental parameters soil pH is negativelycorrelated with the estimated length of the fire cy-cle, while percent organic matter, CEC, K, and Caare positively correlated with fire cycle length.These correlations were significant at the p < 0.05level .

The distributions of some groups of ecologicallyrelated species are strongly correlated with the firstDCA axis moisture gradient (Table I). The meanDCA positions of species within a group, whichindicate the centers of distribution for the groupsalong the moisture gradient, vary markedly andcorroborate the direct correlations with the axisscores. The percentages of the flora composed of

IOO-

.

50.

0 10

W E T MESIC

i

.

.

D R Y

. ..

100 ZOOFi rs t DCA Ax is

/.‘r,y. I. DC‘A ordination of presenqabsence data for non-woody species in 625 rnz quadrats. Open symbols indicate siteswith infrequent fires and filled symbols sites with annual III-es.

167

7’uhlr~ I. Distributions of species groups along the DCA ordina-tion axes. ~The first 2 columns show Spearman rank correlations

between percentages of all species in the 625 rn? quadrats whichbelong to the group and the position of the stand on the DCAaxis(* =p

168

ized height depending on the length of time sincethe last fire, can be present under all moisture con-ditions. The bunch-forming grasses grow up to 50cm in height and are largely responsible for creatingmicrotopographic variation within the communi-ties. Grass clump size varies inversely with fire fre-quency from 5 15 cm basal diameter in annuallyburned sites to 35 40 cm. Wet sites along with lowfire frequency sites have relatively pronounced ver-tical relief, up to 10 cm difference between the topsof grass hummocks and the hollows between them.

Wet sgvunnasWet savannas are found in shallow depressions

and as ecotones between mesic savanna vegetationand the shrub-dominated pocosins. In particular,wet savannas are likely to be found where fires onmesic savannas have spread into the adjacent po-cosin eliminating the dominant shrubs.

Dominant grasses of wet savannas include C’te-n i u m aromaticurn, Muhlenhergiu e.upunsa a n dSporoholus teret(fblius, all of which are also domi-nants in mesic savannas. Dichromena lutjfijliu andCurex verruco.w are sedges restricted to thesecommunities or found only rarely in wet micrositeson drier savannas. Other species which are appar-ently specific to wet savannas in the Green Swampinclude Droseru intermedia, Coreopsis firlcata,Rh~~rwhosporu chuluroc~phulu, Oxlr~/701i.~,fi.i~f~~rmis,/ris tridentutu, Aristidu QYnis, Anthutwuntiu rufb,and the shrubs Vuccinium cor~~mhosum and C,xrillurucem~floru. In contrast to drier savannas wherePinus pulustris is the only tree species present, wetsavannas support occasional individuals of Pinusserotinu. Tuxotlium usccwdens and NJ~SSU ,sJplvutic’uvar. h(f/oru.

Most species in wet savannas are confined to thegrass hummocks; the low depressions are typicallydevoid of vegetation. The extended period of winterand spring inundation in these microsites, coupledwith incomplete combustion of wet litter restrictsthe growth of small plants and inhibits most seed-ling establishment.

Dry savannas typically occur on the high, centralportion of the more dome-shaped islands (up to 1.5m above the surrounding pocosin), or where the soilis coarse textured and well drained. Total plantcover ranges between 50 and 70%, compared to over

100o/O cover in the other savanna communities. Theopen tree canopy contains only Pinuspulustris, andthe grass Aristidu stricta dominates the field layer.Although relatively uncommon on mesic and wetsites, legumes are well represented in dry savannas(Table 1, Appendix). Legumes that are apparentlyconfined to drier sites in the study area includeCussia ,fu.sciculutu, Le.speckzu cupitutu, Clitoriumuriunu, and Amorphu herhaceu. Other character-istic species (IOOYo constancy) are the woody MI,-ricu cerlferu and Smilux glutrc~, and the forbs Eu-phorhiu curtisii and Violu septemlohu. Sedges ,where present, are restricted to wet microsites.

Mesic suvunnusMesic savannas occupy an intermediate position

on the moisture gradient and are especially rich inspecies. Where present, the tree canopy consistsalmost exclusively of Pinuspulustris, though rarelydoes it exceed 40% cover. The few species that arerestricted to mesic savannas include Po!vgu/u UU-ciutu, Pinguiculu spp., Luchnocuulon unceps, Lili-urn czte.shei and .X.l,ris .smulliunu. The only specieswith lOOn/ constancy is the endemic Venus’ flytrap,Dionueu muscipulu. A large group of species withhigh constancy in mesic savannas can be identifiedin the Appendix.

Although hummock and hollow microtopog-raphy occurs on these mesic sites, the relief is lesspronounced than in wet savannas. Graminoids, themost important of which are Sporoholus teret(foli-L A S , Muhlenhergiu expunsu, Ctenium aromaticurn,Anrlropogon spp . and Rh~‘ncho.sporu plumosu,form the hummocks and dominate the herbaceouslayer. Most of the hollows are litter free and areimportant microhabitats occupied by species suchas L~I~coporlium curoliniunrrm, the sedges R!rJsn-chosporu hrevisetu and Rh.r~nchosporu chupmunii,the insectivorous I)ionat~u muscipulu and Droserucupilluris, and many species of composites. Fre-quently burned sites are dominated by numerous,small interdigitating hummocks. As litter accumu-lates in the absence of fire, plants which typicallygrow between the hummocks and even the smallhummocks largely disappear.

Species richness in savannas ranges up to 42 spe-

1 6 9

Tuhle 2. Species richness statistic\ for wet. mesic and dry savan-nas under annual and infrequent fire regimes. Mean (s.d.) spe-cies counts are ‘\hown for quadrats of0.25, I, 2.5,25 and 625 ml.

