Vegetation dynamics and mesophication in response to coniferencroachment within an ultramafic system

J BurgessAF K SzlaveczA N RajakarunaBC S LevD and C Swan E

ADepartment of Earth and Planetary Sciences Johns Hopkins University Baltimore MD 21218 USABCollege of the Atlantic 105 Eden Street Bar Harbor ME 04609 USACUnit for Environmental Sciences and Management North-West University Private Bag X6001Potchefstroom 2520 South Africa

DUrban Environmental Biogeochemistry Laboratory Towson University 8000 York Rd TowsonMD 21252-0001 USA

EDepartment of Geography amp Environmental Systems University of Maryland Baltimore CountyBaltimore MD 21250 USAFCorresponding author Email jerryburgessjhuedu

Abstract The biological ecological and evolutionary significance of serpentine habitats has long been recognisedWe used an integrated physiochemical dataset combining plot spatial data with temporal data from tree cores to evaluatechanges in soils and vegetation Data suggest that this unique habitat is undergoing a transition endangering localbiodiversity and endemic plant species The objective of this work was to analyse the vegetation dynamics of a xericserpentine savanna located in the Mid-Atlantic USA We employed vegetation surveys of 32 10 15m quadrats to obtainwoody species composition density basal area and developed a spatial physiochemical dataset of substrate geochemistry toindependently summarise the data using regression and ordination techniques This information was interpreted alongsidehistorical dendrochronologic and soil stable carbon isotopic data to evaluate successional dynamics Comparisons amonggeologic pedologic and vegetation environmental drivers indicated broad correlations across an environmental gradientcorresponding to a grassland to forest transition Thewoodland communities appear to be part of a complex soilmoisture andchemistry gradient that affects the extent density basal area and species composition of these communities Over thegradient there is an increase in a diversity a decrease in the density of xeric and invasive species and an increase in stemdensity of more mesic species Dendrochronology suggests poor recruitment of xeric species and concomitant increase inmore mesic species The data indicated that former C4-dominated grasslands were initially invaded by conifers and are nowexperiencing mesophication with growing dominance by Acer Nyssa and more mesic Quercus and Fagus species

Additional keywords edaphic plantndashsoil relations serpentine succession

Received 15 September 2014 accepted 2 December 2014 published online 7 April 2015

Introduction

Vegetation dynamics associated with environmental gradientsare strongly linked to the underlying abiotic drivers such as theheterogeneity of resources (Grime 1979 Huston 1979 Tilman1984 1987Robertsonet al 1988WilliamsandHouseman2013)In areaswith strong abiotic influence the traitndashenvironment theoryis expected to predominate as the overall species pool is filtered onthe basis of ecologically relevant life-history traits (Keddy 1992Diacuteaz et al 1998 Lawton 1999 Butaye et al 2002 Fernandez-Going et al 2013)

Research over the past century on the relative strengths of theindividualistic organismal and integrated community conceptshas provided little consensus on the mechanisms shapingcommunity dynamics (Clements 1916 Gleason 1926 Lortie

et al 2004 Goumltzenberger et al 2012) Community assemblyin early succession depends on the inherent size of the speciespool and the dispersal potential of the organisms whereas theestablishment of vascular plants in early succession stages ispartly limited by tolerance to drought stress (Weiher and Keddy1999 Jones and del Moral 2009) However disturbance regimesare also important in the creation and maintenance of vegetationmosaics as well as controlling species dynamics (White 1979Pickett and White 1985)

One landscape that demonstrates the tight coupling of abioticdrivers among plants soils and lithology is the serpentinite orultramafic terrain In the Appalachian Piedmont ultramafic rocksoccur in a variety of geologic settings and often form restrictivehabitats with unusual vegetation chiefly because of certain

CSIRO PUBLISHING

Australian Journal of Botanyhttpdxdoiorg101071BT14241

Journal compilation CSIRO 2015 wwwpublishcsiroaujournalsajb

edaphic factors such as low nutrient content high metalconcentrations and shallow soils (Rajakaruna et al 2009)Such communities are of interest because they harbor endemicspecies and unique community types and serve asmodel systemsto study insular communities (Harrison and Rajakaruna 2011)These edaphic systems often exist as isolated units with distinctearly successional communities that are embedded in the regionallandscape Frequent fire or herbivory is often proffered asnecessary to maintain soil conditions and the unique nativegrassland communities (Latham 1993 Tyndall 1992 2005Arabas 2000) During the past century many grass-dominatedecosystems around the world have been affected by the spread ofwoody plants (Archer et al 1995 VanAuken 2000 2009) In theabsence of frequent disturbance the once expansive xerophyticgrasslands and savannas on ultramafic substrate may succumbto secondary succession and become forests or woodlandshowever remnants of these areas are still present (Tyndall andFarr 1989 1990 Smith and Barnes 2008 Burgess et al 2009Noss 2013) The remaining grasslands exhibit lsquoprairiersquo-typecompositions with shortgrass mixed with tallgrass and werelikelymore expansive formerly than today (Marye 1955a 1955b1955c Tyndall 1992) Tyndall (1992) and Tyndall and Farr(1989 1990) have documented the expansion of conifers inserpentinite grasslands yet the extent and dynamics of theserpentinite closed-canopy forests or woodlands has not beeninvestigated These serpentinite woodland regions of the Mid-Atlantic Piedmont of eastern North America are the focus of thisresearch

In eastern North America a rich history of vegetation changehas resulted in textbook models of succession based onabandoned agricultural fields that follows the general trend ofannual herbs followed by grasses and perennial herbs followedbywoody species colonisation and eventual dominance (Oosting1942 Keever 1950 Bazzaz 1968 Inouye et al 1987) Howeveredaphic succession does not correspond to this textbook modelEdaphic features such as those arising from ultramafic parentmaterial can heavily influence the spatial-temporal distributionof succession (Curtis 1959Whittaker 1960 Callaway and Davis1993 Zavala et al 2000 Wright and Fridley 2010) Areasunderlain by serpentinite and other ultramafic rock types arelimited in calcium and potassium and high in toxic elementsincluding magnesium chromium and nickel typically resultingin harsh soils and sparse plant communities (eg Kruckeberg1984 Proctor andNagy 1992 Alexander et al 2007 Rajakarunaand Boyd 2009 Harris and Rajakaruna 2009)

Serpentinite outcrops in the Appalachians form characteristicecosystems termed grasslands barrens balds and savannasThose of the Mid-Atlantic are characterised by savanna orgrassland if maintained by periodic disturbance howeverforests cover much of the Piedmont upland region underlainby ultramafic bedrock as well Within this landscape there arepatchily arranged xerophytic enclaves surrounded by moremesophytic vegetation The mosaics are prairie-like openingsand pine scrub or greenbrier thickets grading to mixedpinendashdeciduous woodlands and oak-dominated forests Workby Tyndall (2005) and Cumming and Kelly (2007) suggestedthat relict grasslands and pine savanna soils have altered abioticproperties lowered herbaceous diversity and shifted fungalcommunities as a result of recent conifer expansion However

Barton and Wallenstein (1997) found that soils invaded by pineactually had similar or lower nutrient and organic matter contentsthan early successional savanna and exhibited no differencein calcium magnesium ratios a critical ultramafic soil indicator

We investigated how vegetation dynamics relate togeoecological gradients based on soil properties and whether theencroachment of woody species threatens the biological integrityof a globally rare ecosystem by altering the successional trajectory(inter alia as regards the woody species richness compositionand structure) in a traditionally xeric environment The specificobjectives of this research were to (1) examine the dynamics ofwoodlands supported by ultramafic soils (2) reconstruct a partialvegetation history and (3) assess the impact of recent woodyencroachment on soil properties and successional trajectoriesWe hypothesised that changes in vegetation dynamics fromgrassland to woodland would not result in a local edaphicclimax nor will the anomalous flora continue to persist indiscrete patches We expected that this edaphic successionwould be accompanied by an increase in soil organic matter andcalcium (Ca) magnesium (Mg) ratios a decrease in pH in asoil profile and a loss of the traditional serpentine grasslandsignature accompanied by species redistribution Ultimately thismesophication and homogenisation will result in a woodlandcommunity indistinguishable from typical Piedmont soilsubstrates

To achieve these goals we integrated natural history recordstree-ring datawoody species community data stable isotope datafrom soil organic matter and a suite of soil physio-chemicalproperties Geologic pedologic and vegetation propertycomparisons were evaluated using a chronosequence approachto study serpentine succession that makes critical assumptionssuch as each site in the sequence formed from the same substratediffers only in age and that each site has traced the samehistory inboth its abiotic and biotic components (Huggett 1998 Johnsonand Miyanishi 2008) Understanding community assembly ofsuch communities helps explain contemporary vegetationdynamics at multiple scales and thus manage them for futureconditions (Foster et al 2003)

Materials and methodsSite description and backgroundThe study was conducted in a Mid-Atlantic temperate Piedmontarea (Fig 1) The site consists of privately owned and co-operatively protected land (Camp Moshava) with a xeric tomesic upland supporting a gt105-ha oakndashconifer woodlandlocated near Cherry Hill Maryland USA (39640N 76360W)The mean annual temperature and rainfall are 12C and1133mm respectively (Maryland State Climatologist Office2014) Topographic elevation ranges from 90 to 145m abovemean sea level

The Cherry Hill field site (CHS) lies within the largest maficcomplex in the Appalachians known as the BaltimoreUltramaficndashMafic Complex (BMC) Ultramafic components oftheBMCcontain abasal unit of serpentinisedperidotitewith relictwebsterite and dunite These units are likely to represent partsof an ophiolitic island arc or a supra-subduction zone complexthat was obducted onto Laurentia during the closure of Iapetusduring the Taconic orogeny at ~455million years ago (Muller

B Australian Journal of Botany J Burgess et al

et al 1989) TheCHS ismore enigmatic lyingwithin theBoulderGneiss metasedimentary (diamictite) unit and bounded to thenorth by metagraywacke (Fig 1) and is likely to represent afragmented sliver of a once larger complex Vegetation atop thestudy area varies from mature forests to savanna-like areas withopen-grown trees and diverse grassland with a wide variety ofgrasses and forbs

A forest encloses the remnant grasslands of the study area andexhibits a transitional ecotone ranging from 10 to 15m (Fig 2)The dominant trees within the forest are oak species (Quercusspp) whereas at the transition eastern red cedar (Juniperusvirginiana L) and Virginia pine (Pinus virginiana Mill) arealso important On the basis of historical records (eg diaries

and mining and forestry reports such as found in Jefferis 1880Williams and Darton 1892 Abbe 1898) aerial photographs(Maintained at the Maryland Geological Survey and HarfordCounty Soil Survey Libraries) and interviews this study area hasnot been farmed or burned during the past century Shortly after agrassland or savanna survey by Tyndall and Farr (1989) coniferswere eliminated in the grassland in a restoration attempt In thesavanna and grassland patches disjunct populations of C4

bunchgrass species including little bluestem (Schizachyriumscoparium (Michx) Nash var scoparium) big bluestem(Andropogon gerardii Vitman) arrowfeather threeawn(Aristida purpurascens Poir) and prairie dropseed (SporobolusheterolepisAGray) are interspersed with a variety of herbaceousC3 species such as Carex umbellata Willd (parasol sedge) andSmilax spp (greenbrier)

Abiotic characteristicsWecarried out detailed geologicmapping to assess howvegetationcharacteristics were influenced by geologic substrate Aftergeologic mapping 32 quadrats (see Fig 1) were establishedusing stratified random sampling to encompass the variety ofenvironmental conditions in the area Quadrats (10 15m)were oriented up and down slope to cover distinct changes inenvironmental gradients and community types Near the centre ofeach plot latitude longitude elevation slope and aspect weredetermined and recorded with a GPS device

During the mapping phase representative rock samples of thefreshest serpentinites (n= 12) were systematically selected forpetrographic analysis from samples collected (n = 34) duringfield mapping Both major and trace elemental whole-rockanalyses were conducted using X-ray fluorescence (XRF)Major elements were determined on glass beads fused fromignited powders to which Li2B4O7 was added (1 5) in aplatinum crucible at 1150C Trace elements were analysed onpressed powder pellets by the same instrument a Bruker AXS S4Explorer following the procedures outlined by Potts (1987)

Woodland Plot (10 x 15 m)

Relict Barrens Areas

Inset area

Highly Sheared or Fractured Areas

Lithologic Contact

Wissahickon Group (Diamictite)

Scale (m)

0 200

Serpentinite amp Ultramafics

Geologic Sample

PENNSYLVANIA

QUADRANGLE LOCATION

Fig 1 Topographic map of the sampling area in near MarylandPennsylvania boundary USA Note that plot sizes are not to scale

Fig 2 Field photograph showing insular relict barrens and grassland being encroached upon by eastern red cedar (Juniperus virginiana) with surroundingmixed-oak (Quercus) woodlands Photo by J Burgess

Conifer encroachment on ultramafics Australian Journal of Botany C

Counting statistics errors are 01 for major elements andaverage detection limits are 50ndash100mg kgndash1 for major oxidesand 5ndash10 ppm for chromium (Cr) nickel (Ni) titanium (Ti)iron (Fe) manganese (Mn) Ca potassium (K) and phosphorus(P) All XRF samples were quantified using a matrix matchedset of USGS- and NIST-certified soil standards At least oneduplicate one replicate and one certified reference sample(SRM 2709 San Joachin Soil) was run with every 10 samplesanalysed to monitor for external and internal reproducibilityLoss on ignition (LOI) for samples is measured by igniting rockpowders in alumina crucibles to 1200C for 10min in a mufflefurnace

Soil characteristics evaluated for each plot or quadrat includedbulk density soil depth pH soil organic matter content majorand trace element analysis carbon (C) nitrogen (N) and isotopicand textural analysis Surface soil epipedons were evaluated forbulk density (g cmndash3) At each location we composited five soilcores (corners plus centre) that were taken with a 5-cm-diametersoil corer Soil coreswere sectioned into 0ndash5 5ndash10 10ndash15 (and atsome sites 5ndash15) cm increments dried at 40C to a constantweight and then calculated as dryweight of sample per unit of soilvolumewith roots removed Soil parameter valueswere averagedto provide a representative value for each 150-m2 plot

Forest soil depth (depth to obstruction) was measured with apush soil probe at five random locations in each plot For soilmoisture or soil depth comparisons from savanna into forestedareas four transects (35ndash68m long) were established in April2011 Our rationale was that distance from bedrock wouldrepresent a proxy for time and points along these transectswould generally represent soil chronosequence (Jenny 19411980) Because of the inherent low moisture values ofserpentine soils measurements were made during the weekafter a 5-mm or greater rainfall event Soil moisture values at0ndash6-cm depth were measured with a ThetaProbereg soil moisturesensor (Dynamax IncHoustonTXUSA) and soil depthswith apush probe (internal diameter = 47 cm) every 1m

Soil samples for chemical analysis were air-dried and passedthrough a USA Standard No 10 sieve Bedrock thin-sectionsoccasionally contain carbonate (calcite andmagnesite) thereforeinitial soils were evaluated for carbonate by an acid (1Mhydrochloric) test The HCl test showed that the soil samplesdid not contain carbonates therefore total C in the soil representsorganic C only

Moist soil colour was determined in the field using Munsellcolour charts (Munsell Color Grand RapidsMI USA) Sampleswere combined mixed and split to give two samples of200ndash400 g that were lightly crushed and sieved to 2mm andsampled for textural analysis (sieve followed hydrometermethod) (Bouyoucos 1936) Soil pH was measured with a 20 gin 20mL soilndashwater slurry by using a glass electrode Theremainder was oven-dried at 30C and powdered using anagate centrifugal ball mill A powdered subset wassubsequently used for XRF and C N analyses Total N and Cwere measured with an automated Perkin-Elmer 2400 Series IICHN analyser (Perkin-Elmer Cornell Labs Ithaca NY USA)Soil organic matter content was estimated as loss on ignition(LOI) at 550C for 3 h Total C N and isotopic analysis wereconducted at the stable isotope facilities at either theUniversity ofCalifornia at Davis CA or Cornell Laboratories Ithaca NY

using an elemental analyser interfaced with continuous-flowisotope-ratio mass spectrometer Values reported are typicallymeans with the associated standard error All the values werereported as per mil (permil) relative to the international standardVienna Pee Dee Belemnite (V-PDB) and were calculated as

d13 CethpermilTHORN frac14 Rsample Rstandard

Rstandard

1000

where d13C is the difference between 13C 12C ratio (R) of thesample and that of the standard Increase in the d value reflectsincreases in the relative amount of the heavy isotope componentor a reciprocal decrease in the light isotope component

Biotic characteristicsTo monitor the serpentine ecosystem-scale vegetation-boundarymovement we examined a time series of high-resolution imagesfrom Google Earth (1994ndash2013) along with aerial photographyfrom 1937 1952 1971 and 1988 using the grid-square methodThe more recent data serve to complement the work by Tyndall(1992) on the expansion of conifers into barrens areas All woodyplant species were recorded for each sampling plot To identifyplants we followed the key by Brown and Brown (1984) withnomenclature clarification via the integrated taxonomicinformation system Woody plants with diameter-at-breast-height (DBH) of gt10 cm are referred to as trees those withDBH of lt10 cm and gt2 cm or 2m height as saplings and anywoody specieslt2-cmDBHare grouped as seedlings or vines Forspecies where DBH is impractical (shrubs) stemsweremeasuredwith a caliper near the root collar Absolute and relative stemdensity and basal area were derived from these data

Canopy cover was measured at each plot centre with aspherical densiometer Plots were classified as savannawoodland or forest on the basis of canopy cover as followssavanna (0ndash25 canopy cover) woodland (26ndash60 canopycover) and forest (gt60 canopy cover) Ground-layerpercentage cover for soil leaf litter and rock was estimatedvisually

To assess vegetation change over time historical recordswerecombined with tree-core data (n= 163) One increment core wasextracted for each tree species per plot when possible Cores weretaken at breast height mounted and sanded following Stokes(1996) and Stokes and Smiley (1968) Because increment coreswere collected at 137mabove thegroundweconsidered the agesobtained as the minimum age of trees at breast height Repetitivecoring was carried out to facilitate the interception of the pithMissing rings in incomplete cores were estimated using diameterage-regression curves (Lorimer 1980) Ring counts were madeusing a stereomicroscope at 10ndash40 magnification Each coresample was skeleton plotted to identify narrow rings (Speer2010) The resulting age distribution was divided into datethresholds to record pulses or absence of regeneration as cohorts

Data analysisTo quantify forest structure the quadrats were characterised bybasal area (m2 handash1) and density (stems handash1) of live specimens ofthe principal tree species Tree species were grouped on the basisof habitat affinity (xericmesic) using the USDA silvics manual(Walters and Yawney 2004) Diversity indices were calculated

D Australian Journal of Botany J Burgess et al

for each plot and on pooled plot data using the ShannonndashWeaverindex (Shannon and Weaver 1949) and the Simpson index(Simpson 1949) as measures of a diversity (a) along withestimators of total diversity using rarefaction and non-parametric approaches of Chao 1 Chao 2 and ACE (Chao1984 1987 Chazdon et al 1998) The importance value (IV)is a measure of relative abundance and was calculated as theaverage of relative density and dominance

The KolmogorovndashSmirnov test statistic was used to assessnormality of the soil variables Those that failed the normality testwere transformed to normal or near-normal distributionPercentage measurements (soil particle analysis abundance CN cover XRF whole-rock oxide weight) were arcsinetransformed soil elemental data were loge transformed fromppm or ppb concentrations slope and soil depth were natural-log transformed and bulk density was square-root transformed

Site microclimate variability was compared using geographiccoordinates (latitude and longitude) inclination and aspect foreach plot to calculate the heat load and the potential annual directincident radiation by applying the third equation from McCuneand Keon (2002) The effects of topographic (slope inclinationannual direct incident radiation and heat load) and soil variables(elemental analysis in addition to pH soil organic C total NC N depth texture and d13C values) on species were examinedusing principal components analysis (PCA) to explore the mainfactors and generate hypotheses as to the reciprocal effects ofthese properties on vegetation Redundancy analysis (RDA)ordination as constrained by the environmental variables wasperformed to identify the main floristic gradients (basal areadensity percentage xeric species) and the edaphic andtopographic factors explaining those trends Two-wayANOVA was used to test for the differences in soil d13C C Nand C N with respect to vegetation (grassland forest) and soildepth ANOVAs were followed by Tukeyrsquos HSD (chosenbecause of uneven number of replicates) multiple pairwisecomparison to relate differences in elemental content and depth

Differences in site woody-species composition wereinvestigated by calculating the pairwise ecological distance forthe entire species matrix The Mantel test was used to determinewhether site similarity in the tree community was related to siteenvironmental variables Multiple linear regression analysis wasused to identify site factors that had a significant effect on speciesabundance The eight highest-ranked species on the basis of IVwere evaluated in termsof their abundances as responsevariablesand soil and topographic data were used as explanatory variablesOrdination using direct (RDA) and indirect (PCA) gradient

analysis was employed to extract variation in the underlyingstructure of the vegetation data as it relates to environmentalvariables (Ludwig and Reynolds 1988) PCA was performed tofind the main factors determining the reciprocal effects of soiland vegetation

The BrayndashCurtis distance with raw abundances was usedto determine the degree of vegetation dissimilarity an aspectof b (b) diversity between sites and vegetation types (woodlandand forest) based on presenceabsence We used a Mantel test(McCune and Grace 2002) to compare the matrix of geographicdistances between plots with the matrix of ecological distancePotential relationships between abiotic variables and diversityvariables (floristic similarity evenness and Simpsonrsquos diversity)were explored Tree-species compositions of early successionalwoodland and more mature forest designations based onpercentage canopy were compared using non-parametricanalysis of similarities (ANOSIM) based on the hypotheses ofno difference between stages

Gamma (g) or serpentine landscape-area diversity wasestimated using all 32 plots to identify species richness at thelandscape scale by constructing a rarefaction curve for all samplesand all species in the dataset

For all statistical tests significance was determined at thea= 005 level other P-values were reported if lt01 The methodfor evaluating the Mantel-test statistic was the t-distribution(Mantel 1967) and the standardised Mantel statistic (r) as ameasure of effect size (McCune and Grace 2002)

All data were analysed using the statistical computingenvironment R (v 311) and the BiodiversityR package(Kindt and Coe 2005) ANOSIM Mantel and other tests wereperformed with n= 1000 permutations Reported errors arestandard errors (se) unless indicated otherwise

Results

Abiotic substrate and soil composition

Mean whole-rock major-element values were typical of alteredultramafic rocks (Table 1) Calcium and aluminium (Al) contentswere low reflecting disappearance of the calcic pyroxene byserpentinisation Only samples where clinopyroxene wasdetected showed an increase in Ca Samples with elevated Alare generally accompanied by the presence of chlorite in therocks As is typical of ultramafics the immobile refractoryelements Cr and Ni were elevated Ternary diagrams depictingthe measured whole-rock geochemistry of the samples illustratedrelative compositional variations in the systemMgOndashSiO2ndashH2O

Table 1 Summary of the whole-rock geochemistry parameters for serpentinite samples obtained from X-ray fluorescence (XRF) analysisSerpentinite water content based on loss on ignition (LOI) ~124 03 weight Al aluminium Ca calcium Cr chromium Fe iron K potassium Mg

magnesium Na sodium Ni nickel Si silicon Ti titanium V vanadium Zn zinc

Lithology Major element geochemistry (weight ) Trace elements (mg kgndash1)Serpentinite n= 12 Si Al Fe Mg Ca Na K Ti Cr Ni V Zn

Mean 4129 031 954 3669 042 000 007 002 283799 231968 1189 4385se 088 015 078 070 022 000 002 000 28651 15716 334 594Soils (0ndash5 cm) n= 32 3513 091 657 625 024 015 070 048 252651 115744 9678 10705se 132 019 040 060 006 003 005 003 27641 13040 642 463Soils (5ndash15 cm) n= 48 3799 105 744 762 015 015 069 046 234790 124434 9725 8426se 106 020 036 060 004 002 005 003 20726 11024 554 311

Conifer encroachment on ultramafics Australian Journal of Botany E

(Fig 3) Concentrations of soilmacronutrientsCN P andKwerenot related to ultramafic parent material

Soil depths varied from105 cm to52 cmwith variable texture(mean of 134 gravel) across the sites Soil colours ranged fromdark brown (75YR 32) to grey (75YR 61) in the upper 1ndash5 cmof the soil profile to yellowndashbrown (10YR 56) and redndashbrown(25YR 44) with depth Talcose serpentinites yielded moreyellow (10YR Hues) varieties The mean and standard error ofsoils collected from the 32 plots recorded a narrow range of

edaphic characteristics including soil depth 289 20 (cm) pH54 01 bulk density 088 007 (gm3) soil organic matter(SOM) 1251 26 () sand 5103 18 () clay 267 11() and C N 1518 09

Soil moisture and soil depth increased with distance from thebarrens areas Strong positive correlationwith distancewas foundfor both soil moisture and depth (r= 0570 P lt 0001 andr= 0848 Plt 0001) respectively (Fig 4) Other soil physio-chemical parameters such as texture and trace element contentdid not show correlations with distance (P gt 005 Spearmanrsquosrank correlation)

Magnesium concentrations were high and correspondingCa Mg ratios low (lt05) across all sites The Ca Mg ratio isa measure of serpentine cation imbalance Surface soils hadhigher Ca Mg (02 to 05) ratios than did subsurface soils(005 to 02) Chromium and Ni concentrations were elevatedwith a maximum of over 6000 ppm and 2700 ppm respectively

The pH of the soils was typically acidic (surface soils average~54) but ranged from (~4ndash7) Soil pHwas negatively correlatedwith Ca Mg (r= ndash049 P lt 005) and C N ratios (r= ndash062P lt 005) SOM content (based on LOI) had a mean of 166in the upper 5 cm declining to below 7with depth (10ndash15 cm)N content was generally low and C N ratios varied between 5and 24with the higher values on themore developed soils Depthprofiles of SOM displayed increasing d13C isotopic signaturesfrom 2means of ndash2623permil (0ndash5 cm) to ndash2095permil (10ndash15 cm)(Table 2) whereas remnant grassland areas recorded meand13C isotopic signatures of ndash1623permil (0ndash5 cm)

Woody-species characteristics

Using high-resolution Google Earth imagery the current (2013)percentage of the site that contains grassland or savanna is lessthan 2 The forests include xeric woodlands with stands ofP virginiana and J virginiana intermingled with Quercusmarilandica Meunch (blackjack oak) and Quercus stellata

Talc

Enstatite

Serpentine

Olivine

Ca Mg

Si

Diopside

Typical Ultramafic

Fig 3 Ternary diagram detailing the chemography of Cherry Hillserpentinites (open circles) projected onto the plane CaOndashMgOndashSiO2 alsoshowing typical ultramafic rock (filled square) and mineral compositions(solid circles)

60

Soil depth (cm)

Percent moisture

50

40

30

20

Soi

l dep

th (

cm)

and

moi

stur

e (

by

volu

me)

Distance along transect (m)

10

00 10 20 30 40 50 60

Fig 4 Monotonic change of variables along transects The relationship between soil depth percentage moisture (volume) and distance from barrens bedrockwere assessed using correlational analysis (Plt 00001) Data are compiled averages of four transects from rocky barrens to woodlandndashforest ecotone Thecorrelation of the two variables soil depth and moisture is significant (r = 067 Plt 0001)

F Australian Journal of Botany J Burgess et al

Wang (post oak) as well as more mesophytic species Largeopen-crown trees are rare more typical are contorted down-sweeping branches of Q marilandica surrounded by stand-grown trees Aerial photographs from 1937 show conifersadvancing from the dendritic drainage areas Mesophyticforests are composed of these oaks and others such as Quercusrubra L (red oak) Q alba L (white oak) Q montana Willd(chestnut oak) and Q velutina Lam (black oak) along withnumerous other species most notably Prunus spp (blackcherry wild cherry) Pinus spp (eg white pine shortleafpine) Fagus grandifolia Ehrh (American beech) Acer rubrumL (red maple) Sassafras albidum Nutt (sassafras) Populusgrandidentata Michx (bigtooth aspen) Nyssa sylvaticaMarshall (blackgum) Betula lenta L (sSweet birch) andRobinia pseudoacacia L (black locust) On exposed rocksurfaces the serpentine endemic (as reviewed by Harris andRajakaruna 2009) Cerastium velutinum var velutinum Raf(serpentine mouse eared chickweed) and other serpentinophilessuch as Symphyotrichum depauperatum Fern (serpentine aster)and saplings of Q marilandica are common

Biodiversity richness estimators Chao 1 Chao 2 and ACEestimated maximum woody-species richness between 263and 308 species The observed richness was 24 species(Table 3) Overall diversity and mean diversity for the 32 siteswere calculated as 274 (X = 167 s= 029) (Shannon index) and

092 (X = 078 s= 008) (Simpson index) On the basis of relativeIV the most important species were P virginiana andJ virginiana (Table 3) These two species had similar relativeIVs and stood out from all other species A second group ofspecies included A rubrum and Q marilandica At the genuslevel Quercus was the most important contributor to the treecommunity with 39 relative IV On the basis of basal area themost dominant species were Q rubra P virginiana andQ marilandica Constancy values (Table 3) showedA rubrum (078) as a clear break from the next grouping ofspecies (P virginiana and J virginiana) whose constancy valueshovered near 05 followed by Q marilandica F grandifoliaN sylvatica and Q stellate with constancy values near 04

The Mantel test comparing floristic similarity andenvironmental distance showed significant (P lt 005)correlations between environmental distance and ecologicaldistance for the following parameters heat load bulk densitybasal area diversity (H) percentage sand Ni and Cr(Table 4) The ANOSIM statistic was calculated only for thecanopy category (woodlandforest) with a resulting r= 014 andP-value of 005 Both the Mantel and ANOSIM tests yieldedsignificant (Plt 005) but weak correlations

Relationships between vegetation and abiotic properties

Results of multiple regression analysis were evaluated tounderstand the response of species distributions to soilproperties Table 5 shows the regression results for the eightspecies the importance of which met or exceeded 5 withtopographic and edaphic parameters as the predictor variablesPredictor variables that indicated a meaningful contribution toexplaining species abundancewere incident radiation and soilMgcontent for Q marilandica incident radiation for J virginianaand Q stellata soil organic carbon (SOC) for A rubrum soildepth for F grandifolia and bulk density for the conifers

PCA was performed for 16 abiotic parameters across the 32sampled quadrats so as to identify covariation between the solarradiation and soil variables (Fig 5) Themain trend of variation in

Table 2 Descriptive statistics for soil d13C (permil) and soil organic matterin grassland and woodland elements

Means followed by the same letter are not significantly different (ANOVApost hoc Tukeyrsquos HSD P = 005)

Statistic Grassland Forest(0ndash5 cm)

Forest(5ndash10 cm)

Forest(5ndash15 cm)

Forest(10ndash15 cm)

Mean ndash1623a ndash2623b ndash2359bcd ndash2298de ndash2095eMinimum ndash1791 ndash2923 ndash2725 ndash2649 ndash2514Maximum ndash1423 ndash1928 ndash1811 ndash1972 ndash1654se 068 044 067 065 060

Table 3 Constancy importance (average value of a species when present) density dominance andimportance value (average of relative density and relative dominance) measures for woody stems in

serpentinite woodland at Cherry Hill Maryland USA

Species Constancy Importance Dominance(m2 handash1)

Density(stems handash1)

Importancevalue (IV) ()

Acer rubrum L 078 109 5799 7778 988Ailanthus altissima Mill 003 003 075 222 027Betula lenta L 006 006 346 444 060Celtis occidentalis L 006 006 059 444 041Cercis canadensis L 003 003 078 222 022Cornus florida L 006 006 146 444 038Diospyros virginiana L 003 003 185 222 024Fagus grandifolia Ehrh 043 065 5713 4667 619Juniperus virginiana L 050 106 8657 7556 1214Kalmia latifolia L 012 015 163 1111 106Liriodendron tulipifera L 009 009 2372 667 108Nyssa sylvatica Marsh 043 075 5769 5333 720Pinus strobus L 003 003 692 222 090Pinus virginiana Mill 056 078 16498 5556 1343Populus grandidentata Michx 003 006 640 444 058Prunus serotina Ehrh 028 028 2046 2000 271

Conifer encroachment on ultramafics Australian Journal of Botany G

the soil chemical properties and topographic variables asreflected by the PCA Axis 1 (accounting for 30 of variance)was one dominated by significant correlations of soil texture(sand and bulk density) soil depth Mg Cr and Niconcentrations incident radiation and heat load The secondtrend of soil variation as reflected by the PCA Axis 2 (17 ofvariation) can be interpreted as a combined gradient of essentialnutrients (C andN) and SOC The PCA (axes accounting for 44of the variation) revealed strong positive associations betweenboth soil depth and bulk density which were negativelycorrelated with heavy metal concentrations and pH SoilCa Mg ratios and Al concentrations were negativelycorrelated with macronutrients (C N P) and SOC Fig 5bshows the PCA factor map for the abiotic variables and thebiotic covariates for the sampling plots When thecharacteristics of woody-species community are included inthe PCA both basal area and stem density are positivelycorrelated with soil Ca Mg ratios and Al content Diversity(Shannonrsquos index) was strongly correlated with soil depth andbulkdensity andnegatively related toheavymetal concentrations

and heat load As individual-species basal area stem densityand diversity typically respond to environmental gradientsconstrained ordination utilising RDA was used to inferrelationships between the species matrix and theenvironmental matrix using Euclidean distance between sitesArrows of increasing species abundance represent RDA scoresThe sum of the RDA eigenvalues for the first two axes using allabiotic explanatory variables accounted for 368 of the woody-species composition After reduced variable selection based onANOVA tests of the ordinationmodel seven termswere selectedand these collectively explained 271 of the variance (Fig 6)Interestingly those species whose importance was gt4 arepartitioned into the four quadrants of the ordination From theRDAAxes1and2biplot it canbe seen that traditional serpentine-savanna (shade-intolerant)woody species (S albidumQ stellataand Q marilandica) are positively associated with increasingincident radiation and negatively associated with increasingsoil depth Al and Mg soil concentrations Post-disturbanceinitial colonisers (J virginiana and P virginiana) are stronglynegatively correlated with bulk densityQ alba andQ montanawhich are widely distributed across a variety of conditionsfrom wet mesic to subxeric are associated with increasingbulk density Shade-tolerant species such as N sylvaticaA rubrum and F grandifolia are correlated with increasingsoil depth and soils with more Ca and Al

Soils at the site differed in the total amount verticaldistribution and isotopic composition of SOM (Tables 1 2)ANOVA results indicated significant differences in theisotopic means according to their depth (F487 = 276P 00001) Tukeyrsquos HSD test showed that the mean d13C ofgrassland soils (ndash162 07permil) was significantly higher thanthat of shallow forested soils (ndash262 04permil) (Table 5)Shallow forested soils were significantly different from deepersoils (ndash2095 06permil) Total soil C and N and soil C N declinedwith soil depth C N ratios were significantly different for alldepths ranging from 1518 088 on the surface soils to1162 1 in soils 10ndash15 cm below surface (F266 = 59

Table 4 Effects of environmental distance on similarity between plotsMantel statistic calculated by comparing matrix of environmental distancewith BrayndashCurtis distance The standardised Mantel statistic (r) indicatesthe strength of the relationship between sample-point position and forest-floor thickness and the associated P-value indicates the significance of

that relationship

Environmental indicator Correlation coefficient (r) P-value

Heat load 0126 0009Bulk density 0236 0003Basal area 0141 0015Diversity (H) 0109 0034Percentage sand 0203 0002Nickel 0111 0034Chromium 0159 0009Canopy 0150 0009

Table 5 Results of multiple regression with list of factors significantly contributing to species abundanceSOC soil organic carbon Plt 0001 P lt 001 Plt 005 lsquorsquo Plt 01

Pvirginiana J virginiana A rubrum Q marilandica Q rubra N sylvatica Q alba Q stellata F grandifolia Q montana

IV () 1343 1214 988 940 810 720 703 626 619 440Incident radiation 09247 00249 01791 00008 04332 02136 09123 00450 03030 03225Heat load 02516 04869 03661 08494 09552 09203 09825 02522 01128 03502Soil depth 02599 03110 08937 00463 01881 03041 03381 01689 00242 09920Bulk density 00023 00037 08260 08860 05216 02157 00415 02686 09894 01171Texture ( sand) 00976 02800 02942 06247 08124 09313 03207 03451 01454 06054Calcium magnesium 09840 02962 07824 07031 09971 04587 07945 05441 06414 03639Magnesium 06146 05570 05725 00019 01500 06177 00914 03238 07491 03623Aluminium 07988 07323 01491 09176 00673 00583 08508 04152 05380 08685Chromium 05879 02397 02066 01012 06214 09977 06126 03991 06870 00434Nickel 08699 02237 07478 03211 08340 06190 09519 05875 06307 01074Phosphorus 05192 05066 07822 09800 07523 06641 04927 09619 09430 02268SOC 06435 01185 00298 00053 09782 02790 05185 02333 08701 07036Carbon nitrogen 05806 02325 06793 08013 05674 06443 01727 01105 08015 06972Carbon 07737 07963 02099 01231 04380 08148 04280 04885 07995 05145Nitrogen 07506 09537 01179 03581 06363 05329 03124 07375 00153 05744pH 07946 01863 05144 01547 00176 04256 02597 06743 01299 09047Multiple R2 059 068 057 079 057 046 051 055 062 056

H Australian Journal of Botany J Burgess et al

P = 0004) There were significant vegetation depth andvegetation depth effects on d13C of SOM (Table 5 Fig 7)

Age structure

In total 163 core samples were evaluated with results displayedin the stand-age structure plot (Fig 8) Skeleton plot chronologiesvaried among species Older minimum age at breast height wasgenerally seen inQ stellata (114 15 years) andQ marilandica(110 7 years) and the younger chronologies developed inspecies such as A rubrum (53 4 years) F grandifolia(56 8 years) and L tulipifera (43 8 years) Only xeric oakswere present in the oldest age classes The period before 1900 iscomposed of 54 xeric oaks and when all oaks are grouped theycomprise 79 of trees from that era although Q montana isabsent from this group (oldest minimum age is 88 years) Theoldest non-oak was N sylvatica (144 years) The period between

1900 and 1940 records a marked increase in conifers andconcomitant decrease of xeric oaks comprising 39 and 19of all species respectively Between 1950 and 1990 thepercentage of xeric oaks decreased to 3 and conifers to 13whereas Acer spp increased to 19 The number of shade-tolerant species rose from 125 in the decades from 1850 to1900 to 59 during the interval from 1950 to 1990

The age structure pooled from all the plots (Fig 8) showeddiscontinuous patterns with prominent peaks of regenerationpulses and periods of either low or absent recruitment Coniferestablishment and recruitment clearly increased during1900ndash1960 and then receded through the rest of the samplingperiod There was a conspicuous absence of younger xeric oaks(post-1970s) and only two xeric oaks from1950 to the present Atthe same time the percentage of non-oak deciduous treesincreased to 68 post-1950

10

05

00

Dim

2 (

171

7)

Dim

2 (

229

1)

Dim 1 (3038)

Dim 1 (402)

ndash10

ndash05

4

2

0

ndash2

ndash4

ndash6

ndash15

ndash15 ndash10 ndash5 0 5 10

ndash10 ndash05 00 05 10 15

(a)

(b)

Fig 5 (a) Unconstrained principal component analysis (PCA) correlation circle of pedological topological and community variables (b) Forest communityclustering along an edaphic gradient

Conifer encroachment on ultramafics Australian Journal of Botany I

Discussion

The present study examined a moderately sized area in the Mid-Atlantic but the changes ndash encroachment and mesophication ndash

represent potential widespread conservation problems Theserpentinite landscapes were once common components of theancient Piedmont (Carroll et al 2002 Owen 2002 Noss 2013)These unique plant communities have existed as persistentlsquoislandsrsquo embedded in a deciduous forest region for centuriesor longer We documented a state transition from grasslandand xeric oak savanna to conifer-dominated communities

that have given way to eastern broadleaf forests Comparisonsamong geologic pedologic and vegetation properties reflect thischange and indicate that the conversion to forest has been rapidand characterised by large changes in soil properties (pH water-holding capacity SOMcomposition) aswell as plant-communitystructure (physiognomyand stemdensity) and composition (xericto more mesic communities)

Thin-section and petrologic data confirmed that the geologicsubstrate is a serpentinised peridotitewith characteristics typical ofserpentinite terranes in the Appalachians (Misra and Keller 1978Mittwede and Stoddard 1989 Gates 1992 Alexander 2009)Although Ca is depleted in bulk rock samples because of thebreakdownof augite and interstitial plagioclase soil Ca Mg ratiosare higher in surface soils The concentrations of SOC C N andmacronutrients in woodland and forested areas are linked to theturnover of organic matter and not the parent material

Compared with the rock soils are enriched in most tracemetals although the Cr contents were similar This was likelyto be due to resistance of Cr-oxides to weathering and suggeststhat these phases were not likely sources for Cr in soil solutionsand plants of serpentine soils (Oze et al 2004) Bedrock to foresttransects demonstrated a strong correlation between soilmoistureand soil depth Soil nutrient content and soil moisture should beproportional to the cube of soil depth (Belcher et al 1995) thuseach of these parameters figured significantly in the PCA andRDA

Afforestation has been relatively recent in this region (Tyndall1992)Concomitantwith the increase in treedensity andsize is thecontraction of the grassland and savanna component study areafrom 18 in 1937 to 15 in 2013 which is consistent withprevious work documenting the invasion of serpentine savannasby conifers (Tyndall 1992 Barton andWallenstein 1997 Arabas2000)

In landscapeswhereC3 andC4plants coexist stableC isotopiccomposition of SOM represents a powerful method for