~..Moisture class: Wet Mesic Dry

-~

Fire f r e q u e n c y : H i g h 1.0~ H igh Low H igh I.owNumber o f Gtes: 2 2 8 3 3 3

Quadrat six (m?)---_____ .-

0 . 2 5 19.5 20.8 2 2 . 2 14.0 13.2 I I.0

(0.5) (1.7) (3.6) (3.2) (4.9) (5.X)I .O“ 3 2 . 2 33.6 3 5 . 2 2 5 . 9 24.3 2 2 . 4

(2.3) (2.4) (4.X) (7.3) (X.X) (9.1)2.5 43 .5 44 .5 4 6 . 5 3 4 . 7 3 3 . 7 3 1 . 7

(4.9) (2.1) (6.3) (3.2) (12.7) (12.9)25 53 .5 54 .0 5 7 . 3 50.3 4 3 . 0 4 3 . 3

(2 .1) (5 .7) (X.6) (1.1) (14 .4) (11.1)625 73.5 6X.0 79.3 7 1 . 2 5x.0 5 4 . 3

(3 .5 ) (5 .7 ) (5 .3 ) (2 .9 ) (13 .0 ) (X.1)

11 Calculated ft-om specie\ area curves obtained by regression.Species richness : A i H (log area).

ties per 0.25 m2 quadrat, to over 50 per m2, andfrom 63 to 84 (43 in site 1 1) species in 625 m* plots.For all quadrat sizes maximum richness is found inannually burned, mesic savannas (Table 2, Fig. 2).Richness for both 0.25 and 625 m2 samples in-creases with site moisture (decreasing axis 1 score)to a peak in mesic savannas, and levels off or de-creases slightly in wet sites. At the 0.25 m* scale thedry sites are only 60% as rich as annually burned,mesic savannas. This in part reflects the smallerplant size in mesic savannas, but at the 625 m* scalerichness in dry savannas is still only 73% of that onmesic s i tes .

Annually burned mesic savannas average 26%more species per m* than less frequently burnedsavannas. In wet and dry savannas, however, dif-ferences in richness with fire frequency are not sig-nificant.

i /

:

/I /

WET ’ MESIC I DRYI I .

E 4 0 / IY

/ // /

::/ II

l !

,-I .

/I .I

0 .A! ,/ ..-0 100 2 0 0

First DCA Am

Fi,q. 3. Specie\ richness as a function of fire frequency andposition on the DCA Axis I. Open symbols indicate infrequently

burned sites and filled symbols annually burned sites. Squaresindicate richness for 625 1x12 quadrats and circles the mean rich-ness in ten 0.25 rn? quadrats.

Diversit.v indicesFrequency data were used to calculate domi-

nance-weighted diversity statistics for the 6 classesof savannas (3 moisture classes under 2 fire regimes;Table 3). Consistent with the observations on spe-cies richness, the highest diversity values (I / C; ExpN’) occurred in the frequently burned mesic savan-nas. The mesic savannas also showed the greatestdrop in diversity with a shift to low fire frequency.Among infrequently burned savannas diversity washighest in wet sites. No clear differences were ap-parent in Alatalo’s evenness index. Biomass datawere used to calculate diversity indices for the sixharvested sample sites (Table 4). Results for 1979from sites 1 and 3 (selected as replicates) are nearlyidentical.

Diversity( 1 /C; Exp W’) appeared notably higherin 1982 while at the same time the number of species

7irhlcp.I. Diversity indice\ calculated using frquency in ten 0.5 x 0.5 m quadI-ats per stand in the 2 I plots sampled using nested quadrats.

Values al-c‘ means (s.d.).

Numbet- Moistut-e

plots c lass

Fire

ft-equency

Spp., 0.25 m? Hill’s N>

( I : C)

Hill’s N,

(Exp(ff’))

Ala&lo’s El.2

((Nz-l)i(N~-f))

2 wet high 19.5 (0.5) 2X.01 (2.12) 32.14 ( 2.71) 0 . 8 7 (.()I)2 wet low 20.X ( I .7) 2 9 . 9 5 (0.89) 3 3 . 9 6 ( 0.08) 0 . 8 8 1.03)8 m e s i c high 2 2 . 2 (3.6) 3 I .05 (0.27) 3 5 . 4 4 ( 4.74) 0 . 8 7 (.03)

3 me\ic low 14.0 (3.3) 2 I .24 (4.58) 25. I4 ( 4.65) 0 . 8 3 C.03)3 dry high 13.2 (4.9) 2 0 . 9 9 (X.X7) 2 5 . 0 0 (10.52) 0.x3 (.03)3 dry low I I .o (5.X) 18.73 (9.4X) 22.90 (10.52) 0 . 8 0 (.04)

170

Tab/c> 4. Diversity indices calculated using aboveground biomass in 6 IO 0.25 m2 plots. Values for richness are means (sd.).

Plot/ y e a r Moistureclass

Con/rolI;79 m e s i c3179 m e s i c3,X2 m e s i c2/ 79 dry5 79 IllCSiC

F~~rrili~r4 ( I season)4,:79 m e s i c6’79 m e s i c

Ferfilizrd (4 seasons)4,‘82 m e s i c

F i r ef requency

h i g h

h i g hh i g hh i g hl o w

h i g hIOW

h i g h

Spp.10.25 m2

3 3 . 1 (4.0)

3 I .4 ( I .X)2 6 . 0 (6.1)24.0 (3.3)

I I .5 (3.2)

3 2 . 9 (4.9)

1 3 . 3 (2.4)

29.0 (5.7)

Hill’s N2 H i l l ’ s NI Alatalo’s El.2

( 1 i 0 (Exp(W) ((Nz-l):(NI-l))

7 . 4 5 13.76 0.51X.IX 13.41 0 . 5 8

10.36 15.59 0 . 6 45 . 0 0 8 . 0 4 0 . 5 42.43 4.5 I 0 . 4 9

10.00 14.41 0 . 6 74 . 0 2 6 . 2 6 0 . 5 7

7 . 4 0 12.36 0 . 5 6

per 0.25 m* plot had dropped markedly. The datasuggest a post-drought decline in the abundance ofthe typical dominants leading to lower concentra-tion of dominance (note the decline in importanceof grasses shown in Table 6), but also a decline inthe average number of species per plot.