20

15

1

0

10

ndash10

ndash2 ndash1 0 1 2 3

05

ndash05

00

RD

A 2

RDA 1

Fig 6 Ordination diagram resulting from standard redundancy analysis (RDA) on transformed abiotic variables illustrating relationships between woodyspecies and physiochemical attributes The diamonds represent species and the vectors indicate the strength of relationship

ndash29 ndash27

(0ndash5)

(5ndash10)

(5ndash15)

(10ndash15)

Grassland (0ndash10)

Soil δ13 C (permil)

Forests

Grassland

Soi

l dep

th (

cm)

ndash25 ndash23 ndash21 ndash19 ndash17 ndash15

Fig 7 Mean carbon isotopic signature in forest (n= 96) and grassland(n= 8) soils for Cherry Hill Maryland USA Error bars are 1 se Datainterpreted as evidence for expansive C4 or C3 historic grassland Verticallines over 5ndash10- and 5ndash15-cm data indicate group pairs that were notsignificantly different using Tukeyrsquos HSD multiple comparisons

J Australian Journal of Botany J Burgess et al

reconstructing vegetation changes (Deines 1980 Tieszen andArcher 1990 Boutton and Yamasaki 1996 Boutton et al 1998)Spatial patterns of soil d13C were strongly related to type ofvegetation (primarily woody versus grassland cover) Remnantgrasslands are composed of a mixture of C3 grasses and forbsand C4 grasses It is well documented that the d13C signature ofC3 plants (ranging from26permil to28permil) is easily distinguishedfrom that of the C4 plants (ranging from 12permil to 15permil)(DeNiro and Epstein 1977 Cerling et al 1989 Tieszen 1991Ehleringer et al 2000 Fox et al 2012 Throop et al 2013) Otherprocesses besides vegetation change alter isotopic ratios (such aspreferential degradation of some plant materials microbiologicalactivity Suess effect and adsorption processes) and may causeshifts of the order of 1ndash3permil (Friedli et al 1987Becker-HeidmannandScharpenseel 1992Trolier et al 1996Ehleringer et al 2000SantRuCkova et al 2000 Krull et al 2005) d13C enrichmentin temperate forest soils larger than 3ndash4permil is generally attributedto the presence of remaining old C4-derived SOC Following thereasoning of Wynn and Bird (2008) we sampled the upper soils(A horizon) and assumed that the intrinsic metabolic processesoccurring in C3 and C4 vegetation impart an unequivocald13C signature into the SOM and that these signatures persistin decaying organic matter such that increasing depth yields achronologic sequence recording vegetation change At the studyarea d13C values of soils in the grassland matrix averaged ndash16permil(Table 4) to 26permil in forested areas dominated by C3 plantsDepth-integrated SOM d13C significantly reflected the currentvegetation cover (mixed C3ndashC4 and solely C3) and accordinglycould be used as a reference to trace past changes in vegetationrecorded in the SOM at deeper soil layers The d13C values ofSOC consistently increased with depth reflecting a higherpercentage of C4-derived carbon in the older soils that isinterpreted to reflect a site-wide shift in vegetation from C4 toC3 plants

Multiple regression and ordination results supported the roleof soil depth and incident radiation both of which influence soilmoisture content as factors governing the occurrence of xericserpentine savannaQuercus species Abiotic factors in areas withhigher importance of shade-tolerant species were associated withdeeper soils higher bulk density and higher Ca Mg ratios In oursite early successional woodlands were favoured in areas of highincident radiation and broadleaf forests were more developed onslopes containing more clay rich and denser soils Diversity ofwoody specieswas higher in areaswith deeper soilsUnlikemanysystems with heavy metals we found no correlations betweentoxic metals (Cr and Ni) and vegetation dynamics

Historical records show that oak savanna ecosystems wereimportant components of the Mid-Atlantic landscape (Shreveet al 1910) The lack of any oaks before the 1850s suggests thatthey were likely harvested for fuel posts or to support grazingFarming of these serpentinite areas is not documented and isunlikely given the inhospitable nature of the soil Our studyindicated that tree recruitment before the 20th centurywasmostlyoaks perhaps regenerating from root sprouts or propagules in theseedbankXeric oak recruitment decreasedover timeandnoxericoaks younger than 40 years established in the forested areasalthough seedlings and saplings still exist in the remnantgrasslandsavannas This failure to regenerate is likely toreflect the consequences of changing disturbance regimes suchas fire-suppression (Nuzzo 1986 Abrams 1992 2006 PetersonandReich 2001) alongwith competitionwith invasive andmesictree species (Nuzzo 1986) White-tailed deer and squirrelherbivory can reduce recruitment especially in isolatedsystems such as these serpentinite landscapes (Shea andStange 1998 Haas and Heske 2005) A turning point forcommunity development appears to be ~1900 with theencroachment of Pinus virginiana and Juniperus virginianaThese conifer species have a variety of characteristics that can

30

25

Minor species

Mesic species (A rubrum F grandifolia N sylvatica P serotina)

Conifers

Quercus spp (Mixed)

Quercus spp (Xeric)

(n = 163)15

5

20

10

Num

ber

of tr

ees

sam

pled

Decade of origin

01850 1860 1870 1880 1890 1900 1910 1920 1930 1940 1950 1960 1970 1980 1990

Fig 8 Age distribution from trees on the basis of core data (n= 163)Xeric species areQuercusmarilandica andQ stellataMixed oak species includeQ albaQ montana Q rubra and Q velutina

Conifer encroachment on ultramafics Australian Journal of Botany K

potentially alter ecosystem processes for decades as they matureand senesce Both conifers are aggressive invaders because ofrapid growth high seed production high seed-dispersalefficiency tolerance of xeric conditions and opportunistic useof ECM mycorrhizae (Ormsbee et al 1976 Holthuijzen andSharik 1984 1985 Briggs andGibson 1992Axmann andKnapp1993 Thiet and Boerner 2007) Once established these speciescan have a profound effect on soil properties as a result ofaccumulation of organic matter on the forest floor and theeventual decomposition of detritus which leads to higherwater-holding capacity and lowering of soil pH This positivefeedback between soil depth and litter cover exerted by theconifers could promote succession towards non-serpentineconditions

The initial serpentine edaphic signature acts as a filterpreventing species that belong to the regional flora but lack thetraits required to survive in such conditions to successfullycolonise these habitats By the 1970s such filters appear tohave diminished Current site characteristics have shiftedtowards more mesic conditions with changes in soil propertiessuch as bulk density soil depth pH SOM and metalsconcentration Over time the community has changed fromxeric (grassland or oak savanna) to one dominated by moremesic oaks and shade-tolerant Acer Nyssa and Fagus speciesThemajor onset ofAcer establishment began in the 1940sUnlikethePinus and Juniperus establishment pulse which proceeded inan open-canopy situation Acer dominance likely proceededunder closed canopy and canopy gaps because it is shade-tolerant One may speculate that over the past 100ndash200 years

the landscape has been dominated by Quercus with patches ofPinus and Juniperus that historicallywere present at lownumbersbecauseof disturbance (miningfire grazing logging)Hilgartneret al (2009) recorded evidence of the presence of Pinusvirginiana from the early 1800s in a serpentine barren siteonly 50 km distant Over the past 60 years Acer has becomeabundant where it along with other species currently present(N sylvatica Cornus florida and Celtis occidentalis) wouldnormally be considered an aberrant species in a hostile edaphicsetting Age and diameter structure indicated a shift away fromQuercus and towards an A rubrum-dominated landscape Thissame species shift has been widely reported throughout theeastern deciduous forests of North America (Abrams 1998Nowacki and Abrams 2008 McEwan et al 2010)

In the context of serpentinite ecosystems the influxof conifersnearly a century ago has acted as a landscapemodulator (Shachaket al 2008) The resulting build-up of SOM increases the water-holding capacity and other resources such as ground-levelradiation facilitate the influx of new species adapted to thealtered conditions As the difference between the mesicwoodlands and the xeric patches decreases the resourcecontrasts also decrease leading to increasing communitysimilarity Different oak species can grow in a wide variety oflight and moisture conditions However none of the traditionalserpentine oak species is considered shade tolerant to the samedegree as are Acer spp Fagus grandifolia and Nyssa sylvaticaWith progressive development of the soil and a dense-canopyclosure growth and germination of the traditional xeric oakspecies are likely to be suppressed marking a shift from

Colonisation

Fig 9 Conceptual model for serpentine (orangendashred colouring) habitats and species redistribution resulting in similar structure for a typical piedmontgeoecologic system (bluered colouring) Bedrock geology (specifically elemental concentrations and structures) exerts its control on soil fertility with subtlevariations resulting in markedly different above-ground and below-ground communities The change in soils over time (dsdt) equation from Huggett (1995)is a function of the various drivers including b (biology) r (topography) a (atmospheric effects) s (soils) and h (hydrology) Through changing vegetationdynamics and species redistribution the tree functional groups and forest types will see increasing similarity as a result of converging successional trajectoriesRegional species pools stochastic events and many other factors determine ultimate outcomes

L Australian Journal of Botany J Burgess et al

xerothermic to mesic andor nutrient-demanding speciesFigure 9 details a conceptual model for pedologic and ecologicchanges at serpentine sites of theMid-Atlantic with successionaltrajectories converging on a species redistribution and loss oftypical serpentine species such as Cerastium arvense varvillosissimum Andropogon gerardii and the xeric oaks(Q marilandica and Q stellata)

Although the canopy is still oak or pine dominated thetrajectory points to a maplendashbeechndashred oak-dominated forestThe Mid-Atlantic has already lost much of its originalgrasslandndashoak woodlandndashforest mosaic This state transitionmay lead to the loss of a unique ecosystem and thustaxonomic homogenisation with the regional flora Thesechanges result from anthropic influences through changes inland use and fire frequency the loss of Castanea dentata(American chestnut) and changing herbivore populationsNearly a century ago these areas gave way to a mosaic ofsmall woodlands typically characterised by xeric oaksinterspersed with shallow rooting low nutrient-tolerantconifers Released from suppressive disturbance regimes(herbivory and fire for example) Pinus virginiana andJuniperus virginiana quickly became the dominant speciesHowever our data suggest that the landscape is nowundergoing another distinct change as the once-open canopystructure has becomemore closed Edaphic properties such as soildepth chemistry and moisture that once drove communityorganisation across environmental stress gradients are nowbeing modulated by species no longer filtered by theserpentine syndrome This modulation may allow for therecruitment and establishment of drought-sensitive speciesbeyond their typical tolerance limits (Warman and Moles2008) This unique ecosystem which took centuries tomillennia to evolve appears to be fading and may soonevanesce into the past

Sources of funding

Our work was funded by the National Science Foundation(DEB-0423476 and DEB-1027188) and grants from theR Balk and D Elliott Field Fund (Johns Hopkins University)

Conflicts of interest

No conflicts of interest

Acknowledgements

We thank Drs Richard Back Naomi Levin Bruce Marsh and Ben Passey foruse of their elemental and isotopic facilities We also thank Dr David Veblenand two anonymous reviewers for numerous constructive comments on thegeology

References

Abbe C (1898) A general report on the physiography of Maryland JohnsHopkins University (Johns Hopkins University Press Baltimore MD)

Abrams MD (1992) Fire and the development of oak forests Bioscience 42346ndash353 doi1023071311781

Abrams MD (1998) The red maple paradox Bioscience 48 355ndash364doi1023071313374

AbramsMD (2006) Ecological and ecophysiological attributes and responsesto fire in eastern oak forests In lsquoFire in eastern oak forests deliveringscience to landmanagersrsquo (EdMBDickinson) pp 74ndash89 (USDA ForestService General Technical Report NRS-P-1)

Alexander EB (2009) Serpentine geoecology of the eastern and southeasternmargins of North America Northeastern Naturalist 16 223ndash252doi1016560450160518

Alexander EB Ellis CC Burke R (2007) A chronosequence of soils andvegetation on serpentine terraces in the Klamath Mountains USA SoilScience 172 565ndash576 doi101097ss0b013e31804fa22d

Arabas KB (2000) Spatial and temporal relationships among fire frequencyvegetation and soil depth in an easternNorthAmerican serpentine barrenThe Journal of the Torrey Botanical Society 127 51ndash65doi1023073088747

Archer S Schimel DS Holland EA (1995) Mechanisms of shrublandexpansion land use climate or CO2 Climatic Change 29 91ndash99doi101007BF01091640

Axmann BD Knapp AK (1993) Water relations of Juniperus virginiana andAndropogon gerardii in an unburned tallgrass prairie watershed TheSouthwestern Naturalist 38 325ndash330 doi1023073671610

Barton AM Wallenstein MD (1997) Effects of invasion of Pinus virginianaon soil properties in serpentine barrens in southeastern PennsylvaniaThe Journal of the Torrey Botanical Society 124 297ndash305doi1023072997264

Bazzaz FA (1968) Succession on abandoned fields in the Shawnee Hillssouthern Illinois Ecology 49 924 doi1023071936544

Becker-Heidmann P Scharpenseel H-W (1992) Studies of soil organicmatterdynamics using natural carbon isotopes The Science of the TotalEnvironment 117ndash118 305ndash312 doi1010160048-9697(92)90097-C

Belcher JW Keddy PA Twolan-Strutt L (1995) Root and shoot competitionintensity along a soil depth gradient Journal of Ecology 83 673ndash682doi1023072261635

BouttonTWArcher SRMidwoodAJZitzer SFBolR (1998)d13Cvalues ofsoil organic carbon and their use in documenting vegetation change in asubtropical savanna ecosystem Geoderma 82 5ndash41doi101016S0016-7061(97)00095-5

Bouyoucos GJ (1936) Directions for making mechanical analyses of soilsby the hydrometer method Soil Science 42 225ndash230doi10109700010694-193609000-00007

Briggs JMGibsonDJ (1992)Effect offire on tree spatial patterns in a tallgrassprairie landscape Bulletin of the Torrey Botanical Club 119 300ndash307doi1023072996761

BrownML Brown RG (1984) lsquoHerbaceous plants of Marylandrsquo (Universityof Maryland College Park MD)

Burgess JL Lev S Swan CM Szlavecz K (2009) Geologic and edaphiccontrols on a serpentine forest community Northeastern Naturalist 16366ndash384 doi1016560450160527

Butaye J Jacquemyn H Honnay O Hermy M (2002) The species poolconcept applied to forests in a fragmented landscape dispersal limitationversus habitat limitation Journal of Vegetation Science 13 27ndash34doi101111j1654-11032002tb02020x

Callaway RM Davis FW (1993) Vegetation dynamics fire and the physical-environment in coastal central California Rid B-7010-2009 Ecology 741567ndash1578 doi1023071940084

Carroll W D Kapeluck P R Harper R A and Van Lear D H (2002)Background paper historical overview of the southern forest landscapeand associated resources In lsquoSouthern forest resource assessmentrsquo (EdsDNWear JGGreis) pp 583ndash605 (USDepartmentofAgricultureForestService Southern Research Station Asheville NC)

Cerling TE Quade J Wang Y Bowman JR (1989) Carbon isotopes in soilsand palaeosols as ecology and palaeoecology indicators Nature 341138ndash139 doi101038341138a0

Chao A (1984) Nonparametric estimation of the number of classes in apopulation Scandinavian Journal of Statistics 11 265ndash270

ChaoA (1987) Estimating the population size for capturendashrecapture data withunequal catchability Biometrics 43 783ndash791 doi1023072531532

Chazdon RL Colwell RK Denslow JS Guariguata MR (1998) Statisticalmethods for estimating species richness of woody regeneration in primaryand secondary rain forests of Northeastern Costa Rica In lsquoForest

Conifer encroachment on ultramafics Australian Journal of Botany M

et al 1989) TheCHS ismore enigmatic lyingwithin theBoulderGneiss metasedimentary (diamictite) unit and bounded to thenorth by metagraywacke (Fig 1) and is likely to represent afragmented sliver of a once larger complex Vegetation atop thestudy area varies from mature forests to savanna-like areas withopen-grown trees and diverse grassland with a wide variety ofgrasses and forbs

A forest encloses the remnant grasslands of the study area andexhibits a transitional ecotone ranging from 10 to 15m (Fig 2)The dominant trees within the forest are oak species (Quercusspp) whereas at the transition eastern red cedar (Juniperusvirginiana L) and Virginia pine (Pinus virginiana Mill) arealso important On the basis of historical records (eg diaries

and mining and forestry reports such as found in Jefferis 1880Williams and Darton 1892 Abbe 1898) aerial photographs(Maintained at the Maryland Geological Survey and HarfordCounty Soil Survey Libraries) and interviews this study area hasnot been farmed or burned during the past century Shortly after agrassland or savanna survey by Tyndall and Farr (1989) coniferswere eliminated in the grassland in a restoration attempt In thesavanna and grassland patches disjunct populations of C4

bunchgrass species including little bluestem (Schizachyriumscoparium (Michx) Nash var scoparium) big bluestem(Andropogon gerardii Vitman) arrowfeather threeawn(Aristida purpurascens Poir) and prairie dropseed (SporobolusheterolepisAGray) are interspersed with a variety of herbaceousC3 species such as Carex umbellata Willd (parasol sedge) andSmilax spp (greenbrier)

Abiotic characteristicsWecarried out detailed geologicmapping to assess howvegetationcharacteristics were influenced by geologic substrate Aftergeologic mapping 32 quadrats (see Fig 1) were establishedusing stratified random sampling to encompass the variety ofenvironmental conditions in the area Quadrats (10 15m)were oriented up and down slope to cover distinct changes inenvironmental gradients and community types Near the centre ofeach plot latitude longitude elevation slope and aspect weredetermined and recorded with a GPS device

During the mapping phase representative rock samples of thefreshest serpentinites (n= 12) were systematically selected forpetrographic analysis from samples collected (n = 34) duringfield mapping Both major and trace elemental whole-rockanalyses were conducted using X-ray fluorescence (XRF)Major elements were determined on glass beads fused fromignited powders to which Li2B4O7 was added (1 5) in aplatinum crucible at 1150C Trace elements were analysed onpressed powder pellets by the same instrument a Bruker AXS S4Explorer following the procedures outlined by Potts (1987)

Woodland Plot (10 x 15 m)

Relict Barrens Areas

Inset area

Highly Sheared or Fractured Areas

Lithologic Contact

Wissahickon Group (Diamictite)

Scale (m)

0 200

Serpentinite amp Ultramafics

Geologic Sample

PENNSYLVANIA

QUADRANGLE LOCATION

Fig 1 Topographic map of the sampling area in near MarylandPennsylvania boundary USA Note that plot sizes are not to scale

Fig 2 Field photograph showing insular relict barrens and grassland being encroached upon by eastern red cedar (Juniperus virginiana) with surroundingmixed-oak (Quercus) woodlands Photo by J Burgess

Conifer encroachment on ultramafics Australian Journal of Botany C

Counting statistics errors are 01 for major elements andaverage detection limits are 50ndash100mg kgndash1 for major oxidesand 5ndash10 ppm for chromium (Cr) nickel (Ni) titanium (Ti)iron (Fe) manganese (Mn) Ca potassium (K) and phosphorus(P) All XRF samples were quantified using a matrix matchedset of USGS- and NIST-certified soil standards At least oneduplicate one replicate and one certified reference sample(SRM 2709 San Joachin Soil) was run with every 10 samplesanalysed to monitor for external and internal reproducibilityLoss on ignition (LOI) for samples is measured by igniting rockpowders in alumina crucibles to 1200C for 10min in a mufflefurnace

Soil characteristics evaluated for each plot or quadrat includedbulk density soil depth pH soil organic matter content majorand trace element analysis carbon (C) nitrogen (N) and isotopicand textural analysis Surface soil epipedons were evaluated forbulk density (g cmndash3) At each location we composited five soilcores (corners plus centre) that were taken with a 5-cm-diametersoil corer Soil coreswere sectioned into 0ndash5 5ndash10 10ndash15 (and atsome sites 5ndash15) cm increments dried at 40C to a constantweight and then calculated as dryweight of sample per unit of soilvolumewith roots removed Soil parameter valueswere averagedto provide a representative value for each 150-m2 plot

Forest soil depth (depth to obstruction) was measured with apush soil probe at five random locations in each plot For soilmoisture or soil depth comparisons from savanna into forestedareas four transects (35ndash68m long) were established in April2011 Our rationale was that distance from bedrock wouldrepresent a proxy for time and points along these transectswould generally represent soil chronosequence (Jenny 19411980) Because of the inherent low moisture values ofserpentine soils measurements were made during the weekafter a 5-mm or greater rainfall event Soil moisture values at0ndash6-cm depth were measured with a ThetaProbereg soil moisturesensor (Dynamax IncHoustonTXUSA) and soil depthswith apush probe (internal diameter = 47 cm) every 1m

Soil samples for chemical analysis were air-dried and passedthrough a USA Standard No 10 sieve Bedrock thin-sectionsoccasionally contain carbonate (calcite andmagnesite) thereforeinitial soils were evaluated for carbonate by an acid (1Mhydrochloric) test The HCl test showed that the soil samplesdid not contain carbonates therefore total C in the soil representsorganic C only

Moist soil colour was determined in the field using Munsellcolour charts (Munsell Color Grand RapidsMI USA) Sampleswere combined mixed and split to give two samples of200ndash400 g that were lightly crushed and sieved to 2mm andsampled for textural analysis (sieve followed hydrometermethod) (Bouyoucos 1936) Soil pH was measured with a 20 gin 20mL soilndashwater slurry by using a glass electrode Theremainder was oven-dried at 30C and powdered using anagate centrifugal ball mill A powdered subset wassubsequently used for XRF and C N analyses Total N and Cwere measured with an automated Perkin-Elmer 2400 Series IICHN analyser (Perkin-Elmer Cornell Labs Ithaca NY USA)Soil organic matter content was estimated as loss on ignition(LOI) at 550C for 3 h Total C N and isotopic analysis wereconducted at the stable isotope facilities at either theUniversity ofCalifornia at Davis CA or Cornell Laboratories Ithaca NY

using an elemental analyser interfaced with continuous-flowisotope-ratio mass spectrometer Values reported are typicallymeans with the associated standard error All the values werereported as per mil (permil) relative to the international standardVienna Pee Dee Belemnite (V-PDB) and were calculated as

d13 CethpermilTHORN frac14 Rsample Rstandard

Rstandard

1000

where d13C is the difference between 13C 12C ratio (R) of thesample and that of the standard Increase in the d value reflectsincreases in the relative amount of the heavy isotope componentor a reciprocal decrease in the light isotope component

Biotic characteristicsTo monitor the serpentine ecosystem-scale vegetation-boundarymovement we examined a time series of high-resolution imagesfrom Google Earth (1994ndash2013) along with aerial photographyfrom 1937 1952 1971 and 1988 using the grid-square methodThe more recent data serve to complement the work by Tyndall(1992) on the expansion of conifers into barrens areas All woodyplant species were recorded for each sampling plot To identifyplants we followed the key by Brown and Brown (1984) withnomenclature clarification via the integrated taxonomicinformation system Woody plants with diameter-at-breast-height (DBH) of gt10 cm are referred to as trees those withDBH of lt10 cm and gt2 cm or 2m height as saplings and anywoody specieslt2-cmDBHare grouped as seedlings or vines Forspecies where DBH is impractical (shrubs) stemsweremeasuredwith a caliper near the root collar Absolute and relative stemdensity and basal area were derived from these data

Canopy cover was measured at each plot centre with aspherical densiometer Plots were classified as savannawoodland or forest on the basis of canopy cover as followssavanna (0ndash25 canopy cover) woodland (26ndash60 canopycover) and forest (gt60 canopy cover) Ground-layerpercentage cover for soil leaf litter and rock was estimatedvisually

To assess vegetation change over time historical recordswerecombined with tree-core data (n= 163) One increment core wasextracted for each tree species per plot when possible Cores weretaken at breast height mounted and sanded following Stokes(1996) and Stokes and Smiley (1968) Because increment coreswere collected at 137mabove thegroundweconsidered the agesobtained as the minimum age of trees at breast height Repetitivecoring was carried out to facilitate the interception of the pithMissing rings in incomplete cores were estimated using diameterage-regression curves (Lorimer 1980) Ring counts were madeusing a stereomicroscope at 10ndash40 magnification Each coresample was skeleton plotted to identify narrow rings (Speer2010) The resulting age distribution was divided into datethresholds to record pulses or absence of regeneration as cohorts

Data analysisTo quantify forest structure the quadrats were characterised bybasal area (m2 handash1) and density (stems handash1) of live specimens ofthe principal tree species Tree species were grouped on the basisof habitat affinity (xericmesic) using the USDA silvics manual(Walters and Yawney 2004) Diversity indices were calculated

D Australian Journal of Botany J Burgess et al

for each plot and on pooled plot data using the ShannonndashWeaverindex (Shannon and Weaver 1949) and the Simpson index(Simpson 1949) as measures of a diversity (a) along withestimators of total diversity using rarefaction and non-parametric approaches of Chao 1 Chao 2 and ACE (Chao1984 1987 Chazdon et al 1998) The importance value (IV)is a measure of relative abundance and was calculated as theaverage of relative density and dominance

The KolmogorovndashSmirnov test statistic was used to assessnormality of the soil variables Those that failed the normality testwere transformed to normal or near-normal distributionPercentage measurements (soil particle analysis abundance CN cover XRF whole-rock oxide weight) were arcsinetransformed soil elemental data were loge transformed fromppm or ppb concentrations slope and soil depth were natural-log transformed and bulk density was square-root transformed

Site microclimate variability was compared using geographiccoordinates (latitude and longitude) inclination and aspect foreach plot to calculate the heat load and the potential annual directincident radiation by applying the third equation from McCuneand Keon (2002) The effects of topographic (slope inclinationannual direct incident radiation and heat load) and soil variables(elemental analysis in addition to pH soil organic C total NC N depth texture and d13C values) on species were examinedusing principal components analysis (PCA) to explore the mainfactors and generate hypotheses as to the reciprocal effects ofthese properties on vegetation Redundancy analysis (RDA)ordination as constrained by the environmental variables wasperformed to identify the main floristic gradients (basal areadensity percentage xeric species) and the edaphic andtopographic factors explaining those trends Two-wayANOVA was used to test for the differences in soil d13C C Nand C N with respect to vegetation (grassland forest) and soildepth ANOVAs were followed by Tukeyrsquos HSD (chosenbecause of uneven number of replicates) multiple pairwisecomparison to relate differences in elemental content and depth

Differences in site woody-species composition wereinvestigated by calculating the pairwise ecological distance forthe entire species matrix The Mantel test was used to determinewhether site similarity in the tree community was related to siteenvironmental variables Multiple linear regression analysis wasused to identify site factors that had a significant effect on speciesabundance The eight highest-ranked species on the basis of IVwere evaluated in termsof their abundances as responsevariablesand soil and topographic data were used as explanatory variablesOrdination using direct (RDA) and indirect (PCA) gradient

analysis was employed to extract variation in the underlyingstructure of the vegetation data as it relates to environmentalvariables (Ludwig and Reynolds 1988) PCA was performed tofind the main factors determining the reciprocal effects of soiland vegetation

The BrayndashCurtis distance with raw abundances was usedto determine the degree of vegetation dissimilarity an aspectof b (b) diversity between sites and vegetation types (woodlandand forest) based on presenceabsence We used a Mantel test(McCune and Grace 2002) to compare the matrix of geographicdistances between plots with the matrix of ecological distancePotential relationships between abiotic variables and diversityvariables (floristic similarity evenness and Simpsonrsquos diversity)were explored Tree-species compositions of early successionalwoodland and more mature forest designations based onpercentage canopy were compared using non-parametricanalysis of similarities (ANOSIM) based on the hypotheses ofno difference between stages

Gamma (g) or serpentine landscape-area diversity wasestimated using all 32 plots to identify species richness at thelandscape scale by constructing a rarefaction curve for all samplesand all species in the dataset

For all statistical tests significance was determined at thea= 005 level other P-values were reported if lt01 The methodfor evaluating the Mantel-test statistic was the t-distribution(Mantel 1967) and the standardised Mantel statistic (r) as ameasure of effect size (McCune and Grace 2002)

All data were analysed using the statistical computingenvironment R (v 311) and the BiodiversityR package(Kindt and Coe 2005) ANOSIM Mantel and other tests wereperformed with n= 1000 permutations Reported errors arestandard errors (se) unless indicated otherwise

Results

Abiotic substrate and soil composition

Mean whole-rock major-element values were typical of alteredultramafic rocks (Table 1) Calcium and aluminium (Al) contentswere low reflecting disappearance of the calcic pyroxene byserpentinisation Only samples where clinopyroxene wasdetected showed an increase in Ca Samples with elevated Alare generally accompanied by the presence of chlorite in therocks As is typical of ultramafics the immobile refractoryelements Cr and Ni were elevated Ternary diagrams depictingthe measured whole-rock geochemistry of the samples illustratedrelative compositional variations in the systemMgOndashSiO2ndashH2O

Table 1 Summary of the whole-rock geochemistry parameters for serpentinite samples obtained from X-ray fluorescence (XRF) analysisSerpentinite water content based on loss on ignition (LOI) ~124 03 weight Al aluminium Ca calcium Cr chromium Fe iron K potassium Mg

magnesium Na sodium Ni nickel Si silicon Ti titanium V vanadium Zn zinc

Lithology Major element geochemistry (weight ) Trace elements (mg kgndash1)Serpentinite n= 12 Si Al Fe Mg Ca Na K Ti Cr Ni V Zn

Mean 4129 031 954 3669 042 000 007 002 283799 231968 1189 4385se 088 015 078 070 022 000 002 000 28651 15716 334 594Soils (0ndash5 cm) n= 32 3513 091 657 625 024 015 070 048 252651 115744 9678 10705se 132 019 040 060 006 003 005 003 27641 13040 642 463Soils (5ndash15 cm) n= 48 3799 105 744 762 015 015 069 046 234790 124434 9725 8426se 106 020 036 060 004 002 005 003 20726 11024 554 311

Conifer encroachment on ultramafics Australian Journal of Botany E

(Fig 3) Concentrations of soilmacronutrientsCN P andKwerenot related to ultramafic parent material

Soil depths varied from105 cm to52 cmwith variable texture(mean of 134 gravel) across the sites Soil colours ranged fromdark brown (75YR 32) to grey (75YR 61) in the upper 1ndash5 cmof the soil profile to yellowndashbrown (10YR 56) and redndashbrown(25YR 44) with depth Talcose serpentinites yielded moreyellow (10YR Hues) varieties The mean and standard error ofsoils collected from the 32 plots recorded a narrow range of

edaphic characteristics including soil depth 289 20 (cm) pH54 01 bulk density 088 007 (gm3) soil organic matter(SOM) 1251 26 () sand 5103 18 () clay 267 11() and C N 1518 09

Soil moisture and soil depth increased with distance from thebarrens areas Strong positive correlationwith distancewas foundfor both soil moisture and depth (r= 0570 P lt 0001 andr= 0848 Plt 0001) respectively (Fig 4) Other soil physio-chemical parameters such as texture and trace element contentdid not show correlations with distance (P gt 005 Spearmanrsquosrank correlation)

Magnesium concentrations were high and correspondingCa Mg ratios low (lt05) across all sites The Ca Mg ratio isa measure of serpentine cation imbalance Surface soils hadhigher Ca Mg (02 to 05) ratios than did subsurface soils(005 to 02) Chromium and Ni concentrations were elevatedwith a maximum of over 6000 ppm and 2700 ppm respectively

The pH of the soils was typically acidic (surface soils average~54) but ranged from (~4ndash7) Soil pHwas negatively correlatedwith Ca Mg (r= ndash049 P lt 005) and C N ratios (r= ndash062P lt 005) SOM content (based on LOI) had a mean of 166in the upper 5 cm declining to below 7with depth (10ndash15 cm)N content was generally low and C N ratios varied between 5and 24with the higher values on themore developed soils Depthprofiles of SOM displayed increasing d13C isotopic signaturesfrom 2means of ndash2623permil (0ndash5 cm) to ndash2095permil (10ndash15 cm)(Table 2) whereas remnant grassland areas recorded meand13C isotopic signatures of ndash1623permil (0ndash5 cm)

Woody-species characteristics

Using high-resolution Google Earth imagery the current (2013)percentage of the site that contains grassland or savanna is lessthan 2 The forests include xeric woodlands with stands ofP virginiana and J virginiana intermingled with Quercusmarilandica Meunch (blackjack oak) and Quercus stellata

Talc

Enstatite

Serpentine

Olivine

Ca Mg

Si

Diopside

Typical Ultramafic

Fig 3 Ternary diagram detailing the chemography of Cherry Hillserpentinites (open circles) projected onto the plane CaOndashMgOndashSiO2 alsoshowing typical ultramafic rock (filled square) and mineral compositions(solid circles)

60

Soil depth (cm)

Percent moisture

50

40

30

20

Soi

l dep

th (

cm)

and

moi

stur

e (

by

volu

me)

Distance along transect (m)

10

00 10 20 30 40 50 60

Fig 4 Monotonic change of variables along transects The relationship between soil depth percentage moisture (volume) and distance from barrens bedrockwere assessed using correlational analysis (Plt 00001) Data are compiled averages of four transects from rocky barrens to woodlandndashforest ecotone Thecorrelation of the two variables soil depth and moisture is significant (r = 067 Plt 0001)

F Australian Journal of Botany J Burgess et al

Wang (post oak) as well as more mesophytic species Largeopen-crown trees are rare more typical are contorted down-sweeping branches of Q marilandica surrounded by stand-grown trees Aerial photographs from 1937 show conifersadvancing from the dendritic drainage areas Mesophyticforests are composed of these oaks and others such as Quercusrubra L (red oak) Q alba L (white oak) Q montana Willd(chestnut oak) and Q velutina Lam (black oak) along withnumerous other species most notably Prunus spp (blackcherry wild cherry) Pinus spp (eg white pine shortleafpine) Fagus grandifolia Ehrh (American beech) Acer rubrumL (red maple) Sassafras albidum Nutt (sassafras) Populusgrandidentata Michx (bigtooth aspen) Nyssa sylvaticaMarshall (blackgum) Betula lenta L (sSweet birch) andRobinia pseudoacacia L (black locust) On exposed rocksurfaces the serpentine endemic (as reviewed by Harris andRajakaruna 2009) Cerastium velutinum var velutinum Raf(serpentine mouse eared chickweed) and other serpentinophilessuch as Symphyotrichum depauperatum Fern (serpentine aster)and saplings of Q marilandica are common

Biodiversity richness estimators Chao 1 Chao 2 and ACEestimated maximum woody-species richness between 263and 308 species The observed richness was 24 species(Table 3) Overall diversity and mean diversity for the 32 siteswere calculated as 274 (X = 167 s= 029) (Shannon index) and

092 (X = 078 s= 008) (Simpson index) On the basis of relativeIV the most important species were P virginiana andJ virginiana (Table 3) These two species had similar relativeIVs and stood out from all other species A second group ofspecies included A rubrum and Q marilandica At the genuslevel Quercus was the most important contributor to the treecommunity with 39 relative IV On the basis of basal area themost dominant species were Q rubra P virginiana andQ marilandica Constancy values (Table 3) showedA rubrum (078) as a clear break from the next grouping ofspecies (P virginiana and J virginiana) whose constancy valueshovered near 05 followed by Q marilandica F grandifoliaN sylvatica and Q stellate with constancy values near 04

The Mantel test comparing floristic similarity andenvironmental distance showed significant (P lt 005)correlations between environmental distance and ecologicaldistance for the following parameters heat load bulk densitybasal area diversity (H) percentage sand Ni and Cr(Table 4) The ANOSIM statistic was calculated only for thecanopy category (woodlandforest) with a resulting r= 014 andP-value of 005 Both the Mantel and ANOSIM tests yieldedsignificant (Plt 005) but weak correlations

Relationships between vegetation and abiotic properties

Results of multiple regression analysis were evaluated tounderstand the response of species distributions to soilproperties Table 5 shows the regression results for the eightspecies the importance of which met or exceeded 5 withtopographic and edaphic parameters as the predictor variablesPredictor variables that indicated a meaningful contribution toexplaining species abundancewere incident radiation and soilMgcontent for Q marilandica incident radiation for J virginianaand Q stellata soil organic carbon (SOC) for A rubrum soildepth for F grandifolia and bulk density for the conifers

PCA was performed for 16 abiotic parameters across the 32sampled quadrats so as to identify covariation between the solarradiation and soil variables (Fig 5) Themain trend of variation in

Table 2 Descriptive statistics for soil d13C (permil) and soil organic matterin grassland and woodland elements

Means followed by the same letter are not significantly different (ANOVApost hoc Tukeyrsquos HSD P = 005)

Statistic Grassland Forest(0ndash5 cm)

Forest(5ndash10 cm)

Forest(5ndash15 cm)

Forest(10ndash15 cm)

Mean ndash1623a ndash2623b ndash2359bcd ndash2298de ndash2095eMinimum ndash1791 ndash2923 ndash2725 ndash2649 ndash2514Maximum ndash1423 ndash1928 ndash1811 ndash1972 ndash1654se 068 044 067 065 060

Table 3 Constancy importance (average value of a species when present) density dominance andimportance value (average of relative density and relative dominance) measures for woody stems in

serpentinite woodland at Cherry Hill Maryland USA

Species Constancy Importance Dominance(m2 handash1)

Density(stems handash1)

Importancevalue (IV) ()

Acer rubrum L 078 109 5799 7778 988Ailanthus altissima Mill 003 003 075 222 027Betula lenta L 006 006 346 444 060Celtis occidentalis L 006 006 059 444 041Cercis canadensis L 003 003 078 222 022Cornus florida L 006 006 146 444 038Diospyros virginiana L 003 003 185 222 024Fagus grandifolia Ehrh 043 065 5713 4667 619Juniperus virginiana L 050 106 8657 7556 1214Kalmia latifolia L 012 015 163 1111 106Liriodendron tulipifera L 009 009 2372 667 108Nyssa sylvatica Marsh 043 075 5769 5333 720Pinus strobus L 003 003 692 222 090Pinus virginiana Mill 056 078 16498 5556 1343Populus grandidentata Michx 003 006 640 444 058Prunus serotina Ehrh 028 028 2046 2000 271

Conifer encroachment on ultramafics Australian Journal of Botany G

the soil chemical properties and topographic variables asreflected by the PCA Axis 1 (accounting for 30 of variance)was one dominated by significant correlations of soil texture(sand and bulk density) soil depth Mg Cr and Niconcentrations incident radiation and heat load The secondtrend of soil variation as reflected by the PCA Axis 2 (17 ofvariation) can be interpreted as a combined gradient of essentialnutrients (C andN) and SOC The PCA (axes accounting for 44of the variation) revealed strong positive associations betweenboth soil depth and bulk density which were negativelycorrelated with heavy metal concentrations and pH SoilCa Mg ratios and Al concentrations were negativelycorrelated with macronutrients (C N P) and SOC Fig 5bshows the PCA factor map for the abiotic variables and thebiotic covariates for the sampling plots When thecharacteristics of woody-species community are included inthe PCA both basal area and stem density are positivelycorrelated with soil Ca Mg ratios and Al content Diversity(Shannonrsquos index) was strongly correlated with soil depth andbulkdensity andnegatively related toheavymetal concentrations

and heat load As individual-species basal area stem densityand diversity typically respond to environmental gradientsconstrained ordination utilising RDA was used to inferrelationships between the species matrix and theenvironmental matrix using Euclidean distance between sitesArrows of increasing species abundance represent RDA scoresThe sum of the RDA eigenvalues for the first two axes using allabiotic explanatory variables accounted for 368 of the woody-species composition After reduced variable selection based onANOVA tests of the ordinationmodel seven termswere selectedand these collectively explained 271 of the variance (Fig 6)Interestingly those species whose importance was gt4 arepartitioned into the four quadrants of the ordination From theRDAAxes1and2biplot it canbe seen that traditional serpentine-savanna (shade-intolerant)woody species (S albidumQ stellataand Q marilandica) are positively associated with increasingincident radiation and negatively associated with increasingsoil depth Al and Mg soil concentrations Post-disturbanceinitial colonisers (J virginiana and P virginiana) are stronglynegatively correlated with bulk densityQ alba andQ montanawhich are widely distributed across a variety of conditionsfrom wet mesic to subxeric are associated with increasingbulk density Shade-tolerant species such as N sylvaticaA rubrum and F grandifolia are correlated with increasingsoil depth and soils with more Ca and Al

Soils at the site differed in the total amount verticaldistribution and isotopic composition of SOM (Tables 1 2)ANOVA results indicated significant differences in theisotopic means according to their depth (F487 = 276P 00001) Tukeyrsquos HSD test showed that the mean d13C ofgrassland soils (ndash162 07permil) was significantly higher thanthat of shallow forested soils (ndash262 04permil) (Table 5)Shallow forested soils were significantly different from deepersoils (ndash2095 06permil) Total soil C and N and soil C N declinedwith soil depth C N ratios were significantly different for alldepths ranging from 1518 088 on the surface soils to1162 1 in soils 10ndash15 cm below surface (F266 = 59

Table 4 Effects of environmental distance on similarity between plotsMantel statistic calculated by comparing matrix of environmental distancewith BrayndashCurtis distance The standardised Mantel statistic (r) indicatesthe strength of the relationship between sample-point position and forest-floor thickness and the associated P-value indicates the significance of

that relationship

Environmental indicator Correlation coefficient (r) P-value

Heat load 0126 0009Bulk density 0236 0003Basal area 0141 0015Diversity (H) 0109 0034Percentage sand 0203 0002Nickel 0111 0034Chromium 0159 0009Canopy 0150 0009

Table 5 Results of multiple regression with list of factors significantly contributing to species abundanceSOC soil organic carbon Plt 0001 P lt 001 Plt 005 lsquorsquo Plt 01