Fertilization resulted in no decrease in species per0.25 m* during the first year. After four years thedrop in richness was no more than on the controlplot where drought apparently caused a modestdecline.

Spatial homogeneit,y

We have observed conspicuous spatial heteroge-neity in savannas. On particularly wet or dry sites,or on sites with only infrequent fire, the dominantgrasses form large, discrete tussocks. In contrast,frequently burned mesic sites are characterized bysmall, anastomosing clumps of grasses and sedges.We attempted to quantify this difference in ho-mogeneity by calculating the mean pairwise coeffi-cient of community (see Whittaker & Gauch, 1973)between the ten 0.25 m* quadrats for each plot. ASshown in Figure 3, the mesic, frequently burnedsites had the highest homogeneit ies and the dry si tes(with one exception notable for its low diversity)had low values. Thus, on mesic sites with frequentfires, the between-patch diversity at the 0.25 m2scale is small relative to that of other savanna types.

A hoveground prou’uc’t ion

Peak aboveground hiomuss

Peak live, aboveground biomass for the six harv-

First D C A AXIS

Fix. 3. Within-community homogeneity as a function of firefrequency and position on the DCA Axis I. Homogeneity isrepresented by the mean pairwise s i m i l a r i t y o f t e n 0 . 2 5 m? quad-r a t s . O p e n s y m b o l s i n d i c a t e i n f r e q u e n t l y b u r n e d s i t e s a n d f i l l e ds y m b o l s a n n u a l l y b u r n e d s i t e s .

ested savanna plots is shown in Table 5. Peak stand-ing crop in the two annually burned mesic savannaswas estimated at 293 and 236 g/ m* for 1979, whilethe dry site had a lower value of 2 16 g/ m*. Themesic, low fire frequency site supported a signifi-cantly greater living biomass than the mesic sitesthat were annually burned.

Grasses contributed about 70% of the biomass onall sites, whereas the importance of other groupsvaried with moisture and fire frequency (Table 6).With low fire frequency, sedges, other monocots,and composites declined in relative biomass, whileother dicot herbs and shrubs increased. Differences

DRY

.

:> / l

2 0 0

Moisture cla\\: Mcsic Me\iC Dry

Fire frequency: Annual Infrequent Annual

YlX1r Plot B iomass Plot Biomass Plot Biomass

~‘~lrl/rolI979 I 73.15 5 94.18 2 5 4 . 0 0

(I 1.5) (35.4) (17.2)

1979 3 5 8 . 9 9(44.9)

1982 3 75.0x

(06.7)f-~

172

84P O A C E A E

,’oromallcum

/’

iAndropogo” spp

4/7 515 6110 7/14 912

Sampling Dote

6’ C Y P E R A C E A E

: .Sc’er’a5- :

pauc/fioro

G-

<:

:c” :

z4 i0

4/7 515 6/10 ,/I4 P/2

S a m p l i n g D o t e

?-A S T E R A C E A E I

Chondrophoro“udoro

:6-

ering but ahead of the peak for most species. Puni-cum chumuelonche, which flowers twice each sea-son, similarly peaks early.

Sedges also show variable seasonal patterns ofgrowth, but generally peak before the dominantgrasses. For these species flowering and the asso-ciated peak biomass proceeds from the small stat-ured species like Rh~~nchosporu chupmanii early inthe season to the taller canopy species like R. hre-\,i.seta and R. plur?~o.vcr in the autumn.

Discussion

Published descriptions of savanna vegetation in-clude the physiognomic sketches provided by earlybotanical explorers (Bartram, 179 1; Elliot,1 8 2 1 1824; Nash, 1895), as well as floristic analysesfeaturing species lists (e.g. Thorne, 1954; Pullen &Plummer, 1964; Eleutarius & Jones, 1969; Gaddy,1982). Only a few studies systematically describesavannas in the context of regional vegetation orinterpret variation in savanna vegetation relative toenvironmental gradients. Among these few reports,all suggest the importance of a moisture gradientand several also emphasize the significance of firefrequency (e.g. Lemon, 1949, 1967; Vogl, 1973).

Woodwell (1956) described savannas as coastalplain wetland communities. He recognized threedistinct herbaceous components which he foundassociated with Pinus pulustris or P. .serotinu onwetter sites. His Pinus/ Aristidu communities,which were found on sandy flats and ridges, overlapwith the dry savannas described here. He reportedherb layers dominated by Muhlenhergiu and An-dropogon, and compositionally similar to mesicand wet savannas occurring on clay substrates.

Both Snyder (1980) and Kologiski (1977) de-scribed geographically more restricted vegetationunits. It is clear that the sites Snyder studied in theCroatan National Forest (150 km northeast of theGreen Swamp) were not as species-rich as the GreenSwamp counterparts, resulting perhaps from drierconditions at his sites, and in part from the lack ofpropagules of typical savanna species, many ofwhich reach their northern limits in the area. Kolo-giski’s Green Swamp P. pulustris Aristidu stric-tu Rh~~nchosporu community covers the range of

173

mesic and dry sites in our study. His P. serotinu-Arist ida stricta-Rh~~nchospora type generallycoincides with our wet sites, but the name is some-what misleading considering that Arisridu strictadominates only on dry sites in the study area.