Pvirginiana J virginiana A rubrum Q marilandica Q rubra N sylvatica Q alba Q stellata F grandifolia Q montana

IV () 1343 1214 988 940 810 720 703 626 619 440Incident radiation 09247 00249 01791 00008 04332 02136 09123 00450 03030 03225Heat load 02516 04869 03661 08494 09552 09203 09825 02522 01128 03502Soil depth 02599 03110 08937 00463 01881 03041 03381 01689 00242 09920Bulk density 00023 00037 08260 08860 05216 02157 00415 02686 09894 01171Texture ( sand) 00976 02800 02942 06247 08124 09313 03207 03451 01454 06054Calcium magnesium 09840 02962 07824 07031 09971 04587 07945 05441 06414 03639Magnesium 06146 05570 05725 00019 01500 06177 00914 03238 07491 03623Aluminium 07988 07323 01491 09176 00673 00583 08508 04152 05380 08685Chromium 05879 02397 02066 01012 06214 09977 06126 03991 06870 00434Nickel 08699 02237 07478 03211 08340 06190 09519 05875 06307 01074Phosphorus 05192 05066 07822 09800 07523 06641 04927 09619 09430 02268SOC 06435 01185 00298 00053 09782 02790 05185 02333 08701 07036Carbon nitrogen 05806 02325 06793 08013 05674 06443 01727 01105 08015 06972Carbon 07737 07963 02099 01231 04380 08148 04280 04885 07995 05145Nitrogen 07506 09537 01179 03581 06363 05329 03124 07375 00153 05744pH 07946 01863 05144 01547 00176 04256 02597 06743 01299 09047Multiple R2 059 068 057 079 057 046 051 055 062 056

H Australian Journal of Botany J Burgess et al

P = 0004) There were significant vegetation depth andvegetation depth effects on d13C of SOM (Table 5 Fig 7)

Age structure

In total 163 core samples were evaluated with results displayedin the stand-age structure plot (Fig 8) Skeleton plot chronologiesvaried among species Older minimum age at breast height wasgenerally seen inQ stellata (114 15 years) andQ marilandica(110 7 years) and the younger chronologies developed inspecies such as A rubrum (53 4 years) F grandifolia(56 8 years) and L tulipifera (43 8 years) Only xeric oakswere present in the oldest age classes The period before 1900 iscomposed of 54 xeric oaks and when all oaks are grouped theycomprise 79 of trees from that era although Q montana isabsent from this group (oldest minimum age is 88 years) Theoldest non-oak was N sylvatica (144 years) The period between

1900 and 1940 records a marked increase in conifers andconcomitant decrease of xeric oaks comprising 39 and 19of all species respectively Between 1950 and 1990 thepercentage of xeric oaks decreased to 3 and conifers to 13whereas Acer spp increased to 19 The number of shade-tolerant species rose from 125 in the decades from 1850 to1900 to 59 during the interval from 1950 to 1990

The age structure pooled from all the plots (Fig 8) showeddiscontinuous patterns with prominent peaks of regenerationpulses and periods of either low or absent recruitment Coniferestablishment and recruitment clearly increased during1900ndash1960 and then receded through the rest of the samplingperiod There was a conspicuous absence of younger xeric oaks(post-1970s) and only two xeric oaks from1950 to the present Atthe same time the percentage of non-oak deciduous treesincreased to 68 post-1950

10

05

00

Dim

2 (

171

7)

Dim

2 (

229

1)

Dim 1 (3038)

Dim 1 (402)

ndash10

ndash05

4

2

0

ndash2

ndash4

ndash6

ndash15

ndash15 ndash10 ndash5 0 5 10

ndash10 ndash05 00 05 10 15

(a)

(b)

Fig 5 (a) Unconstrained principal component analysis (PCA) correlation circle of pedological topological and community variables (b) Forest communityclustering along an edaphic gradient

Conifer encroachment on ultramafics Australian Journal of Botany I

Discussion

The present study examined a moderately sized area in the Mid-Atlantic but the changes ndash encroachment and mesophication ndash

represent potential widespread conservation problems Theserpentinite landscapes were once common components of theancient Piedmont (Carroll et al 2002 Owen 2002 Noss 2013)These unique plant communities have existed as persistentlsquoislandsrsquo embedded in a deciduous forest region for centuriesor longer We documented a state transition from grasslandand xeric oak savanna to conifer-dominated communities

that have given way to eastern broadleaf forests Comparisonsamong geologic pedologic and vegetation properties reflect thischange and indicate that the conversion to forest has been rapidand characterised by large changes in soil properties (pH water-holding capacity SOMcomposition) aswell as plant-communitystructure (physiognomyand stemdensity) and composition (xericto more mesic communities)

Thin-section and petrologic data confirmed that the geologicsubstrate is a serpentinised peridotitewith characteristics typical ofserpentinite terranes in the Appalachians (Misra and Keller 1978Mittwede and Stoddard 1989 Gates 1992 Alexander 2009)Although Ca is depleted in bulk rock samples because of thebreakdownof augite and interstitial plagioclase soil Ca Mg ratiosare higher in surface soils The concentrations of SOC C N andmacronutrients in woodland and forested areas are linked to theturnover of organic matter and not the parent material

Compared with the rock soils are enriched in most tracemetals although the Cr contents were similar This was likelyto be due to resistance of Cr-oxides to weathering and suggeststhat these phases were not likely sources for Cr in soil solutionsand plants of serpentine soils (Oze et al 2004) Bedrock to foresttransects demonstrated a strong correlation between soilmoistureand soil depth Soil nutrient content and soil moisture should beproportional to the cube of soil depth (Belcher et al 1995) thuseach of these parameters figured significantly in the PCA andRDA

Afforestation has been relatively recent in this region (Tyndall1992)Concomitantwith the increase in treedensity andsize is thecontraction of the grassland and savanna component study areafrom 18 in 1937 to 15 in 2013 which is consistent withprevious work documenting the invasion of serpentine savannasby conifers (Tyndall 1992 Barton andWallenstein 1997 Arabas2000)

In landscapeswhereC3 andC4plants coexist stableC isotopiccomposition of SOM represents a powerful method for

20

15

1

0

10

ndash10

ndash2 ndash1 0 1 2 3

05

ndash05

00

RD

A 2

RDA 1

Fig 6 Ordination diagram resulting from standard redundancy analysis (RDA) on transformed abiotic variables illustrating relationships between woodyspecies and physiochemical attributes The diamonds represent species and the vectors indicate the strength of relationship

ndash29 ndash27

(0ndash5)

(5ndash10)

(5ndash15)

(10ndash15)

Grassland (0ndash10)

Soil δ13 C (permil)

Forests

Grassland

Soi

l dep

th (

cm)

ndash25 ndash23 ndash21 ndash19 ndash17 ndash15

Fig 7 Mean carbon isotopic signature in forest (n= 96) and grassland(n= 8) soils for Cherry Hill Maryland USA Error bars are 1 se Datainterpreted as evidence for expansive C4 or C3 historic grassland Verticallines over 5ndash10- and 5ndash15-cm data indicate group pairs that were notsignificantly different using Tukeyrsquos HSD multiple comparisons

J Australian Journal of Botany J Burgess et al

reconstructing vegetation changes (Deines 1980 Tieszen andArcher 1990 Boutton and Yamasaki 1996 Boutton et al 1998)Spatial patterns of soil d13C were strongly related to type ofvegetation (primarily woody versus grassland cover) Remnantgrasslands are composed of a mixture of C3 grasses and forbsand C4 grasses It is well documented that the d13C signature ofC3 plants (ranging from26permil to28permil) is easily distinguishedfrom that of the C4 plants (ranging from 12permil to 15permil)(DeNiro and Epstein 1977 Cerling et al 1989 Tieszen 1991Ehleringer et al 2000 Fox et al 2012 Throop et al 2013) Otherprocesses besides vegetation change alter isotopic ratios (such aspreferential degradation of some plant materials microbiologicalactivity Suess effect and adsorption processes) and may causeshifts of the order of 1ndash3permil (Friedli et al 1987Becker-HeidmannandScharpenseel 1992Trolier et al 1996Ehleringer et al 2000SantRuCkova et al 2000 Krull et al 2005) d13C enrichmentin temperate forest soils larger than 3ndash4permil is generally attributedto the presence of remaining old C4-derived SOC Following thereasoning of Wynn and Bird (2008) we sampled the upper soils(A horizon) and assumed that the intrinsic metabolic processesoccurring in C3 and C4 vegetation impart an unequivocald13C signature into the SOM and that these signatures persistin decaying organic matter such that increasing depth yields achronologic sequence recording vegetation change At the studyarea d13C values of soils in the grassland matrix averaged ndash16permil(Table 4) to 26permil in forested areas dominated by C3 plantsDepth-integrated SOM d13C significantly reflected the currentvegetation cover (mixed C3ndashC4 and solely C3) and accordinglycould be used as a reference to trace past changes in vegetationrecorded in the SOM at deeper soil layers The d13C values ofSOC consistently increased with depth reflecting a higherpercentage of C4-derived carbon in the older soils that isinterpreted to reflect a site-wide shift in vegetation from C4 toC3 plants

Multiple regression and ordination results supported the roleof soil depth and incident radiation both of which influence soilmoisture content as factors governing the occurrence of xericserpentine savannaQuercus species Abiotic factors in areas withhigher importance of shade-tolerant species were associated withdeeper soils higher bulk density and higher Ca Mg ratios In oursite early successional woodlands were favoured in areas of highincident radiation and broadleaf forests were more developed onslopes containing more clay rich and denser soils Diversity ofwoody specieswas higher in areaswith deeper soilsUnlikemanysystems with heavy metals we found no correlations betweentoxic metals (Cr and Ni) and vegetation dynamics

Historical records show that oak savanna ecosystems wereimportant components of the Mid-Atlantic landscape (Shreveet al 1910) The lack of any oaks before the 1850s suggests thatthey were likely harvested for fuel posts or to support grazingFarming of these serpentinite areas is not documented and isunlikely given the inhospitable nature of the soil Our studyindicated that tree recruitment before the 20th centurywasmostlyoaks perhaps regenerating from root sprouts or propagules in theseedbankXeric oak recruitment decreasedover timeandnoxericoaks younger than 40 years established in the forested areasalthough seedlings and saplings still exist in the remnantgrasslandsavannas This failure to regenerate is likely toreflect the consequences of changing disturbance regimes suchas fire-suppression (Nuzzo 1986 Abrams 1992 2006 PetersonandReich 2001) alongwith competitionwith invasive andmesictree species (Nuzzo 1986) White-tailed deer and squirrelherbivory can reduce recruitment especially in isolatedsystems such as these serpentinite landscapes (Shea andStange 1998 Haas and Heske 2005) A turning point forcommunity development appears to be ~1900 with theencroachment of Pinus virginiana and Juniperus virginianaThese conifer species have a variety of characteristics that can

30

25

Minor species

Mesic species (A rubrum F grandifolia N sylvatica P serotina)

Conifers

Quercus spp (Mixed)

Quercus spp (Xeric)

(n = 163)15

5

20

10

Num

ber

of tr

ees

sam

pled

Decade of origin

01850 1860 1870 1880 1890 1900 1910 1920 1930 1940 1950 1960 1970 1980 1990

Fig 8 Age distribution from trees on the basis of core data (n= 163)Xeric species areQuercusmarilandica andQ stellataMixed oak species includeQ albaQ montana Q rubra and Q velutina

Conifer encroachment on ultramafics Australian Journal of Botany K

potentially alter ecosystem processes for decades as they matureand senesce Both conifers are aggressive invaders because ofrapid growth high seed production high seed-dispersalefficiency tolerance of xeric conditions and opportunistic useof ECM mycorrhizae (Ormsbee et al 1976 Holthuijzen andSharik 1984 1985 Briggs andGibson 1992Axmann andKnapp1993 Thiet and Boerner 2007) Once established these speciescan have a profound effect on soil properties as a result ofaccumulation of organic matter on the forest floor and theeventual decomposition of detritus which leads to higherwater-holding capacity and lowering of soil pH This positivefeedback between soil depth and litter cover exerted by theconifers could promote succession towards non-serpentineconditions

The initial serpentine edaphic signature acts as a filterpreventing species that belong to the regional flora but lack thetraits required to survive in such conditions to successfullycolonise these habitats By the 1970s such filters appear tohave diminished Current site characteristics have shiftedtowards more mesic conditions with changes in soil propertiessuch as bulk density soil depth pH SOM and metalsconcentration Over time the community has changed fromxeric (grassland or oak savanna) to one dominated by moremesic oaks and shade-tolerant Acer Nyssa and Fagus speciesThemajor onset ofAcer establishment began in the 1940sUnlikethePinus and Juniperus establishment pulse which proceeded inan open-canopy situation Acer dominance likely proceededunder closed canopy and canopy gaps because it is shade-tolerant One may speculate that over the past 100ndash200 years

the landscape has been dominated by Quercus with patches ofPinus and Juniperus that historicallywere present at lownumbersbecauseof disturbance (miningfire grazing logging)Hilgartneret al (2009) recorded evidence of the presence of Pinusvirginiana from the early 1800s in a serpentine barren siteonly 50 km distant Over the past 60 years Acer has becomeabundant where it along with other species currently present(N sylvatica Cornus florida and Celtis occidentalis) wouldnormally be considered an aberrant species in a hostile edaphicsetting Age and diameter structure indicated a shift away fromQuercus and towards an A rubrum-dominated landscape Thissame species shift has been widely reported throughout theeastern deciduous forests of North America (Abrams 1998Nowacki and Abrams 2008 McEwan et al 2010)

In the context of serpentinite ecosystems the influxof conifersnearly a century ago has acted as a landscapemodulator (Shachaket al 2008) The resulting build-up of SOM increases the water-holding capacity and other resources such as ground-levelradiation facilitate the influx of new species adapted to thealtered conditions As the difference between the mesicwoodlands and the xeric patches decreases the resourcecontrasts also decrease leading to increasing communitysimilarity Different oak species can grow in a wide variety oflight and moisture conditions However none of the traditionalserpentine oak species is considered shade tolerant to the samedegree as are Acer spp Fagus grandifolia and Nyssa sylvaticaWith progressive development of the soil and a dense-canopyclosure growth and germination of the traditional xeric oakspecies are likely to be suppressed marking a shift from

Colonisation

Fig 9 Conceptual model for serpentine (orangendashred colouring) habitats and species redistribution resulting in similar structure for a typical piedmontgeoecologic system (bluered colouring) Bedrock geology (specifically elemental concentrations and structures) exerts its control on soil fertility with subtlevariations resulting in markedly different above-ground and below-ground communities The change in soils over time (dsdt) equation from Huggett (1995)is a function of the various drivers including b (biology) r (topography) a (atmospheric effects) s (soils) and h (hydrology) Through changing vegetationdynamics and species redistribution the tree functional groups and forest types will see increasing similarity as a result of converging successional trajectoriesRegional species pools stochastic events and many other factors determine ultimate outcomes

L Australian Journal of Botany J Burgess et al

xerothermic to mesic andor nutrient-demanding speciesFigure 9 details a conceptual model for pedologic and ecologicchanges at serpentine sites of theMid-Atlantic with successionaltrajectories converging on a species redistribution and loss oftypical serpentine species such as Cerastium arvense varvillosissimum Andropogon gerardii and the xeric oaks(Q marilandica and Q stellata)

Although the canopy is still oak or pine dominated thetrajectory points to a maplendashbeechndashred oak-dominated forestThe Mid-Atlantic has already lost much of its originalgrasslandndashoak woodlandndashforest mosaic This state transitionmay lead to the loss of a unique ecosystem and thustaxonomic homogenisation with the regional flora Thesechanges result from anthropic influences through changes inland use and fire frequency the loss of Castanea dentata(American chestnut) and changing herbivore populationsNearly a century ago these areas gave way to a mosaic ofsmall woodlands typically characterised by xeric oaksinterspersed with shallow rooting low nutrient-tolerantconifers Released from suppressive disturbance regimes(herbivory and fire for example) Pinus virginiana andJuniperus virginiana quickly became the dominant speciesHowever our data suggest that the landscape is nowundergoing another distinct change as the once-open canopystructure has becomemore closed Edaphic properties such as soildepth chemistry and moisture that once drove communityorganisation across environmental stress gradients are nowbeing modulated by species no longer filtered by theserpentine syndrome This modulation may allow for therecruitment and establishment of drought-sensitive speciesbeyond their typical tolerance limits (Warman and Moles2008) This unique ecosystem which took centuries tomillennia to evolve appears to be fading and may soonevanesce into the past

Sources of funding

Our work was funded by the National Science Foundation(DEB-0423476 and DEB-1027188) and grants from theR Balk and D Elliott Field Fund (Johns Hopkins University)

Conflicts of interest

No conflicts of interest

Acknowledgements

We thank Drs Richard Back Naomi Levin Bruce Marsh and Ben Passey foruse of their elemental and isotopic facilities We also thank Dr David Veblenand two anonymous reviewers for numerous constructive comments on thegeology

References

Abbe C (1898) A general report on the physiography of Maryland JohnsHopkins University (Johns Hopkins University Press Baltimore MD)

Abrams MD (1992) Fire and the development of oak forests Bioscience 42346ndash353 doi1023071311781

Abrams MD (1998) The red maple paradox Bioscience 48 355ndash364doi1023071313374

AbramsMD (2006) Ecological and ecophysiological attributes and responsesto fire in eastern oak forests In lsquoFire in eastern oak forests deliveringscience to landmanagersrsquo (EdMBDickinson) pp 74ndash89 (USDA ForestService General Technical Report NRS-P-1)

Alexander EB (2009) Serpentine geoecology of the eastern and southeasternmargins of North America Northeastern Naturalist 16 223ndash252doi1016560450160518

Alexander EB Ellis CC Burke R (2007) A chronosequence of soils andvegetation on serpentine terraces in the Klamath Mountains USA SoilScience 172 565ndash576 doi101097ss0b013e31804fa22d

Arabas KB (2000) Spatial and temporal relationships among fire frequencyvegetation and soil depth in an easternNorthAmerican serpentine barrenThe Journal of the Torrey Botanical Society 127 51ndash65doi1023073088747

Archer S Schimel DS Holland EA (1995) Mechanisms of shrublandexpansion land use climate or CO2 Climatic Change 29 91ndash99doi101007BF01091640

Axmann BD Knapp AK (1993) Water relations of Juniperus virginiana andAndropogon gerardii in an unburned tallgrass prairie watershed TheSouthwestern Naturalist 38 325ndash330 doi1023073671610

Barton AM Wallenstein MD (1997) Effects of invasion of Pinus virginianaon soil properties in serpentine barrens in southeastern PennsylvaniaThe Journal of the Torrey Botanical Society 124 297ndash305doi1023072997264

Bazzaz FA (1968) Succession on abandoned fields in the Shawnee Hillssouthern Illinois Ecology 49 924 doi1023071936544

Becker-Heidmann P Scharpenseel H-W (1992) Studies of soil organicmatterdynamics using natural carbon isotopes The Science of the TotalEnvironment 117ndash118 305ndash312 doi1010160048-9697(92)90097-C

Belcher JW Keddy PA Twolan-Strutt L (1995) Root and shoot competitionintensity along a soil depth gradient Journal of Ecology 83 673ndash682doi1023072261635

BouttonTWArcher SRMidwoodAJZitzer SFBolR (1998)d13Cvalues ofsoil organic carbon and their use in documenting vegetation change in asubtropical savanna ecosystem Geoderma 82 5ndash41doi101016S0016-7061(97)00095-5

Bouyoucos GJ (1936) Directions for making mechanical analyses of soilsby the hydrometer method Soil Science 42 225ndash230doi10109700010694-193609000-00007

Briggs JMGibsonDJ (1992)Effect offire on tree spatial patterns in a tallgrassprairie landscape Bulletin of the Torrey Botanical Club 119 300ndash307doi1023072996761

BrownML Brown RG (1984) lsquoHerbaceous plants of Marylandrsquo (Universityof Maryland College Park MD)

Burgess JL Lev S Swan CM Szlavecz K (2009) Geologic and edaphiccontrols on a serpentine forest community Northeastern Naturalist 16366ndash384 doi1016560450160527

Butaye J Jacquemyn H Honnay O Hermy M (2002) The species poolconcept applied to forests in a fragmented landscape dispersal limitationversus habitat limitation Journal of Vegetation Science 13 27ndash34doi101111j1654-11032002tb02020x

Callaway RM Davis FW (1993) Vegetation dynamics fire and the physical-environment in coastal central California Rid B-7010-2009 Ecology 741567ndash1578 doi1023071940084

Carroll W D Kapeluck P R Harper R A and Van Lear D H (2002)Background paper historical overview of the southern forest landscapeand associated resources In lsquoSouthern forest resource assessmentrsquo (EdsDNWear JGGreis) pp 583ndash605 (USDepartmentofAgricultureForestService Southern Research Station Asheville NC)

Cerling TE Quade J Wang Y Bowman JR (1989) Carbon isotopes in soilsand palaeosols as ecology and palaeoecology indicators Nature 341138ndash139 doi101038341138a0

Chao A (1984) Nonparametric estimation of the number of classes in apopulation Scandinavian Journal of Statistics 11 265ndash270

ChaoA (1987) Estimating the population size for capturendashrecapture data withunequal catchability Biometrics 43 783ndash791 doi1023072531532

Chazdon RL Colwell RK Denslow JS Guariguata MR (1998) Statisticalmethods for estimating species richness of woody regeneration in primaryand secondary rain forests of Northeastern Costa Rica In lsquoForest

Conifer encroachment on ultramafics Australian Journal of Botany M

Counting statistics errors are 01 for major elements andaverage detection limits are 50ndash100mg kgndash1 for major oxidesand 5ndash10 ppm for chromium (Cr) nickel (Ni) titanium (Ti)iron (Fe) manganese (Mn) Ca potassium (K) and phosphorus(P) All XRF samples were quantified using a matrix matchedset of USGS- and NIST-certified soil standards At least oneduplicate one replicate and one certified reference sample(SRM 2709 San Joachin Soil) was run with every 10 samplesanalysed to monitor for external and internal reproducibilityLoss on ignition (LOI) for samples is measured by igniting rockpowders in alumina crucibles to 1200C for 10min in a mufflefurnace

Soil characteristics evaluated for each plot or quadrat includedbulk density soil depth pH soil organic matter content majorand trace element analysis carbon (C) nitrogen (N) and isotopicand textural analysis Surface soil epipedons were evaluated forbulk density (g cmndash3) At each location we composited five soilcores (corners plus centre) that were taken with a 5-cm-diametersoil corer Soil coreswere sectioned into 0ndash5 5ndash10 10ndash15 (and atsome sites 5ndash15) cm increments dried at 40C to a constantweight and then calculated as dryweight of sample per unit of soilvolumewith roots removed Soil parameter valueswere averagedto provide a representative value for each 150-m2 plot

Forest soil depth (depth to obstruction) was measured with apush soil probe at five random locations in each plot For soilmoisture or soil depth comparisons from savanna into forestedareas four transects (35ndash68m long) were established in April2011 Our rationale was that distance from bedrock wouldrepresent a proxy for time and points along these transectswould generally represent soil chronosequence (Jenny 19411980) Because of the inherent low moisture values ofserpentine soils measurements were made during the weekafter a 5-mm or greater rainfall event Soil moisture values at0ndash6-cm depth were measured with a ThetaProbereg soil moisturesensor (Dynamax IncHoustonTXUSA) and soil depthswith apush probe (internal diameter = 47 cm) every 1m

Soil samples for chemical analysis were air-dried and passedthrough a USA Standard No 10 sieve Bedrock thin-sectionsoccasionally contain carbonate (calcite andmagnesite) thereforeinitial soils were evaluated for carbonate by an acid (1Mhydrochloric) test The HCl test showed that the soil samplesdid not contain carbonates therefore total C in the soil representsorganic C only

Moist soil colour was determined in the field using Munsellcolour charts (Munsell Color Grand RapidsMI USA) Sampleswere combined mixed and split to give two samples of200ndash400 g that were lightly crushed and sieved to 2mm andsampled for textural analysis (sieve followed hydrometermethod) (Bouyoucos 1936) Soil pH was measured with a 20 gin 20mL soilndashwater slurry by using a glass electrode Theremainder was oven-dried at 30C and powdered using anagate centrifugal ball mill A powdered subset wassubsequently used for XRF and C N analyses Total N and Cwere measured with an automated Perkin-Elmer 2400 Series IICHN analyser (Perkin-Elmer Cornell Labs Ithaca NY USA)Soil organic matter content was estimated as loss on ignition(LOI) at 550C for 3 h Total C N and isotopic analysis wereconducted at the stable isotope facilities at either theUniversity ofCalifornia at Davis CA or Cornell Laboratories Ithaca NY

using an elemental analyser interfaced with continuous-flowisotope-ratio mass spectrometer Values reported are typicallymeans with the associated standard error All the values werereported as per mil (permil) relative to the international standardVienna Pee Dee Belemnite (V-PDB) and were calculated as

d13 CethpermilTHORN frac14 Rsample Rstandard

Rstandard

1000

where d13C is the difference between 13C 12C ratio (R) of thesample and that of the standard Increase in the d value reflectsincreases in the relative amount of the heavy isotope componentor a reciprocal decrease in the light isotope component

Biotic characteristicsTo monitor the serpentine ecosystem-scale vegetation-boundarymovement we examined a time series of high-resolution imagesfrom Google Earth (1994ndash2013) along with aerial photographyfrom 1937 1952 1971 and 1988 using the grid-square methodThe more recent data serve to complement the work by Tyndall(1992) on the expansion of conifers into barrens areas All woodyplant species were recorded for each sampling plot To identifyplants we followed the key by Brown and Brown (1984) withnomenclature clarification via the integrated taxonomicinformation system Woody plants with diameter-at-breast-height (DBH) of gt10 cm are referred to as trees those withDBH of lt10 cm and gt2 cm or 2m height as saplings and anywoody specieslt2-cmDBHare grouped as seedlings or vines Forspecies where DBH is impractical (shrubs) stemsweremeasuredwith a caliper near the root collar Absolute and relative stemdensity and basal area were derived from these data

Canopy cover was measured at each plot centre with aspherical densiometer Plots were classified as savannawoodland or forest on the basis of canopy cover as followssavanna (0ndash25 canopy cover) woodland (26ndash60 canopycover) and forest (gt60 canopy cover) Ground-layerpercentage cover for soil leaf litter and rock was estimatedvisually

To assess vegetation change over time historical recordswerecombined with tree-core data (n= 163) One increment core wasextracted for each tree species per plot when possible Cores weretaken at breast height mounted and sanded following Stokes(1996) and Stokes and Smiley (1968) Because increment coreswere collected at 137mabove thegroundweconsidered the agesobtained as the minimum age of trees at breast height Repetitivecoring was carried out to facilitate the interception of the pithMissing rings in incomplete cores were estimated using diameterage-regression curves (Lorimer 1980) Ring counts were madeusing a stereomicroscope at 10ndash40 magnification Each coresample was skeleton plotted to identify narrow rings (Speer2010) The resulting age distribution was divided into datethresholds to record pulses or absence of regeneration as cohorts

Data analysisTo quantify forest structure the quadrats were characterised bybasal area (m2 handash1) and density (stems handash1) of live specimens ofthe principal tree species Tree species were grouped on the basisof habitat affinity (xericmesic) using the USDA silvics manual(Walters and Yawney 2004) Diversity indices were calculated

D Australian Journal of Botany J Burgess et al

for each plot and on pooled plot data using the ShannonndashWeaverindex (Shannon and Weaver 1949) and the Simpson index(Simpson 1949) as measures of a diversity (a) along withestimators of total diversity using rarefaction and non-parametric approaches of Chao 1 Chao 2 and ACE (Chao1984 1987 Chazdon et al 1998) The importance value (IV)is a measure of relative abundance and was calculated as theaverage of relative density and dominance

The KolmogorovndashSmirnov test statistic was used to assessnormality of the soil variables Those that failed the normality testwere transformed to normal or near-normal distributionPercentage measurements (soil particle analysis abundance CN cover XRF whole-rock oxide weight) were arcsinetransformed soil elemental data were loge transformed fromppm or ppb concentrations slope and soil depth were natural-log transformed and bulk density was square-root transformed

Site microclimate variability was compared using geographiccoordinates (latitude and longitude) inclination and aspect foreach plot to calculate the heat load and the potential annual directincident radiation by applying the third equation from McCuneand Keon (2002) The effects of topographic (slope inclinationannual direct incident radiation and heat load) and soil variables(elemental analysis in addition to pH soil organic C total NC N depth texture and d13C values) on species were examinedusing principal components analysis (PCA) to explore the mainfactors and generate hypotheses as to the reciprocal effects ofthese properties on vegetation Redundancy analysis (RDA)ordination as constrained by the environmental variables wasperformed to identify the main floristic gradients (basal areadensity percentage xeric species) and the edaphic andtopographic factors explaining those trends Two-wayANOVA was used to test for the differences in soil d13C C Nand C N with respect to vegetation (grassland forest) and soildepth ANOVAs were followed by Tukeyrsquos HSD (chosenbecause of uneven number of replicates) multiple pairwisecomparison to relate differences in elemental content and depth

Differences in site woody-species composition wereinvestigated by calculating the pairwise ecological distance forthe entire species matrix The Mantel test was used to determinewhether site similarity in the tree community was related to siteenvironmental variables Multiple linear regression analysis wasused to identify site factors that had a significant effect on speciesabundance The eight highest-ranked species on the basis of IVwere evaluated in termsof their abundances as responsevariablesand soil and topographic data were used as explanatory variablesOrdination using direct (RDA) and indirect (PCA) gradient

analysis was employed to extract variation in the underlyingstructure of the vegetation data as it relates to environmentalvariables (Ludwig and Reynolds 1988) PCA was performed tofind the main factors determining the reciprocal effects of soiland vegetation

The BrayndashCurtis distance with raw abundances was usedto determine the degree of vegetation dissimilarity an aspectof b (b) diversity between sites and vegetation types (woodlandand forest) based on presenceabsence We used a Mantel test(McCune and Grace 2002) to compare the matrix of geographicdistances between plots with the matrix of ecological distancePotential relationships between abiotic variables and diversityvariables (floristic similarity evenness and Simpsonrsquos diversity)were explored Tree-species compositions of early successionalwoodland and more mature forest designations based onpercentage canopy were compared using non-parametricanalysis of similarities (ANOSIM) based on the hypotheses ofno difference between stages

Gamma (g) or serpentine landscape-area diversity wasestimated using all 32 plots to identify species richness at thelandscape scale by constructing a rarefaction curve for all samplesand all species in the dataset

For all statistical tests significance was determined at thea= 005 level other P-values were reported if lt01 The methodfor evaluating the Mantel-test statistic was the t-distribution(Mantel 1967) and the standardised Mantel statistic (r) as ameasure of effect size (McCune and Grace 2002)

All data were analysed using the statistical computingenvironment R (v 311) and the BiodiversityR package(Kindt and Coe 2005) ANOSIM Mantel and other tests wereperformed with n= 1000 permutations Reported errors arestandard errors (se) unless indicated otherwise

Results

Abiotic substrate and soil composition

Mean whole-rock major-element values were typical of alteredultramafic rocks (Table 1) Calcium and aluminium (Al) contentswere low reflecting disappearance of the calcic pyroxene byserpentinisation Only samples where clinopyroxene wasdetected showed an increase in Ca Samples with elevated Alare generally accompanied by the presence of chlorite in therocks As is typical of ultramafics the immobile refractoryelements Cr and Ni were elevated Ternary diagrams depictingthe measured whole-rock geochemistry of the samples illustratedrelative compositional variations in the systemMgOndashSiO2ndashH2O

Table 1 Summary of the whole-rock geochemistry parameters for serpentinite samples obtained from X-ray fluorescence (XRF) analysisSerpentinite water content based on loss on ignition (LOI) ~124 03 weight Al aluminium Ca calcium Cr chromium Fe iron K potassium Mg

magnesium Na sodium Ni nickel Si silicon Ti titanium V vanadium Zn zinc

Lithology Major element geochemistry (weight ) Trace elements (mg kgndash1)Serpentinite n= 12 Si Al Fe Mg Ca Na K Ti Cr Ni V Zn

Mean 4129 031 954 3669 042 000 007 002 283799 231968 1189 4385se 088 015 078 070 022 000 002 000 28651 15716 334 594Soils (0ndash5 cm) n= 32 3513 091 657 625 024 015 070 048 252651 115744 9678 10705se 132 019 040 060 006 003 005 003 27641 13040 642 463Soils (5ndash15 cm) n= 48 3799 105 744 762 015 015 069 046 234790 124434 9725 8426se 106 020 036 060 004 002 005 003 20726 11024 554 311

Conifer encroachment on ultramafics Australian Journal of Botany E

(Fig 3) Concentrations of soilmacronutrientsCN P andKwerenot related to ultramafic parent material

Soil depths varied from105 cm to52 cmwith variable texture(mean of 134 gravel) across the sites Soil colours ranged fromdark brown (75YR 32) to grey (75YR 61) in the upper 1ndash5 cmof the soil profile to yellowndashbrown (10YR 56) and redndashbrown(25YR 44) with depth Talcose serpentinites yielded moreyellow (10YR Hues) varieties The mean and standard error ofsoils collected from the 32 plots recorded a narrow range of

edaphic characteristics including soil depth 289 20 (cm) pH54 01 bulk density 088 007 (gm3) soil organic matter(SOM) 1251 26 () sand 5103 18 () clay 267 11() and C N 1518 09

Soil moisture and soil depth increased with distance from thebarrens areas Strong positive correlationwith distancewas foundfor both soil moisture and depth (r= 0570 P lt 0001 andr= 0848 Plt 0001) respectively (Fig 4) Other soil physio-chemical parameters such as texture and trace element contentdid not show correlations with distance (P gt 005 Spearmanrsquosrank correlation)

Magnesium concentrations were high and correspondingCa Mg ratios low (lt05) across all sites The Ca Mg ratio isa measure of serpentine cation imbalance Surface soils hadhigher Ca Mg (02 to 05) ratios than did subsurface soils(005 to 02) Chromium and Ni concentrations were elevatedwith a maximum of over 6000 ppm and 2700 ppm respectively

The pH of the soils was typically acidic (surface soils average~54) but ranged from (~4ndash7) Soil pHwas negatively correlatedwith Ca Mg (r= ndash049 P lt 005) and C N ratios (r= ndash062P lt 005) SOM content (based on LOI) had a mean of 166in the upper 5 cm declining to below 7with depth (10ndash15 cm)N content was generally low and C N ratios varied between 5and 24with the higher values on themore developed soils Depthprofiles of SOM displayed increasing d13C isotopic signaturesfrom 2means of ndash2623permil (0ndash5 cm) to ndash2095permil (10ndash15 cm)(Table 2) whereas remnant grassland areas recorded meand13C isotopic signatures of ndash1623permil (0ndash5 cm)

Woody-species characteristics

Using high-resolution Google Earth imagery the current (2013)percentage of the site that contains grassland or savanna is lessthan 2 The forests include xeric woodlands with stands ofP virginiana and J virginiana intermingled with Quercusmarilandica Meunch (blackjack oak) and Quercus stellata

Talc

Enstatite

Serpentine

Olivine

Ca Mg

Si

Diopside

Typical Ultramafic

Fig 3 Ternary diagram detailing the chemography of Cherry Hillserpentinites (open circles) projected onto the plane CaOndashMgOndashSiO2 alsoshowing typical ultramafic rock (filled square) and mineral compositions(solid circles)

60

Soil depth (cm)

Percent moisture

50

40

30

20

Soi

l dep

th (

cm)

and

moi

stur

e (

by

volu

me)

Distance along transect (m)

10

00 10 20 30 40 50 60

Fig 4 Monotonic change of variables along transects The relationship between soil depth percentage moisture (volume) and distance from barrens bedrockwere assessed using correlational analysis (Plt 00001) Data are compiled averages of four transects from rocky barrens to woodlandndashforest ecotone Thecorrelation of the two variables soil depth and moisture is significant (r = 067 Plt 0001)

F Australian Journal of Botany J Burgess et al

Wang (post oak) as well as more mesophytic species Largeopen-crown trees are rare more typical are contorted down-sweeping branches of Q marilandica surrounded by stand-grown trees Aerial photographs from 1937 show conifersadvancing from the dendritic drainage areas Mesophyticforests are composed of these oaks and others such as Quercusrubra L (red oak) Q alba L (white oak) Q montana Willd(chestnut oak) and Q velutina Lam (black oak) along withnumerous other species most notably Prunus spp (blackcherry wild cherry) Pinus spp (eg white pine shortleafpine) Fagus grandifolia Ehrh (American beech) Acer rubrumL (red maple) Sassafras albidum Nutt (sassafras) Populusgrandidentata Michx (bigtooth aspen) Nyssa sylvaticaMarshall (blackgum) Betula lenta L (sSweet birch) andRobinia pseudoacacia L (black locust) On exposed rocksurfaces the serpentine endemic (as reviewed by Harris andRajakaruna 2009) Cerastium velutinum var velutinum Raf(serpentine mouse eared chickweed) and other serpentinophilessuch as Symphyotrichum depauperatum Fern (serpentine aster)and saplings of Q marilandica are common

Biodiversity richness estimators Chao 1 Chao 2 and ACEestimated maximum woody-species richness between 263and 308 species The observed richness was 24 species(Table 3) Overall diversity and mean diversity for the 32 siteswere calculated as 274 (X = 167 s= 029) (Shannon index) and

092 (X = 078 s= 008) (Simpson index) On the basis of relativeIV the most important species were P virginiana andJ virginiana (Table 3) These two species had similar relativeIVs and stood out from all other species A second group ofspecies included A rubrum and Q marilandica At the genuslevel Quercus was the most important contributor to the treecommunity with 39 relative IV On the basis of basal area themost dominant species were Q rubra P virginiana andQ marilandica Constancy values (Table 3) showedA rubrum (078) as a clear break from the next grouping ofspecies (P virginiana and J virginiana) whose constancy valueshovered near 05 followed by Q marilandica F grandifoliaN sylvatica and Q stellate with constancy values near 04

The Mantel test comparing floristic similarity andenvironmental distance showed significant (P lt 005)correlations between environmental distance and ecologicaldistance for the following parameters heat load bulk densitybasal area diversity (H) percentage sand Ni and Cr(Table 4) The ANOSIM statistic was calculated only for thecanopy category (woodlandforest) with a resulting r= 014 andP-value of 005 Both the Mantel and ANOSIM tests yieldedsignificant (Plt 005) but weak correlations

Relationships between vegetation and abiotic properties

Results of multiple regression analysis were evaluated tounderstand the response of species distributions to soilproperties Table 5 shows the regression results for the eightspecies the importance of which met or exceeded 5 withtopographic and edaphic parameters as the predictor variablesPredictor variables that indicated a meaningful contribution toexplaining species abundancewere incident radiation and soilMgcontent for Q marilandica incident radiation for J virginianaand Q stellata soil organic carbon (SOC) for A rubrum soildepth for F grandifolia and bulk density for the conifers

PCA was performed for 16 abiotic parameters across the 32sampled quadrats so as to identify covariation between the solarradiation and soil variables (Fig 5) Themain trend of variation in

Table 2 Descriptive statistics for soil d13C (permil) and soil organic matterin grassland and woodland elements

Means followed by the same letter are not significantly different (ANOVApost hoc Tukeyrsquos HSD P = 005)

Statistic Grassland Forest(0ndash5 cm)

Forest(5ndash10 cm)

Forest(5ndash15 cm)

Forest(10ndash15 cm)

Mean ndash1623a ndash2623b ndash2359bcd ndash2298de ndash2095eMinimum ndash1791 ndash2923 ndash2725 ndash2649 ndash2514Maximum ndash1423 ndash1928 ndash1811 ndash1972 ndash1654se 068 044 067 065 060

Table 3 Constancy importance (average value of a species when present) density dominance andimportance value (average of relative density and relative dominance) measures for woody stems in

serpentinite woodland at Cherry Hill Maryland USA

Species Constancy Importance Dominance(m2 handash1)

Density(stems handash1)

Importancevalue (IV) ()

Acer rubrum L 078 109 5799 7778 988Ailanthus altissima Mill 003 003 075 222 027Betula lenta L 006 006 346 444 060Celtis occidentalis L 006 006 059 444 041Cercis canadensis L 003 003 078 222 022Cornus florida L 006 006 146 444 038Diospyros virginiana L 003 003 185 222 024Fagus grandifolia Ehrh 043 065 5713 4667 619Juniperus virginiana L 050 106 8657 7556 1214Kalmia latifolia L 012 015 163 1111 106Liriodendron tulipifera L 009 009 2372 667 108Nyssa sylvatica Marsh 043 075 5769 5333 720Pinus strobus L 003 003 692 222 090Pinus virginiana Mill 056 078 16498 5556 1343Populus grandidentata Michx 003 006 640 444 058Prunus serotina Ehrh 028 028 2046 2000 271

Conifer encroachment on ultramafics Australian Journal of Botany G

the soil chemical properties and topographic variables asreflected by the PCA Axis 1 (accounting for 30 of variance)was one dominated by significant correlations of soil texture(sand and bulk density) soil depth Mg Cr and Niconcentrations incident radiation and heat load The secondtrend of soil variation as reflected by the PCA Axis 2 (17 ofvariation) can be interpreted as a combined gradient of essentialnutrients (C andN) and SOC The PCA (axes accounting for 44of the variation) revealed strong positive associations betweenboth soil depth and bulk density which were negativelycorrelated with heavy metal concentrations and pH SoilCa Mg ratios and Al concentrations were negativelycorrelated with macronutrients (C N P) and SOC Fig 5bshows the PCA factor map for the abiotic variables and thebiotic covariates for the sampling plots When thecharacteristics of woody-species community are included inthe PCA both basal area and stem density are positivelycorrelated with soil Ca Mg ratios and Al content Diversity(Shannonrsquos index) was strongly correlated with soil depth andbulkdensity andnegatively related toheavymetal concentrations

and heat load As individual-species basal area stem densityand diversity typically respond to environmental gradientsconstrained ordination utilising RDA was used to inferrelationships between the species matrix and theenvironmental matrix using Euclidean distance between sitesArrows of increasing species abundance represent RDA scoresThe sum of the RDA eigenvalues for the first two axes using allabiotic explanatory variables accounted for 368 of the woody-species composition After reduced variable selection based onANOVA tests of the ordinationmodel seven termswere selectedand these collectively explained 271 of the variance (Fig 6)Interestingly those species whose importance was gt4 arepartitioned into the four quadrants of the ordination From theRDAAxes1and2biplot it canbe seen that traditional serpentine-savanna (shade-intolerant)woody species (S albidumQ stellataand Q marilandica) are positively associated with increasingincident radiation and negatively associated with increasingsoil depth Al and Mg soil concentrations Post-disturbanceinitial colonisers (J virginiana and P virginiana) are stronglynegatively correlated with bulk densityQ alba andQ montanawhich are widely distributed across a variety of conditionsfrom wet mesic to subxeric are associated with increasingbulk density Shade-tolerant species such as N sylvaticaA rubrum and F grandifolia are correlated with increasingsoil depth and soils with more Ca and Al

Soils at the site differed in the total amount verticaldistribution and isotopic composition of SOM (Tables 1 2)ANOVA results indicated significant differences in theisotopic means according to their depth (F487 = 276P 00001) Tukeyrsquos HSD test showed that the mean d13C ofgrassland soils (ndash162 07permil) was significantly higher thanthat of shallow forested soils (ndash262 04permil) (Table 5)Shallow forested soils were significantly different from deepersoils (ndash2095 06permil) Total soil C and N and soil C N declinedwith soil depth C N ratios were significantly different for alldepths ranging from 1518 088 on the surface soils to1162 1 in soils 10ndash15 cm below surface (F266 = 59

Table 4 Effects of environmental distance on similarity between plotsMantel statistic calculated by comparing matrix of environmental distancewith BrayndashCurtis distance The standardised Mantel statistic (r) indicatesthe strength of the relationship between sample-point position and forest-floor thickness and the associated P-value indicates the significance of

that relationship

Environmental indicator Correlation coefficient (r) P-value

Heat load 0126 0009Bulk density 0236 0003Basal area 0141 0015Diversity (H) 0109 0034Percentage sand 0203 0002Nickel 0111 0034Chromium 0159 0009Canopy 0150 0009

Table 5 Results of multiple regression with list of factors significantly contributing to species abundanceSOC soil organic carbon Plt 0001 P lt 001 Plt 005 lsquorsquo Plt 01

Pvirginiana J virginiana A rubrum Q marilandica Q rubra N sylvatica Q alba Q stellata F grandifolia Q montana

IV () 1343 1214 988 940 810 720 703 626 619 440Incident radiation 09247 00249 01791 00008 04332 02136 09123 00450 03030 03225Heat load 02516 04869 03661 08494 09552 09203 09825 02522 01128 03502Soil depth 02599 03110 08937 00463 01881 03041 03381 01689 00242 09920Bulk density 00023 00037 08260 08860 05216 02157 00415 02686 09894 01171Texture ( sand) 00976 02800 02942 06247 08124 09313 03207 03451 01454 06054Calcium magnesium 09840 02962 07824 07031 09971 04587 07945 05441 06414 03639Magnesium 06146 05570 05725 00019 01500 06177 00914 03238 07491 03623Aluminium 07988 07323 01491 09176 00673 00583 08508 04152 05380 08685Chromium 05879 02397 02066 01012 06214 09977 06126 03991 06870 00434Nickel 08699 02237 07478 03211 08340 06190 09519 05875 06307 01074Phosphorus 05192 05066 07822 09800 07523 06641 04927 09619 09430 02268SOC 06435 01185 00298 00053 09782 02790 05185 02333 08701 07036Carbon nitrogen 05806 02325 06793 08013 05674 06443 01727 01105 08015 06972Carbon 07737 07963 02099 01231 04380 08148 04280 04885 07995 05145Nitrogen 07506 09537 01179 03581 06363 05329 03124 07375 00153 05744pH 07946 01863 05144 01547 00176 04256 02597 06743 01299 09047Multiple R2 059 068 057 079 057 046 051 055 062 056

H Australian Journal of Botany J Burgess et al

P = 0004) There were significant vegetation depth andvegetation depth effects on d13C of SOM (Table 5 Fig 7)

Age structure

In total 163 core samples were evaluated with results displayedin the stand-age structure plot (Fig 8) Skeleton plot chronologiesvaried among species Older minimum age at breast height wasgenerally seen inQ stellata (114 15 years) andQ marilandica(110 7 years) and the younger chronologies developed inspecies such as A rubrum (53 4 years) F grandifolia(56 8 years) and L tulipifera (43 8 years) Only xeric oakswere present in the oldest age classes The period before 1900 iscomposed of 54 xeric oaks and when all oaks are grouped theycomprise 79 of trees from that era although Q montana isabsent from this group (oldest minimum age is 88 years) Theoldest non-oak was N sylvatica (144 years) The period between

1900 and 1940 records a marked increase in conifers andconcomitant decrease of xeric oaks comprising 39 and 19of all species respectively Between 1950 and 1990 thepercentage of xeric oaks decreased to 3 and conifers to 13whereas Acer spp increased to 19 The number of shade-tolerant species rose from 125 in the decades from 1850 to1900 to 59 during the interval from 1950 to 1990

The age structure pooled from all the plots (Fig 8) showeddiscontinuous patterns with prominent peaks of regenerationpulses and periods of either low or absent recruitment Coniferestablishment and recruitment clearly increased during1900ndash1960 and then receded through the rest of the samplingperiod There was a conspicuous absence of younger xeric oaks(post-1970s) and only two xeric oaks from1950 to the present Atthe same time the percentage of non-oak deciduous treesincreased to 68 post-1950

10

05

00

Dim

2 (

171

7)

Dim

2 (

229

1)

Dim 1 (3038)

Dim 1 (402)

ndash10

ndash05

4

2

0

ndash2

ndash4

ndash6

ndash15

ndash15 ndash10 ndash5 0 5 10

ndash10 ndash05 00 05 10 15

(a)

(b)

Fig 5 (a) Unconstrained principal component analysis (PCA) correlation circle of pedological topological and community variables (b) Forest communityclustering along an edaphic gradient

Conifer encroachment on ultramafics Australian Journal of Botany I

Discussion

The present study examined a moderately sized area in the Mid-Atlantic but the changes ndash encroachment and mesophication ndash

represent potential widespread conservation problems Theserpentinite landscapes were once common components of theancient Piedmont (Carroll et al 2002 Owen 2002 Noss 2013)These unique plant communities have existed as persistentlsquoislandsrsquo embedded in a deciduous forest region for centuriesor longer We documented a state transition from grasslandand xeric oak savanna to conifer-dominated communities

that have given way to eastern broadleaf forests Comparisonsamong geologic pedologic and vegetation properties reflect thischange and indicate that the conversion to forest has been rapidand characterised by large changes in soil properties (pH water-holding capacity SOMcomposition) aswell as plant-communitystructure (physiognomyand stemdensity) and composition (xericto more mesic communities)

Thin-section and petrologic data confirmed that the geologicsubstrate is a serpentinised peridotitewith characteristics typical ofserpentinite terranes in the Appalachians (Misra and Keller 1978Mittwede and Stoddard 1989 Gates 1992 Alexander 2009)Although Ca is depleted in bulk rock samples because of thebreakdownof augite and interstitial plagioclase soil Ca Mg ratiosare higher in surface soils The concentrations of SOC C N andmacronutrients in woodland and forested areas are linked to theturnover of organic matter and not the parent material

Compared with the rock soils are enriched in most tracemetals although the Cr contents were similar This was likelyto be due to resistance of Cr-oxides to weathering and suggeststhat these phases were not likely sources for Cr in soil solutionsand plants of serpentine soils (Oze et al 2004) Bedrock to foresttransects demonstrated a strong correlation between soilmoistureand soil depth Soil nutrient content and soil moisture should beproportional to the cube of soil depth (Belcher et al 1995) thuseach of these parameters figured significantly in the PCA andRDA

Afforestation has been relatively recent in this region (Tyndall1992)Concomitantwith the increase in treedensity andsize is thecontraction of the grassland and savanna component study areafrom 18 in 1937 to 15 in 2013 which is consistent withprevious work documenting the invasion of serpentine savannasby conifers (Tyndall 1992 Barton andWallenstein 1997 Arabas2000)

In landscapeswhereC3 andC4plants coexist stableC isotopiccomposition of SOM represents a powerful method for

20

15

1

0

10

ndash10

ndash2 ndash1 0 1 2 3

05

ndash05

00

RD

A 2

RDA 1

Fig 6 Ordination diagram resulting from standard redundancy analysis (RDA) on transformed abiotic variables illustrating relationships between woodyspecies and physiochemical attributes The diamonds represent species and the vectors indicate the strength of relationship

ndash29 ndash27

(0ndash5)

(5ndash10)

(5ndash15)

(10ndash15)

Grassland (0ndash10)

Soil δ13 C (permil)

Forests

Grassland

Soi

l dep

th (

cm)

ndash25 ndash23 ndash21 ndash19 ndash17 ndash15

Fig 7 Mean carbon isotopic signature in forest (n= 96) and grassland(n= 8) soils for Cherry Hill Maryland USA Error bars are 1 se Datainterpreted as evidence for expansive C4 or C3 historic grassland Verticallines over 5ndash10- and 5ndash15-cm data indicate group pairs that were notsignificantly different using Tukeyrsquos HSD multiple comparisons

J Australian Journal of Botany J Burgess et al

reconstructing vegetation changes (Deines 1980 Tieszen andArcher 1990 Boutton and Yamasaki 1996 Boutton et al 1998)Spatial patterns of soil d13C were strongly related to type ofvegetation (primarily woody versus grassland cover) Remnantgrasslands are composed of a mixture of C3 grasses and forbsand C4 grasses It is well documented that the d13C signature ofC3 plants (ranging from26permil to28permil) is easily distinguishedfrom that of the C4 plants (ranging from 12permil to 15permil)(DeNiro and Epstein 1977 Cerling et al 1989 Tieszen 1991Ehleringer et al 2000 Fox et al 2012 Throop et al 2013) Otherprocesses besides vegetation change alter isotopic ratios (such aspreferential degradation of some plant materials microbiologicalactivity Suess effect and adsorption processes) and may causeshifts of the order of 1ndash3permil (Friedli et al 1987Becker-HeidmannandScharpenseel 1992Trolier et al 1996Ehleringer et al 2000SantRuCkova et al 2000 Krull et al 2005) d13C enrichmentin temperate forest soils larger than 3ndash4permil is generally attributedto the presence of remaining old C4-derived SOC Following thereasoning of Wynn and Bird (2008) we sampled the upper soils(A horizon) and assumed that the intrinsic metabolic processesoccurring in C3 and C4 vegetation impart an unequivocald13C signature into the SOM and that these signatures persistin decaying organic matter such that increasing depth yields achronologic sequence recording vegetation change At the studyarea d13C values of soils in the grassland matrix averaged ndash16permil(Table 4) to 26permil in forested areas dominated by C3 plantsDepth-integrated SOM d13C significantly reflected the currentvegetation cover (mixed C3ndashC4 and solely C3) and accordinglycould be used as a reference to trace past changes in vegetationrecorded in the SOM at deeper soil layers The d13C values ofSOC consistently increased with depth reflecting a higherpercentage of C4-derived carbon in the older soils that isinterpreted to reflect a site-wide shift in vegetation from C4 toC3 plants

Multiple regression and ordination results supported the roleof soil depth and incident radiation both of which influence soilmoisture content as factors governing the occurrence of xericserpentine savannaQuercus species Abiotic factors in areas withhigher importance of shade-tolerant species were associated withdeeper soils higher bulk density and higher Ca Mg ratios In oursite early successional woodlands were favoured in areas of highincident radiation and broadleaf forests were more developed onslopes containing more clay rich and denser soils Diversity ofwoody specieswas higher in areaswith deeper soilsUnlikemanysystems with heavy metals we found no correlations betweentoxic metals (Cr and Ni) and vegetation dynamics

Historical records show that oak savanna ecosystems wereimportant components of the Mid-Atlantic landscape (Shreveet al 1910) The lack of any oaks before the 1850s suggests thatthey were likely harvested for fuel posts or to support grazingFarming of these serpentinite areas is not documented and isunlikely given the inhospitable nature of the soil Our studyindicated that tree recruitment before the 20th centurywasmostlyoaks perhaps regenerating from root sprouts or propagules in theseedbankXeric oak recruitment decreasedover timeandnoxericoaks younger than 40 years established in the forested areasalthough seedlings and saplings still exist in the remnantgrasslandsavannas This failure to regenerate is likely toreflect the consequences of changing disturbance regimes suchas fire-suppression (Nuzzo 1986 Abrams 1992 2006 PetersonandReich 2001) alongwith competitionwith invasive andmesictree species (Nuzzo 1986) White-tailed deer and squirrelherbivory can reduce recruitment especially in isolatedsystems such as these serpentinite landscapes (Shea andStange 1998 Haas and Heske 2005) A turning point forcommunity development appears to be ~1900 with theencroachment of Pinus virginiana and Juniperus virginianaThese conifer species have a variety of characteristics that can

30

25

Minor species

Mesic species (A rubrum F grandifolia N sylvatica P serotina)

Conifers

Quercus spp (Mixed)

Quercus spp (Xeric)

(n = 163)15

5

20

10

Num

ber

of tr

ees

sam

pled

Decade of origin

01850 1860 1870 1880 1890 1900 1910 1920 1930 1940 1950 1960 1970 1980 1990

Fig 8 Age distribution from trees on the basis of core data (n= 163)Xeric species areQuercusmarilandica andQ stellataMixed oak species includeQ albaQ montana Q rubra and Q velutina

Conifer encroachment on ultramafics Australian Journal of Botany K

potentially alter ecosystem processes for decades as they matureand senesce Both conifers are aggressive invaders because ofrapid growth high seed production high seed-dispersalefficiency tolerance of xeric conditions and opportunistic useof ECM mycorrhizae (Ormsbee et al 1976 Holthuijzen andSharik 1984 1985 Briggs andGibson 1992Axmann andKnapp1993 Thiet and Boerner 2007) Once established these speciescan have a profound effect on soil properties as a result ofaccumulation of organic matter on the forest floor and theeventual decomposition of detritus which leads to higherwater-holding capacity and lowering of soil pH This positivefeedback between soil depth and litter cover exerted by theconifers could promote succession towards non-serpentineconditions

The initial serpentine edaphic signature acts as a filterpreventing species that belong to the regional flora but lack thetraits required to survive in such conditions to successfullycolonise these habitats By the 1970s such filters appear tohave diminished Current site characteristics have shiftedtowards more mesic conditions with changes in soil propertiessuch as bulk density soil depth pH SOM and metalsconcentration Over time the community has changed fromxeric (grassland or oak savanna) to one dominated by moremesic oaks and shade-tolerant Acer Nyssa and Fagus speciesThemajor onset ofAcer establishment began in the 1940sUnlikethePinus and Juniperus establishment pulse which proceeded inan open-canopy situation Acer dominance likely proceededunder closed canopy and canopy gaps because it is shade-tolerant One may speculate that over the past 100ndash200 years

the landscape has been dominated by Quercus with patches ofPinus and Juniperus that historicallywere present at lownumbersbecauseof disturbance (miningfire grazing logging)Hilgartneret al (2009) recorded evidence of the presence of Pinusvirginiana from the early 1800s in a serpentine barren siteonly 50 km distant Over the past 60 years Acer has becomeabundant where it along with other species currently present(N sylvatica Cornus florida and Celtis occidentalis) wouldnormally be considered an aberrant species in a hostile edaphicsetting Age and diameter structure indicated a shift away fromQuercus and towards an A rubrum-dominated landscape Thissame species shift has been widely reported throughout theeastern deciduous forests of North America (Abrams 1998Nowacki and Abrams 2008 McEwan et al 2010)

In the context of serpentinite ecosystems the influxof conifersnearly a century ago has acted as a landscapemodulator (Shachaket al 2008) The resulting build-up of SOM increases the water-holding capacity and other resources such as ground-levelradiation facilitate the influx of new species adapted to thealtered conditions As the difference between the mesicwoodlands and the xeric patches decreases the resourcecontrasts also decrease leading to increasing communitysimilarity Different oak species can grow in a wide variety oflight and moisture conditions However none of the traditionalserpentine oak species is considered shade tolerant to the samedegree as are Acer spp Fagus grandifolia and Nyssa sylvaticaWith progressive development of the soil and a dense-canopyclosure growth and germination of the traditional xeric oakspecies are likely to be suppressed marking a shift from

Colonisation

Fig 9 Conceptual model for serpentine (orangendashred colouring) habitats and species redistribution resulting in similar structure for a typical piedmontgeoecologic system (bluered colouring) Bedrock geology (specifically elemental concentrations and structures) exerts its control on soil fertility with subtlevariations resulting in markedly different above-ground and below-ground communities The change in soils over time (dsdt) equation from Huggett (1995)is a function of the various drivers including b (biology) r (topography) a (atmospheric effects) s (soils) and h (hydrology) Through changing vegetationdynamics and species redistribution the tree functional groups and forest types will see increasing similarity as a result of converging successional trajectoriesRegional species pools stochastic events and many other factors determine ultimate outcomes

L Australian Journal of Botany J Burgess et al

xerothermic to mesic andor nutrient-demanding speciesFigure 9 details a conceptual model for pedologic and ecologicchanges at serpentine sites of theMid-Atlantic with successionaltrajectories converging on a species redistribution and loss oftypical serpentine species such as Cerastium arvense varvillosissimum Andropogon gerardii and the xeric oaks(Q marilandica and Q stellata)

Although the canopy is still oak or pine dominated thetrajectory points to a maplendashbeechndashred oak-dominated forestThe Mid-Atlantic has already lost much of its originalgrasslandndashoak woodlandndashforest mosaic This state transitionmay lead to the loss of a unique ecosystem and thustaxonomic homogenisation with the regional flora Thesechanges result from anthropic influences through changes inland use and fire frequency the loss of Castanea dentata(American chestnut) and changing herbivore populationsNearly a century ago these areas gave way to a mosaic ofsmall woodlands typically characterised by xeric oaksinterspersed with shallow rooting low nutrient-tolerantconifers Released from suppressive disturbance regimes(herbivory and fire for example) Pinus virginiana andJuniperus virginiana quickly became the dominant speciesHowever our data suggest that the landscape is nowundergoing another distinct change as the once-open canopystructure has becomemore closed Edaphic properties such as soildepth chemistry and moisture that once drove communityorganisation across environmental stress gradients are nowbeing modulated by species no longer filtered by theserpentine syndrome This modulation may allow for therecruitment and establishment of drought-sensitive speciesbeyond their typical tolerance limits (Warman and Moles2008) This unique ecosystem which took centuries tomillennia to evolve appears to be fading and may soonevanesce into the past

Sources of funding

Our work was funded by the National Science Foundation(DEB-0423476 and DEB-1027188) and grants from theR Balk and D Elliott Field Fund (Johns Hopkins University)

Conflicts of interest

No conflicts of interest

Acknowledgements

We thank Drs Richard Back Naomi Levin Bruce Marsh and Ben Passey foruse of their elemental and isotopic facilities We also thank Dr David Veblenand two anonymous reviewers for numerous constructive comments on thegeology

References

Abbe C (1898) A general report on the physiography of Maryland JohnsHopkins University (Johns Hopkins University Press Baltimore MD)

Abrams MD (1992) Fire and the development of oak forests Bioscience 42346ndash353 doi1023071311781

Abrams MD (1998) The red maple paradox Bioscience 48 355ndash364doi1023071313374

AbramsMD (2006) Ecological and ecophysiological attributes and responsesto fire in eastern oak forests In lsquoFire in eastern oak forests deliveringscience to landmanagersrsquo (EdMBDickinson) pp 74ndash89 (USDA ForestService General Technical Report NRS-P-1)

Alexander EB (2009) Serpentine geoecology of the eastern and southeasternmargins of North America Northeastern Naturalist 16 223ndash252doi1016560450160518

Alexander EB Ellis CC Burke R (2007) A chronosequence of soils andvegetation on serpentine terraces in the Klamath Mountains USA SoilScience 172 565ndash576 doi101097ss0b013e31804fa22d

Arabas KB (2000) Spatial and temporal relationships among fire frequencyvegetation and soil depth in an easternNorthAmerican serpentine barrenThe Journal of the Torrey Botanical Society 127 51ndash65doi1023073088747

Archer S Schimel DS Holland EA (1995) Mechanisms of shrublandexpansion land use climate or CO2 Climatic Change 29 91ndash99doi101007BF01091640

Axmann BD Knapp AK (1993) Water relations of Juniperus virginiana andAndropogon gerardii in an unburned tallgrass prairie watershed TheSouthwestern Naturalist 38 325ndash330 doi1023073671610

Barton AM Wallenstein MD (1997) Effects of invasion of Pinus virginianaon soil properties in serpentine barrens in southeastern PennsylvaniaThe Journal of the Torrey Botanical Society 124 297ndash305doi1023072997264

Bazzaz FA (1968) Succession on abandoned fields in the Shawnee Hillssouthern Illinois Ecology 49 924 doi1023071936544

Becker-Heidmann P Scharpenseel H-W (1992) Studies of soil organicmatterdynamics using natural carbon isotopes The Science of the TotalEnvironment 117ndash118 305ndash312 doi1010160048-9697(92)90097-C

Belcher JW Keddy PA Twolan-Strutt L (1995) Root and shoot competitionintensity along a soil depth gradient Journal of Ecology 83 673ndash682doi1023072261635

BouttonTWArcher SRMidwoodAJZitzer SFBolR (1998)d13Cvalues ofsoil organic carbon and their use in documenting vegetation change in asubtropical savanna ecosystem Geoderma 82 5ndash41doi101016S0016-7061(97)00095-5

Bouyoucos GJ (1936) Directions for making mechanical analyses of soilsby the hydrometer method Soil Science 42 225ndash230doi10109700010694-193609000-00007

Briggs JMGibsonDJ (1992)Effect offire on tree spatial patterns in a tallgrassprairie landscape Bulletin of the Torrey Botanical Club 119 300ndash307doi1023072996761

BrownML Brown RG (1984) lsquoHerbaceous plants of Marylandrsquo (Universityof Maryland College Park MD)

Burgess JL Lev S Swan CM Szlavecz K (2009) Geologic and edaphiccontrols on a serpentine forest community Northeastern Naturalist 16366ndash384 doi1016560450160527

Butaye J Jacquemyn H Honnay O Hermy M (2002) The species poolconcept applied to forests in a fragmented landscape dispersal limitationversus habitat limitation Journal of Vegetation Science 13 27ndash34doi101111j1654-11032002tb02020x

Callaway RM Davis FW (1993) Vegetation dynamics fire and the physical-environment in coastal central California Rid B-7010-2009 Ecology 741567ndash1578 doi1023071940084

Carroll W D Kapeluck P R Harper R A and Van Lear D H (2002)Background paper historical overview of the southern forest landscapeand associated resources In lsquoSouthern forest resource assessmentrsquo (EdsDNWear JGGreis) pp 583ndash605 (USDepartmentofAgricultureForestService Southern Research Station Asheville NC)

Cerling TE Quade J Wang Y Bowman JR (1989) Carbon isotopes in soilsand palaeosols as ecology and palaeoecology indicators Nature 341138ndash139 doi101038341138a0

Chao A (1984) Nonparametric estimation of the number of classes in apopulation Scandinavian Journal of Statistics 11 265ndash270

ChaoA (1987) Estimating the population size for capturendashrecapture data withunequal catchability Biometrics 43 783ndash791 doi1023072531532

Chazdon RL Colwell RK Denslow JS Guariguata MR (1998) Statisticalmethods for estimating species richness of woody regeneration in primaryand secondary rain forests of Northeastern Costa Rica In lsquoForest

Conifer encroachment on ultramafics Australian Journal of Botany M

for each plot and on pooled plot data using the ShannonndashWeaverindex (Shannon and Weaver 1949) and the Simpson index(Simpson 1949) as measures of a diversity (a) along withestimators of total diversity using rarefaction and non-parametric approaches of Chao 1 Chao 2 and ACE (Chao1984 1987 Chazdon et al 1998) The importance value (IV)is a measure of relative abundance and was calculated as theaverage of relative density and dominance

The KolmogorovndashSmirnov test statistic was used to assessnormality of the soil variables Those that failed the normality testwere transformed to normal or near-normal distributionPercentage measurements (soil particle analysis abundance CN cover XRF whole-rock oxide weight) were arcsinetransformed soil elemental data were loge transformed fromppm or ppb concentrations slope and soil depth were natural-log transformed and bulk density was square-root transformed

Site microclimate variability was compared using geographiccoordinates (latitude and longitude) inclination and aspect foreach plot to calculate the heat load and the potential annual directincident radiation by applying the third equation from McCuneand Keon (2002) The effects of topographic (slope inclinationannual direct incident radiation and heat load) and soil variables(elemental analysis in addition to pH soil organic C total NC N depth texture and d13C values) on species were examinedusing principal components analysis (PCA) to explore the mainfactors and generate hypotheses as to the reciprocal effects ofthese properties on vegetation Redundancy analysis (RDA)ordination as constrained by the environmental variables wasperformed to identify the main floristic gradients (basal areadensity percentage xeric species) and the edaphic andtopographic factors explaining those trends Two-wayANOVA was used to test for the differences in soil d13C C Nand C N with respect to vegetation (grassland forest) and soildepth ANOVAs were followed by Tukeyrsquos HSD (chosenbecause of uneven number of replicates) multiple pairwisecomparison to relate differences in elemental content and depth

Differences in site woody-species composition wereinvestigated by calculating the pairwise ecological distance forthe entire species matrix The Mantel test was used to determinewhether site similarity in the tree community was related to siteenvironmental variables Multiple linear regression analysis wasused to identify site factors that had a significant effect on speciesabundance The eight highest-ranked species on the basis of IVwere evaluated in termsof their abundances as responsevariablesand soil and topographic data were used as explanatory variablesOrdination using direct (RDA) and indirect (PCA) gradient

analysis was employed to extract variation in the underlyingstructure of the vegetation data as it relates to environmentalvariables (Ludwig and Reynolds 1988) PCA was performed tofind the main factors determining the reciprocal effects of soiland vegetation

The BrayndashCurtis distance with raw abundances was usedto determine the degree of vegetation dissimilarity an aspectof b (b) diversity between sites and vegetation types (woodlandand forest) based on presenceabsence We used a Mantel test(McCune and Grace 2002) to compare the matrix of geographicdistances between plots with the matrix of ecological distancePotential relationships between abiotic variables and diversityvariables (floristic similarity evenness and Simpsonrsquos diversity)were explored Tree-species compositions of early successionalwoodland and more mature forest designations based onpercentage canopy were compared using non-parametricanalysis of similarities (ANOSIM) based on the hypotheses ofno difference between stages

Gamma (g) or serpentine landscape-area diversity wasestimated using all 32 plots to identify species richness at thelandscape scale by constructing a rarefaction curve for all samplesand all species in the dataset

For all statistical tests significance was determined at thea= 005 level other P-values were reported if lt01 The methodfor evaluating the Mantel-test statistic was the t-distribution(Mantel 1967) and the standardised Mantel statistic (r) as ameasure of effect size (McCune and Grace 2002)

All data were analysed using the statistical computingenvironment R (v 311) and the BiodiversityR package(Kindt and Coe 2005) ANOSIM Mantel and other tests wereperformed with n= 1000 permutations Reported errors arestandard errors (se) unless indicated otherwise

Results

Abiotic substrate and soil composition

Mean whole-rock major-element values were typical of alteredultramafic rocks (Table 1) Calcium and aluminium (Al) contentswere low reflecting disappearance of the calcic pyroxene byserpentinisation Only samples where clinopyroxene wasdetected showed an increase in Ca Samples with elevated Alare generally accompanied by the presence of chlorite in therocks As is typical of ultramafics the immobile refractoryelements Cr and Ni were elevated Ternary diagrams depictingthe measured whole-rock geochemistry of the samples illustratedrelative compositional variations in the systemMgOndashSiO2ndashH2O

Table 1 Summary of the whole-rock geochemistry parameters for serpentinite samples obtained from X-ray fluorescence (XRF) analysisSerpentinite water content based on loss on ignition (LOI) ~124 03 weight Al aluminium Ca calcium Cr chromium Fe iron K potassium Mg

magnesium Na sodium Ni nickel Si silicon Ti titanium V vanadium Zn zinc

Lithology Major element geochemistry (weight ) Trace elements (mg kgndash1)Serpentinite n= 12 Si Al Fe Mg Ca Na K Ti Cr Ni V Zn

Mean 4129 031 954 3669 042 000 007 002 283799 231968 1189 4385se 088 015 078 070 022 000 002 000 28651 15716 334 594Soils (0ndash5 cm) n= 32 3513 091 657 625 024 015 070 048 252651 115744 9678 10705se 132 019 040 060 006 003 005 003 27641 13040 642 463Soils (5ndash15 cm) n= 48 3799 105 744 762 015 015 069 046 234790 124434 9725 8426se 106 020 036 060 004 002 005 003 20726 11024 554 311

Conifer encroachment on ultramafics Australian Journal of Botany E

(Fig 3) Concentrations of soilmacronutrientsCN P andKwerenot related to ultramafic parent material

Soil depths varied from105 cm to52 cmwith variable texture(mean of 134 gravel) across the sites Soil colours ranged fromdark brown (75YR 32) to grey (75YR 61) in the upper 1ndash5 cmof the soil profile to yellowndashbrown (10YR 56) and redndashbrown(25YR 44) with depth Talcose serpentinites yielded moreyellow (10YR Hues) varieties The mean and standard error ofsoils collected from the 32 plots recorded a narrow range of

edaphic characteristics including soil depth 289 20 (cm) pH54 01 bulk density 088 007 (gm3) soil organic matter(SOM) 1251 26 () sand 5103 18 () clay 267 11() and C N 1518 09

Soil moisture and soil depth increased with distance from thebarrens areas Strong positive correlationwith distancewas foundfor both soil moisture and depth (r= 0570 P lt 0001 andr= 0848 Plt 0001) respectively (Fig 4) Other soil physio-chemical parameters such as texture and trace element contentdid not show correlations with distance (P gt 005 Spearmanrsquosrank correlation)

Magnesium concentrations were high and correspondingCa Mg ratios low (lt05) across all sites The Ca Mg ratio isa measure of serpentine cation imbalance Surface soils hadhigher Ca Mg (02 to 05) ratios than did subsurface soils(005 to 02) Chromium and Ni concentrations were elevatedwith a maximum of over 6000 ppm and 2700 ppm respectively

The pH of the soils was typically acidic (surface soils average~54) but ranged from (~4ndash7) Soil pHwas negatively correlatedwith Ca Mg (r= ndash049 P lt 005) and C N ratios (r= ndash062P lt 005) SOM content (based on LOI) had a mean of 166in the upper 5 cm declining to below 7with depth (10ndash15 cm)N content was generally low and C N ratios varied between 5and 24with the higher values on themore developed soils Depthprofiles of SOM displayed increasing d13C isotopic signaturesfrom 2means of ndash2623permil (0ndash5 cm) to ndash2095permil (10ndash15 cm)(Table 2) whereas remnant grassland areas recorded meand13C isotopic signatures of ndash1623permil (0ndash5 cm)

Woody-species characteristics

Using high-resolution Google Earth imagery the current (2013)percentage of the site that contains grassland or savanna is lessthan 2 The forests include xeric woodlands with stands ofP virginiana and J virginiana intermingled with Quercusmarilandica Meunch (blackjack oak) and Quercus stellata

Talc

Enstatite

Serpentine

Olivine

Ca Mg

Si

Diopside

Typical Ultramafic

Fig 3 Ternary diagram detailing the chemography of Cherry Hillserpentinites (open circles) projected onto the plane CaOndashMgOndashSiO2 alsoshowing typical ultramafic rock (filled square) and mineral compositions(solid circles)

60

Soil depth (cm)

Percent moisture

50

40

30

20

Soi

l dep

th (

cm)

and

moi

stur

e (

by

volu

me)

Distance along transect (m)

10

00 10 20 30 40 50 60

Fig 4 Monotonic change of variables along transects The relationship between soil depth percentage moisture (volume) and distance from barrens bedrockwere assessed using correlational analysis (Plt 00001) Data are compiled averages of four transects from rocky barrens to woodlandndashforest ecotone Thecorrelation of the two variables soil depth and moisture is significant (r = 067 Plt 0001)

F Australian Journal of Botany J Burgess et al

Wang (post oak) as well as more mesophytic species Largeopen-crown trees are rare more typical are contorted down-sweeping branches of Q marilandica surrounded by stand-grown trees Aerial photographs from 1937 show conifersadvancing from the dendritic drainage areas Mesophyticforests are composed of these oaks and others such as Quercusrubra L (red oak) Q alba L (white oak) Q montana Willd(chestnut oak) and Q velutina Lam (black oak) along withnumerous other species most notably Prunus spp (blackcherry wild cherry) Pinus spp (eg white pine shortleafpine) Fagus grandifolia Ehrh (American beech) Acer rubrumL (red maple) Sassafras albidum Nutt (sassafras) Populusgrandidentata Michx (bigtooth aspen) Nyssa sylvaticaMarshall (blackgum) Betula lenta L (sSweet birch) andRobinia pseudoacacia L (black locust) On exposed rocksurfaces the serpentine endemic (as reviewed by Harris andRajakaruna 2009) Cerastium velutinum var velutinum Raf(serpentine mouse eared chickweed) and other serpentinophilessuch as Symphyotrichum depauperatum Fern (serpentine aster)and saplings of Q marilandica are common

Biodiversity richness estimators Chao 1 Chao 2 and ACEestimated maximum woody-species richness between 263and 308 species The observed richness was 24 species(Table 3) Overall diversity and mean diversity for the 32 siteswere calculated as 274 (X = 167 s= 029) (Shannon index) and

092 (X = 078 s= 008) (Simpson index) On the basis of relativeIV the most important species were P virginiana andJ virginiana (Table 3) These two species had similar relativeIVs and stood out from all other species A second group ofspecies included A rubrum and Q marilandica At the genuslevel Quercus was the most important contributor to the treecommunity with 39 relative IV On the basis of basal area themost dominant species were Q rubra P virginiana andQ marilandica Constancy values (Table 3) showedA rubrum (078) as a clear break from the next grouping ofspecies (P virginiana and J virginiana) whose constancy valueshovered near 05 followed by Q marilandica F grandifoliaN sylvatica and Q stellate with constancy values near 04

The Mantel test comparing floristic similarity andenvironmental distance showed significant (P lt 005)correlations between environmental distance and ecologicaldistance for the following parameters heat load bulk densitybasal area diversity (H) percentage sand Ni and Cr(Table 4) The ANOSIM statistic was calculated only for thecanopy category (woodlandforest) with a resulting r= 014 andP-value of 005 Both the Mantel and ANOSIM tests yieldedsignificant (Plt 005) but weak correlations

Relationships between vegetation and abiotic properties

Results of multiple regression analysis were evaluated tounderstand the response of species distributions to soilproperties Table 5 shows the regression results for the eightspecies the importance of which met or exceeded 5 withtopographic and edaphic parameters as the predictor variablesPredictor variables that indicated a meaningful contribution toexplaining species abundancewere incident radiation and soilMgcontent for Q marilandica incident radiation for J virginianaand Q stellata soil organic carbon (SOC) for A rubrum soildepth for F grandifolia and bulk density for the conifers

PCA was performed for 16 abiotic parameters across the 32sampled quadrats so as to identify covariation between the solarradiation and soil variables (Fig 5) Themain trend of variation in

Table 2 Descriptive statistics for soil d13C (permil) and soil organic matterin grassland and woodland elements

Means followed by the same letter are not significantly different (ANOVApost hoc Tukeyrsquos HSD P = 005)

Statistic Grassland Forest(0ndash5 cm)

Forest(5ndash10 cm)

Forest(5ndash15 cm)

Forest(10ndash15 cm)

Mean ndash1623a ndash2623b ndash2359bcd ndash2298de ndash2095eMinimum ndash1791 ndash2923 ndash2725 ndash2649 ndash2514Maximum ndash1423 ndash1928 ndash1811 ndash1972 ndash1654se 068 044 067 065 060

Table 3 Constancy importance (average value of a species when present) density dominance andimportance value (average of relative density and relative dominance) measures for woody stems in

serpentinite woodland at Cherry Hill Maryland USA

Species Constancy Importance Dominance(m2 handash1)

Density(stems handash1)

Importancevalue (IV) ()

Acer rubrum L 078 109 5799 7778 988Ailanthus altissima Mill 003 003 075 222 027Betula lenta L 006 006 346 444 060Celtis occidentalis L 006 006 059 444 041Cercis canadensis L 003 003 078 222 022Cornus florida L 006 006 146 444 038Diospyros virginiana L 003 003 185 222 024Fagus grandifolia Ehrh 043 065 5713 4667 619Juniperus virginiana L 050 106 8657 7556 1214Kalmia latifolia L 012 015 163 1111 106Liriodendron tulipifera L 009 009 2372 667 108Nyssa sylvatica Marsh 043 075 5769 5333 720Pinus strobus L 003 003 692 222 090Pinus virginiana Mill 056 078 16498 5556 1343Populus grandidentata Michx 003 006 640 444 058Prunus serotina Ehrh 028 028 2046 2000 271

Conifer encroachment on ultramafics Australian Journal of Botany G

the soil chemical properties and topographic variables asreflected by the PCA Axis 1 (accounting for 30 of variance)was one dominated by significant correlations of soil texture(sand and bulk density) soil depth Mg Cr and Niconcentrations incident radiation and heat load The secondtrend of soil variation as reflected by the PCA Axis 2 (17 ofvariation) can be interpreted as a combined gradient of essentialnutrients (C andN) and SOC The PCA (axes accounting for 44of the variation) revealed strong positive associations betweenboth soil depth and bulk density which were negativelycorrelated with heavy metal concentrations and pH SoilCa Mg ratios and Al concentrations were negativelycorrelated with macronutrients (C N P) and SOC Fig 5bshows the PCA factor map for the abiotic variables and thebiotic covariates for the sampling plots When thecharacteristics of woody-species community are included inthe PCA both basal area and stem density are positivelycorrelated with soil Ca Mg ratios and Al content Diversity(Shannonrsquos index) was strongly correlated with soil depth andbulkdensity andnegatively related toheavymetal concentrations

and heat load As individual-species basal area stem densityand diversity typically respond to environmental gradientsconstrained ordination utilising RDA was used to inferrelationships between the species matrix and theenvironmental matrix using Euclidean distance between sitesArrows of increasing species abundance represent RDA scoresThe sum of the RDA eigenvalues for the first two axes using allabiotic explanatory variables accounted for 368 of the woody-species composition After reduced variable selection based onANOVA tests of the ordinationmodel seven termswere selectedand these collectively explained 271 of the variance (Fig 6)Interestingly those species whose importance was gt4 arepartitioned into the four quadrants of the ordination From theRDAAxes1and2biplot it canbe seen that traditional serpentine-savanna (shade-intolerant)woody species (S albidumQ stellataand Q marilandica) are positively associated with increasingincident radiation and negatively associated with increasingsoil depth Al and Mg soil concentrations Post-disturbanceinitial colonisers (J virginiana and P virginiana) are stronglynegatively correlated with bulk densityQ alba andQ montanawhich are widely distributed across a variety of conditionsfrom wet mesic to subxeric are associated with increasingbulk density Shade-tolerant species such as N sylvaticaA rubrum and F grandifolia are correlated with increasingsoil depth and soils with more Ca and Al

Soils at the site differed in the total amount verticaldistribution and isotopic composition of SOM (Tables 1 2)ANOVA results indicated significant differences in theisotopic means according to their depth (F487 = 276P 00001) Tukeyrsquos HSD test showed that the mean d13C ofgrassland soils (ndash162 07permil) was significantly higher thanthat of shallow forested soils (ndash262 04permil) (Table 5)Shallow forested soils were significantly different from deepersoils (ndash2095 06permil) Total soil C and N and soil C N declinedwith soil depth C N ratios were significantly different for alldepths ranging from 1518 088 on the surface soils to1162 1 in soils 10ndash15 cm below surface (F266 = 59

Table 4 Effects of environmental distance on similarity between plotsMantel statistic calculated by comparing matrix of environmental distancewith BrayndashCurtis distance The standardised Mantel statistic (r) indicatesthe strength of the relationship between sample-point position and forest-floor thickness and the associated P-value indicates the significance of

that relationship

Environmental indicator Correlation coefficient (r) P-value

Heat load 0126 0009Bulk density 0236 0003Basal area 0141 0015Diversity (H) 0109 0034Percentage sand 0203 0002Nickel 0111 0034Chromium 0159 0009Canopy 0150 0009

Table 5 Results of multiple regression with list of factors significantly contributing to species abundanceSOC soil organic carbon Plt 0001 P lt 001 Plt 005 lsquorsquo Plt 01

Pvirginiana J virginiana A rubrum Q marilandica Q rubra N sylvatica Q alba Q stellata F grandifolia Q montana