Wells & Shunk (1928) related species composi-tion to habitat factors such as seasonal variation inthe water table, soil aeration, and soil nutrient sta-tus. They described Campulosus (Ctenium), Pani-cum, and Andropogon consocies which represent agradient from the wet to the dry-mesic savannas ofour study. Our dry savannas correspond closelywith a hydric or Aristida semi-bog phase of Wells’QuercusvAristia’a associes (Wells & Shunk, 1931).

Species richness

The Green Swamp savannas are remarkable fortheir high species diversity. At a 625 m2 scale, therange of 70 to 84 species places mesic savannasamong the more species-rich communities in tem-perate North America (see Peet, 1978). At the 1 m*and 0.25 m2 scales, however, the Green Swampsavannas appear to surpass all other North Ameri-can plant communities. In over 20 000 1 m2 plotsexamined for temperate North American wood-lands and forests, species richness never exceeded17 species, and in extensive studies of North Ameri-can tall-grass prairie, richness averaged 18 m 2 andnever exceeded 28 m * (Peet et al., 1983). The aver-age richness of 35 m * in frequently burned mesicsavannas of the Green Swamp provides a strikingcontrast that begs explanation.

Species richness on a 1 .O or 0.25 m* scale equal tothat in the Green Swamp savannas has been report-ed in the world literature only in areas chronicallygrazed by domestic stock or wildlife, or regularlymowed for long periods. These communities in-clude chalk grasslands in Britain (Rorison, 1971;Grime, 1973; 1979), oligotrophic grasslands in Ja-pan (Itow, 1963) and in the Netherlands (van denBergh, 1979), the alvars of Sweden (Sjogren, 1971;Rosen & Sjogren, 1973) dune slacks in Holland(van der Maarel, 1971; van der Maarel & Leer-touwer, 1967; Thalen, 1971), Mediterranean scrubin lsreal (Whittaker, 1977b; Naveh & Whittaker,1979; Shmida & Ellner, 1984), calcareous fens inSweden (RegnCll, 1980), and wildlife-grazed grass-land in Sri Lanka (Mueller-Dombois, 1977). All ofthe examples appear to be from communities where

soil conditions are less than optimal for plantgrowth.

Non-equilibrium theories qf species diversit)>

Several authors have advanced non-equilibriumtheories of species diversity which predict that amoderate amount of disturbance maximizes speciesdiversity (e.g. Loucks, 1970; Grime, 1973; 1979;Whittaker, 1977b; Connell, 1978; Huston, 1979;Abugov, 1982). Grime (1973; 1979) proposed thatspecies richness can be interpreted as a function of‘environmental stress’ and/ or intensity of manage-ment ‘such as grazing, mowing, burning and tram-pling, which tend to prevent potentially competitivespecies from attaining maximum size and vigor’. Hehypothesized that species richness will peak at someintermediate value of these two factors. He attrib-uted low species richness in environmentally stress-ful or frequently disturbed sites to the small numberof species adapted to the imposed stress, and hehypothesized that low richness in stable and/orhighly productive sites results from the ability of afew species to pre-empt resources, thereby sup-pressing richness by competitive exclusion.

A refinement of Grime’s model has been pro-posed by Huston (1979). Huston’s model differs inthat he explicitly described an interaction betweendisturbance and rate of competitive exclusion. InHuston’s view, the frequency of population reduc-tion or disturbance which maximizes richness in aparticular community is dependent on potentialgrowth rate and thus competitive exclusion by thepotential dominants. The rate of competitive dis-placement is correlated with site favorability: goodsites yield more rapid growth and consequentlyhave more rapid extinction of competitors. Thus,rather than a single, complex gradient, Hustonproposed that two gradients are necessary to de-termine a richness response surface.

Our observations on species richness in the GreenSwamp savannas are consistent with the Grime andHuston models. Mesic, annually burned sites typi-cally have higher species richness than lessfrequent-ly burned sites (Fig. 2), a result consistent with theprediction that when the disturbance regime is re-laxed, richness will drop. As would be predictedfrom Huston’s model, the influence of fire is lesspronounced at the extremes of the moisture gra-dient where the potential for competitive exclusion

is lower. None of the savannas burn with less than al-year cycle, so we cannot directly assess the impactof more frequent disturbance. The frequency ofdisturbance in savannas contrasts with the continu-al biomass reduction typical of most other species-rich communities. The low fertility of savanna soilsenforces a slow recovery from disturbance and it isthus likely that any increase in frequency or intensi-ty of disturbance would lead to a reduction in spe-cies r ichness.

Al-Mufti et al. (1977) scaled Grime’s one-dimen-sional gradient of stress and disturbance usingaboveground standing crop. Among a variety ofherbaceous communities, they found maximumrichness to be limited to sites supporting biomass inthe range of 350 to 750 g . m *. in the Green Swampsavannas maximum richness occurs at roughly 280g . m * with high values occurring when standingcrop is as low as 165 g . m 2 (dry sites). Similarly,Willems (1980) reported richness to peak in north-western European chalk grasslands with biomassvalues between 150 and 350 g . m *. Willems attrib-uted this deviation from the Al-Mufti et al. resultsto his own harvesting at 2 cm rather than groundlevel .

The high savanna diversity at a somewhat lowerbiomass than observed by Al-Mufti ef al. can beinterpreted in an evolutionary context. Evolution-ary history should lead to a local pool of speciesadapted to prevailing ecological conditions (seeTerborgh, 1973). The savanna flora evolved underconditions of lower fertility than did the Britishcommunities where Al-Mufti ef al. worked. Whit-taker( 1977b; Naveh& Whittaker, 1980) proposed asimilar explanation when confronted with differingrichness responses to grazing intensity in differentecosystems.