IV () 1343 1214 988 940 810 720 703 626 619 440Incident radiation 09247 00249 01791 00008 04332 02136 09123 00450 03030 03225Heat load 02516 04869 03661 08494 09552 09203 09825 02522 01128 03502Soil depth 02599 03110 08937 00463 01881 03041 03381 01689 00242 09920Bulk density 00023 00037 08260 08860 05216 02157 00415 02686 09894 01171Texture ( sand) 00976 02800 02942 06247 08124 09313 03207 03451 01454 06054Calcium magnesium 09840 02962 07824 07031 09971 04587 07945 05441 06414 03639Magnesium 06146 05570 05725 00019 01500 06177 00914 03238 07491 03623Aluminium 07988 07323 01491 09176 00673 00583 08508 04152 05380 08685Chromium 05879 02397 02066 01012 06214 09977 06126 03991 06870 00434Nickel 08699 02237 07478 03211 08340 06190 09519 05875 06307 01074Phosphorus 05192 05066 07822 09800 07523 06641 04927 09619 09430 02268SOC 06435 01185 00298 00053 09782 02790 05185 02333 08701 07036Carbon nitrogen 05806 02325 06793 08013 05674 06443 01727 01105 08015 06972Carbon 07737 07963 02099 01231 04380 08148 04280 04885 07995 05145Nitrogen 07506 09537 01179 03581 06363 05329 03124 07375 00153 05744pH 07946 01863 05144 01547 00176 04256 02597 06743 01299 09047Multiple R2 059 068 057 079 057 046 051 055 062 056

H Australian Journal of Botany J Burgess et al

P = 0004) There were significant vegetation depth andvegetation depth effects on d13C of SOM (Table 5 Fig 7)

Age structure

In total 163 core samples were evaluated with results displayedin the stand-age structure plot (Fig 8) Skeleton plot chronologiesvaried among species Older minimum age at breast height wasgenerally seen inQ stellata (114 15 years) andQ marilandica(110 7 years) and the younger chronologies developed inspecies such as A rubrum (53 4 years) F grandifolia(56 8 years) and L tulipifera (43 8 years) Only xeric oakswere present in the oldest age classes The period before 1900 iscomposed of 54 xeric oaks and when all oaks are grouped theycomprise 79 of trees from that era although Q montana isabsent from this group (oldest minimum age is 88 years) Theoldest non-oak was N sylvatica (144 years) The period between

1900 and 1940 records a marked increase in conifers andconcomitant decrease of xeric oaks comprising 39 and 19of all species respectively Between 1950 and 1990 thepercentage of xeric oaks decreased to 3 and conifers to 13whereas Acer spp increased to 19 The number of shade-tolerant species rose from 125 in the decades from 1850 to1900 to 59 during the interval from 1950 to 1990

The age structure pooled from all the plots (Fig 8) showeddiscontinuous patterns with prominent peaks of regenerationpulses and periods of either low or absent recruitment Coniferestablishment and recruitment clearly increased during1900ndash1960 and then receded through the rest of the samplingperiod There was a conspicuous absence of younger xeric oaks(post-1970s) and only two xeric oaks from1950 to the present Atthe same time the percentage of non-oak deciduous treesincreased to 68 post-1950

10

05

00

Dim

2 (

171

7)

Dim

2 (

229

1)

Dim 1 (3038)

Dim 1 (402)

ndash10

ndash05

4

2

0

ndash2

ndash4

ndash6

ndash15

ndash15 ndash10 ndash5 0 5 10

ndash10 ndash05 00 05 10 15

(a)

(b)

Fig 5 (a) Unconstrained principal component analysis (PCA) correlation circle of pedological topological and community variables (b) Forest communityclustering along an edaphic gradient

Conifer encroachment on ultramafics Australian Journal of Botany I

Discussion

The present study examined a moderately sized area in the Mid-Atlantic but the changes ndash encroachment and mesophication ndash

represent potential widespread conservation problems Theserpentinite landscapes were once common components of theancient Piedmont (Carroll et al 2002 Owen 2002 Noss 2013)These unique plant communities have existed as persistentlsquoislandsrsquo embedded in a deciduous forest region for centuriesor longer We documented a state transition from grasslandand xeric oak savanna to conifer-dominated communities

that have given way to eastern broadleaf forests Comparisonsamong geologic pedologic and vegetation properties reflect thischange and indicate that the conversion to forest has been rapidand characterised by large changes in soil properties (pH water-holding capacity SOMcomposition) aswell as plant-communitystructure (physiognomyand stemdensity) and composition (xericto more mesic communities)

Thin-section and petrologic data confirmed that the geologicsubstrate is a serpentinised peridotitewith characteristics typical ofserpentinite terranes in the Appalachians (Misra and Keller 1978Mittwede and Stoddard 1989 Gates 1992 Alexander 2009)Although Ca is depleted in bulk rock samples because of thebreakdownof augite and interstitial plagioclase soil Ca Mg ratiosare higher in surface soils The concentrations of SOC C N andmacronutrients in woodland and forested areas are linked to theturnover of organic matter and not the parent material

Compared with the rock soils are enriched in most tracemetals although the Cr contents were similar This was likelyto be due to resistance of Cr-oxides to weathering and suggeststhat these phases were not likely sources for Cr in soil solutionsand plants of serpentine soils (Oze et al 2004) Bedrock to foresttransects demonstrated a strong correlation between soilmoistureand soil depth Soil nutrient content and soil moisture should beproportional to the cube of soil depth (Belcher et al 1995) thuseach of these parameters figured significantly in the PCA andRDA

Afforestation has been relatively recent in this region (Tyndall1992)Concomitantwith the increase in treedensity andsize is thecontraction of the grassland and savanna component study areafrom 18 in 1937 to 15 in 2013 which is consistent withprevious work documenting the invasion of serpentine savannasby conifers (Tyndall 1992 Barton andWallenstein 1997 Arabas2000)

In landscapeswhereC3 andC4plants coexist stableC isotopiccomposition of SOM represents a powerful method for

20

15

1

0

10

ndash10

ndash2 ndash1 0 1 2 3

05

ndash05

00

RD

A 2

RDA 1

Fig 6 Ordination diagram resulting from standard redundancy analysis (RDA) on transformed abiotic variables illustrating relationships between woodyspecies and physiochemical attributes The diamonds represent species and the vectors indicate the strength of relationship

ndash29 ndash27

(0ndash5)

(5ndash10)

(5ndash15)

(10ndash15)

Grassland (0ndash10)

Soil δ13 C (permil)

Forests

Grassland

Soi

l dep

th (

cm)

ndash25 ndash23 ndash21 ndash19 ndash17 ndash15

Fig 7 Mean carbon isotopic signature in forest (n= 96) and grassland(n= 8) soils for Cherry Hill Maryland USA Error bars are 1 se Datainterpreted as evidence for expansive C4 or C3 historic grassland Verticallines over 5ndash10- and 5ndash15-cm data indicate group pairs that were notsignificantly different using Tukeyrsquos HSD multiple comparisons

J Australian Journal of Botany J Burgess et al

reconstructing vegetation changes (Deines 1980 Tieszen andArcher 1990 Boutton and Yamasaki 1996 Boutton et al 1998)Spatial patterns of soil d13C were strongly related to type ofvegetation (primarily woody versus grassland cover) Remnantgrasslands are composed of a mixture of C3 grasses and forbsand C4 grasses It is well documented that the d13C signature ofC3 plants (ranging from26permil to28permil) is easily distinguishedfrom that of the C4 plants (ranging from 12permil to 15permil)(DeNiro and Epstein 1977 Cerling et al 1989 Tieszen 1991Ehleringer et al 2000 Fox et al 2012 Throop et al 2013) Otherprocesses besides vegetation change alter isotopic ratios (such aspreferential degradation of some plant materials microbiologicalactivity Suess effect and adsorption processes) and may causeshifts of the order of 1ndash3permil (Friedli et al 1987Becker-HeidmannandScharpenseel 1992Trolier et al 1996Ehleringer et al 2000SantRuCkova et al 2000 Krull et al 2005) d13C enrichmentin temperate forest soils larger than 3ndash4permil is generally attributedto the presence of remaining old C4-derived SOC Following thereasoning of Wynn and Bird (2008) we sampled the upper soils(A horizon) and assumed that the intrinsic metabolic processesoccurring in C3 and C4 vegetation impart an unequivocald13C signature into the SOM and that these signatures persistin decaying organic matter such that increasing depth yields achronologic sequence recording vegetation change At the studyarea d13C values of soils in the grassland matrix averaged ndash16permil(Table 4) to 26permil in forested areas dominated by C3 plantsDepth-integrated SOM d13C significantly reflected the currentvegetation cover (mixed C3ndashC4 and solely C3) and accordinglycould be used as a reference to trace past changes in vegetationrecorded in the SOM at deeper soil layers The d13C values ofSOC consistently increased with depth reflecting a higherpercentage of C4-derived carbon in the older soils that isinterpreted to reflect a site-wide shift in vegetation from C4 toC3 plants

Multiple regression and ordination results supported the roleof soil depth and incident radiation both of which influence soilmoisture content as factors governing the occurrence of xericserpentine savannaQuercus species Abiotic factors in areas withhigher importance of shade-tolerant species were associated withdeeper soils higher bulk density and higher Ca Mg ratios In oursite early successional woodlands were favoured in areas of highincident radiation and broadleaf forests were more developed onslopes containing more clay rich and denser soils Diversity ofwoody specieswas higher in areaswith deeper soilsUnlikemanysystems with heavy metals we found no correlations betweentoxic metals (Cr and Ni) and vegetation dynamics

Historical records show that oak savanna ecosystems wereimportant components of the Mid-Atlantic landscape (Shreveet al 1910) The lack of any oaks before the 1850s suggests thatthey were likely harvested for fuel posts or to support grazingFarming of these serpentinite areas is not documented and isunlikely given the inhospitable nature of the soil Our studyindicated that tree recruitment before the 20th centurywasmostlyoaks perhaps regenerating from root sprouts or propagules in theseedbankXeric oak recruitment decreasedover timeandnoxericoaks younger than 40 years established in the forested areasalthough seedlings and saplings still exist in the remnantgrasslandsavannas This failure to regenerate is likely toreflect the consequences of changing disturbance regimes suchas fire-suppression (Nuzzo 1986 Abrams 1992 2006 PetersonandReich 2001) alongwith competitionwith invasive andmesictree species (Nuzzo 1986) White-tailed deer and squirrelherbivory can reduce recruitment especially in isolatedsystems such as these serpentinite landscapes (Shea andStange 1998 Haas and Heske 2005) A turning point forcommunity development appears to be ~1900 with theencroachment of Pinus virginiana and Juniperus virginianaThese conifer species have a variety of characteristics that can

30

25

Minor species

Mesic species (A rubrum F grandifolia N sylvatica P serotina)

Conifers

Quercus spp (Mixed)

Quercus spp (Xeric)

(n = 163)15

5

20

10

Num

ber

of tr

ees

sam

pled

Decade of origin

01850 1860 1870 1880 1890 1900 1910 1920 1930 1940 1950 1960 1970 1980 1990

Fig 8 Age distribution from trees on the basis of core data (n= 163)Xeric species areQuercusmarilandica andQ stellataMixed oak species includeQ albaQ montana Q rubra and Q velutina

Conifer encroachment on ultramafics Australian Journal of Botany K

potentially alter ecosystem processes for decades as they matureand senesce Both conifers are aggressive invaders because ofrapid growth high seed production high seed-dispersalefficiency tolerance of xeric conditions and opportunistic useof ECM mycorrhizae (Ormsbee et al 1976 Holthuijzen andSharik 1984 1985 Briggs andGibson 1992Axmann andKnapp1993 Thiet and Boerner 2007) Once established these speciescan have a profound effect on soil properties as a result ofaccumulation of organic matter on the forest floor and theeventual decomposition of detritus which leads to higherwater-holding capacity and lowering of soil pH This positivefeedback between soil depth and litter cover exerted by theconifers could promote succession towards non-serpentineconditions

The initial serpentine edaphic signature acts as a filterpreventing species that belong to the regional flora but lack thetraits required to survive in such conditions to successfullycolonise these habitats By the 1970s such filters appear tohave diminished Current site characteristics have shiftedtowards more mesic conditions with changes in soil propertiessuch as bulk density soil depth pH SOM and metalsconcentration Over time the community has changed fromxeric (grassland or oak savanna) to one dominated by moremesic oaks and shade-tolerant Acer Nyssa and Fagus speciesThemajor onset ofAcer establishment began in the 1940sUnlikethePinus and Juniperus establishment pulse which proceeded inan open-canopy situation Acer dominance likely proceededunder closed canopy and canopy gaps because it is shade-tolerant One may speculate that over the past 100ndash200 years

the landscape has been dominated by Quercus with patches ofPinus and Juniperus that historicallywere present at lownumbersbecauseof disturbance (miningfire grazing logging)Hilgartneret al (2009) recorded evidence of the presence of Pinusvirginiana from the early 1800s in a serpentine barren siteonly 50 km distant Over the past 60 years Acer has becomeabundant where it along with other species currently present(N sylvatica Cornus florida and Celtis occidentalis) wouldnormally be considered an aberrant species in a hostile edaphicsetting Age and diameter structure indicated a shift away fromQuercus and towards an A rubrum-dominated landscape Thissame species shift has been widely reported throughout theeastern deciduous forests of North America (Abrams 1998Nowacki and Abrams 2008 McEwan et al 2010)

In the context of serpentinite ecosystems the influxof conifersnearly a century ago has acted as a landscapemodulator (Shachaket al 2008) The resulting build-up of SOM increases the water-holding capacity and other resources such as ground-levelradiation facilitate the influx of new species adapted to thealtered conditions As the difference between the mesicwoodlands and the xeric patches decreases the resourcecontrasts also decrease leading to increasing communitysimilarity Different oak species can grow in a wide variety oflight and moisture conditions However none of the traditionalserpentine oak species is considered shade tolerant to the samedegree as are Acer spp Fagus grandifolia and Nyssa sylvaticaWith progressive development of the soil and a dense-canopyclosure growth and germination of the traditional xeric oakspecies are likely to be suppressed marking a shift from

Colonisation

Fig 9 Conceptual model for serpentine (orangendashred colouring) habitats and species redistribution resulting in similar structure for a typical piedmontgeoecologic system (bluered colouring) Bedrock geology (specifically elemental concentrations and structures) exerts its control on soil fertility with subtlevariations resulting in markedly different above-ground and below-ground communities The change in soils over time (dsdt) equation from Huggett (1995)is a function of the various drivers including b (biology) r (topography) a (atmospheric effects) s (soils) and h (hydrology) Through changing vegetationdynamics and species redistribution the tree functional groups and forest types will see increasing similarity as a result of converging successional trajectoriesRegional species pools stochastic events and many other factors determine ultimate outcomes

L Australian Journal of Botany J Burgess et al

xerothermic to mesic andor nutrient-demanding speciesFigure 9 details a conceptual model for pedologic and ecologicchanges at serpentine sites of theMid-Atlantic with successionaltrajectories converging on a species redistribution and loss oftypical serpentine species such as Cerastium arvense varvillosissimum Andropogon gerardii and the xeric oaks(Q marilandica and Q stellata)

Although the canopy is still oak or pine dominated thetrajectory points to a maplendashbeechndashred oak-dominated forestThe Mid-Atlantic has already lost much of its originalgrasslandndashoak woodlandndashforest mosaic This state transitionmay lead to the loss of a unique ecosystem and thustaxonomic homogenisation with the regional flora Thesechanges result from anthropic influences through changes inland use and fire frequency the loss of Castanea dentata(American chestnut) and changing herbivore populationsNearly a century ago these areas gave way to a mosaic ofsmall woodlands typically characterised by xeric oaksinterspersed with shallow rooting low nutrient-tolerantconifers Released from suppressive disturbance regimes(herbivory and fire for example) Pinus virginiana andJuniperus virginiana quickly became the dominant speciesHowever our data suggest that the landscape is nowundergoing another distinct change as the once-open canopystructure has becomemore closed Edaphic properties such as soildepth chemistry and moisture that once drove communityorganisation across environmental stress gradients are nowbeing modulated by species no longer filtered by theserpentine syndrome This modulation may allow for therecruitment and establishment of drought-sensitive speciesbeyond their typical tolerance limits (Warman and Moles2008) This unique ecosystem which took centuries tomillennia to evolve appears to be fading and may soonevanesce into the past

Sources of funding

Our work was funded by the National Science Foundation(DEB-0423476 and DEB-1027188) and grants from theR Balk and D Elliott Field Fund (Johns Hopkins University)

Conflicts of interest

No conflicts of interest

Acknowledgements

We thank Drs Richard Back Naomi Levin Bruce Marsh and Ben Passey foruse of their elemental and isotopic facilities We also thank Dr David Veblenand two anonymous reviewers for numerous constructive comments on thegeology

References

Abbe C (1898) A general report on the physiography of Maryland JohnsHopkins University (Johns Hopkins University Press Baltimore MD)

Abrams MD (1992) Fire and the development of oak forests Bioscience 42346ndash353 doi1023071311781

Abrams MD (1998) The red maple paradox Bioscience 48 355ndash364doi1023071313374

AbramsMD (2006) Ecological and ecophysiological attributes and responsesto fire in eastern oak forests In lsquoFire in eastern oak forests deliveringscience to landmanagersrsquo (EdMBDickinson) pp 74ndash89 (USDA ForestService General Technical Report NRS-P-1)

Alexander EB (2009) Serpentine geoecology of the eastern and southeasternmargins of North America Northeastern Naturalist 16 223ndash252doi1016560450160518

Alexander EB Ellis CC Burke R (2007) A chronosequence of soils andvegetation on serpentine terraces in the Klamath Mountains USA SoilScience 172 565ndash576 doi101097ss0b013e31804fa22d

Arabas KB (2000) Spatial and temporal relationships among fire frequencyvegetation and soil depth in an easternNorthAmerican serpentine barrenThe Journal of the Torrey Botanical Society 127 51ndash65doi1023073088747

Archer S Schimel DS Holland EA (1995) Mechanisms of shrublandexpansion land use climate or CO2 Climatic Change 29 91ndash99doi101007BF01091640

Axmann BD Knapp AK (1993) Water relations of Juniperus virginiana andAndropogon gerardii in an unburned tallgrass prairie watershed TheSouthwestern Naturalist 38 325ndash330 doi1023073671610

Barton AM Wallenstein MD (1997) Effects of invasion of Pinus virginianaon soil properties in serpentine barrens in southeastern PennsylvaniaThe Journal of the Torrey Botanical Society 124 297ndash305doi1023072997264

Bazzaz FA (1968) Succession on abandoned fields in the Shawnee Hillssouthern Illinois Ecology 49 924 doi1023071936544

Becker-Heidmann P Scharpenseel H-W (1992) Studies of soil organicmatterdynamics using natural carbon isotopes The Science of the TotalEnvironment 117ndash118 305ndash312 doi1010160048-9697(92)90097-C

Belcher JW Keddy PA Twolan-Strutt L (1995) Root and shoot competitionintensity along a soil depth gradient Journal of Ecology 83 673ndash682doi1023072261635

BouttonTWArcher SRMidwoodAJZitzer SFBolR (1998)d13Cvalues ofsoil organic carbon and their use in documenting vegetation change in asubtropical savanna ecosystem Geoderma 82 5ndash41doi101016S0016-7061(97)00095-5

Bouyoucos GJ (1936) Directions for making mechanical analyses of soilsby the hydrometer method Soil Science 42 225ndash230doi10109700010694-193609000-00007

Briggs JMGibsonDJ (1992)Effect offire on tree spatial patterns in a tallgrassprairie landscape Bulletin of the Torrey Botanical Club 119 300ndash307doi1023072996761

BrownML Brown RG (1984) lsquoHerbaceous plants of Marylandrsquo (Universityof Maryland College Park MD)

Burgess JL Lev S Swan CM Szlavecz K (2009) Geologic and edaphiccontrols on a serpentine forest community Northeastern Naturalist 16366ndash384 doi1016560450160527

Butaye J Jacquemyn H Honnay O Hermy M (2002) The species poolconcept applied to forests in a fragmented landscape dispersal limitationversus habitat limitation Journal of Vegetation Science 13 27ndash34doi101111j1654-11032002tb02020x

Callaway RM Davis FW (1993) Vegetation dynamics fire and the physical-environment in coastal central California Rid B-7010-2009 Ecology 741567ndash1578 doi1023071940084

Carroll W D Kapeluck P R Harper R A and Van Lear D H (2002)Background paper historical overview of the southern forest landscapeand associated resources In lsquoSouthern forest resource assessmentrsquo (EdsDNWear JGGreis) pp 583ndash605 (USDepartmentofAgricultureForestService Southern Research Station Asheville NC)

Cerling TE Quade J Wang Y Bowman JR (1989) Carbon isotopes in soilsand palaeosols as ecology and palaeoecology indicators Nature 341138ndash139 doi101038341138a0

Chao A (1984) Nonparametric estimation of the number of classes in apopulation Scandinavian Journal of Statistics 11 265ndash270

ChaoA (1987) Estimating the population size for capturendashrecapture data withunequal catchability Biometrics 43 783ndash791 doi1023072531532

Chazdon RL Colwell RK Denslow JS Guariguata MR (1998) Statisticalmethods for estimating species richness of woody regeneration in primaryand secondary rain forests of Northeastern Costa Rica In lsquoForest

Conifer encroachment on ultramafics Australian Journal of Botany M

(Fig 3) Concentrations of soilmacronutrientsCN P andKwerenot related to ultramafic parent material

Soil depths varied from105 cm to52 cmwith variable texture(mean of 134 gravel) across the sites Soil colours ranged fromdark brown (75YR 32) to grey (75YR 61) in the upper 1ndash5 cmof the soil profile to yellowndashbrown (10YR 56) and redndashbrown(25YR 44) with depth Talcose serpentinites yielded moreyellow (10YR Hues) varieties The mean and standard error ofsoils collected from the 32 plots recorded a narrow range of

edaphic characteristics including soil depth 289 20 (cm) pH54 01 bulk density 088 007 (gm3) soil organic matter(SOM) 1251 26 () sand 5103 18 () clay 267 11() and C N 1518 09

Soil moisture and soil depth increased with distance from thebarrens areas Strong positive correlationwith distancewas foundfor both soil moisture and depth (r= 0570 P lt 0001 andr= 0848 Plt 0001) respectively (Fig 4) Other soil physio-chemical parameters such as texture and trace element contentdid not show correlations with distance (P gt 005 Spearmanrsquosrank correlation)

Magnesium concentrations were high and correspondingCa Mg ratios low (lt05) across all sites The Ca Mg ratio isa measure of serpentine cation imbalance Surface soils hadhigher Ca Mg (02 to 05) ratios than did subsurface soils(005 to 02) Chromium and Ni concentrations were elevatedwith a maximum of over 6000 ppm and 2700 ppm respectively

The pH of the soils was typically acidic (surface soils average~54) but ranged from (~4ndash7) Soil pHwas negatively correlatedwith Ca Mg (r= ndash049 P lt 005) and C N ratios (r= ndash062P lt 005) SOM content (based on LOI) had a mean of 166in the upper 5 cm declining to below 7with depth (10ndash15 cm)N content was generally low and C N ratios varied between 5and 24with the higher values on themore developed soils Depthprofiles of SOM displayed increasing d13C isotopic signaturesfrom 2means of ndash2623permil (0ndash5 cm) to ndash2095permil (10ndash15 cm)(Table 2) whereas remnant grassland areas recorded meand13C isotopic signatures of ndash1623permil (0ndash5 cm)

Woody-species characteristics

Using high-resolution Google Earth imagery the current (2013)percentage of the site that contains grassland or savanna is lessthan 2 The forests include xeric woodlands with stands ofP virginiana and J virginiana intermingled with Quercusmarilandica Meunch (blackjack oak) and Quercus stellata

Talc

Enstatite

Serpentine

Olivine

Ca Mg

Si

Diopside

Typical Ultramafic

Fig 3 Ternary diagram detailing the chemography of Cherry Hillserpentinites (open circles) projected onto the plane CaOndashMgOndashSiO2 alsoshowing typical ultramafic rock (filled square) and mineral compositions(solid circles)

60

Soil depth (cm)

Percent moisture

50

40

30

20

Soi

l dep

th (

cm)

and

moi

stur

e (

by

volu

me)

Distance along transect (m)

10

00 10 20 30 40 50 60

Fig 4 Monotonic change of variables along transects The relationship between soil depth percentage moisture (volume) and distance from barrens bedrockwere assessed using correlational analysis (Plt 00001) Data are compiled averages of four transects from rocky barrens to woodlandndashforest ecotone Thecorrelation of the two variables soil depth and moisture is significant (r = 067 Plt 0001)

F Australian Journal of Botany J Burgess et al

Wang (post oak) as well as more mesophytic species Largeopen-crown trees are rare more typical are contorted down-sweeping branches of Q marilandica surrounded by stand-grown trees Aerial photographs from 1937 show conifersadvancing from the dendritic drainage areas Mesophyticforests are composed of these oaks and others such as Quercusrubra L (red oak) Q alba L (white oak) Q montana Willd(chestnut oak) and Q velutina Lam (black oak) along withnumerous other species most notably Prunus spp (blackcherry wild cherry) Pinus spp (eg white pine shortleafpine) Fagus grandifolia Ehrh (American beech) Acer rubrumL (red maple) Sassafras albidum Nutt (sassafras) Populusgrandidentata Michx (bigtooth aspen) Nyssa sylvaticaMarshall (blackgum) Betula lenta L (sSweet birch) andRobinia pseudoacacia L (black locust) On exposed rocksurfaces the serpentine endemic (as reviewed by Harris andRajakaruna 2009) Cerastium velutinum var velutinum Raf(serpentine mouse eared chickweed) and other serpentinophilessuch as Symphyotrichum depauperatum Fern (serpentine aster)and saplings of Q marilandica are common

Biodiversity richness estimators Chao 1 Chao 2 and ACEestimated maximum woody-species richness between 263and 308 species The observed richness was 24 species(Table 3) Overall diversity and mean diversity for the 32 siteswere calculated as 274 (X = 167 s= 029) (Shannon index) and

092 (X = 078 s= 008) (Simpson index) On the basis of relativeIV the most important species were P virginiana andJ virginiana (Table 3) These two species had similar relativeIVs and stood out from all other species A second group ofspecies included A rubrum and Q marilandica At the genuslevel Quercus was the most important contributor to the treecommunity with 39 relative IV On the basis of basal area themost dominant species were Q rubra P virginiana andQ marilandica Constancy values (Table 3) showedA rubrum (078) as a clear break from the next grouping ofspecies (P virginiana and J virginiana) whose constancy valueshovered near 05 followed by Q marilandica F grandifoliaN sylvatica and Q stellate with constancy values near 04

The Mantel test comparing floristic similarity andenvironmental distance showed significant (P lt 005)correlations between environmental distance and ecologicaldistance for the following parameters heat load bulk densitybasal area diversity (H) percentage sand Ni and Cr(Table 4) The ANOSIM statistic was calculated only for thecanopy category (woodlandforest) with a resulting r= 014 andP-value of 005 Both the Mantel and ANOSIM tests yieldedsignificant (Plt 005) but weak correlations

Relationships between vegetation and abiotic properties

Results of multiple regression analysis were evaluated tounderstand the response of species distributions to soilproperties Table 5 shows the regression results for the eightspecies the importance of which met or exceeded 5 withtopographic and edaphic parameters as the predictor variablesPredictor variables that indicated a meaningful contribution toexplaining species abundancewere incident radiation and soilMgcontent for Q marilandica incident radiation for J virginianaand Q stellata soil organic carbon (SOC) for A rubrum soildepth for F grandifolia and bulk density for the conifers

PCA was performed for 16 abiotic parameters across the 32sampled quadrats so as to identify covariation between the solarradiation and soil variables (Fig 5) Themain trend of variation in

Table 2 Descriptive statistics for soil d13C (permil) and soil organic matterin grassland and woodland elements

Means followed by the same letter are not significantly different (ANOVApost hoc Tukeyrsquos HSD P = 005)

Statistic Grassland Forest(0ndash5 cm)

Forest(5ndash10 cm)

Forest(5ndash15 cm)

Forest(10ndash15 cm)

Mean ndash1623a ndash2623b ndash2359bcd ndash2298de ndash2095eMinimum ndash1791 ndash2923 ndash2725 ndash2649 ndash2514Maximum ndash1423 ndash1928 ndash1811 ndash1972 ndash1654se 068 044 067 065 060

Table 3 Constancy importance (average value of a species when present) density dominance andimportance value (average of relative density and relative dominance) measures for woody stems in

serpentinite woodland at Cherry Hill Maryland USA

Species Constancy Importance Dominance(m2 handash1)

Density(stems handash1)

Importancevalue (IV) ()

Acer rubrum L 078 109 5799 7778 988Ailanthus altissima Mill 003 003 075 222 027Betula lenta L 006 006 346 444 060Celtis occidentalis L 006 006 059 444 041Cercis canadensis L 003 003 078 222 022Cornus florida L 006 006 146 444 038Diospyros virginiana L 003 003 185 222 024Fagus grandifolia Ehrh 043 065 5713 4667 619Juniperus virginiana L 050 106 8657 7556 1214Kalmia latifolia L 012 015 163 1111 106Liriodendron tulipifera L 009 009 2372 667 108Nyssa sylvatica Marsh 043 075 5769 5333 720Pinus strobus L 003 003 692 222 090Pinus virginiana Mill 056 078 16498 5556 1343Populus grandidentata Michx 003 006 640 444 058Prunus serotina Ehrh 028 028 2046 2000 271

Conifer encroachment on ultramafics Australian Journal of Botany G

the soil chemical properties and topographic variables asreflected by the PCA Axis 1 (accounting for 30 of variance)was one dominated by significant correlations of soil texture(sand and bulk density) soil depth Mg Cr and Niconcentrations incident radiation and heat load The secondtrend of soil variation as reflected by the PCA Axis 2 (17 ofvariation) can be interpreted as a combined gradient of essentialnutrients (C andN) and SOC The PCA (axes accounting for 44of the variation) revealed strong positive associations betweenboth soil depth and bulk density which were negativelycorrelated with heavy metal concentrations and pH SoilCa Mg ratios and Al concentrations were negativelycorrelated with macronutrients (C N P) and SOC Fig 5bshows the PCA factor map for the abiotic variables and thebiotic covariates for the sampling plots When thecharacteristics of woody-species community are included inthe PCA both basal area and stem density are positivelycorrelated with soil Ca Mg ratios and Al content Diversity(Shannonrsquos index) was strongly correlated with soil depth andbulkdensity andnegatively related toheavymetal concentrations

and heat load As individual-species basal area stem densityand diversity typically respond to environmental gradientsconstrained ordination utilising RDA was used to inferrelationships between the species matrix and theenvironmental matrix using Euclidean distance between sitesArrows of increasing species abundance represent RDA scoresThe sum of the RDA eigenvalues for the first two axes using allabiotic explanatory variables accounted for 368 of the woody-species composition After reduced variable selection based onANOVA tests of the ordinationmodel seven termswere selectedand these collectively explained 271 of the variance (Fig 6)Interestingly those species whose importance was gt4 arepartitioned into the four quadrants of the ordination From theRDAAxes1and2biplot it canbe seen that traditional serpentine-savanna (shade-intolerant)woody species (S albidumQ stellataand Q marilandica) are positively associated with increasingincident radiation and negatively associated with increasingsoil depth Al and Mg soil concentrations Post-disturbanceinitial colonisers (J virginiana and P virginiana) are stronglynegatively correlated with bulk densityQ alba andQ montanawhich are widely distributed across a variety of conditionsfrom wet mesic to subxeric are associated with increasingbulk density Shade-tolerant species such as N sylvaticaA rubrum and F grandifolia are correlated with increasingsoil depth and soils with more Ca and Al

Soils at the site differed in the total amount verticaldistribution and isotopic composition of SOM (Tables 1 2)ANOVA results indicated significant differences in theisotopic means according to their depth (F487 = 276P 00001) Tukeyrsquos HSD test showed that the mean d13C ofgrassland soils (ndash162 07permil) was significantly higher thanthat of shallow forested soils (ndash262 04permil) (Table 5)Shallow forested soils were significantly different from deepersoils (ndash2095 06permil) Total soil C and N and soil C N declinedwith soil depth C N ratios were significantly different for alldepths ranging from 1518 088 on the surface soils to1162 1 in soils 10ndash15 cm below surface (F266 = 59

Table 4 Effects of environmental distance on similarity between plotsMantel statistic calculated by comparing matrix of environmental distancewith BrayndashCurtis distance The standardised Mantel statistic (r) indicatesthe strength of the relationship between sample-point position and forest-floor thickness and the associated P-value indicates the significance of

that relationship

Environmental indicator Correlation coefficient (r) P-value

Heat load 0126 0009Bulk density 0236 0003Basal area 0141 0015Diversity (H) 0109 0034Percentage sand 0203 0002Nickel 0111 0034Chromium 0159 0009Canopy 0150 0009

Table 5 Results of multiple regression with list of factors significantly contributing to species abundanceSOC soil organic carbon Plt 0001 P lt 001 Plt 005 lsquorsquo Plt 01

Pvirginiana J virginiana A rubrum Q marilandica Q rubra N sylvatica Q alba Q stellata F grandifolia Q montana

IV () 1343 1214 988 940 810 720 703 626 619 440Incident radiation 09247 00249 01791 00008 04332 02136 09123 00450 03030 03225Heat load 02516 04869 03661 08494 09552 09203 09825 02522 01128 03502Soil depth 02599 03110 08937 00463 01881 03041 03381 01689 00242 09920Bulk density 00023 00037 08260 08860 05216 02157 00415 02686 09894 01171Texture ( sand) 00976 02800 02942 06247 08124 09313 03207 03451 01454 06054Calcium magnesium 09840 02962 07824 07031 09971 04587 07945 05441 06414 03639Magnesium 06146 05570 05725 00019 01500 06177 00914 03238 07491 03623Aluminium 07988 07323 01491 09176 00673 00583 08508 04152 05380 08685Chromium 05879 02397 02066 01012 06214 09977 06126 03991 06870 00434Nickel 08699 02237 07478 03211 08340 06190 09519 05875 06307 01074Phosphorus 05192 05066 07822 09800 07523 06641 04927 09619 09430 02268SOC 06435 01185 00298 00053 09782 02790 05185 02333 08701 07036Carbon nitrogen 05806 02325 06793 08013 05674 06443 01727 01105 08015 06972Carbon 07737 07963 02099 01231 04380 08148 04280 04885 07995 05145Nitrogen 07506 09537 01179 03581 06363 05329 03124 07375 00153 05744pH 07946 01863 05144 01547 00176 04256 02597 06743 01299 09047Multiple R2 059 068 057 079 057 046 051 055 062 056

H Australian Journal of Botany J Burgess et al

P = 0004) There were significant vegetation depth andvegetation depth effects on d13C of SOM (Table 5 Fig 7)

Age structure

In total 163 core samples were evaluated with results displayedin the stand-age structure plot (Fig 8) Skeleton plot chronologiesvaried among species Older minimum age at breast height wasgenerally seen inQ stellata (114 15 years) andQ marilandica(110 7 years) and the younger chronologies developed inspecies such as A rubrum (53 4 years) F grandifolia(56 8 years) and L tulipifera (43 8 years) Only xeric oakswere present in the oldest age classes The period before 1900 iscomposed of 54 xeric oaks and when all oaks are grouped theycomprise 79 of trees from that era although Q montana isabsent from this group (oldest minimum age is 88 years) Theoldest non-oak was N sylvatica (144 years) The period between

1900 and 1940 records a marked increase in conifers andconcomitant decrease of xeric oaks comprising 39 and 19of all species respectively Between 1950 and 1990 thepercentage of xeric oaks decreased to 3 and conifers to 13whereas Acer spp increased to 19 The number of shade-tolerant species rose from 125 in the decades from 1850 to1900 to 59 during the interval from 1950 to 1990

The age structure pooled from all the plots (Fig 8) showeddiscontinuous patterns with prominent peaks of regenerationpulses and periods of either low or absent recruitment Coniferestablishment and recruitment clearly increased during1900ndash1960 and then receded through the rest of the samplingperiod There was a conspicuous absence of younger xeric oaks(post-1970s) and only two xeric oaks from1950 to the present Atthe same time the percentage of non-oak deciduous treesincreased to 68 post-1950

10

05

00

Dim

2 (

171

7)

Dim

2 (

229

1)

Dim 1 (3038)

Dim 1 (402)

ndash10

ndash05

4

2

0

ndash2

ndash4

ndash6

ndash15

ndash15 ndash10 ndash5 0 5 10

ndash10 ndash05 00 05 10 15

(a)

(b)

Fig 5 (a) Unconstrained principal component analysis (PCA) correlation circle of pedological topological and community variables (b) Forest communityclustering along an edaphic gradient

Conifer encroachment on ultramafics Australian Journal of Botany I

Discussion

The present study examined a moderately sized area in the Mid-Atlantic but the changes ndash encroachment and mesophication ndash

represent potential widespread conservation problems Theserpentinite landscapes were once common components of theancient Piedmont (Carroll et al 2002 Owen 2002 Noss 2013)These unique plant communities have existed as persistentlsquoislandsrsquo embedded in a deciduous forest region for centuriesor longer We documented a state transition from grasslandand xeric oak savanna to conifer-dominated communities

that have given way to eastern broadleaf forests Comparisonsamong geologic pedologic and vegetation properties reflect thischange and indicate that the conversion to forest has been rapidand characterised by large changes in soil properties (pH water-holding capacity SOMcomposition) aswell as plant-communitystructure (physiognomyand stemdensity) and composition (xericto more mesic communities)

Thin-section and petrologic data confirmed that the geologicsubstrate is a serpentinised peridotitewith characteristics typical ofserpentinite terranes in the Appalachians (Misra and Keller 1978Mittwede and Stoddard 1989 Gates 1992 Alexander 2009)Although Ca is depleted in bulk rock samples because of thebreakdownof augite and interstitial plagioclase soil Ca Mg ratiosare higher in surface soils The concentrations of SOC C N andmacronutrients in woodland and forested areas are linked to theturnover of organic matter and not the parent material

Compared with the rock soils are enriched in most tracemetals although the Cr contents were similar This was likelyto be due to resistance of Cr-oxides to weathering and suggeststhat these phases were not likely sources for Cr in soil solutionsand plants of serpentine soils (Oze et al 2004) Bedrock to foresttransects demonstrated a strong correlation between soilmoistureand soil depth Soil nutrient content and soil moisture should beproportional to the cube of soil depth (Belcher et al 1995) thuseach of these parameters figured significantly in the PCA andRDA

Afforestation has been relatively recent in this region (Tyndall1992)Concomitantwith the increase in treedensity andsize is thecontraction of the grassland and savanna component study areafrom 18 in 1937 to 15 in 2013 which is consistent withprevious work documenting the invasion of serpentine savannasby conifers (Tyndall 1992 Barton andWallenstein 1997 Arabas2000)

In landscapeswhereC3 andC4plants coexist stableC isotopiccomposition of SOM represents a powerful method for

20

15

1

0

10

ndash10

ndash2 ndash1 0 1 2 3

05

ndash05

00

RD

A 2

RDA 1

Fig 6 Ordination diagram resulting from standard redundancy analysis (RDA) on transformed abiotic variables illustrating relationships between woodyspecies and physiochemical attributes The diamonds represent species and the vectors indicate the strength of relationship

ndash29 ndash27

(0ndash5)

(5ndash10)

(5ndash15)

(10ndash15)

Grassland (0ndash10)

Soil δ13 C (permil)

Forests

Grassland

Soi

l dep

th (

cm)

ndash25 ndash23 ndash21 ndash19 ndash17 ndash15

Fig 7 Mean carbon isotopic signature in forest (n= 96) and grassland(n= 8) soils for Cherry Hill Maryland USA Error bars are 1 se Datainterpreted as evidence for expansive C4 or C3 historic grassland Verticallines over 5ndash10- and 5ndash15-cm data indicate group pairs that were notsignificantly different using Tukeyrsquos HSD multiple comparisons

J Australian Journal of Botany J Burgess et al

reconstructing vegetation changes (Deines 1980 Tieszen andArcher 1990 Boutton and Yamasaki 1996 Boutton et al 1998)Spatial patterns of soil d13C were strongly related to type ofvegetation (primarily woody versus grassland cover) Remnantgrasslands are composed of a mixture of C3 grasses and forbsand C4 grasses It is well documented that the d13C signature ofC3 plants (ranging from26permil to28permil) is easily distinguishedfrom that of the C4 plants (ranging from 12permil to 15permil)(DeNiro and Epstein 1977 Cerling et al 1989 Tieszen 1991Ehleringer et al 2000 Fox et al 2012 Throop et al 2013) Otherprocesses besides vegetation change alter isotopic ratios (such aspreferential degradation of some plant materials microbiologicalactivity Suess effect and adsorption processes) and may causeshifts of the order of 1ndash3permil (Friedli et al 1987Becker-HeidmannandScharpenseel 1992Trolier et al 1996Ehleringer et al 2000SantRuCkova et al 2000 Krull et al 2005) d13C enrichmentin temperate forest soils larger than 3ndash4permil is generally attributedto the presence of remaining old C4-derived SOC Following thereasoning of Wynn and Bird (2008) we sampled the upper soils(A horizon) and assumed that the intrinsic metabolic processesoccurring in C3 and C4 vegetation impart an unequivocald13C signature into the SOM and that these signatures persistin decaying organic matter such that increasing depth yields achronologic sequence recording vegetation change At the studyarea d13C values of soils in the grassland matrix averaged ndash16permil(Table 4) to 26permil in forested areas dominated by C3 plantsDepth-integrated SOM d13C significantly reflected the currentvegetation cover (mixed C3ndashC4 and solely C3) and accordinglycould be used as a reference to trace past changes in vegetationrecorded in the SOM at deeper soil layers The d13C values ofSOC consistently increased with depth reflecting a higherpercentage of C4-derived carbon in the older soils that isinterpreted to reflect a site-wide shift in vegetation from C4 toC3 plants