The role qf.fi;re

The species-rich savannas of the Green Swampappear distinctive among communities currentlyknown to have species richness greater than 35 m *in that the only chronic disturbance is fire. Thisassociation with fire is an ancient one. It is likelythat man has burned the savannas of southeasternNorth America for more than 3 000 years (Rights,1947; Vogl, 1973; Christensen, 1981), and lightningis likely to have ignited savanna fires long before thefirst human occupation (Vogl, 1973). The large

number of species endemic to these communitiesattests to the antiquity of the assemblage.

Fire can function in several ways to enhancesavanna diversity. The most conspicuous effect isthe reduction in woody plants which would other-wise increase in size and density until the savannaswere replaced by forest or shrubland (Wells &Shunk, 1928; Lemon, 1949; Eleutarius & Jones,1969; Wells & Whitford, 1976; Christensen, 1981;Folkerts, 1982). Fire also removes grass and sedgefoliage which casts a heavy shade and which canlead to a loss of the smaller grasses and forbs whichgrow between the clumps. As many savanna speciesgrow under the grass canopy and conduct much oftheir photosynthesis by means of rosettes of leaves,failure to burn the litter layer for more than one ortwo seasons may contribute to a decline in thenumber of species occupying a typical 1 m2 area.Low light levels as found under heavy litter accum-ulations have been implicated in the loss of Dionaeutnusc~ipulu from such communities (Roberts &Oosting, 1958).

In addition to stimulating flower and seed pro-duction in many savanna species (Greene, 1935;Lemon, 1949; Parrot, 1967; Vogl, 1973), fire alsoopens microsites in which new seedlings can be-come established (Grubb, 1977). The importance offire for seedling establishment in the Green Swampsavannas has not yet been ascertained, but Lemon(1949) has described a group of readily dispersedspecies which colonize savanna sites madeavailableby fire.

Species which become established following dis-turbance are likely to experience less rigourouscompetition than might otherwise be the case. Con-sequently, species typically found in adjacent habi-tats may develop substantial populations in com-munities which they could not invade withoutperiodic fire. Peet e/ ul. (1983) used this argument topredict that average niche breadth along the mois-ture gradient should be greater in annually burnedsavannas than in the less frequently burned savan-nas. By similar reasoning beta diversity, composi-tional changes along the gradient, should decreasewith increased fire frequency. Our observation thatmany species common on annually burned savan-nas appear confined to mesic ecotones between sa-vanna and shrub bog in the absence of fire is con-sistent with this prediction.

175

Equilibrium theoric).s ofspecies richness

According to equilibrium theories, speciesshould co-exist in a community only if significantdifferences exist in their use of resources such thatcompetitive exclusion does not occur. Niche differ-entiation among species should be the rule.

The phenological patterns in savannas, in bothflowering (Wells & Shunk, 1928; Caddy, 1982) andaboveground growth (Fig. 5). may represent a formof niche differentiation resulting in reduced compe-tition. It has been suggested that in a similar wayseasonal partitioning of primary production servesto maintain species richness in North Americanprairies (Williams & Markley, 1973).

Seasonal differences in aboveground growth areparticularly pronounced among the sedges. Thisobservation could be interpreted to be the result ofcompetition, with the remaining sedges having di-vergent phenologies. However, coupling these ob-servations with species morphological data sug-gests an alternative interpretation. The sedgesappear to have adopted different strategies for co-existence with the dominant grasses. Small-stat-ured sedges which are likely to be overtopped bygrasses grow rapidly early in the season thus avoid-ing competition (e.g. RhJnchospora chapmanii, R.ciliaris), whereas larger species which effectivelycompete with the grasses for canopy space (e.g.Rh.~lncho.spora hreviseta), peak concurrently withthe grasses.

Marked phenological variation in productioncan also be seen among the composites, but this islargely due to timing of bolting and may be unrelat-ed to photosynthesis, much of which could takeplace during the winter months in the basal rosetteleaves.

A second likely form of species differentiation isin response to soil moisture. Our ordination analy-sis of the Green Swamp savannas shows a composi-tional gradient correlated with soil moisture. Sig-nificant species turnover occurs along this gradientas indicated in the Appendix. In addition, the rela-tive contributions of various species groups showpronounced trends (Table 1).

We have reported only the most prominentmechanisms of potential species differentiation insavannas. Numerous other factors arc likely to beimportant, some perhaps peculiar to one or a fewspecies. For example, a number of woody species

appear largely confined to growth on old, rottingpine logs, and Cur&us virginiana has a conspicu-ous affinity for ant hills (Wells & Shunk, 1928).Other possible factors such as root distribution andadaptations for uptake of scarce nutrients and dif-fering nutrient requirements (see Tilman, 1982) stillneed to be examined.

Conclusions

Our investigations confirm that the GreenSwamp savannas are remarkably species-rich, es-pecially at small scales, and that this richness ismaximal in sites with high fire frequency near themiddle of the moisture gradient. Both non-equili-brium and equilibrium processes appear to con-tribute to the establishment and maintenance ofhigh species diversity in these savannas.

Appendix

Constancy of species in wet, mesic and &y savan-n a s

Herbaceous species which occur in more than one and less

t h a n I7 o f the 2 I s tands are ordered by increas ing DCA f i r s t a x i sSC”ES.