Multiple regression and ordination results supported the roleof soil depth and incident radiation both of which influence soilmoisture content as factors governing the occurrence of xericserpentine savannaQuercus species Abiotic factors in areas withhigher importance of shade-tolerant species were associated withdeeper soils higher bulk density and higher Ca Mg ratios In oursite early successional woodlands were favoured in areas of highincident radiation and broadleaf forests were more developed onslopes containing more clay rich and denser soils Diversity ofwoody specieswas higher in areaswith deeper soilsUnlikemanysystems with heavy metals we found no correlations betweentoxic metals (Cr and Ni) and vegetation dynamics

Historical records show that oak savanna ecosystems wereimportant components of the Mid-Atlantic landscape (Shreveet al 1910) The lack of any oaks before the 1850s suggests thatthey were likely harvested for fuel posts or to support grazingFarming of these serpentinite areas is not documented and isunlikely given the inhospitable nature of the soil Our studyindicated that tree recruitment before the 20th centurywasmostlyoaks perhaps regenerating from root sprouts or propagules in theseedbankXeric oak recruitment decreasedover timeandnoxericoaks younger than 40 years established in the forested areasalthough seedlings and saplings still exist in the remnantgrasslandsavannas This failure to regenerate is likely toreflect the consequences of changing disturbance regimes suchas fire-suppression (Nuzzo 1986 Abrams 1992 2006 PetersonandReich 2001) alongwith competitionwith invasive andmesictree species (Nuzzo 1986) White-tailed deer and squirrelherbivory can reduce recruitment especially in isolatedsystems such as these serpentinite landscapes (Shea andStange 1998 Haas and Heske 2005) A turning point forcommunity development appears to be ~1900 with theencroachment of Pinus virginiana and Juniperus virginianaThese conifer species have a variety of characteristics that can

30

25

Minor species

Mesic species (A rubrum F grandifolia N sylvatica P serotina)

Conifers

Quercus spp (Mixed)

Quercus spp (Xeric)

(n = 163)15

5

20

10

Num

ber

of tr

ees

sam

pled

Decade of origin

01850 1860 1870 1880 1890 1900 1910 1920 1930 1940 1950 1960 1970 1980 1990

Fig 8 Age distribution from trees on the basis of core data (n= 163)Xeric species areQuercusmarilandica andQ stellataMixed oak species includeQ albaQ montana Q rubra and Q velutina

Conifer encroachment on ultramafics Australian Journal of Botany K

potentially alter ecosystem processes for decades as they matureand senesce Both conifers are aggressive invaders because ofrapid growth high seed production high seed-dispersalefficiency tolerance of xeric conditions and opportunistic useof ECM mycorrhizae (Ormsbee et al 1976 Holthuijzen andSharik 1984 1985 Briggs andGibson 1992Axmann andKnapp1993 Thiet and Boerner 2007) Once established these speciescan have a profound effect on soil properties as a result ofaccumulation of organic matter on the forest floor and theeventual decomposition of detritus which leads to higherwater-holding capacity and lowering of soil pH This positivefeedback between soil depth and litter cover exerted by theconifers could promote succession towards non-serpentineconditions

The initial serpentine edaphic signature acts as a filterpreventing species that belong to the regional flora but lack thetraits required to survive in such conditions to successfullycolonise these habitats By the 1970s such filters appear tohave diminished Current site characteristics have shiftedtowards more mesic conditions with changes in soil propertiessuch as bulk density soil depth pH SOM and metalsconcentration Over time the community has changed fromxeric (grassland or oak savanna) to one dominated by moremesic oaks and shade-tolerant Acer Nyssa and Fagus speciesThemajor onset ofAcer establishment began in the 1940sUnlikethePinus and Juniperus establishment pulse which proceeded inan open-canopy situation Acer dominance likely proceededunder closed canopy and canopy gaps because it is shade-tolerant One may speculate that over the past 100ndash200 years

the landscape has been dominated by Quercus with patches ofPinus and Juniperus that historicallywere present at lownumbersbecauseof disturbance (miningfire grazing logging)Hilgartneret al (2009) recorded evidence of the presence of Pinusvirginiana from the early 1800s in a serpentine barren siteonly 50 km distant Over the past 60 years Acer has becomeabundant where it along with other species currently present(N sylvatica Cornus florida and Celtis occidentalis) wouldnormally be considered an aberrant species in a hostile edaphicsetting Age and diameter structure indicated a shift away fromQuercus and towards an A rubrum-dominated landscape Thissame species shift has been widely reported throughout theeastern deciduous forests of North America (Abrams 1998Nowacki and Abrams 2008 McEwan et al 2010)

In the context of serpentinite ecosystems the influxof conifersnearly a century ago has acted as a landscapemodulator (Shachaket al 2008) The resulting build-up of SOM increases the water-holding capacity and other resources such as ground-levelradiation facilitate the influx of new species adapted to thealtered conditions As the difference between the mesicwoodlands and the xeric patches decreases the resourcecontrasts also decrease leading to increasing communitysimilarity Different oak species can grow in a wide variety oflight and moisture conditions However none of the traditionalserpentine oak species is considered shade tolerant to the samedegree as are Acer spp Fagus grandifolia and Nyssa sylvaticaWith progressive development of the soil and a dense-canopyclosure growth and germination of the traditional xeric oakspecies are likely to be suppressed marking a shift from

Colonisation

Fig 9 Conceptual model for serpentine (orangendashred colouring) habitats and species redistribution resulting in similar structure for a typical piedmontgeoecologic system (bluered colouring) Bedrock geology (specifically elemental concentrations and structures) exerts its control on soil fertility with subtlevariations resulting in markedly different above-ground and below-ground communities The change in soils over time (dsdt) equation from Huggett (1995)is a function of the various drivers including b (biology) r (topography) a (atmospheric effects) s (soils) and h (hydrology) Through changing vegetationdynamics and species redistribution the tree functional groups and forest types will see increasing similarity as a result of converging successional trajectoriesRegional species pools stochastic events and many other factors determine ultimate outcomes

L Australian Journal of Botany J Burgess et al

xerothermic to mesic andor nutrient-demanding speciesFigure 9 details a conceptual model for pedologic and ecologicchanges at serpentine sites of theMid-Atlantic with successionaltrajectories converging on a species redistribution and loss oftypical serpentine species such as Cerastium arvense varvillosissimum Andropogon gerardii and the xeric oaks(Q marilandica and Q stellata)

Although the canopy is still oak or pine dominated thetrajectory points to a maplendashbeechndashred oak-dominated forestThe Mid-Atlantic has already lost much of its originalgrasslandndashoak woodlandndashforest mosaic This state transitionmay lead to the loss of a unique ecosystem and thustaxonomic homogenisation with the regional flora Thesechanges result from anthropic influences through changes inland use and fire frequency the loss of Castanea dentata(American chestnut) and changing herbivore populationsNearly a century ago these areas gave way to a mosaic ofsmall woodlands typically characterised by xeric oaksinterspersed with shallow rooting low nutrient-tolerantconifers Released from suppressive disturbance regimes(herbivory and fire for example) Pinus virginiana andJuniperus virginiana quickly became the dominant speciesHowever our data suggest that the landscape is nowundergoing another distinct change as the once-open canopystructure has becomemore closed Edaphic properties such as soildepth chemistry and moisture that once drove communityorganisation across environmental stress gradients are nowbeing modulated by species no longer filtered by theserpentine syndrome This modulation may allow for therecruitment and establishment of drought-sensitive speciesbeyond their typical tolerance limits (Warman and Moles2008) This unique ecosystem which took centuries tomillennia to evolve appears to be fading and may soonevanesce into the past

Sources of funding

Our work was funded by the National Science Foundation(DEB-0423476 and DEB-1027188) and grants from theR Balk and D Elliott Field Fund (Johns Hopkins University)

Conflicts of interest

No conflicts of interest

Acknowledgements

We thank Drs Richard Back Naomi Levin Bruce Marsh and Ben Passey foruse of their elemental and isotopic facilities We also thank Dr David Veblenand two anonymous reviewers for numerous constructive comments on thegeology

References

Abbe C (1898) A general report on the physiography of Maryland JohnsHopkins University (Johns Hopkins University Press Baltimore MD)

Abrams MD (1992) Fire and the development of oak forests Bioscience 42346ndash353 doi1023071311781

Abrams MD (1998) The red maple paradox Bioscience 48 355ndash364doi1023071313374

AbramsMD (2006) Ecological and ecophysiological attributes and responsesto fire in eastern oak forests In lsquoFire in eastern oak forests deliveringscience to landmanagersrsquo (EdMBDickinson) pp 74ndash89 (USDA ForestService General Technical Report NRS-P-1)

Alexander EB (2009) Serpentine geoecology of the eastern and southeasternmargins of North America Northeastern Naturalist 16 223ndash252doi1016560450160518

Alexander EB Ellis CC Burke R (2007) A chronosequence of soils andvegetation on serpentine terraces in the Klamath Mountains USA SoilScience 172 565ndash576 doi101097ss0b013e31804fa22d

Arabas KB (2000) Spatial and temporal relationships among fire frequencyvegetation and soil depth in an easternNorthAmerican serpentine barrenThe Journal of the Torrey Botanical Society 127 51ndash65doi1023073088747

Archer S Schimel DS Holland EA (1995) Mechanisms of shrublandexpansion land use climate or CO2 Climatic Change 29 91ndash99doi101007BF01091640

Axmann BD Knapp AK (1993) Water relations of Juniperus virginiana andAndropogon gerardii in an unburned tallgrass prairie watershed TheSouthwestern Naturalist 38 325ndash330 doi1023073671610

Barton AM Wallenstein MD (1997) Effects of invasion of Pinus virginianaon soil properties in serpentine barrens in southeastern PennsylvaniaThe Journal of the Torrey Botanical Society 124 297ndash305doi1023072997264

Bazzaz FA (1968) Succession on abandoned fields in the Shawnee Hillssouthern Illinois Ecology 49 924 doi1023071936544

Becker-Heidmann P Scharpenseel H-W (1992) Studies of soil organicmatterdynamics using natural carbon isotopes The Science of the TotalEnvironment 117ndash118 305ndash312 doi1010160048-9697(92)90097-C

Belcher JW Keddy PA Twolan-Strutt L (1995) Root and shoot competitionintensity along a soil depth gradient Journal of Ecology 83 673ndash682doi1023072261635

BouttonTWArcher SRMidwoodAJZitzer SFBolR (1998)d13Cvalues ofsoil organic carbon and their use in documenting vegetation change in asubtropical savanna ecosystem Geoderma 82 5ndash41doi101016S0016-7061(97)00095-5

Bouyoucos GJ (1936) Directions for making mechanical analyses of soilsby the hydrometer method Soil Science 42 225ndash230doi10109700010694-193609000-00007

Briggs JMGibsonDJ (1992)Effect offire on tree spatial patterns in a tallgrassprairie landscape Bulletin of the Torrey Botanical Club 119 300ndash307doi1023072996761

BrownML Brown RG (1984) lsquoHerbaceous plants of Marylandrsquo (Universityof Maryland College Park MD)

Burgess JL Lev S Swan CM Szlavecz K (2009) Geologic and edaphiccontrols on a serpentine forest community Northeastern Naturalist 16366ndash384 doi1016560450160527

Butaye J Jacquemyn H Honnay O Hermy M (2002) The species poolconcept applied to forests in a fragmented landscape dispersal limitationversus habitat limitation Journal of Vegetation Science 13 27ndash34doi101111j1654-11032002tb02020x

Callaway RM Davis FW (1993) Vegetation dynamics fire and the physical-environment in coastal central California Rid B-7010-2009 Ecology 741567ndash1578 doi1023071940084

Carroll W D Kapeluck P R Harper R A and Van Lear D H (2002)Background paper historical overview of the southern forest landscapeand associated resources In lsquoSouthern forest resource assessmentrsquo (EdsDNWear JGGreis) pp 583ndash605 (USDepartmentofAgricultureForestService Southern Research Station Asheville NC)

Cerling TE Quade J Wang Y Bowman JR (1989) Carbon isotopes in soilsand palaeosols as ecology and palaeoecology indicators Nature 341138ndash139 doi101038341138a0

Chao A (1984) Nonparametric estimation of the number of classes in apopulation Scandinavian Journal of Statistics 11 265ndash270

ChaoA (1987) Estimating the population size for capturendashrecapture data withunequal catchability Biometrics 43 783ndash791 doi1023072531532

Chazdon RL Colwell RK Denslow JS Guariguata MR (1998) Statisticalmethods for estimating species richness of woody regeneration in primaryand secondary rain forests of Northeastern Costa Rica In lsquoForest

Conifer encroachment on ultramafics Australian Journal of Botany M

Wang (post oak) as well as more mesophytic species Largeopen-crown trees are rare more typical are contorted down-sweeping branches of Q marilandica surrounded by stand-grown trees Aerial photographs from 1937 show conifersadvancing from the dendritic drainage areas Mesophyticforests are composed of these oaks and others such as Quercusrubra L (red oak) Q alba L (white oak) Q montana Willd(chestnut oak) and Q velutina Lam (black oak) along withnumerous other species most notably Prunus spp (blackcherry wild cherry) Pinus spp (eg white pine shortleafpine) Fagus grandifolia Ehrh (American beech) Acer rubrumL (red maple) Sassafras albidum Nutt (sassafras) Populusgrandidentata Michx (bigtooth aspen) Nyssa sylvaticaMarshall (blackgum) Betula lenta L (sSweet birch) andRobinia pseudoacacia L (black locust) On exposed rocksurfaces the serpentine endemic (as reviewed by Harris andRajakaruna 2009) Cerastium velutinum var velutinum Raf(serpentine mouse eared chickweed) and other serpentinophilessuch as Symphyotrichum depauperatum Fern (serpentine aster)and saplings of Q marilandica are common

Biodiversity richness estimators Chao 1 Chao 2 and ACEestimated maximum woody-species richness between 263and 308 species The observed richness was 24 species(Table 3) Overall diversity and mean diversity for the 32 siteswere calculated as 274 (X = 167 s= 029) (Shannon index) and

092 (X = 078 s= 008) (Simpson index) On the basis of relativeIV the most important species were P virginiana andJ virginiana (Table 3) These two species had similar relativeIVs and stood out from all other species A second group ofspecies included A rubrum and Q marilandica At the genuslevel Quercus was the most important contributor to the treecommunity with 39 relative IV On the basis of basal area themost dominant species were Q rubra P virginiana andQ marilandica Constancy values (Table 3) showedA rubrum (078) as a clear break from the next grouping ofspecies (P virginiana and J virginiana) whose constancy valueshovered near 05 followed by Q marilandica F grandifoliaN sylvatica and Q stellate with constancy values near 04

The Mantel test comparing floristic similarity andenvironmental distance showed significant (P lt 005)correlations between environmental distance and ecologicaldistance for the following parameters heat load bulk densitybasal area diversity (H) percentage sand Ni and Cr(Table 4) The ANOSIM statistic was calculated only for thecanopy category (woodlandforest) with a resulting r= 014 andP-value of 005 Both the Mantel and ANOSIM tests yieldedsignificant (Plt 005) but weak correlations

Relationships between vegetation and abiotic properties

Results of multiple regression analysis were evaluated tounderstand the response of species distributions to soilproperties Table 5 shows the regression results for the eightspecies the importance of which met or exceeded 5 withtopographic and edaphic parameters as the predictor variablesPredictor variables that indicated a meaningful contribution toexplaining species abundancewere incident radiation and soilMgcontent for Q marilandica incident radiation for J virginianaand Q stellata soil organic carbon (SOC) for A rubrum soildepth for F grandifolia and bulk density for the conifers

PCA was performed for 16 abiotic parameters across the 32sampled quadrats so as to identify covariation between the solarradiation and soil variables (Fig 5) Themain trend of variation in

Table 2 Descriptive statistics for soil d13C (permil) and soil organic matterin grassland and woodland elements

Means followed by the same letter are not significantly different (ANOVApost hoc Tukeyrsquos HSD P = 005)

Statistic Grassland Forest(0ndash5 cm)

Forest(5ndash10 cm)

Forest(5ndash15 cm)

Forest(10ndash15 cm)

Mean ndash1623a ndash2623b ndash2359bcd ndash2298de ndash2095eMinimum ndash1791 ndash2923 ndash2725 ndash2649 ndash2514Maximum ndash1423 ndash1928 ndash1811 ndash1972 ndash1654se 068 044 067 065 060

Table 3 Constancy importance (average value of a species when present) density dominance andimportance value (average of relative density and relative dominance) measures for woody stems in

serpentinite woodland at Cherry Hill Maryland USA

Species Constancy Importance Dominance(m2 handash1)

Density(stems handash1)

Importancevalue (IV) ()

Acer rubrum L 078 109 5799 7778 988Ailanthus altissima Mill 003 003 075 222 027Betula lenta L 006 006 346 444 060Celtis occidentalis L 006 006 059 444 041Cercis canadensis L 003 003 078 222 022Cornus florida L 006 006 146 444 038Diospyros virginiana L 003 003 185 222 024Fagus grandifolia Ehrh 043 065 5713 4667 619Juniperus virginiana L 050 106 8657 7556 1214Kalmia latifolia L 012 015 163 1111 106Liriodendron tulipifera L 009 009 2372 667 108Nyssa sylvatica Marsh 043 075 5769 5333 720Pinus strobus L 003 003 692 222 090Pinus virginiana Mill 056 078 16498 5556 1343Populus grandidentata Michx 003 006 640 444 058Prunus serotina Ehrh 028 028 2046 2000 271

Conifer encroachment on ultramafics Australian Journal of Botany G

the soil chemical properties and topographic variables asreflected by the PCA Axis 1 (accounting for 30 of variance)was one dominated by significant correlations of soil texture(sand and bulk density) soil depth Mg Cr and Niconcentrations incident radiation and heat load The secondtrend of soil variation as reflected by the PCA Axis 2 (17 ofvariation) can be interpreted as a combined gradient of essentialnutrients (C andN) and SOC The PCA (axes accounting for 44of the variation) revealed strong positive associations betweenboth soil depth and bulk density which were negativelycorrelated with heavy metal concentrations and pH SoilCa Mg ratios and Al concentrations were negativelycorrelated with macronutrients (C N P) and SOC Fig 5bshows the PCA factor map for the abiotic variables and thebiotic covariates for the sampling plots When thecharacteristics of woody-species community are included inthe PCA both basal area and stem density are positivelycorrelated with soil Ca Mg ratios and Al content Diversity(Shannonrsquos index) was strongly correlated with soil depth andbulkdensity andnegatively related toheavymetal concentrations

and heat load As individual-species basal area stem densityand diversity typically respond to environmental gradientsconstrained ordination utilising RDA was used to inferrelationships between the species matrix and theenvironmental matrix using Euclidean distance between sitesArrows of increasing species abundance represent RDA scoresThe sum of the RDA eigenvalues for the first two axes using allabiotic explanatory variables accounted for 368 of the woody-species composition After reduced variable selection based onANOVA tests of the ordinationmodel seven termswere selectedand these collectively explained 271 of the variance (Fig 6)Interestingly those species whose importance was gt4 arepartitioned into the four quadrants of the ordination From theRDAAxes1and2biplot it canbe seen that traditional serpentine-savanna (shade-intolerant)woody species (S albidumQ stellataand Q marilandica) are positively associated with increasingincident radiation and negatively associated with increasingsoil depth Al and Mg soil concentrations Post-disturbanceinitial colonisers (J virginiana and P virginiana) are stronglynegatively correlated with bulk densityQ alba andQ montanawhich are widely distributed across a variety of conditionsfrom wet mesic to subxeric are associated with increasingbulk density Shade-tolerant species such as N sylvaticaA rubrum and F grandifolia are correlated with increasingsoil depth and soils with more Ca and Al

Soils at the site differed in the total amount verticaldistribution and isotopic composition of SOM (Tables 1 2)ANOVA results indicated significant differences in theisotopic means according to their depth (F487 = 276P 00001) Tukeyrsquos HSD test showed that the mean d13C ofgrassland soils (ndash162 07permil) was significantly higher thanthat of shallow forested soils (ndash262 04permil) (Table 5)Shallow forested soils were significantly different from deepersoils (ndash2095 06permil) Total soil C and N and soil C N declinedwith soil depth C N ratios were significantly different for alldepths ranging from 1518 088 on the surface soils to1162 1 in soils 10ndash15 cm below surface (F266 = 59

Table 4 Effects of environmental distance on similarity between plotsMantel statistic calculated by comparing matrix of environmental distancewith BrayndashCurtis distance The standardised Mantel statistic (r) indicatesthe strength of the relationship between sample-point position and forest-floor thickness and the associated P-value indicates the significance of

that relationship

Environmental indicator Correlation coefficient (r) P-value

Heat load 0126 0009Bulk density 0236 0003Basal area 0141 0015Diversity (H) 0109 0034Percentage sand 0203 0002Nickel 0111 0034Chromium 0159 0009Canopy 0150 0009

Table 5 Results of multiple regression with list of factors significantly contributing to species abundanceSOC soil organic carbon Plt 0001 P lt 001 Plt 005 lsquorsquo Plt 01

Pvirginiana J virginiana A rubrum Q marilandica Q rubra N sylvatica Q alba Q stellata F grandifolia Q montana

IV () 1343 1214 988 940 810 720 703 626 619 440Incident radiation 09247 00249 01791 00008 04332 02136 09123 00450 03030 03225Heat load 02516 04869 03661 08494 09552 09203 09825 02522 01128 03502Soil depth 02599 03110 08937 00463 01881 03041 03381 01689 00242 09920Bulk density 00023 00037 08260 08860 05216 02157 00415 02686 09894 01171Texture ( sand) 00976 02800 02942 06247 08124 09313 03207 03451 01454 06054Calcium magnesium 09840 02962 07824 07031 09971 04587 07945 05441 06414 03639Magnesium 06146 05570 05725 00019 01500 06177 00914 03238 07491 03623Aluminium 07988 07323 01491 09176 00673 00583 08508 04152 05380 08685Chromium 05879 02397 02066 01012 06214 09977 06126 03991 06870 00434Nickel 08699 02237 07478 03211 08340 06190 09519 05875 06307 01074Phosphorus 05192 05066 07822 09800 07523 06641 04927 09619 09430 02268SOC 06435 01185 00298 00053 09782 02790 05185 02333 08701 07036Carbon nitrogen 05806 02325 06793 08013 05674 06443 01727 01105 08015 06972Carbon 07737 07963 02099 01231 04380 08148 04280 04885 07995 05145Nitrogen 07506 09537 01179 03581 06363 05329 03124 07375 00153 05744pH 07946 01863 05144 01547 00176 04256 02597 06743 01299 09047Multiple R2 059 068 057 079 057 046 051 055 062 056

H Australian Journal of Botany J Burgess et al

P = 0004) There were significant vegetation depth andvegetation depth effects on d13C of SOM (Table 5 Fig 7)

Age structure

In total 163 core samples were evaluated with results displayedin the stand-age structure plot (Fig 8) Skeleton plot chronologiesvaried among species Older minimum age at breast height wasgenerally seen inQ stellata (114 15 years) andQ marilandica(110 7 years) and the younger chronologies developed inspecies such as A rubrum (53 4 years) F grandifolia(56 8 years) and L tulipifera (43 8 years) Only xeric oakswere present in the oldest age classes The period before 1900 iscomposed of 54 xeric oaks and when all oaks are grouped theycomprise 79 of trees from that era although Q montana isabsent from this group (oldest minimum age is 88 years) Theoldest non-oak was N sylvatica (144 years) The period between

1900 and 1940 records a marked increase in conifers andconcomitant decrease of xeric oaks comprising 39 and 19of all species respectively Between 1950 and 1990 thepercentage of xeric oaks decreased to 3 and conifers to 13whereas Acer spp increased to 19 The number of shade-tolerant species rose from 125 in the decades from 1850 to1900 to 59 during the interval from 1950 to 1990

The age structure pooled from all the plots (Fig 8) showeddiscontinuous patterns with prominent peaks of regenerationpulses and periods of either low or absent recruitment Coniferestablishment and recruitment clearly increased during1900ndash1960 and then receded through the rest of the samplingperiod There was a conspicuous absence of younger xeric oaks(post-1970s) and only two xeric oaks from1950 to the present Atthe same time the percentage of non-oak deciduous treesincreased to 68 post-1950

10

05

00

Dim

2 (

171

7)

Dim

2 (

229

1)

Dim 1 (3038)

Dim 1 (402)

ndash10

ndash05

4

2

0

ndash2

ndash4

ndash6

ndash15

ndash15 ndash10 ndash5 0 5 10

ndash10 ndash05 00 05 10 15

(a)

(b)

Fig 5 (a) Unconstrained principal component analysis (PCA) correlation circle of pedological topological and community variables (b) Forest communityclustering along an edaphic gradient

Conifer encroachment on ultramafics Australian Journal of Botany I

Discussion

The present study examined a moderately sized area in the Mid-Atlantic but the changes ndash encroachment and mesophication ndash

represent potential widespread conservation problems Theserpentinite landscapes were once common components of theancient Piedmont (Carroll et al 2002 Owen 2002 Noss 2013)These unique plant communities have existed as persistentlsquoislandsrsquo embedded in a deciduous forest region for centuriesor longer We documented a state transition from grasslandand xeric oak savanna to conifer-dominated communities

that have given way to eastern broadleaf forests Comparisonsamong geologic pedologic and vegetation properties reflect thischange and indicate that the conversion to forest has been rapidand characterised by large changes in soil properties (pH water-holding capacity SOMcomposition) aswell as plant-communitystructure (physiognomyand stemdensity) and composition (xericto more mesic communities)

Thin-section and petrologic data confirmed that the geologicsubstrate is a serpentinised peridotitewith characteristics typical ofserpentinite terranes in the Appalachians (Misra and Keller 1978Mittwede and Stoddard 1989 Gates 1992 Alexander 2009)Although Ca is depleted in bulk rock samples because of thebreakdownof augite and interstitial plagioclase soil Ca Mg ratiosare higher in surface soils The concentrations of SOC C N andmacronutrients in woodland and forested areas are linked to theturnover of organic matter and not the parent material

Compared with the rock soils are enriched in most tracemetals although the Cr contents were similar This was likelyto be due to resistance of Cr-oxides to weathering and suggeststhat these phases were not likely sources for Cr in soil solutionsand plants of serpentine soils (Oze et al 2004) Bedrock to foresttransects demonstrated a strong correlation between soilmoistureand soil depth Soil nutrient content and soil moisture should beproportional to the cube of soil depth (Belcher et al 1995) thuseach of these parameters figured significantly in the PCA andRDA

Afforestation has been relatively recent in this region (Tyndall1992)Concomitantwith the increase in treedensity andsize is thecontraction of the grassland and savanna component study areafrom 18 in 1937 to 15 in 2013 which is consistent withprevious work documenting the invasion of serpentine savannasby conifers (Tyndall 1992 Barton andWallenstein 1997 Arabas2000)

In landscapeswhereC3 andC4plants coexist stableC isotopiccomposition of SOM represents a powerful method for

20

15

1

0

10

ndash10

ndash2 ndash1 0 1 2 3

05

ndash05

00

RD

A 2

RDA 1

Fig 6 Ordination diagram resulting from standard redundancy analysis (RDA) on transformed abiotic variables illustrating relationships between woodyspecies and physiochemical attributes The diamonds represent species and the vectors indicate the strength of relationship

ndash29 ndash27

(0ndash5)

(5ndash10)

(5ndash15)

(10ndash15)

Grassland (0ndash10)

Soil δ13 C (permil)

Forests

Grassland

Soi

l dep

th (

cm)

ndash25 ndash23 ndash21 ndash19 ndash17 ndash15

Fig 7 Mean carbon isotopic signature in forest (n= 96) and grassland(n= 8) soils for Cherry Hill Maryland USA Error bars are 1 se Datainterpreted as evidence for expansive C4 or C3 historic grassland Verticallines over 5ndash10- and 5ndash15-cm data indicate group pairs that were notsignificantly different using Tukeyrsquos HSD multiple comparisons

J Australian Journal of Botany J Burgess et al

reconstructing vegetation changes (Deines 1980 Tieszen andArcher 1990 Boutton and Yamasaki 1996 Boutton et al 1998)Spatial patterns of soil d13C were strongly related to type ofvegetation (primarily woody versus grassland cover) Remnantgrasslands are composed of a mixture of C3 grasses and forbsand C4 grasses It is well documented that the d13C signature ofC3 plants (ranging from26permil to28permil) is easily distinguishedfrom that of the C4 plants (ranging from 12permil to 15permil)(DeNiro and Epstein 1977 Cerling et al 1989 Tieszen 1991Ehleringer et al 2000 Fox et al 2012 Throop et al 2013) Otherprocesses besides vegetation change alter isotopic ratios (such aspreferential degradation of some plant materials microbiologicalactivity Suess effect and adsorption processes) and may causeshifts of the order of 1ndash3permil (Friedli et al 1987Becker-HeidmannandScharpenseel 1992Trolier et al 1996Ehleringer et al 2000SantRuCkova et al 2000 Krull et al 2005) d13C enrichmentin temperate forest soils larger than 3ndash4permil is generally attributedto the presence of remaining old C4-derived SOC Following thereasoning of Wynn and Bird (2008) we sampled the upper soils(A horizon) and assumed that the intrinsic metabolic processesoccurring in C3 and C4 vegetation impart an unequivocald13C signature into the SOM and that these signatures persistin decaying organic matter such that increasing depth yields achronologic sequence recording vegetation change At the studyarea d13C values of soils in the grassland matrix averaged ndash16permil(Table 4) to 26permil in forested areas dominated by C3 plantsDepth-integrated SOM d13C significantly reflected the currentvegetation cover (mixed C3ndashC4 and solely C3) and accordinglycould be used as a reference to trace past changes in vegetationrecorded in the SOM at deeper soil layers The d13C values ofSOC consistently increased with depth reflecting a higherpercentage of C4-derived carbon in the older soils that isinterpreted to reflect a site-wide shift in vegetation from C4 toC3 plants

Multiple regression and ordination results supported the roleof soil depth and incident radiation both of which influence soilmoisture content as factors governing the occurrence of xericserpentine savannaQuercus species Abiotic factors in areas withhigher importance of shade-tolerant species were associated withdeeper soils higher bulk density and higher Ca Mg ratios In oursite early successional woodlands were favoured in areas of highincident radiation and broadleaf forests were more developed onslopes containing more clay rich and denser soils Diversity ofwoody specieswas higher in areaswith deeper soilsUnlikemanysystems with heavy metals we found no correlations betweentoxic metals (Cr and Ni) and vegetation dynamics

Historical records show that oak savanna ecosystems wereimportant components of the Mid-Atlantic landscape (Shreveet al 1910) The lack of any oaks before the 1850s suggests thatthey were likely harvested for fuel posts or to support grazingFarming of these serpentinite areas is not documented and isunlikely given the inhospitable nature of the soil Our studyindicated that tree recruitment before the 20th centurywasmostlyoaks perhaps regenerating from root sprouts or propagules in theseedbankXeric oak recruitment decreasedover timeandnoxericoaks younger than 40 years established in the forested areasalthough seedlings and saplings still exist in the remnantgrasslandsavannas This failure to regenerate is likely toreflect the consequences of changing disturbance regimes suchas fire-suppression (Nuzzo 1986 Abrams 1992 2006 PetersonandReich 2001) alongwith competitionwith invasive andmesictree species (Nuzzo 1986) White-tailed deer and squirrelherbivory can reduce recruitment especially in isolatedsystems such as these serpentinite landscapes (Shea andStange 1998 Haas and Heske 2005) A turning point forcommunity development appears to be ~1900 with theencroachment of Pinus virginiana and Juniperus virginianaThese conifer species have a variety of characteristics that can

30

25

Minor species

Mesic species (A rubrum F grandifolia N sylvatica P serotina)

Conifers

Quercus spp (Mixed)

Quercus spp (Xeric)

(n = 163)15

5

20

10

Num

ber

of tr

ees

sam

pled

Decade of origin

01850 1860 1870 1880 1890 1900 1910 1920 1930 1940 1950 1960 1970 1980 1990

Fig 8 Age distribution from trees on the basis of core data (n= 163)Xeric species areQuercusmarilandica andQ stellataMixed oak species includeQ albaQ montana Q rubra and Q velutina

Conifer encroachment on ultramafics Australian Journal of Botany K

potentially alter ecosystem processes for decades as they matureand senesce Both conifers are aggressive invaders because ofrapid growth high seed production high seed-dispersalefficiency tolerance of xeric conditions and opportunistic useof ECM mycorrhizae (Ormsbee et al 1976 Holthuijzen andSharik 1984 1985 Briggs andGibson 1992Axmann andKnapp1993 Thiet and Boerner 2007) Once established these speciescan have a profound effect on soil properties as a result ofaccumulation of organic matter on the forest floor and theeventual decomposition of detritus which leads to higherwater-holding capacity and lowering of soil pH This positivefeedback between soil depth and litter cover exerted by theconifers could promote succession towards non-serpentineconditions

The initial serpentine edaphic signature acts as a filterpreventing species that belong to the regional flora but lack thetraits required to survive in such conditions to successfullycolonise these habitats By the 1970s such filters appear tohave diminished Current site characteristics have shiftedtowards more mesic conditions with changes in soil propertiessuch as bulk density soil depth pH SOM and metalsconcentration Over time the community has changed fromxeric (grassland or oak savanna) to one dominated by moremesic oaks and shade-tolerant Acer Nyssa and Fagus speciesThemajor onset ofAcer establishment began in the 1940sUnlikethePinus and Juniperus establishment pulse which proceeded inan open-canopy situation Acer dominance likely proceededunder closed canopy and canopy gaps because it is shade-tolerant One may speculate that over the past 100ndash200 years

the landscape has been dominated by Quercus with patches ofPinus and Juniperus that historicallywere present at lownumbersbecauseof disturbance (miningfire grazing logging)Hilgartneret al (2009) recorded evidence of the presence of Pinusvirginiana from the early 1800s in a serpentine barren siteonly 50 km distant Over the past 60 years Acer has becomeabundant where it along with other species currently present(N sylvatica Cornus florida and Celtis occidentalis) wouldnormally be considered an aberrant species in a hostile edaphicsetting Age and diameter structure indicated a shift away fromQuercus and towards an A rubrum-dominated landscape Thissame species shift has been widely reported throughout theeastern deciduous forests of North America (Abrams 1998Nowacki and Abrams 2008 McEwan et al 2010)

In the context of serpentinite ecosystems the influxof conifersnearly a century ago has acted as a landscapemodulator (Shachaket al 2008) The resulting build-up of SOM increases the water-holding capacity and other resources such as ground-levelradiation facilitate the influx of new species adapted to thealtered conditions As the difference between the mesicwoodlands and the xeric patches decreases the resourcecontrasts also decrease leading to increasing communitysimilarity Different oak species can grow in a wide variety oflight and moisture conditions However none of the traditionalserpentine oak species is considered shade tolerant to the samedegree as are Acer spp Fagus grandifolia and Nyssa sylvaticaWith progressive development of the soil and a dense-canopyclosure growth and germination of the traditional xeric oakspecies are likely to be suppressed marking a shift from

Colonisation

Fig 9 Conceptual model for serpentine (orangendashred colouring) habitats and species redistribution resulting in similar structure for a typical piedmontgeoecologic system (bluered colouring) Bedrock geology (specifically elemental concentrations and structures) exerts its control on soil fertility with subtlevariations resulting in markedly different above-ground and below-ground communities The change in soils over time (dsdt) equation from Huggett (1995)is a function of the various drivers including b (biology) r (topography) a (atmospheric effects) s (soils) and h (hydrology) Through changing vegetationdynamics and species redistribution the tree functional groups and forest types will see increasing similarity as a result of converging successional trajectoriesRegional species pools stochastic events and many other factors determine ultimate outcomes

L Australian Journal of Botany J Burgess et al

xerothermic to mesic andor nutrient-demanding speciesFigure 9 details a conceptual model for pedologic and ecologicchanges at serpentine sites of theMid-Atlantic with successionaltrajectories converging on a species redistribution and loss oftypical serpentine species such as Cerastium arvense varvillosissimum Andropogon gerardii and the xeric oaks(Q marilandica and Q stellata)

Although the canopy is still oak or pine dominated thetrajectory points to a maplendashbeechndashred oak-dominated forestThe Mid-Atlantic has already lost much of its originalgrasslandndashoak woodlandndashforest mosaic This state transitionmay lead to the loss of a unique ecosystem and thustaxonomic homogenisation with the regional flora Thesechanges result from anthropic influences through changes inland use and fire frequency the loss of Castanea dentata(American chestnut) and changing herbivore populationsNearly a century ago these areas gave way to a mosaic ofsmall woodlands typically characterised by xeric oaksinterspersed with shallow rooting low nutrient-tolerantconifers Released from suppressive disturbance regimes(herbivory and fire for example) Pinus virginiana andJuniperus virginiana quickly became the dominant speciesHowever our data suggest that the landscape is nowundergoing another distinct change as the once-open canopystructure has becomemore closed Edaphic properties such as soildepth chemistry and moisture that once drove communityorganisation across environmental stress gradients are nowbeing modulated by species no longer filtered by theserpentine syndrome This modulation may allow for therecruitment and establishment of drought-sensitive speciesbeyond their typical tolerance limits (Warman and Moles2008) This unique ecosystem which took centuries tomillennia to evolve appears to be fading and may soonevanesce into the past

Sources of funding

Our work was funded by the National Science Foundation(DEB-0423476 and DEB-1027188) and grants from theR Balk and D Elliott Field Fund (Johns Hopkins University)

Conflicts of interest

No conflicts of interest

Acknowledgements

We thank Drs Richard Back Naomi Levin Bruce Marsh and Ben Passey foruse of their elemental and isotopic facilities We also thank Dr David Veblenand two anonymous reviewers for numerous constructive comments on thegeology

References

Abbe C (1898) A general report on the physiography of Maryland JohnsHopkins University (Johns Hopkins University Press Baltimore MD)

Abrams MD (1992) Fire and the development of oak forests Bioscience 42346ndash353 doi1023071311781

Abrams MD (1998) The red maple paradox Bioscience 48 355ndash364doi1023071313374

AbramsMD (2006) Ecological and ecophysiological attributes and responsesto fire in eastern oak forests In lsquoFire in eastern oak forests deliveringscience to landmanagersrsquo (EdMBDickinson) pp 74ndash89 (USDA ForestService General Technical Report NRS-P-1)

Alexander EB (2009) Serpentine geoecology of the eastern and southeasternmargins of North America Northeastern Naturalist 16 223ndash252doi1016560450160518

Alexander EB Ellis CC Burke R (2007) A chronosequence of soils andvegetation on serpentine terraces in the Klamath Mountains USA SoilScience 172 565ndash576 doi101097ss0b013e31804fa22d

Arabas KB (2000) Spatial and temporal relationships among fire frequencyvegetation and soil depth in an easternNorthAmerican serpentine barrenThe Journal of the Torrey Botanical Society 127 51ndash65doi1023073088747

Archer S Schimel DS Holland EA (1995) Mechanisms of shrublandexpansion land use climate or CO2 Climatic Change 29 91ndash99doi101007BF01091640

Axmann BD Knapp AK (1993) Water relations of Juniperus virginiana andAndropogon gerardii in an unburned tallgrass prairie watershed TheSouthwestern Naturalist 38 325ndash330 doi1023073671610

Barton AM Wallenstein MD (1997) Effects of invasion of Pinus virginianaon soil properties in serpentine barrens in southeastern PennsylvaniaThe Journal of the Torrey Botanical Society 124 297ndash305doi1023072997264

Bazzaz FA (1968) Succession on abandoned fields in the Shawnee Hillssouthern Illinois Ecology 49 924 doi1023071936544

Becker-Heidmann P Scharpenseel H-W (1992) Studies of soil organicmatterdynamics using natural carbon isotopes The Science of the TotalEnvironment 117ndash118 305ndash312 doi1010160048-9697(92)90097-C

Belcher JW Keddy PA Twolan-Strutt L (1995) Root and shoot competitionintensity along a soil depth gradient Journal of Ecology 83 673ndash682doi1023072261635

BouttonTWArcher SRMidwoodAJZitzer SFBolR (1998)d13Cvalues ofsoil organic carbon and their use in documenting vegetation change in asubtropical savanna ecosystem Geoderma 82 5ndash41doi101016S0016-7061(97)00095-5

Bouyoucos GJ (1936) Directions for making mechanical analyses of soilsby the hydrometer method Soil Science 42 225ndash230doi10109700010694-193609000-00007

Briggs JMGibsonDJ (1992)Effect offire on tree spatial patterns in a tallgrassprairie landscape Bulletin of the Torrey Botanical Club 119 300ndash307doi1023072996761

BrownML Brown RG (1984) lsquoHerbaceous plants of Marylandrsquo (Universityof Maryland College Park MD)

Burgess JL Lev S Swan CM Szlavecz K (2009) Geologic and edaphiccontrols on a serpentine forest community Northeastern Naturalist 16366ndash384 doi1016560450160527

Butaye J Jacquemyn H Honnay O Hermy M (2002) The species poolconcept applied to forests in a fragmented landscape dispersal limitationversus habitat limitation Journal of Vegetation Science 13 27ndash34doi101111j1654-11032002tb02020x

Callaway RM Davis FW (1993) Vegetation dynamics fire and the physical-environment in coastal central California Rid B-7010-2009 Ecology 741567ndash1578 doi1023071940084

Carroll W D Kapeluck P R Harper R A and Van Lear D H (2002)Background paper historical overview of the southern forest landscapeand associated resources In lsquoSouthern forest resource assessmentrsquo (EdsDNWear JGGreis) pp 583ndash605 (USDepartmentofAgricultureForestService Southern Research Station Asheville NC)

Cerling TE Quade J Wang Y Bowman JR (1989) Carbon isotopes in soilsand palaeosols as ecology and palaeoecology indicators Nature 341138ndash139 doi101038341138a0

Chao A (1984) Nonparametric estimation of the number of classes in apopulation Scandinavian Journal of Statistics 11 265ndash270

ChaoA (1987) Estimating the population size for capturendashrecapture data withunequal catchability Biometrics 43 783ndash791 doi1023072531532

Chazdon RL Colwell RK Denslow JS Guariguata MR (1998) Statisticalmethods for estimating species richness of woody regeneration in primaryand secondary rain forests of Northeastern Costa Rica In lsquoForest

Conifer encroachment on ultramafics Australian Journal of Botany M

the soil chemical properties and topographic variables asreflected by the PCA Axis 1 (accounting for 30 of variance)was one dominated by significant correlations of soil texture(sand and bulk density) soil depth Mg Cr and Niconcentrations incident radiation and heat load The secondtrend of soil variation as reflected by the PCA Axis 2 (17 ofvariation) can be interpreted as a combined gradient of essentialnutrients (C andN) and SOC The PCA (axes accounting for 44of the variation) revealed strong positive associations betweenboth soil depth and bulk density which were negativelycorrelated with heavy metal concentrations and pH SoilCa Mg ratios and Al concentrations were negativelycorrelated with macronutrients (C N P) and SOC Fig 5bshows the PCA factor map for the abiotic variables and thebiotic covariates for the sampling plots When thecharacteristics of woody-species community are included inthe PCA both basal area and stem density are positivelycorrelated with soil Ca Mg ratios and Al content Diversity(Shannonrsquos index) was strongly correlated with soil depth andbulkdensity andnegatively related toheavymetal concentrations

and heat load As individual-species basal area stem densityand diversity typically respond to environmental gradientsconstrained ordination utilising RDA was used to inferrelationships between the species matrix and theenvironmental matrix using Euclidean distance between sitesArrows of increasing species abundance represent RDA scoresThe sum of the RDA eigenvalues for the first two axes using allabiotic explanatory variables accounted for 368 of the woody-species composition After reduced variable selection based onANOVA tests of the ordinationmodel seven termswere selectedand these collectively explained 271 of the variance (Fig 6)Interestingly those species whose importance was gt4 arepartitioned into the four quadrants of the ordination From theRDAAxes1and2biplot it canbe seen that traditional serpentine-savanna (shade-intolerant)woody species (S albidumQ stellataand Q marilandica) are positively associated with increasingincident radiation and negatively associated with increasingsoil depth Al and Mg soil concentrations Post-disturbanceinitial colonisers (J virginiana and P virginiana) are stronglynegatively correlated with bulk densityQ alba andQ montanawhich are widely distributed across a variety of conditionsfrom wet mesic to subxeric are associated with increasingbulk density Shade-tolerant species such as N sylvaticaA rubrum and F grandifolia are correlated with increasingsoil depth and soils with more Ca and Al

Soils at the site differed in the total amount verticaldistribution and isotopic composition of SOM (Tables 1 2)ANOVA results indicated significant differences in theisotopic means according to their depth (F487 = 276P 00001) Tukeyrsquos HSD test showed that the mean d13C ofgrassland soils (ndash162 07permil) was significantly higher thanthat of shallow forested soils (ndash262 04permil) (Table 5)Shallow forested soils were significantly different from deepersoils (ndash2095 06permil) Total soil C and N and soil C N declinedwith soil depth C N ratios were significantly different for alldepths ranging from 1518 088 on the surface soils to1162 1 in soils 10ndash15 cm below surface (F266 = 59

Table 4 Effects of environmental distance on similarity between plotsMantel statistic calculated by comparing matrix of environmental distancewith BrayndashCurtis distance The standardised Mantel statistic (r) indicatesthe strength of the relationship between sample-point position and forest-floor thickness and the associated P-value indicates the significance of

that relationship

Environmental indicator Correlation coefficient (r) P-value

Heat load 0126 0009Bulk density 0236 0003Basal area 0141 0015Diversity (H) 0109 0034Percentage sand 0203 0002Nickel 0111 0034Chromium 0159 0009Canopy 0150 0009

Table 5 Results of multiple regression with list of factors significantly contributing to species abundanceSOC soil organic carbon Plt 0001 P lt 001 Plt 005 lsquorsquo Plt 01

Pvirginiana J virginiana A rubrum Q marilandica Q rubra N sylvatica Q alba Q stellata F grandifolia Q montana

IV () 1343 1214 988 940 810 720 703 626 619 440Incident radiation 09247 00249 01791 00008 04332 02136 09123 00450 03030 03225Heat load 02516 04869 03661 08494 09552 09203 09825 02522 01128 03502Soil depth 02599 03110 08937 00463 01881 03041 03381 01689 00242 09920Bulk density 00023 00037 08260 08860 05216 02157 00415 02686 09894 01171Texture ( sand) 00976 02800 02942 06247 08124 09313 03207 03451 01454 06054Calcium magnesium 09840 02962 07824 07031 09971 04587 07945 05441 06414 03639Magnesium 06146 05570 05725 00019 01500 06177 00914 03238 07491 03623Aluminium 07988 07323 01491 09176 00673 00583 08508 04152 05380 08685Chromium 05879 02397 02066 01012 06214 09977 06126 03991 06870 00434Nickel 08699 02237 07478 03211 08340 06190 09519 05875 06307 01074Phosphorus 05192 05066 07822 09800 07523 06641 04927 09619 09430 02268SOC 06435 01185 00298 00053 09782 02790 05185 02333 08701 07036Carbon nitrogen 05806 02325 06793 08013 05674 06443 01727 01105 08015 06972Carbon 07737 07963 02099 01231 04380 08148 04280 04885 07995 05145Nitrogen 07506 09537 01179 03581 06363 05329 03124 07375 00153 05744pH 07946 01863 05144 01547 00176 04256 02597 06743 01299 09047Multiple R2 059 068 057 079 057 046 051 055 062 056

H Australian Journal of Botany J Burgess et al

P = 0004) There were significant vegetation depth andvegetation depth effects on d13C of SOM (Table 5 Fig 7)

Age structure

In total 163 core samples were evaluated with results displayedin the stand-age structure plot (Fig 8) Skeleton plot chronologiesvaried among species Older minimum age at breast height wasgenerally seen inQ stellata (114 15 years) andQ marilandica(110 7 years) and the younger chronologies developed inspecies such as A rubrum (53 4 years) F grandifolia(56 8 years) and L tulipifera (43 8 years) Only xeric oakswere present in the oldest age classes The period before 1900 iscomposed of 54 xeric oaks and when all oaks are grouped theycomprise 79 of trees from that era although Q montana isabsent from this group (oldest minimum age is 88 years) Theoldest non-oak was N sylvatica (144 years) The period between

1900 and 1940 records a marked increase in conifers andconcomitant decrease of xeric oaks comprising 39 and 19of all species respectively Between 1950 and 1990 thepercentage of xeric oaks decreased to 3 and conifers to 13whereas Acer spp increased to 19 The number of shade-tolerant species rose from 125 in the decades from 1850 to1900 to 59 during the interval from 1950 to 1990

The age structure pooled from all the plots (Fig 8) showeddiscontinuous patterns with prominent peaks of regenerationpulses and periods of either low or absent recruitment Coniferestablishment and recruitment clearly increased during1900ndash1960 and then receded through the rest of the samplingperiod There was a conspicuous absence of younger xeric oaks(post-1970s) and only two xeric oaks from1950 to the present Atthe same time the percentage of non-oak deciduous treesincreased to 68 post-1950

10

05

00

Dim

2 (

171

7)

Dim

2 (

229

1)

Dim 1 (3038)

Dim 1 (402)

ndash10

ndash05

4

2

0

ndash2

ndash4

ndash6

ndash15

ndash15 ndash10 ndash5 0 5 10

ndash10 ndash05 00 05 10 15

(a)

(b)

Fig 5 (a) Unconstrained principal component analysis (PCA) correlation circle of pedological topological and community variables (b) Forest communityclustering along an edaphic gradient

Conifer encroachment on ultramafics Australian Journal of Botany I

Discussion

The present study examined a moderately sized area in the Mid-Atlantic but the changes ndash encroachment and mesophication ndash

represent potential widespread conservation problems Theserpentinite landscapes were once common components of theancient Piedmont (Carroll et al 2002 Owen 2002 Noss 2013)These unique plant communities have existed as persistentlsquoislandsrsquo embedded in a deciduous forest region for centuriesor longer We documented a state transition from grasslandand xeric oak savanna to conifer-dominated communities

that have given way to eastern broadleaf forests Comparisonsamong geologic pedologic and vegetation properties reflect thischange and indicate that the conversion to forest has been rapidand characterised by large changes in soil properties (pH water-holding capacity SOMcomposition) aswell as plant-communitystructure (physiognomyand stemdensity) and composition (xericto more mesic communities)

Thin-section and petrologic data confirmed that the geologicsubstrate is a serpentinised peridotitewith characteristics typical ofserpentinite terranes in the Appalachians (Misra and Keller 1978Mittwede and Stoddard 1989 Gates 1992 Alexander 2009)Although Ca is depleted in bulk rock samples because of thebreakdownof augite and interstitial plagioclase soil Ca Mg ratiosare higher in surface soils The concentrations of SOC C N andmacronutrients in woodland and forested areas are linked to theturnover of organic matter and not the parent material

Compared with the rock soils are enriched in most tracemetals although the Cr contents were similar This was likelyto be due to resistance of Cr-oxides to weathering and suggeststhat these phases were not likely sources for Cr in soil solutionsand plants of serpentine soils (Oze et al 2004) Bedrock to foresttransects demonstrated a strong correlation between soilmoistureand soil depth Soil nutrient content and soil moisture should beproportional to the cube of soil depth (Belcher et al 1995) thuseach of these parameters figured significantly in the PCA andRDA

Afforestation has been relatively recent in this region (Tyndall1992)Concomitantwith the increase in treedensity andsize is thecontraction of the grassland and savanna component study areafrom 18 in 1937 to 15 in 2013 which is consistent withprevious work documenting the invasion of serpentine savannasby conifers (Tyndall 1992 Barton andWallenstein 1997 Arabas2000)

In landscapeswhereC3 andC4plants coexist stableC isotopiccomposition of SOM represents a powerful method for

20

15

1

0

10

ndash10

ndash2 ndash1 0 1 2 3

05

ndash05

00

RD

A 2

RDA 1

Fig 6 Ordination diagram resulting from standard redundancy analysis (RDA) on transformed abiotic variables illustrating relationships between woodyspecies and physiochemical attributes The diamonds represent species and the vectors indicate the strength of relationship

ndash29 ndash27

(0ndash5)

(5ndash10)

(5ndash15)

(10ndash15)

Grassland (0ndash10)

Soil δ13 C (permil)

Forests

Grassland

Soi

l dep

th (

cm)

ndash25 ndash23 ndash21 ndash19 ndash17 ndash15

Fig 7 Mean carbon isotopic signature in forest (n= 96) and grassland(n= 8) soils for Cherry Hill Maryland USA Error bars are 1 se Datainterpreted as evidence for expansive C4 or C3 historic grassland Verticallines over 5ndash10- and 5ndash15-cm data indicate group pairs that were notsignificantly different using Tukeyrsquos HSD multiple comparisons

J Australian Journal of Botany J Burgess et al

reconstructing vegetation changes (Deines 1980 Tieszen andArcher 1990 Boutton and Yamasaki 1996 Boutton et al 1998)Spatial patterns of soil d13C were strongly related to type ofvegetation (primarily woody versus grassland cover) Remnantgrasslands are composed of a mixture of C3 grasses and forbsand C4 grasses It is well documented that the d13C signature ofC3 plants (ranging from26permil to28permil) is easily distinguishedfrom that of the C4 plants (ranging from 12permil to 15permil)(DeNiro and Epstein 1977 Cerling et al 1989 Tieszen 1991Ehleringer et al 2000 Fox et al 2012 Throop et al 2013) Otherprocesses besides vegetation change alter isotopic ratios (such aspreferential degradation of some plant materials microbiologicalactivity Suess effect and adsorption processes) and may causeshifts of the order of 1ndash3permil (Friedli et al 1987Becker-HeidmannandScharpenseel 1992Trolier et al 1996Ehleringer et al 2000SantRuCkova et al 2000 Krull et al 2005) d13C enrichmentin temperate forest soils larger than 3ndash4permil is generally attributedto the presence of remaining old C4-derived SOC Following thereasoning of Wynn and Bird (2008) we sampled the upper soils(A horizon) and assumed that the intrinsic metabolic processesoccurring in C3 and C4 vegetation impart an unequivocald13C signature into the SOM and that these signatures persistin decaying organic matter such that increasing depth yields achronologic sequence recording vegetation change At the studyarea d13C values of soils in the grassland matrix averaged ndash16permil(Table 4) to 26permil in forested areas dominated by C3 plantsDepth-integrated SOM d13C significantly reflected the currentvegetation cover (mixed C3ndashC4 and solely C3) and accordinglycould be used as a reference to trace past changes in vegetationrecorded in the SOM at deeper soil layers The d13C values ofSOC consistently increased with depth reflecting a higherpercentage of C4-derived carbon in the older soils that isinterpreted to reflect a site-wide shift in vegetation from C4 toC3 plants

Multiple regression and ordination results supported the roleof soil depth and incident radiation both of which influence soilmoisture content as factors governing the occurrence of xericserpentine savannaQuercus species Abiotic factors in areas withhigher importance of shade-tolerant species were associated withdeeper soils higher bulk density and higher Ca Mg ratios In oursite early successional woodlands were favoured in areas of highincident radiation and broadleaf forests were more developed onslopes containing more clay rich and denser soils Diversity ofwoody specieswas higher in areaswith deeper soilsUnlikemanysystems with heavy metals we found no correlations betweentoxic metals (Cr and Ni) and vegetation dynamics

Historical records show that oak savanna ecosystems wereimportant components of the Mid-Atlantic landscape (Shreveet al 1910) The lack of any oaks before the 1850s suggests thatthey were likely harvested for fuel posts or to support grazingFarming of these serpentinite areas is not documented and isunlikely given the inhospitable nature of the soil Our studyindicated that tree recruitment before the 20th centurywasmostlyoaks perhaps regenerating from root sprouts or propagules in theseedbankXeric oak recruitment decreasedover timeandnoxericoaks younger than 40 years established in the forested areasalthough seedlings and saplings still exist in the remnantgrasslandsavannas This failure to regenerate is likely toreflect the consequences of changing disturbance regimes suchas fire-suppression (Nuzzo 1986 Abrams 1992 2006 PetersonandReich 2001) alongwith competitionwith invasive andmesictree species (Nuzzo 1986) White-tailed deer and squirrelherbivory can reduce recruitment especially in isolatedsystems such as these serpentinite landscapes (Shea andStange 1998 Haas and Heske 2005) A turning point forcommunity development appears to be ~1900 with theencroachment of Pinus virginiana and Juniperus virginianaThese conifer species have a variety of characteristics that can

30

25

Minor species

Mesic species (A rubrum F grandifolia N sylvatica P serotina)

Conifers

Quercus spp (Mixed)

Quercus spp (Xeric)

(n = 163)15

5

20

10

Num

ber

of tr

ees

sam

pled

Decade of origin

01850 1860 1870 1880 1890 1900 1910 1920 1930 1940 1950 1960 1970 1980 1990

Fig 8 Age distribution from trees on the basis of core data (n= 163)Xeric species areQuercusmarilandica andQ stellataMixed oak species includeQ albaQ montana Q rubra and Q velutina

Conifer encroachment on ultramafics Australian Journal of Botany K

potentially alter ecosystem processes for decades as they matureand senesce Both conifers are aggressive invaders because ofrapid growth high seed production high seed-dispersalefficiency tolerance of xeric conditions and opportunistic useof ECM mycorrhizae (Ormsbee et al 1976 Holthuijzen andSharik 1984 1985 Briggs andGibson 1992Axmann andKnapp1993 Thiet and Boerner 2007) Once established these speciescan have a profound effect on soil properties as a result ofaccumulation of organic matter on the forest floor and theeventual decomposition of detritus which leads to higherwater-holding capacity and lowering of soil pH This positivefeedback between soil depth and litter cover exerted by theconifers could promote succession towards non-serpentineconditions

The initial serpentine edaphic signature acts as a filterpreventing species that belong to the regional flora but lack thetraits required to survive in such conditions to successfullycolonise these habitats By the 1970s such filters appear tohave diminished Current site characteristics have shiftedtowards more mesic conditions with changes in soil propertiessuch as bulk density soil depth pH SOM and metalsconcentration Over time the community has changed fromxeric (grassland or oak savanna) to one dominated by moremesic oaks and shade-tolerant Acer Nyssa and Fagus speciesThemajor onset ofAcer establishment began in the 1940sUnlikethePinus and Juniperus establishment pulse which proceeded inan open-canopy situation Acer dominance likely proceededunder closed canopy and canopy gaps because it is shade-tolerant One may speculate that over the past 100ndash200 years

the landscape has been dominated by Quercus with patches ofPinus and Juniperus that historicallywere present at lownumbersbecauseof disturbance (miningfire grazing logging)Hilgartneret al (2009) recorded evidence of the presence of Pinusvirginiana from the early 1800s in a serpentine barren siteonly 50 km distant Over the past 60 years Acer has becomeabundant where it along with other species currently present(N sylvatica Cornus florida and Celtis occidentalis) wouldnormally be considered an aberrant species in a hostile edaphicsetting Age and diameter structure indicated a shift away fromQuercus and towards an A rubrum-dominated landscape Thissame species shift has been widely reported throughout theeastern deciduous forests of North America (Abrams 1998Nowacki and Abrams 2008 McEwan et al 2010)

In the context of serpentinite ecosystems the influxof conifersnearly a century ago has acted as a landscapemodulator (Shachaket al 2008) The resulting build-up of SOM increases the water-holding capacity and other resources such as ground-levelradiation facilitate the influx of new species adapted to thealtered conditions As the difference between the mesicwoodlands and the xeric patches decreases the resourcecontrasts also decrease leading to increasing communitysimilarity Different oak species can grow in a wide variety oflight and moisture conditions However none of the traditionalserpentine oak species is considered shade tolerant to the samedegree as are Acer spp Fagus grandifolia and Nyssa sylvaticaWith progressive development of the soil and a dense-canopyclosure growth and germination of the traditional xeric oakspecies are likely to be suppressed marking a shift from

Colonisation

Fig 9 Conceptual model for serpentine (orangendashred colouring) habitats and species redistribution resulting in similar structure for a typical piedmontgeoecologic system (bluered colouring) Bedrock geology (specifically elemental concentrations and structures) exerts its control on soil fertility with subtlevariations resulting in markedly different above-ground and below-ground communities The change in soils over time (dsdt) equation from Huggett (1995)is a function of the various drivers including b (biology) r (topography) a (atmospheric effects) s (soils) and h (hydrology) Through changing vegetationdynamics and species redistribution the tree functional groups and forest types will see increasing similarity as a result of converging successional trajectoriesRegional species pools stochastic events and many other factors determine ultimate outcomes

L Australian Journal of Botany J Burgess et al

xerothermic to mesic andor nutrient-demanding speciesFigure 9 details a conceptual model for pedologic and ecologicchanges at serpentine sites of theMid-Atlantic with successionaltrajectories converging on a species redistribution and loss oftypical serpentine species such as Cerastium arvense varvillosissimum Andropogon gerardii and the xeric oaks(Q marilandica and Q stellata)

Although the canopy is still oak or pine dominated thetrajectory points to a maplendashbeechndashred oak-dominated forestThe Mid-Atlantic has already lost much of its originalgrasslandndashoak woodlandndashforest mosaic This state transitionmay lead to the loss of a unique ecosystem and thustaxonomic homogenisation with the regional flora Thesechanges result from anthropic influences through changes inland use and fire frequency the loss of Castanea dentata(American chestnut) and changing herbivore populationsNearly a century ago these areas gave way to a mosaic ofsmall woodlands typically characterised by xeric oaksinterspersed with shallow rooting low nutrient-tolerantconifers Released from suppressive disturbance regimes(herbivory and fire for example) Pinus virginiana andJuniperus virginiana quickly became the dominant speciesHowever our data suggest that the landscape is nowundergoing another distinct change as the once-open canopystructure has becomemore closed Edaphic properties such as soildepth chemistry and moisture that once drove communityorganisation across environmental stress gradients are nowbeing modulated by species no longer filtered by theserpentine syndrome This modulation may allow for therecruitment and establishment of drought-sensitive speciesbeyond their typical tolerance limits (Warman and Moles2008) This unique ecosystem which took centuries tomillennia to evolve appears to be fading and may soonevanesce into the past

Sources of funding

Our work was funded by the National Science Foundation(DEB-0423476 and DEB-1027188) and grants from theR Balk and D Elliott Field Fund (Johns Hopkins University)

Conflicts of interest

No conflicts of interest

Acknowledgements

We thank Drs Richard Back Naomi Levin Bruce Marsh and Ben Passey foruse of their elemental and isotopic facilities We also thank Dr David Veblenand two anonymous reviewers for numerous constructive comments on thegeology

References

Abbe C (1898) A general report on the physiography of Maryland JohnsHopkins University (Johns Hopkins University Press Baltimore MD)

Abrams MD (1992) Fire and the development of oak forests Bioscience 42346ndash353 doi1023071311781

Abrams MD (1998) The red maple paradox Bioscience 48 355ndash364doi1023071313374

AbramsMD (2006) Ecological and ecophysiological attributes and responsesto fire in eastern oak forests In lsquoFire in eastern oak forests deliveringscience to landmanagersrsquo (EdMBDickinson) pp 74ndash89 (USDA ForestService General Technical Report NRS-P-1)

Alexander EB (2009) Serpentine geoecology of the eastern and southeasternmargins of North America Northeastern Naturalist 16 223ndash252doi1016560450160518

Alexander EB Ellis CC Burke R (2007) A chronosequence of soils andvegetation on serpentine terraces in the Klamath Mountains USA SoilScience 172 565ndash576 doi101097ss0b013e31804fa22d

Arabas KB (2000) Spatial and temporal relationships among fire frequencyvegetation and soil depth in an easternNorthAmerican serpentine barrenThe Journal of the Torrey Botanical Society 127 51ndash65doi1023073088747

Archer S Schimel DS Holland EA (1995) Mechanisms of shrublandexpansion land use climate or CO2 Climatic Change 29 91ndash99doi101007BF01091640

Axmann BD Knapp AK (1993) Water relations of Juniperus virginiana andAndropogon gerardii in an unburned tallgrass prairie watershed TheSouthwestern Naturalist 38 325ndash330 doi1023073671610

Barton AM Wallenstein MD (1997) Effects of invasion of Pinus virginianaon soil properties in serpentine barrens in southeastern PennsylvaniaThe Journal of the Torrey Botanical Society 124 297ndash305doi1023072997264

Bazzaz FA (1968) Succession on abandoned fields in the Shawnee Hillssouthern Illinois Ecology 49 924 doi1023071936544

Becker-Heidmann P Scharpenseel H-W (1992) Studies of soil organicmatterdynamics using natural carbon isotopes The Science of the TotalEnvironment 117ndash118 305ndash312 doi1010160048-9697(92)90097-C

Belcher JW Keddy PA Twolan-Strutt L (1995) Root and shoot competitionintensity along a soil depth gradient Journal of Ecology 83 673ndash682doi1023072261635

BouttonTWArcher SRMidwoodAJZitzer SFBolR (1998)d13Cvalues ofsoil organic carbon and their use in documenting vegetation change in asubtropical savanna ecosystem Geoderma 82 5ndash41doi101016S0016-7061(97)00095-5

Bouyoucos GJ (1936) Directions for making mechanical analyses of soilsby the hydrometer method Soil Science 42 225ndash230doi10109700010694-193609000-00007

Briggs JMGibsonDJ (1992)Effect offire on tree spatial patterns in a tallgrassprairie landscape Bulletin of the Torrey Botanical Club 119 300ndash307doi1023072996761

BrownML Brown RG (1984) lsquoHerbaceous plants of Marylandrsquo (Universityof Maryland College Park MD)

Burgess JL Lev S Swan CM Szlavecz K (2009) Geologic and edaphiccontrols on a serpentine forest community Northeastern Naturalist 16366ndash384 doi1016560450160527

Butaye J Jacquemyn H Honnay O Hermy M (2002) The species poolconcept applied to forests in a fragmented landscape dispersal limitationversus habitat limitation Journal of Vegetation Science 13 27ndash34doi101111j1654-11032002tb02020x

Callaway RM Davis FW (1993) Vegetation dynamics fire and the physical-environment in coastal central California Rid B-7010-2009 Ecology 741567ndash1578 doi1023071940084

Carroll W D Kapeluck P R Harper R A and Van Lear D H (2002)Background paper historical overview of the southern forest landscapeand associated resources In lsquoSouthern forest resource assessmentrsquo (EdsDNWear JGGreis) pp 583ndash605 (USDepartmentofAgricultureForestService Southern Research Station Asheville NC)

Cerling TE Quade J Wang Y Bowman JR (1989) Carbon isotopes in soilsand palaeosols as ecology and palaeoecology indicators Nature 341138ndash139 doi101038341138a0

Chao A (1984) Nonparametric estimation of the number of classes in apopulation Scandinavian Journal of Statistics 11 265ndash270

ChaoA (1987) Estimating the population size for capturendashrecapture data withunequal catchability Biometrics 43 783ndash791 doi1023072531532

Chazdon RL Colwell RK Denslow JS Guariguata MR (1998) Statisticalmethods for estimating species richness of woody regeneration in primaryand secondary rain forests of Northeastern Costa Rica In lsquoForest

Conifer encroachment on ultramafics Australian Journal of Botany M

P = 0004) There were significant vegetation depth andvegetation depth effects on d13C of SOM (Table 5 Fig 7)

Age structure

In total 163 core samples were evaluated with results displayedin the stand-age structure plot (Fig 8) Skeleton plot chronologiesvaried among species Older minimum age at breast height wasgenerally seen inQ stellata (114 15 years) andQ marilandica(110 7 years) and the younger chronologies developed inspecies such as A rubrum (53 4 years) F grandifolia(56 8 years) and L tulipifera (43 8 years) Only xeric oakswere present in the oldest age classes The period before 1900 iscomposed of 54 xeric oaks and when all oaks are grouped theycomprise 79 of trees from that era although Q montana isabsent from this group (oldest minimum age is 88 years) Theoldest non-oak was N sylvatica (144 years) The period between

1900 and 1940 records a marked increase in conifers andconcomitant decrease of xeric oaks comprising 39 and 19of all species respectively Between 1950 and 1990 thepercentage of xeric oaks decreased to 3 and conifers to 13whereas Acer spp increased to 19 The number of shade-tolerant species rose from 125 in the decades from 1850 to1900 to 59 during the interval from 1950 to 1990

The age structure pooled from all the plots (Fig 8) showeddiscontinuous patterns with prominent peaks of regenerationpulses and periods of either low or absent recruitment Coniferestablishment and recruitment clearly increased during1900ndash1960 and then receded through the rest of the samplingperiod There was a conspicuous absence of younger xeric oaks(post-1970s) and only two xeric oaks from1950 to the present Atthe same time the percentage of non-oak deciduous treesincreased to 68 post-1950

10

05

00

Dim

2 (

171

7)

Dim

2 (

229

1)

Dim 1 (3038)

Dim 1 (402)

ndash10

ndash05

4

2

0

ndash2

ndash4

ndash6

ndash15

ndash15 ndash10 ndash5 0 5 10

ndash10 ndash05 00 05 10 15

(a)

(b)

Fig 5 (a) Unconstrained principal component analysis (PCA) correlation circle of pedological topological and community variables (b) Forest communityclustering along an edaphic gradient

Conifer encroachment on ultramafics Australian Journal of Botany I

Discussion

The present study examined a moderately sized area in the Mid-Atlantic but the changes ndash encroachment and mesophication ndash

represent potential widespread conservation problems Theserpentinite landscapes were once common components of theancient Piedmont (Carroll et al 2002 Owen 2002 Noss 2013)These unique plant communities have existed as persistentlsquoislandsrsquo embedded in a deciduous forest region for centuriesor longer We documented a state transition from grasslandand xeric oak savanna to conifer-dominated communities

that have given way to eastern broadleaf forests Comparisonsamong geologic pedologic and vegetation properties reflect thischange and indicate that the conversion to forest has been rapidand characterised by large changes in soil properties (pH water-holding capacity SOMcomposition) aswell as plant-communitystructure (physiognomyand stemdensity) and composition (xericto more mesic communities)

Thin-section and petrologic data confirmed that the geologicsubstrate is a serpentinised peridotitewith characteristics typical ofserpentinite terranes in the Appalachians (Misra and Keller 1978Mittwede and Stoddard 1989 Gates 1992 Alexander 2009)Although Ca is depleted in bulk rock samples because of thebreakdownof augite and interstitial plagioclase soil Ca Mg ratiosare higher in surface soils The concentrations of SOC C N andmacronutrients in woodland and forested areas are linked to theturnover of organic matter and not the parent material

Compared with the rock soils are enriched in most tracemetals although the Cr contents were similar This was likelyto be due to resistance of Cr-oxides to weathering and suggeststhat these phases were not likely sources for Cr in soil solutionsand plants of serpentine soils (Oze et al 2004) Bedrock to foresttransects demonstrated a strong correlation between soilmoistureand soil depth Soil nutrient content and soil moisture should beproportional to the cube of soil depth (Belcher et al 1995) thuseach of these parameters figured significantly in the PCA andRDA

Afforestation has been relatively recent in this region (Tyndall1992)Concomitantwith the increase in treedensity andsize is thecontraction of the grassland and savanna component study areafrom 18 in 1937 to 15 in 2013 which is consistent withprevious work documenting the invasion of serpentine savannasby conifers (Tyndall 1992 Barton andWallenstein 1997 Arabas2000)

In landscapeswhereC3 andC4plants coexist stableC isotopiccomposition of SOM represents a powerful method for

20

15

1

0

10

ndash10

ndash2 ndash1 0 1 2 3

05

ndash05

00

RD

A 2

RDA 1

Fig 6 Ordination diagram resulting from standard redundancy analysis (RDA) on transformed abiotic variables illustrating relationships between woodyspecies and physiochemical attributes The diamonds represent species and the vectors indicate the strength of relationship

ndash29 ndash27

(0ndash5)

(5ndash10)

(5ndash15)

(10ndash15)

Grassland (0ndash10)

Soil δ13 C (permil)

Forests

Grassland

Soi

l dep

th (

cm)

ndash25 ndash23 ndash21 ndash19 ndash17 ndash15

Fig 7 Mean carbon isotopic signature in forest (n= 96) and grassland(n= 8) soils for Cherry Hill Maryland USA Error bars are 1 se Datainterpreted as evidence for expansive C4 or C3 historic grassland Verticallines over 5ndash10- and 5ndash15-cm data indicate group pairs that were notsignificantly different using Tukeyrsquos HSD multiple comparisons

J Australian Journal of Botany J Burgess et al

reconstructing vegetation changes (Deines 1980 Tieszen andArcher 1990 Boutton and Yamasaki 1996 Boutton et al 1998)Spatial patterns of soil d13C were strongly related to type ofvegetation (primarily woody versus grassland cover) Remnantgrasslands are composed of a mixture of C3 grasses and forbsand C4 grasses It is well documented that the d13C signature ofC3 plants (ranging from26permil to28permil) is easily distinguishedfrom that of the C4 plants (ranging from 12permil to 15permil)(DeNiro and Epstein 1977 Cerling et al 1989 Tieszen 1991Ehleringer et al 2000 Fox et al 2012 Throop et al 2013) Otherprocesses besides vegetation change alter isotopic ratios (such aspreferential degradation of some plant materials microbiologicalactivity Suess effect and adsorption processes) and may causeshifts of the order of 1ndash3permil (Friedli et al 1987Becker-HeidmannandScharpenseel 1992Trolier et al 1996Ehleringer et al 2000SantRuCkova et al 2000 Krull et al 2005) d13C enrichmentin temperate forest soils larger than 3ndash4permil is generally attributedto the presence of remaining old C4-derived SOC Following thereasoning of Wynn and Bird (2008) we sampled the upper soils(A horizon) and assumed that the intrinsic metabolic processesoccurring in C3 and C4 vegetation impart an unequivocald13C signature into the SOM and that these signatures persistin decaying organic matter such that increasing depth yields achronologic sequence recording vegetation change At the studyarea d13C values of soils in the grassland matrix averaged ndash16permil(Table 4) to 26permil in forested areas dominated by C3 plantsDepth-integrated SOM d13C significantly reflected the currentvegetation cover (mixed C3ndashC4 and solely C3) and accordinglycould be used as a reference to trace past changes in vegetationrecorded in the SOM at deeper soil layers The d13C values ofSOC consistently increased with depth reflecting a higherpercentage of C4-derived carbon in the older soils that isinterpreted to reflect a site-wide shift in vegetation from C4 toC3 plants

Multiple regression and ordination results supported the roleof soil depth and incident radiation both of which influence soilmoisture content as factors governing the occurrence of xericserpentine savannaQuercus species Abiotic factors in areas withhigher importance of shade-tolerant species were associated withdeeper soils higher bulk density and higher Ca Mg ratios In oursite early successional woodlands were favoured in areas of highincident radiation and broadleaf forests were more developed onslopes containing more clay rich and denser soils Diversity ofwoody specieswas higher in areaswith deeper soilsUnlikemanysystems with heavy metals we found no correlations betweentoxic metals (Cr and Ni) and vegetation dynamics

Historical records show that oak savanna ecosystems wereimportant components of the Mid-Atlantic landscape (Shreveet al 1910) The lack of any oaks before the 1850s suggests thatthey were likely harvested for fuel posts or to support grazingFarming of these serpentinite areas is not documented and isunlikely given the inhospitable nature of the soil Our studyindicated that tree recruitment before the 20th centurywasmostlyoaks perhaps regenerating from root sprouts or propagules in theseedbankXeric oak recruitment decreasedover timeandnoxericoaks younger than 40 years established in the forested areasalthough seedlings and saplings still exist in the remnantgrasslandsavannas This failure to regenerate is likely toreflect the consequences of changing disturbance regimes suchas fire-suppression (Nuzzo 1986 Abrams 1992 2006 PetersonandReich 2001) alongwith competitionwith invasive andmesictree species (Nuzzo 1986) White-tailed deer and squirrelherbivory can reduce recruitment especially in isolatedsystems such as these serpentinite landscapes (Shea andStange 1998 Haas and Heske 2005) A turning point forcommunity development appears to be ~1900 with theencroachment of Pinus virginiana and Juniperus virginianaThese conifer species have a variety of characteristics that can

30

25

Minor species

Mesic species (A rubrum F grandifolia N sylvatica P serotina)

Conifers

Quercus spp (Mixed)

Quercus spp (Xeric)

(n = 163)15

5

20

10

Num

ber

of tr

ees

sam

pled

Decade of origin

01850 1860 1870 1880 1890 1900 1910 1920 1930 1940 1950 1960 1970 1980 1990

Fig 8 Age distribution from trees on the basis of core data (n= 163)Xeric species areQuercusmarilandica andQ stellataMixed oak species includeQ albaQ montana Q rubra and Q velutina

Conifer encroachment on ultramafics Australian Journal of Botany K

potentially alter ecosystem processes for decades as they matureand senesce Both conifers are aggressive invaders because ofrapid growth high seed production high seed-dispersalefficiency tolerance of xeric conditions and opportunistic useof ECM mycorrhizae (Ormsbee et al 1976 Holthuijzen andSharik 1984 1985 Briggs andGibson 1992Axmann andKnapp1993 Thiet and Boerner 2007) Once established these speciescan have a profound effect on soil properties as a result ofaccumulation of organic matter on the forest floor and theeventual decomposition of detritus which leads to higherwater-holding capacity and lowering of soil pH This positivefeedback between soil depth and litter cover exerted by theconifers could promote succession towards non-serpentineconditions

The initial serpentine edaphic signature acts as a filterpreventing species that belong to the regional flora but lack thetraits required to survive in such conditions to successfullycolonise these habitats By the 1970s such filters appear tohave diminished Current site characteristics have shiftedtowards more mesic conditions with changes in soil propertiessuch as bulk density soil depth pH SOM and metalsconcentration Over time the community has changed fromxeric (grassland or oak savanna) to one dominated by moremesic oaks and shade-tolerant Acer Nyssa and Fagus speciesThemajor onset ofAcer establishment began in the 1940sUnlikethePinus and Juniperus establishment pulse which proceeded inan open-canopy situation Acer dominance likely proceededunder closed canopy and canopy gaps because it is shade-tolerant One may speculate that over the past 100ndash200 years

the landscape has been dominated by Quercus with patches ofPinus and Juniperus that historicallywere present at lownumbersbecauseof disturbance (miningfire grazing logging)Hilgartneret al (2009) recorded evidence of the presence of Pinusvirginiana from the early 1800s in a serpentine barren siteonly 50 km distant Over the past 60 years Acer has becomeabundant where it along with other species currently present(N sylvatica Cornus florida and Celtis occidentalis) wouldnormally be considered an aberrant species in a hostile edaphicsetting Age and diameter structure indicated a shift away fromQuercus and towards an A rubrum-dominated landscape Thissame species shift has been widely reported throughout theeastern deciduous forests of North America (Abrams 1998Nowacki and Abrams 2008 McEwan et al 2010)

In the context of serpentinite ecosystems the influxof conifersnearly a century ago has acted as a landscapemodulator (Shachaket al 2008) The resulting build-up of SOM increases the water-holding capacity and other resources such as ground-levelradiation facilitate the influx of new species adapted to thealtered conditions As the difference between the mesicwoodlands and the xeric patches decreases the resourcecontrasts also decrease leading to increasing communitysimilarity Different oak species can grow in a wide variety oflight and moisture conditions However none of the traditionalserpentine oak species is considered shade tolerant to the samedegree as are Acer spp Fagus grandifolia and Nyssa sylvaticaWith progressive development of the soil and a dense-canopyclosure growth and germination of the traditional xeric oakspecies are likely to be suppressed marking a shift from

Colonisation

Fig 9 Conceptual model for serpentine (orangendashred colouring) habitats and species redistribution resulting in similar structure for a typical piedmontgeoecologic system (bluered colouring) Bedrock geology (specifically elemental concentrations and structures) exerts its control on soil fertility with subtlevariations resulting in markedly different above-ground and below-ground communities The change in soils over time (dsdt) equation from Huggett (1995)is a function of the various drivers including b (biology) r (topography) a (atmospheric effects) s (soils) and h (hydrology) Through changing vegetationdynamics and species redistribution the tree functional groups and forest types will see increasing similarity as a result of converging successional trajectoriesRegional species pools stochastic events and many other factors determine ultimate outcomes

L Australian Journal of Botany J Burgess et al

xerothermic to mesic andor nutrient-demanding speciesFigure 9 details a conceptual model for pedologic and ecologicchanges at serpentine sites of theMid-Atlantic with successionaltrajectories converging on a species redistribution and loss oftypical serpentine species such as Cerastium arvense varvillosissimum Andropogon gerardii and the xeric oaks(Q marilandica and Q stellata)

Although the canopy is still oak or pine dominated thetrajectory points to a maplendashbeechndashred oak-dominated forestThe Mid-Atlantic has already lost much of its originalgrasslandndashoak woodlandndashforest mosaic This state transitionmay lead to the loss of a unique ecosystem and thustaxonomic homogenisation with the regional flora Thesechanges result from anthropic influences through changes inland use and fire frequency the loss of Castanea dentata(American chestnut) and changing herbivore populationsNearly a century ago these areas gave way to a mosaic ofsmall woodlands typically characterised by xeric oaksinterspersed with shallow rooting low nutrient-tolerantconifers Released from suppressive disturbance regimes(herbivory and fire for example) Pinus virginiana andJuniperus virginiana quickly became the dominant speciesHowever our data suggest that the landscape is nowundergoing another distinct change as the once-open canopystructure has becomemore closed Edaphic properties such as soildepth chemistry and moisture that once drove communityorganisation across environmental stress gradients are nowbeing modulated by species no longer filtered by theserpentine syndrome This modulation may allow for therecruitment and establishment of drought-sensitive speciesbeyond their typical tolerance limits (Warman and Moles2008) This unique ecosystem which took centuries tomillennia to evolve appears to be fading and may soonevanesce into the past

Sources of funding

Our work was funded by the National Science Foundation(DEB-0423476 and DEB-1027188) and grants from theR Balk and D Elliott Field Fund (Johns Hopkins University)

Conflicts of interest

No conflicts of interest

Acknowledgements

We thank Drs Richard Back Naomi Levin Bruce Marsh and Ben Passey foruse of their elemental and isotopic facilities We also thank Dr David Veblenand two anonymous reviewers for numerous constructive comments on thegeology

References

Abbe C (1898) A general report on the physiography of Maryland JohnsHopkins University (Johns Hopkins University Press Baltimore MD)

Abrams MD (1992) Fire and the development of oak forests Bioscience 42346ndash353 doi1023071311781

Abrams MD (1998) The red maple paradox Bioscience 48 355ndash364doi1023071313374

AbramsMD (2006) Ecological and ecophysiological attributes and responsesto fire in eastern oak forests In lsquoFire in eastern oak forests deliveringscience to landmanagersrsquo (EdMBDickinson) pp 74ndash89 (USDA ForestService General Technical Report NRS-P-1)