Savanna type: wet Mesic DryN u m b e r o f s t a n d s : 4 II 6

Species

7 5 0 9 0 050 0 0 0 05 0 0 0 0 02 5 0 9 0 05 0 0 9 0 0

I 00 I8 0 0

100 36 0 05 0 0 9 I7

IO0 4 5 0 0100 4 5 0 00 0 I8 0 05 0 4 5 0 0

2 5 I8 0 0100 7 3 0 015 8 2 0 02 5 6 4 0 07 5 7 3 0 0

7 5 8 2 0 0100 5 5 3 3

7 5 7 5 0 07 5 2 7 17

176

A/~/xwc/iu (Continued)

Savanna type: wet McsicNumber of stands: 4 I I

Species

0 0 IX 0 07 5 91 I70 0 2 7 0 05 0 IO0 17

I 00 x 2 332 5 3 6 I7

0 0 4 5 0 00 0 7 5 0 05 0 3 6 I77 5 8 2 3 32 5 3 6 1 7

100 0 9 I7I 00 5 5 5 0

2 5 91 17

0 0 1 8 0 050 7 3 3 3

100 64 6 72.5 3 6 I75 0 4 5 I70 0 4 5 0 05 0 5 5 I7

5 0 2 1 5 0

I 00 x2 I72 5 8 2 5 0

0 0 3 6 I72 5 64 I77 5 7 3 x 3

0 0 x 2 3 3IO0 6 4 67

7 5 7 3 67

0 0 4 5 5 00 0 4 5 S O2 5 55 505 0 3 6 x 30 0 IX I7

100 4 5 6 70 0 3 6 5 00 0 8 2 6 7

0 0 36 5 00 0 0 9 5 00 0 3 6 6 7

0 0 2 7 5 00 0 55 3 35 0 0 9 160 0 2 7 6 70 0 13 I 000 0 2 7 x30 0 2 7 6 72 5 5 5 8 32 5 4 5 6 7

2 5 0 0 5 00 0 5.5 I 000 0 4 5 IO0

0 0 0 0 3 3

-

Dry6

177

References

Abugov. R., 1982. Species diversity and phasing ofdisturbance.Ecology 63: 2X9 293.

Alatalo, R.V., 1981. Problemsinthemeasurementofevennessin

ecologp. Oikos 37: 199 204.Al-Mufti, M. M., Sydes, C. I-.. Furness, S. B., Grime, J. P. &

Band, S. R., 1977. A quantitative analysis of shoot phenol-

ogy and dominance in herbaceous vegetation. .I. Ecol. 65:759 791.

Barnes, I’. W., Ticslen, L. L. & Ode, D. J., 1983. Distribution,production, and diversity of Cx- and Cd-dominated com-munities in a mixed prairie. Can. J. Bot. 61: 741 75 I

Bartram, W., 1791. Travels through North and South Carolina.Georgia, East and West Florida, etc., Facsimile ed., DoverPubl., Inc., New York.

Bergh, J. I’. van den. 1979. Changes in thecomposition of mixedpopulations of grassland species. In: M. J. A. Werger (ed.).The Study of Vegetation, pp. 59 80. Junk, The Hague.

Bowen, H. J. M.. 1968. The Flora of Berkshire, Holywell Press,

Oxford. 389 pp.Buol, S. W., 1973. Soils ofthcsouthern Statesand Puerto Rico.

USDA South. Coop. Ser. Bull. 174. I05 pp.

Christensen, N. L., 1977. Fire and soil-plant nutrient relations ina pine wiregrass savanna on the coastal plain of North Ca-rolina. Oecologia 3 I : 27 44.

Christensen, N. I.., 1979. The xcricsandhill and savanna ecosys-tems of the southeastern Atlantic Coastal Plain, U.S.A. In:H. Lieth & E. Landolt (cds.). Contributions to the Knowl-edge of Flora and Vegetation in the Carolinas. Vol. 68, pp.

246-262. VerGff. Geobot. Inst. ETH, Stiftung Riibel, Ziirich.Christensen, N. L., 1981. Fire regimes in southeastern ecosys-

tems. In: I-f. A. Mooney (~1 nl. (eds.). Fire Regimes andEcosystem Properties, pp. 112 136. USDA For. Ser. Gen.

Tech. Rpt. WO-26.Clapham, A. R., Tutin. T. G. & Warburg, E. F., 1962. Flora of

the Br-itish Isles, 2nd ed. Cambridge Univ. Press, Cambridge.

I269 pp.Connell, J. H., 1978. Diversity in tropical rain forests and coral

reefs. Science 199: 1302 1310.

Connell. .I. H., 1980. Diversity and the coevolution of competi-tors, or the ghost of competition past. Oikos 35: I31 138.

Eleutarius, 1.. N. &Jones. S. B., 1969. A floristic and ecological

study of pitcher plant bogs in south Mississippi. Rhodora71:29 34.

Elliot, S., IX21 1824. A Sketch ofthe BotanyofSouth-Carolinaand Geol-gia. .I. R. Schenk, Charleston. S. C.

Fernald, M. L., 1937. Local plants of the inner coastal plain ofsoutheastern Virginia. III. Phytogeographical considera-

tions. Ii hodora 39: 465 49 IFolkerts, G. W., 1982. The Gulf coast pitcher plant bogs. Am.

Sci. 70: 260 267.

Caddy, L. I-., 1982. The floristics of three South Carolina pinesavannahs. Castanea 47: 393 402.

Gano, I.., 1917. A study in physiographic ecology in northernFlorida. Bot. Ga7. 63: 337 372.

Gauch, H. G., 1982. Multivariate Analysis in Community Ecol-ogy. Carnbl-idge Univ. PI-ess, Cambridge. 29X pp.

Gauch. H. G. & Whittaker, R. H., 1981. Hierarchical classifica-

tion of community data. J. Ecol. 69: 537 557.Greene. S. W., 1935. Relation between winter grass fires and

cattle graring in the longleaf pine belt. J. For. 33: 339 341.

Grime, .I. P., 1973. Control of species density in herbaccousvegetation. J. Environ. Manage. I: 151.-167.

Grime, J. I’., 1979. Plant Strategies and Vegetation Processes.

Wiley, New York. 222 pp.Grubb. I’. J., 1977. The maintenance ofspecies-richness in plant

communities: the importance ofthe regeneration niche. Biol.Rev. 52: 107 145.

Harper, R. M., 1906. A phytogeographical sketch of the Alta-maha Grit Region of the Coastal Plain of Georgia. Ann.N.Y. Acad. 17: I 415.

Harper, R. M., 1914. Geography and vegetation of northernFlorida. Ann. Rep. Florida State Geol. Surv. 6: 167 -451.

Harper, R. M., 1943. Forests of Alabama. Geol. Survey ofAlabama Monogr. 10. Univ. of Alabama.

Heyward, F. & Barnette, R. M., 1934. Effect offrequent fires on

chemical composition of forest soils in the longleaf pineregion. Florida Ag. Expt. Sta. Tech. Bull. 265.

Hill, M. O., 1973. Diversity and evenness: a unifying notation

and its consequences. Ecology 54: 427 432.Hill, M. O., 1979a. DECORANA. Cornell Ecology Program,

Cornell Univ., Ithaca, New York.Hill, M. O., 1979b. TWINSPAN. Cornell Ecology Program,

Cornell Univ., Ithaca, New York.

Hill, M. 0. & Gauch, H. G., 1980. Detrended correspondenceanalysis: an improved ordination technique. Vegetatio 42:47 58.

Huston, M., 1979. Ageneral hypothesis ofspeciesdiversity. Am.Nat. 113: 81 101.

Itow, S., 1963. Grassland vegetation in uplands of western Hon-shu, Japan. Part II. Succession and grazing indicators. Jap.J. Bot. IX: 133 167.

Kologiski, R. L., 1977. The phytosociology of the Green

Swamp, North Carolina. N.C. Ag. Expt. Sta. Tech. Bull. No.250. 100 pp.

Lawson, .I., 1714. I.awson’s history ol’h’orth Carolina. Garrett

and Massie, Inc., Richmond, VA. 259 pp. (2nd ed. printed1952.)

I,emon, I’. C., I949. Successional responses of herbs in longleaf-

slash pine forest after f’irc. Ecology 30: I35 145.ILemon. I’. C., 1967. Effects of fire on herbs of the southeastern

United States and central Africa. Proc. Tall Timbers Ecol.Conf. 6: I I2 127.

Loucks, 0. I.., 1970. Evolution of diversity, efficiency, and

community stability. Am. 2001. 10: I7 25.Maarel, E. van der, 1971. Plant species diversity in relation to

management. In: E. Duf fcy& A. S. Watt (eds.). The Scientif-ic Management of Animal and Plant Communities for Con-

servation. Symp. Brit. Ecol. Sot. I I: 45 63.Maat-cl, E. van der& Leertouwer, J., 1967. Variation in vcgeta-

tion and species diversity along a local environmental gra-

178

dient. Acta Hot. Neerl. 16: 21 I 221.Maxwell, H., 1910. The use and abuse of forests by the Virginia

Indians. William & Mary Coil. Q. Hist. Mag. 19: 73 103.Mets. 1.. J., Lotti. T. & Klawitter, R. A., 1961. Some effects of

prescribed burning on Coastal Plain forest soil. USDA For.Ser. Res. Pap. SE-I 13. Asheville. N.C. 10 pp.

Mueller-Dombois, I)., 1977. Comment on Whittaker’, paper.

In: R. Tiixen (ed.). Vegetation und Fauna. pp. 421 422. .I.Cramer, Vadul.

Nash, G. V., 1895. Notes on some Florida plants. Bull. Torrcy

Bot. Club 2: I41 161.Naveh, Z. & Whittaker, R. H., 1979. Measurements and rela-

tionships of plant species diversity in Mediterranean shrub-

landsand woodlands. In: F. Grassle, G. P. Patil, W. Smith&C. Taillee (eds.). Ecological Diversity in Theory and Prac-tice, Statistical Ecol. Vol. S6, pp. 219 239. Internat. Coop.Publ. House, F:ait-land, Md.

Naveh, Z. & Whittaker, R. H.. 1980. Structural and floristic

diversity of shrublands and woodlands in northern Israel andother Mediterranean areas. Vegetatio 41: I71 190.

Ode, D. J.,Tieszen, I.. 1.. & Lerman, .I. C., 19X0. The seasonal

contribution of Ci and C4 plant species to primary produc-tion in a mixed prairie. Ecology 61: 1304 131 I.

Olsson, H., 1979. Vegetation of the New Jersey pine barrens: aphytosociological classification. In: R. ‘f’. T. Forman (ed.).

Pine Barrens Ecosystem and Landscape, pp. 245 263. Aca-demic Press. New York.

Paine, R. T., 1966. Food web complexity and species diversity.

Am. Nat. 100: 65 75.Parrot, R. 7-., 1967. A study of wiregl-ass (Aristida stricta

Michx.) with particular refer-ence to fire. M.A. Thesis, DukeUniv., Durham, N.C. 137 pp.

Peet, R. K., 1974. The measurement of species diversity. Ann.

Rev. Ecol. Sys. 5: 285 307.Peet, R. K., 1978. Forest vegetation of the Colorado Front

Range: patterns of species diversity. Vegetatio 37: 65 78.Peet, R. K., Glenn-lxwin, D. C.& Wolf,J. W., 1983. Prediction

of man’s impact on plant species diversity. In: W. Holzner,M. J. A. Wcrger & I. lkusima (eds.). Man’s Impact on

Vegetation, pp. 41 54. Junk, The Hague.Plummer. G. I-.. 1963. Soils of the pitcher plant habitats in the

Georgia coastal plain. Ecology 44: 727 734.Pullen, 7. & Plummer, G. l.., 1964. Floristic changes within

pitcher plant habitats in Georgia. Rhodora 66: 375 381.Radford, A. E., Ahles, H. E. & Bell, C. R.. 1968. Manual of the

Vascular Flora of the Carolinas. liniv. N.C. Press, ChapelHill. I IX3 pp.

RegnCII, G., 1980. A numerical study of successions in an aban-doned, damp calcareoua meadow in South Sweden. Vegeta-tio 43: 123 130.

Rights. D. I.., 1947. The American Indian in North Car-olina.

Duke Univ. Press, Durham, N.C.Roberts, I-‘. R. & Oosting, H. .I., 1958. Responses of Venus’

flytrap(Dionaea muscipula) to factors involved in itsendcm-

ism. Fcol. Monogr. 2X: 193 218.Roriaon, I. H., 1971. The use of nutrients in the control of the

lloristic composition of grassland. In: E. Duffey & A. S.Watt (eds.). The Scientific Management of Animal and PlantCommunities for Conservation. Symp. Brit. Ecol. Sot. I I:

65 7X.

RosL:n. E. & S.jBgren, E., 1973. Sheep grafing and changes of

vegetation on the limestone heath of eland. Zoon. suppl. I:137 151.

Schoener, T. W.. 1974. Resource partitioning in ecological

communities. Science 1X5: 27 39.Schocner, T. W.. 1982. 7he controvet-sy over interspecific com-

petition. Am. Sci. 70: 5X7 595.Shmida, A. & Ellner, S. I’., 1984. Coexistence of plant species

with similar niches. Vegetatio (in press).S.jiigren, E., 1971. The influence of sheep gra7ing on limestone

heath vegetation on the Baltic island ofoland. In: E. Duffey& A. S. Watt (eds.). The Scientific Management of Animaland Plant Communities for Conservation. Symp. Brit. Ecol.

sot. I I: 487 495.Snyder, J. R., 19X0. Analysis of coastal plain vegetation, Croa-

tan National Forest, North Carolina. In: H. Licth & E.Landolt (eds.). Contributions to the Knowledge of Flora andVegetation in the Carolinas, Vol. 69, pp. 40-l 13. Veriiff.Gebot. Inst. ETH, Stiftung, Riibel, Ztirich.

Stoddard, H. l... 1931. ?-he Bobwhite Quail, Its Habits, Preser-vation and increase. Charles Scribner’s Sons, New York.

Streng. D. II. & Harcombe, P. A., 1982. Why don’t Texas

savannas grow up to forest? Am. Mid. Nat. 108: 278.294.‘Terborgh. J., 1973. On the notion of favorableness in plant

ecology. Am. Nat. 107: 481 501.Thalen, D. C. I’., 197 I, Variation in some saltmarsh and dune

vegetations in the Netherlands with special reference to gra-dient situations. Acta Hot. Neerl. 20: 327 342.

Thorne, R. F., 1954. The vascular plants of southwestern Geor-

gia. Am. Mid. Nat. 52: 257-327.Tilman, D., 1982. ResourceCompetitionandCommunityStruc-

ture. Princeton Monogr. Pop. Biol. 17, Princeton Univ.Press, Princeton, New Jersey. 296 pp.

lJ.S. Dept. Commerce, National Oceanic and AtmosphericAdm., 1977 83. Climatological Data North Carolina, Vol.X2 8X. National Climatic Data Center, Asheville. N.C.

Vogl. R. J.. 1973. Fire in the southeastern grasslands. TallTimbers Fire Ecol. Conf. Proc. 12: I75 198.

Ward, T., 1979. Memo to Thomas Masscngale, North Carolina

Nature Conservancy, Chapel Hill, N.C. April 13, 1979.Wells, B. W., 1928. Plant communities of the coastal plain 01

North Carolina and their successional relations. Ecology 9:

230 242.Wells. B. W., 1932. The Natut-al Gardens of North Carolina.

Univ. of N.C. Press, Chapel Hill. 458 pp.Wells. B. W. & Shunk, I. V., 1928. A southern upland go-ass-

sedge bog: an ecological study. N.C. Ag. Expt. Sta. Tech.Bull. 32.

Well,, B. W. & Shunk. I. V., 1931. The vegetation and habitat

factors of the coarser sands of the North Carolina coastalplain: an ecological study. Ecol. Monogr. I: 465 520.

Wells, 13. W. & Whitford, I.. A., 1976. Histor-y of stream-headswamp foresta, pocosins, and savannahs in the southeast. J.Elisha Mitchell Sci. Sot. 92: 148-150.

Wharton, C. H., 1978. The Natural Environments of Georgia.Off. of Planning and Res., Ga. Dept. Natl. Resources, Atlan-ta. 227 pp.

Whittaker, I

179

Publ. Corp., New York.Whittaker, R. H., 1977b. Animaleffectsonplant speciesdiversi-

ty. In: R. Ttixen(ed.). Vegetation und Fauna, pp. 409.-425. J.

Cramer, Vadu7.Whittaker, R. H. & Gauch, H. G., 1973. Evaluation of ordina-

tion techniques. In: R. H. Whittaker (ed.). Ordination and

Classification of Communities, Hndbk. of Veg. Sci. 5, pp.287 321. Junk. The Hague.

Willems. J. H., 1980. Observations on north-west Europeanlimestone grassland and communities. Proc. K. Ned. Akad.

Wet. Series C 83: 279-306.Williams, G. J. & Markley, J. I..., 1973. The photosynthetic

pathway type of North American shortgrass prairie speciesand some ecological implications. Photosynthetica 7:262 270.

Woodwell, G. M., 1956. Phytosociology of coastal plain wet-lands of the Carolinas. M.A. Thesis, Duke Univ., Durham,N.C. 51 pp.

Accepted 17.10.1983.