Alexander EB (2009) Serpentine geoecology of the eastern and southeasternmargins of North America Northeastern Naturalist 16 223ndash252doi1016560450160518

Alexander EB Ellis CC Burke R (2007) A chronosequence of soils andvegetation on serpentine terraces in the Klamath Mountains USA SoilScience 172 565ndash576 doi101097ss0b013e31804fa22d

Arabas KB (2000) Spatial and temporal relationships among fire frequencyvegetation and soil depth in an easternNorthAmerican serpentine barrenThe Journal of the Torrey Botanical Society 127 51ndash65doi1023073088747

Archer S Schimel DS Holland EA (1995) Mechanisms of shrublandexpansion land use climate or CO2 Climatic Change 29 91ndash99doi101007BF01091640

Axmann BD Knapp AK (1993) Water relations of Juniperus virginiana andAndropogon gerardii in an unburned tallgrass prairie watershed TheSouthwestern Naturalist 38 325ndash330 doi1023073671610

Barton AM Wallenstein MD (1997) Effects of invasion of Pinus virginianaon soil properties in serpentine barrens in southeastern PennsylvaniaThe Journal of the Torrey Botanical Society 124 297ndash305doi1023072997264

Bazzaz FA (1968) Succession on abandoned fields in the Shawnee Hillssouthern Illinois Ecology 49 924 doi1023071936544

Becker-Heidmann P Scharpenseel H-W (1992) Studies of soil organicmatterdynamics using natural carbon isotopes The Science of the TotalEnvironment 117ndash118 305ndash312 doi1010160048-9697(92)90097-C

Belcher JW Keddy PA Twolan-Strutt L (1995) Root and shoot competitionintensity along a soil depth gradient Journal of Ecology 83 673ndash682doi1023072261635

BouttonTWArcher SRMidwoodAJZitzer SFBolR (1998)d13Cvalues ofsoil organic carbon and their use in documenting vegetation change in asubtropical savanna ecosystem Geoderma 82 5ndash41doi101016S0016-7061(97)00095-5

Bouyoucos GJ (1936) Directions for making mechanical analyses of soilsby the hydrometer method Soil Science 42 225ndash230doi10109700010694-193609000-00007

Briggs JMGibsonDJ (1992)Effect offire on tree spatial patterns in a tallgrassprairie landscape Bulletin of the Torrey Botanical Club 119 300ndash307doi1023072996761

BrownML Brown RG (1984) lsquoHerbaceous plants of Marylandrsquo (Universityof Maryland College Park MD)

Burgess JL Lev S Swan CM Szlavecz K (2009) Geologic and edaphiccontrols on a serpentine forest community Northeastern Naturalist 16366ndash384 doi1016560450160527

Butaye J Jacquemyn H Honnay O Hermy M (2002) The species poolconcept applied to forests in a fragmented landscape dispersal limitationversus habitat limitation Journal of Vegetation Science 13 27ndash34doi101111j1654-11032002tb02020x

Callaway RM Davis FW (1993) Vegetation dynamics fire and the physical-environment in coastal central California Rid B-7010-2009 Ecology 741567ndash1578 doi1023071940084

Carroll W D Kapeluck P R Harper R A and Van Lear D H (2002)Background paper historical overview of the southern forest landscapeand associated resources In lsquoSouthern forest resource assessmentrsquo (EdsDNWear JGGreis) pp 583ndash605 (USDepartmentofAgricultureForestService Southern Research Station Asheville NC)

Cerling TE Quade J Wang Y Bowman JR (1989) Carbon isotopes in soilsand palaeosols as ecology and palaeoecology indicators Nature 341138ndash139 doi101038341138a0

Chao A (1984) Nonparametric estimation of the number of classes in apopulation Scandinavian Journal of Statistics 11 265ndash270

ChaoA (1987) Estimating the population size for capturendashrecapture data withunequal catchability Biometrics 43 783ndash791 doi1023072531532

Chazdon RL Colwell RK Denslow JS Guariguata MR (1998) Statisticalmethods for estimating species richness of woody regeneration in primaryand secondary rain forests of Northeastern Costa Rica In lsquoForest

Conifer encroachment on ultramafics Australian Journal of Botany M

Discussion

The present study examined a moderately sized area in the Mid-Atlantic but the changes ndash encroachment and mesophication ndash

represent potential widespread conservation problems Theserpentinite landscapes were once common components of theancient Piedmont (Carroll et al 2002 Owen 2002 Noss 2013)These unique plant communities have existed as persistentlsquoislandsrsquo embedded in a deciduous forest region for centuriesor longer We documented a state transition from grasslandand xeric oak savanna to conifer-dominated communities

that have given way to eastern broadleaf forests Comparisonsamong geologic pedologic and vegetation properties reflect thischange and indicate that the conversion to forest has been rapidand characterised by large changes in soil properties (pH water-holding capacity SOMcomposition) aswell as plant-communitystructure (physiognomyand stemdensity) and composition (xericto more mesic communities)

Thin-section and petrologic data confirmed that the geologicsubstrate is a serpentinised peridotitewith characteristics typical ofserpentinite terranes in the Appalachians (Misra and Keller 1978Mittwede and Stoddard 1989 Gates 1992 Alexander 2009)Although Ca is depleted in bulk rock samples because of thebreakdownof augite and interstitial plagioclase soil Ca Mg ratiosare higher in surface soils The concentrations of SOC C N andmacronutrients in woodland and forested areas are linked to theturnover of organic matter and not the parent material

Compared with the rock soils are enriched in most tracemetals although the Cr contents were similar This was likelyto be due to resistance of Cr-oxides to weathering and suggeststhat these phases were not likely sources for Cr in soil solutionsand plants of serpentine soils (Oze et al 2004) Bedrock to foresttransects demonstrated a strong correlation between soilmoistureand soil depth Soil nutrient content and soil moisture should beproportional to the cube of soil depth (Belcher et al 1995) thuseach of these parameters figured significantly in the PCA andRDA

Afforestation has been relatively recent in this region (Tyndall1992)Concomitantwith the increase in treedensity andsize is thecontraction of the grassland and savanna component study areafrom 18 in 1937 to 15 in 2013 which is consistent withprevious work documenting the invasion of serpentine savannasby conifers (Tyndall 1992 Barton andWallenstein 1997 Arabas2000)

In landscapeswhereC3 andC4plants coexist stableC isotopiccomposition of SOM represents a powerful method for

20

15

1

0

10

ndash10

ndash2 ndash1 0 1 2 3

05

ndash05

00

RD

A 2

RDA 1

Fig 6 Ordination diagram resulting from standard redundancy analysis (RDA) on transformed abiotic variables illustrating relationships between woodyspecies and physiochemical attributes The diamonds represent species and the vectors indicate the strength of relationship

ndash29 ndash27

(0ndash5)

(5ndash10)

(5ndash15)

(10ndash15)

Grassland (0ndash10)

Soil δ13 C (permil)

Forests

Grassland

Soi

l dep

th (

cm)

ndash25 ndash23 ndash21 ndash19 ndash17 ndash15

Fig 7 Mean carbon isotopic signature in forest (n= 96) and grassland(n= 8) soils for Cherry Hill Maryland USA Error bars are 1 se Datainterpreted as evidence for expansive C4 or C3 historic grassland Verticallines over 5ndash10- and 5ndash15-cm data indicate group pairs that were notsignificantly different using Tukeyrsquos HSD multiple comparisons

J Australian Journal of Botany J Burgess et al

reconstructing vegetation changes (Deines 1980 Tieszen andArcher 1990 Boutton and Yamasaki 1996 Boutton et al 1998)Spatial patterns of soil d13C were strongly related to type ofvegetation (primarily woody versus grassland cover) Remnantgrasslands are composed of a mixture of C3 grasses and forbsand C4 grasses It is well documented that the d13C signature ofC3 plants (ranging from26permil to28permil) is easily distinguishedfrom that of the C4 plants (ranging from 12permil to 15permil)(DeNiro and Epstein 1977 Cerling et al 1989 Tieszen 1991Ehleringer et al 2000 Fox et al 2012 Throop et al 2013) Otherprocesses besides vegetation change alter isotopic ratios (such aspreferential degradation of some plant materials microbiologicalactivity Suess effect and adsorption processes) and may causeshifts of the order of 1ndash3permil (Friedli et al 1987Becker-HeidmannandScharpenseel 1992Trolier et al 1996Ehleringer et al 2000SantRuCkova et al 2000 Krull et al 2005) d13C enrichmentin temperate forest soils larger than 3ndash4permil is generally attributedto the presence of remaining old C4-derived SOC Following thereasoning of Wynn and Bird (2008) we sampled the upper soils(A horizon) and assumed that the intrinsic metabolic processesoccurring in C3 and C4 vegetation impart an unequivocald13C signature into the SOM and that these signatures persistin decaying organic matter such that increasing depth yields achronologic sequence recording vegetation change At the studyarea d13C values of soils in the grassland matrix averaged ndash16permil(Table 4) to 26permil in forested areas dominated by C3 plantsDepth-integrated SOM d13C significantly reflected the currentvegetation cover (mixed C3ndashC4 and solely C3) and accordinglycould be used as a reference to trace past changes in vegetationrecorded in the SOM at deeper soil layers The d13C values ofSOC consistently increased with depth reflecting a higherpercentage of C4-derived carbon in the older soils that isinterpreted to reflect a site-wide shift in vegetation from C4 toC3 plants

Multiple regression and ordination results supported the roleof soil depth and incident radiation both of which influence soilmoisture content as factors governing the occurrence of xericserpentine savannaQuercus species Abiotic factors in areas withhigher importance of shade-tolerant species were associated withdeeper soils higher bulk density and higher Ca Mg ratios In oursite early successional woodlands were favoured in areas of highincident radiation and broadleaf forests were more developed onslopes containing more clay rich and denser soils Diversity ofwoody specieswas higher in areaswith deeper soilsUnlikemanysystems with heavy metals we found no correlations betweentoxic metals (Cr and Ni) and vegetation dynamics

Historical records show that oak savanna ecosystems wereimportant components of the Mid-Atlantic landscape (Shreveet al 1910) The lack of any oaks before the 1850s suggests thatthey were likely harvested for fuel posts or to support grazingFarming of these serpentinite areas is not documented and isunlikely given the inhospitable nature of the soil Our studyindicated that tree recruitment before the 20th centurywasmostlyoaks perhaps regenerating from root sprouts or propagules in theseedbankXeric oak recruitment decreasedover timeandnoxericoaks younger than 40 years established in the forested areasalthough seedlings and saplings still exist in the remnantgrasslandsavannas This failure to regenerate is likely toreflect the consequences of changing disturbance regimes suchas fire-suppression (Nuzzo 1986 Abrams 1992 2006 PetersonandReich 2001) alongwith competitionwith invasive andmesictree species (Nuzzo 1986) White-tailed deer and squirrelherbivory can reduce recruitment especially in isolatedsystems such as these serpentinite landscapes (Shea andStange 1998 Haas and Heske 2005) A turning point forcommunity development appears to be ~1900 with theencroachment of Pinus virginiana and Juniperus virginianaThese conifer species have a variety of characteristics that can

30

25

Minor species

Mesic species (A rubrum F grandifolia N sylvatica P serotina)

Conifers

Quercus spp (Mixed)

Quercus spp (Xeric)

(n = 163)15

5

20

10

Num

ber

of tr

ees

sam

pled

Decade of origin

01850 1860 1870 1880 1890 1900 1910 1920 1930 1940 1950 1960 1970 1980 1990

Fig 8 Age distribution from trees on the basis of core data (n= 163)Xeric species areQuercusmarilandica andQ stellataMixed oak species includeQ albaQ montana Q rubra and Q velutina

Conifer encroachment on ultramafics Australian Journal of Botany K

potentially alter ecosystem processes for decades as they matureand senesce Both conifers are aggressive invaders because ofrapid growth high seed production high seed-dispersalefficiency tolerance of xeric conditions and opportunistic useof ECM mycorrhizae (Ormsbee et al 1976 Holthuijzen andSharik 1984 1985 Briggs andGibson 1992Axmann andKnapp1993 Thiet and Boerner 2007) Once established these speciescan have a profound effect on soil properties as a result ofaccumulation of organic matter on the forest floor and theeventual decomposition of detritus which leads to higherwater-holding capacity and lowering of soil pH This positivefeedback between soil depth and litter cover exerted by theconifers could promote succession towards non-serpentineconditions

The initial serpentine edaphic signature acts as a filterpreventing species that belong to the regional flora but lack thetraits required to survive in such conditions to successfullycolonise these habitats By the 1970s such filters appear tohave diminished Current site characteristics have shiftedtowards more mesic conditions with changes in soil propertiessuch as bulk density soil depth pH SOM and metalsconcentration Over time the community has changed fromxeric (grassland or oak savanna) to one dominated by moremesic oaks and shade-tolerant Acer Nyssa and Fagus speciesThemajor onset ofAcer establishment began in the 1940sUnlikethePinus and Juniperus establishment pulse which proceeded inan open-canopy situation Acer dominance likely proceededunder closed canopy and canopy gaps because it is shade-tolerant One may speculate that over the past 100ndash200 years

the landscape has been dominated by Quercus with patches ofPinus and Juniperus that historicallywere present at lownumbersbecauseof disturbance (miningfire grazing logging)Hilgartneret al (2009) recorded evidence of the presence of Pinusvirginiana from the early 1800s in a serpentine barren siteonly 50 km distant Over the past 60 years Acer has becomeabundant where it along with other species currently present(N sylvatica Cornus florida and Celtis occidentalis) wouldnormally be considered an aberrant species in a hostile edaphicsetting Age and diameter structure indicated a shift away fromQuercus and towards an A rubrum-dominated landscape Thissame species shift has been widely reported throughout theeastern deciduous forests of North America (Abrams 1998Nowacki and Abrams 2008 McEwan et al 2010)

In the context of serpentinite ecosystems the influxof conifersnearly a century ago has acted as a landscapemodulator (Shachaket al 2008) The resulting build-up of SOM increases the water-holding capacity and other resources such as ground-levelradiation facilitate the influx of new species adapted to thealtered conditions As the difference between the mesicwoodlands and the xeric patches decreases the resourcecontrasts also decrease leading to increasing communitysimilarity Different oak species can grow in a wide variety oflight and moisture conditions However none of the traditionalserpentine oak species is considered shade tolerant to the samedegree as are Acer spp Fagus grandifolia and Nyssa sylvaticaWith progressive development of the soil and a dense-canopyclosure growth and germination of the traditional xeric oakspecies are likely to be suppressed marking a shift from

Colonisation

Fig 9 Conceptual model for serpentine (orangendashred colouring) habitats and species redistribution resulting in similar structure for a typical piedmontgeoecologic system (bluered colouring) Bedrock geology (specifically elemental concentrations and structures) exerts its control on soil fertility with subtlevariations resulting in markedly different above-ground and below-ground communities The change in soils over time (dsdt) equation from Huggett (1995)is a function of the various drivers including b (biology) r (topography) a (atmospheric effects) s (soils) and h (hydrology) Through changing vegetationdynamics and species redistribution the tree functional groups and forest types will see increasing similarity as a result of converging successional trajectoriesRegional species pools stochastic events and many other factors determine ultimate outcomes

L Australian Journal of Botany J Burgess et al

xerothermic to mesic andor nutrient-demanding speciesFigure 9 details a conceptual model for pedologic and ecologicchanges at serpentine sites of theMid-Atlantic with successionaltrajectories converging on a species redistribution and loss oftypical serpentine species such as Cerastium arvense varvillosissimum Andropogon gerardii and the xeric oaks(Q marilandica and Q stellata)

Although the canopy is still oak or pine dominated thetrajectory points to a maplendashbeechndashred oak-dominated forestThe Mid-Atlantic has already lost much of its originalgrasslandndashoak woodlandndashforest mosaic This state transitionmay lead to the loss of a unique ecosystem and thustaxonomic homogenisation with the regional flora Thesechanges result from anthropic influences through changes inland use and fire frequency the loss of Castanea dentata(American chestnut) and changing herbivore populationsNearly a century ago these areas gave way to a mosaic ofsmall woodlands typically characterised by xeric oaksinterspersed with shallow rooting low nutrient-tolerantconifers Released from suppressive disturbance regimes(herbivory and fire for example) Pinus virginiana andJuniperus virginiana quickly became the dominant speciesHowever our data suggest that the landscape is nowundergoing another distinct change as the once-open canopystructure has becomemore closed Edaphic properties such as soildepth chemistry and moisture that once drove communityorganisation across environmental stress gradients are nowbeing modulated by species no longer filtered by theserpentine syndrome This modulation may allow for therecruitment and establishment of drought-sensitive speciesbeyond their typical tolerance limits (Warman and Moles2008) This unique ecosystem which took centuries tomillennia to evolve appears to be fading and may soonevanesce into the past

Sources of funding

Our work was funded by the National Science Foundation(DEB-0423476 and DEB-1027188) and grants from theR Balk and D Elliott Field Fund (Johns Hopkins University)

Conflicts of interest

No conflicts of interest

Acknowledgements

We thank Drs Richard Back Naomi Levin Bruce Marsh and Ben Passey foruse of their elemental and isotopic facilities We also thank Dr David Veblenand two anonymous reviewers for numerous constructive comments on thegeology

References

Abbe C (1898) A general report on the physiography of Maryland JohnsHopkins University (Johns Hopkins University Press Baltimore MD)

Abrams MD (1992) Fire and the development of oak forests Bioscience 42346ndash353 doi1023071311781

Abrams MD (1998) The red maple paradox Bioscience 48 355ndash364doi1023071313374

AbramsMD (2006) Ecological and ecophysiological attributes and responsesto fire in eastern oak forests In lsquoFire in eastern oak forests deliveringscience to landmanagersrsquo (EdMBDickinson) pp 74ndash89 (USDA ForestService General Technical Report NRS-P-1)

Alexander EB (2009) Serpentine geoecology of the eastern and southeasternmargins of North America Northeastern Naturalist 16 223ndash252doi1016560450160518

Alexander EB Ellis CC Burke R (2007) A chronosequence of soils andvegetation on serpentine terraces in the Klamath Mountains USA SoilScience 172 565ndash576 doi101097ss0b013e31804fa22d

Arabas KB (2000) Spatial and temporal relationships among fire frequencyvegetation and soil depth in an easternNorthAmerican serpentine barrenThe Journal of the Torrey Botanical Society 127 51ndash65doi1023073088747

Archer S Schimel DS Holland EA (1995) Mechanisms of shrublandexpansion land use climate or CO2 Climatic Change 29 91ndash99doi101007BF01091640

Axmann BD Knapp AK (1993) Water relations of Juniperus virginiana andAndropogon gerardii in an unburned tallgrass prairie watershed TheSouthwestern Naturalist 38 325ndash330 doi1023073671610

Barton AM Wallenstein MD (1997) Effects of invasion of Pinus virginianaon soil properties in serpentine barrens in southeastern PennsylvaniaThe Journal of the Torrey Botanical Society 124 297ndash305doi1023072997264

Bazzaz FA (1968) Succession on abandoned fields in the Shawnee Hillssouthern Illinois Ecology 49 924 doi1023071936544

Becker-Heidmann P Scharpenseel H-W (1992) Studies of soil organicmatterdynamics using natural carbon isotopes The Science of the TotalEnvironment 117ndash118 305ndash312 doi1010160048-9697(92)90097-C

Belcher JW Keddy PA Twolan-Strutt L (1995) Root and shoot competitionintensity along a soil depth gradient Journal of Ecology 83 673ndash682doi1023072261635

BouttonTWArcher SRMidwoodAJZitzer SFBolR (1998)d13Cvalues ofsoil organic carbon and their use in documenting vegetation change in asubtropical savanna ecosystem Geoderma 82 5ndash41doi101016S0016-7061(97)00095-5

Bouyoucos GJ (1936) Directions for making mechanical analyses of soilsby the hydrometer method Soil Science 42 225ndash230doi10109700010694-193609000-00007

Briggs JMGibsonDJ (1992)Effect offire on tree spatial patterns in a tallgrassprairie landscape Bulletin of the Torrey Botanical Club 119 300ndash307doi1023072996761

BrownML Brown RG (1984) lsquoHerbaceous plants of Marylandrsquo (Universityof Maryland College Park MD)

Burgess JL Lev S Swan CM Szlavecz K (2009) Geologic and edaphiccontrols on a serpentine forest community Northeastern Naturalist 16366ndash384 doi1016560450160527

Butaye J Jacquemyn H Honnay O Hermy M (2002) The species poolconcept applied to forests in a fragmented landscape dispersal limitationversus habitat limitation Journal of Vegetation Science 13 27ndash34doi101111j1654-11032002tb02020x

Callaway RM Davis FW (1993) Vegetation dynamics fire and the physical-environment in coastal central California Rid B-7010-2009 Ecology 741567ndash1578 doi1023071940084

Carroll W D Kapeluck P R Harper R A and Van Lear D H (2002)Background paper historical overview of the southern forest landscapeand associated resources In lsquoSouthern forest resource assessmentrsquo (EdsDNWear JGGreis) pp 583ndash605 (USDepartmentofAgricultureForestService Southern Research Station Asheville NC)

Cerling TE Quade J Wang Y Bowman JR (1989) Carbon isotopes in soilsand palaeosols as ecology and palaeoecology indicators Nature 341138ndash139 doi101038341138a0

Chao A (1984) Nonparametric estimation of the number of classes in apopulation Scandinavian Journal of Statistics 11 265ndash270

ChaoA (1987) Estimating the population size for capturendashrecapture data withunequal catchability Biometrics 43 783ndash791 doi1023072531532

Chazdon RL Colwell RK Denslow JS Guariguata MR (1998) Statisticalmethods for estimating species richness of woody regeneration in primaryand secondary rain forests of Northeastern Costa Rica In lsquoForest

Conifer encroachment on ultramafics Australian Journal of Botany M

reconstructing vegetation changes (Deines 1980 Tieszen andArcher 1990 Boutton and Yamasaki 1996 Boutton et al 1998)Spatial patterns of soil d13C were strongly related to type ofvegetation (primarily woody versus grassland cover) Remnantgrasslands are composed of a mixture of C3 grasses and forbsand C4 grasses It is well documented that the d13C signature ofC3 plants (ranging from26permil to28permil) is easily distinguishedfrom that of the C4 plants (ranging from 12permil to 15permil)(DeNiro and Epstein 1977 Cerling et al 1989 Tieszen 1991Ehleringer et al 2000 Fox et al 2012 Throop et al 2013) Otherprocesses besides vegetation change alter isotopic ratios (such aspreferential degradation of some plant materials microbiologicalactivity Suess effect and adsorption processes) and may causeshifts of the order of 1ndash3permil (Friedli et al 1987Becker-HeidmannandScharpenseel 1992Trolier et al 1996Ehleringer et al 2000SantRuCkova et al 2000 Krull et al 2005) d13C enrichmentin temperate forest soils larger than 3ndash4permil is generally attributedto the presence of remaining old C4-derived SOC Following thereasoning of Wynn and Bird (2008) we sampled the upper soils(A horizon) and assumed that the intrinsic metabolic processesoccurring in C3 and C4 vegetation impart an unequivocald13C signature into the SOM and that these signatures persistin decaying organic matter such that increasing depth yields achronologic sequence recording vegetation change At the studyarea d13C values of soils in the grassland matrix averaged ndash16permil(Table 4) to 26permil in forested areas dominated by C3 plantsDepth-integrated SOM d13C significantly reflected the currentvegetation cover (mixed C3ndashC4 and solely C3) and accordinglycould be used as a reference to trace past changes in vegetationrecorded in the SOM at deeper soil layers The d13C values ofSOC consistently increased with depth reflecting a higherpercentage of C4-derived carbon in the older soils that isinterpreted to reflect a site-wide shift in vegetation from C4 toC3 plants

Multiple regression and ordination results supported the roleof soil depth and incident radiation both of which influence soilmoisture content as factors governing the occurrence of xericserpentine savannaQuercus species Abiotic factors in areas withhigher importance of shade-tolerant species were associated withdeeper soils higher bulk density and higher Ca Mg ratios In oursite early successional woodlands were favoured in areas of highincident radiation and broadleaf forests were more developed onslopes containing more clay rich and denser soils Diversity ofwoody specieswas higher in areaswith deeper soilsUnlikemanysystems with heavy metals we found no correlations betweentoxic metals (Cr and Ni) and vegetation dynamics

Historical records show that oak savanna ecosystems wereimportant components of the Mid-Atlantic landscape (Shreveet al 1910) The lack of any oaks before the 1850s suggests thatthey were likely harvested for fuel posts or to support grazingFarming of these serpentinite areas is not documented and isunlikely given the inhospitable nature of the soil Our studyindicated that tree recruitment before the 20th centurywasmostlyoaks perhaps regenerating from root sprouts or propagules in theseedbankXeric oak recruitment decreasedover timeandnoxericoaks younger than 40 years established in the forested areasalthough seedlings and saplings still exist in the remnantgrasslandsavannas This failure to regenerate is likely toreflect the consequences of changing disturbance regimes suchas fire-suppression (Nuzzo 1986 Abrams 1992 2006 PetersonandReich 2001) alongwith competitionwith invasive andmesictree species (Nuzzo 1986) White-tailed deer and squirrelherbivory can reduce recruitment especially in isolatedsystems such as these serpentinite landscapes (Shea andStange 1998 Haas and Heske 2005) A turning point forcommunity development appears to be ~1900 with theencroachment of Pinus virginiana and Juniperus virginianaThese conifer species have a variety of characteristics that can

30

25

Minor species

Mesic species (A rubrum F grandifolia N sylvatica P serotina)

Conifers

Quercus spp (Mixed)

Quercus spp (Xeric)

(n = 163)15

5

20

10

Num

ber

of tr

ees

sam

pled

Decade of origin

01850 1860 1870 1880 1890 1900 1910 1920 1930 1940 1950 1960 1970 1980 1990

Fig 8 Age distribution from trees on the basis of core data (n= 163)Xeric species areQuercusmarilandica andQ stellataMixed oak species includeQ albaQ montana Q rubra and Q velutina

Conifer encroachment on ultramafics Australian Journal of Botany K

potentially alter ecosystem processes for decades as they matureand senesce Both conifers are aggressive invaders because ofrapid growth high seed production high seed-dispersalefficiency tolerance of xeric conditions and opportunistic useof ECM mycorrhizae (Ormsbee et al 1976 Holthuijzen andSharik 1984 1985 Briggs andGibson 1992Axmann andKnapp1993 Thiet and Boerner 2007) Once established these speciescan have a profound effect on soil properties as a result ofaccumulation of organic matter on the forest floor and theeventual decomposition of detritus which leads to higherwater-holding capacity and lowering of soil pH This positivefeedback between soil depth and litter cover exerted by theconifers could promote succession towards non-serpentineconditions

The initial serpentine edaphic signature acts as a filterpreventing species that belong to the regional flora but lack thetraits required to survive in such conditions to successfullycolonise these habitats By the 1970s such filters appear tohave diminished Current site characteristics have shiftedtowards more mesic conditions with changes in soil propertiessuch as bulk density soil depth pH SOM and metalsconcentration Over time the community has changed fromxeric (grassland or oak savanna) to one dominated by moremesic oaks and shade-tolerant Acer Nyssa and Fagus speciesThemajor onset ofAcer establishment began in the 1940sUnlikethePinus and Juniperus establishment pulse which proceeded inan open-canopy situation Acer dominance likely proceededunder closed canopy and canopy gaps because it is shade-tolerant One may speculate that over the past 100ndash200 years

the landscape has been dominated by Quercus with patches ofPinus and Juniperus that historicallywere present at lownumbersbecauseof disturbance (miningfire grazing logging)Hilgartneret al (2009) recorded evidence of the presence of Pinusvirginiana from the early 1800s in a serpentine barren siteonly 50 km distant Over the past 60 years Acer has becomeabundant where it along with other species currently present(N sylvatica Cornus florida and Celtis occidentalis) wouldnormally be considered an aberrant species in a hostile edaphicsetting Age and diameter structure indicated a shift away fromQuercus and towards an A rubrum-dominated landscape Thissame species shift has been widely reported throughout theeastern deciduous forests of North America (Abrams 1998Nowacki and Abrams 2008 McEwan et al 2010)

In the context of serpentinite ecosystems the influxof conifersnearly a century ago has acted as a landscapemodulator (Shachaket al 2008) The resulting build-up of SOM increases the water-holding capacity and other resources such as ground-levelradiation facilitate the influx of new species adapted to thealtered conditions As the difference between the mesicwoodlands and the xeric patches decreases the resourcecontrasts also decrease leading to increasing communitysimilarity Different oak species can grow in a wide variety oflight and moisture conditions However none of the traditionalserpentine oak species is considered shade tolerant to the samedegree as are Acer spp Fagus grandifolia and Nyssa sylvaticaWith progressive development of the soil and a dense-canopyclosure growth and germination of the traditional xeric oakspecies are likely to be suppressed marking a shift from

Colonisation

Fig 9 Conceptual model for serpentine (orangendashred colouring) habitats and species redistribution resulting in similar structure for a typical piedmontgeoecologic system (bluered colouring) Bedrock geology (specifically elemental concentrations and structures) exerts its control on soil fertility with subtlevariations resulting in markedly different above-ground and below-ground communities The change in soils over time (dsdt) equation from Huggett (1995)is a function of the various drivers including b (biology) r (topography) a (atmospheric effects) s (soils) and h (hydrology) Through changing vegetationdynamics and species redistribution the tree functional groups and forest types will see increasing similarity as a result of converging successional trajectoriesRegional species pools stochastic events and many other factors determine ultimate outcomes

L Australian Journal of Botany J Burgess et al

xerothermic to mesic andor nutrient-demanding speciesFigure 9 details a conceptual model for pedologic and ecologicchanges at serpentine sites of theMid-Atlantic with successionaltrajectories converging on a species redistribution and loss oftypical serpentine species such as Cerastium arvense varvillosissimum Andropogon gerardii and the xeric oaks(Q marilandica and Q stellata)

Although the canopy is still oak or pine dominated thetrajectory points to a maplendashbeechndashred oak-dominated forestThe Mid-Atlantic has already lost much of its originalgrasslandndashoak woodlandndashforest mosaic This state transitionmay lead to the loss of a unique ecosystem and thustaxonomic homogenisation with the regional flora Thesechanges result from anthropic influences through changes inland use and fire frequency the loss of Castanea dentata(American chestnut) and changing herbivore populationsNearly a century ago these areas gave way to a mosaic ofsmall woodlands typically characterised by xeric oaksinterspersed with shallow rooting low nutrient-tolerantconifers Released from suppressive disturbance regimes(herbivory and fire for example) Pinus virginiana andJuniperus virginiana quickly became the dominant speciesHowever our data suggest that the landscape is nowundergoing another distinct change as the once-open canopystructure has becomemore closed Edaphic properties such as soildepth chemistry and moisture that once drove communityorganisation across environmental stress gradients are nowbeing modulated by species no longer filtered by theserpentine syndrome This modulation may allow for therecruitment and establishment of drought-sensitive speciesbeyond their typical tolerance limits (Warman and Moles2008) This unique ecosystem which took centuries tomillennia to evolve appears to be fading and may soonevanesce into the past

Sources of funding

Our work was funded by the National Science Foundation(DEB-0423476 and DEB-1027188) and grants from theR Balk and D Elliott Field Fund (Johns Hopkins University)

Conflicts of interest

No conflicts of interest

Acknowledgements

We thank Drs Richard Back Naomi Levin Bruce Marsh and Ben Passey foruse of their elemental and isotopic facilities We also thank Dr David Veblenand two anonymous reviewers for numerous constructive comments on thegeology

References

Abbe C (1898) A general report on the physiography of Maryland JohnsHopkins University (Johns Hopkins University Press Baltimore MD)

Abrams MD (1992) Fire and the development of oak forests Bioscience 42346ndash353 doi1023071311781

Abrams MD (1998) The red maple paradox Bioscience 48 355ndash364doi1023071313374

AbramsMD (2006) Ecological and ecophysiological attributes and responsesto fire in eastern oak forests In lsquoFire in eastern oak forests deliveringscience to landmanagersrsquo (EdMBDickinson) pp 74ndash89 (USDA ForestService General Technical Report NRS-P-1)

Alexander EB (2009) Serpentine geoecology of the eastern and southeasternmargins of North America Northeastern Naturalist 16 223ndash252doi1016560450160518

Alexander EB Ellis CC Burke R (2007) A chronosequence of soils andvegetation on serpentine terraces in the Klamath Mountains USA SoilScience 172 565ndash576 doi101097ss0b013e31804fa22d

Arabas KB (2000) Spatial and temporal relationships among fire frequencyvegetation and soil depth in an easternNorthAmerican serpentine barrenThe Journal of the Torrey Botanical Society 127 51ndash65doi1023073088747

Archer S Schimel DS Holland EA (1995) Mechanisms of shrublandexpansion land use climate or CO2 Climatic Change 29 91ndash99doi101007BF01091640

Axmann BD Knapp AK (1993) Water relations of Juniperus virginiana andAndropogon gerardii in an unburned tallgrass prairie watershed TheSouthwestern Naturalist 38 325ndash330 doi1023073671610

Barton AM Wallenstein MD (1997) Effects of invasion of Pinus virginianaon soil properties in serpentine barrens in southeastern PennsylvaniaThe Journal of the Torrey Botanical Society 124 297ndash305doi1023072997264

Bazzaz FA (1968) Succession on abandoned fields in the Shawnee Hillssouthern Illinois Ecology 49 924 doi1023071936544

Becker-Heidmann P Scharpenseel H-W (1992) Studies of soil organicmatterdynamics using natural carbon isotopes The Science of the TotalEnvironment 117ndash118 305ndash312 doi1010160048-9697(92)90097-C

Belcher JW Keddy PA Twolan-Strutt L (1995) Root and shoot competitionintensity along a soil depth gradient Journal of Ecology 83 673ndash682doi1023072261635

BouttonTWArcher SRMidwoodAJZitzer SFBolR (1998)d13Cvalues ofsoil organic carbon and their use in documenting vegetation change in asubtropical savanna ecosystem Geoderma 82 5ndash41doi101016S0016-7061(97)00095-5

Bouyoucos GJ (1936) Directions for making mechanical analyses of soilsby the hydrometer method Soil Science 42 225ndash230doi10109700010694-193609000-00007

Briggs JMGibsonDJ (1992)Effect offire on tree spatial patterns in a tallgrassprairie landscape Bulletin of the Torrey Botanical Club 119 300ndash307doi1023072996761

BrownML Brown RG (1984) lsquoHerbaceous plants of Marylandrsquo (Universityof Maryland College Park MD)

Burgess JL Lev S Swan CM Szlavecz K (2009) Geologic and edaphiccontrols on a serpentine forest community Northeastern Naturalist 16366ndash384 doi1016560450160527

Butaye J Jacquemyn H Honnay O Hermy M (2002) The species poolconcept applied to forests in a fragmented landscape dispersal limitationversus habitat limitation Journal of Vegetation Science 13 27ndash34doi101111j1654-11032002tb02020x

Callaway RM Davis FW (1993) Vegetation dynamics fire and the physical-environment in coastal central California Rid B-7010-2009 Ecology 741567ndash1578 doi1023071940084

Carroll W D Kapeluck P R Harper R A and Van Lear D H (2002)Background paper historical overview of the southern forest landscapeand associated resources In lsquoSouthern forest resource assessmentrsquo (EdsDNWear JGGreis) pp 583ndash605 (USDepartmentofAgricultureForestService Southern Research Station Asheville NC)

Cerling TE Quade J Wang Y Bowman JR (1989) Carbon isotopes in soilsand palaeosols as ecology and palaeoecology indicators Nature 341138ndash139 doi101038341138a0

Chao A (1984) Nonparametric estimation of the number of classes in apopulation Scandinavian Journal of Statistics 11 265ndash270

ChaoA (1987) Estimating the population size for capturendashrecapture data withunequal catchability Biometrics 43 783ndash791 doi1023072531532

Chazdon RL Colwell RK Denslow JS Guariguata MR (1998) Statisticalmethods for estimating species richness of woody regeneration in primaryand secondary rain forests of Northeastern Costa Rica In lsquoForest

Conifer encroachment on ultramafics Australian Journal of Botany M

potentially alter ecosystem processes for decades as they matureand senesce Both conifers are aggressive invaders because ofrapid growth high seed production high seed-dispersalefficiency tolerance of xeric conditions and opportunistic useof ECM mycorrhizae (Ormsbee et al 1976 Holthuijzen andSharik 1984 1985 Briggs andGibson 1992Axmann andKnapp1993 Thiet and Boerner 2007) Once established these speciescan have a profound effect on soil properties as a result ofaccumulation of organic matter on the forest floor and theeventual decomposition of detritus which leads to higherwater-holding capacity and lowering of soil pH This positivefeedback between soil depth and litter cover exerted by theconifers could promote succession towards non-serpentineconditions

The initial serpentine edaphic signature acts as a filterpreventing species that belong to the regional flora but lack thetraits required to survive in such conditions to successfullycolonise these habitats By the 1970s such filters appear tohave diminished Current site characteristics have shiftedtowards more mesic conditions with changes in soil propertiessuch as bulk density soil depth pH SOM and metalsconcentration Over time the community has changed fromxeric (grassland or oak savanna) to one dominated by moremesic oaks and shade-tolerant Acer Nyssa and Fagus speciesThemajor onset ofAcer establishment began in the 1940sUnlikethePinus and Juniperus establishment pulse which proceeded inan open-canopy situation Acer dominance likely proceededunder closed canopy and canopy gaps because it is shade-tolerant One may speculate that over the past 100ndash200 years

the landscape has been dominated by Quercus with patches ofPinus and Juniperus that historicallywere present at lownumbersbecauseof disturbance (miningfire grazing logging)Hilgartneret al (2009) recorded evidence of the presence of Pinusvirginiana from the early 1800s in a serpentine barren siteonly 50 km distant Over the past 60 years Acer has becomeabundant where it along with other species currently present(N sylvatica Cornus florida and Celtis occidentalis) wouldnormally be considered an aberrant species in a hostile edaphicsetting Age and diameter structure indicated a shift away fromQuercus and towards an A rubrum-dominated landscape Thissame species shift has been widely reported throughout theeastern deciduous forests of North America (Abrams 1998Nowacki and Abrams 2008 McEwan et al 2010)

In the context of serpentinite ecosystems the influxof conifersnearly a century ago has acted as a landscapemodulator (Shachaket al 2008) The resulting build-up of SOM increases the water-holding capacity and other resources such as ground-levelradiation facilitate the influx of new species adapted to thealtered conditions As the difference between the mesicwoodlands and the xeric patches decreases the resourcecontrasts also decrease leading to increasing communitysimilarity Different oak species can grow in a wide variety oflight and moisture conditions However none of the traditionalserpentine oak species is considered shade tolerant to the samedegree as are Acer spp Fagus grandifolia and Nyssa sylvaticaWith progressive development of the soil and a dense-canopyclosure growth and germination of the traditional xeric oakspecies are likely to be suppressed marking a shift from

Colonisation

Fig 9 Conceptual model for serpentine (orangendashred colouring) habitats and species redistribution resulting in similar structure for a typical piedmontgeoecologic system (bluered colouring) Bedrock geology (specifically elemental concentrations and structures) exerts its control on soil fertility with subtlevariations resulting in markedly different above-ground and below-ground communities The change in soils over time (dsdt) equation from Huggett (1995)is a function of the various drivers including b (biology) r (topography) a (atmospheric effects) s (soils) and h (hydrology) Through changing vegetationdynamics and species redistribution the tree functional groups and forest types will see increasing similarity as a result of converging successional trajectoriesRegional species pools stochastic events and many other factors determine ultimate outcomes

L Australian Journal of Botany J Burgess et al

xerothermic to mesic andor nutrient-demanding speciesFigure 9 details a conceptual model for pedologic and ecologicchanges at serpentine sites of theMid-Atlantic with successionaltrajectories converging on a species redistribution and loss oftypical serpentine species such as Cerastium arvense varvillosissimum Andropogon gerardii and the xeric oaks(Q marilandica and Q stellata)

Although the canopy is still oak or pine dominated thetrajectory points to a maplendashbeechndashred oak-dominated forestThe Mid-Atlantic has already lost much of its originalgrasslandndashoak woodlandndashforest mosaic This state transitionmay lead to the loss of a unique ecosystem and thustaxonomic homogenisation with the regional flora Thesechanges result from anthropic influences through changes inland use and fire frequency the loss of Castanea dentata(American chestnut) and changing herbivore populationsNearly a century ago these areas gave way to a mosaic ofsmall woodlands typically characterised by xeric oaksinterspersed with shallow rooting low nutrient-tolerantconifers Released from suppressive disturbance regimes(herbivory and fire for example) Pinus virginiana andJuniperus virginiana quickly became the dominant speciesHowever our data suggest that the landscape is nowundergoing another distinct change as the once-open canopystructure has becomemore closed Edaphic properties such as soildepth chemistry and moisture that once drove communityorganisation across environmental stress gradients are nowbeing modulated by species no longer filtered by theserpentine syndrome This modulation may allow for therecruitment and establishment of drought-sensitive speciesbeyond their typical tolerance limits (Warman and Moles2008) This unique ecosystem which took centuries tomillennia to evolve appears to be fading and may soonevanesce into the past

Sources of funding

Our work was funded by the National Science Foundation(DEB-0423476 and DEB-1027188) and grants from theR Balk and D Elliott Field Fund (Johns Hopkins University)

Conflicts of interest

No conflicts of interest

Acknowledgements

We thank Drs Richard Back Naomi Levin Bruce Marsh and Ben Passey foruse of their elemental and isotopic facilities We also thank Dr David Veblenand two anonymous reviewers for numerous constructive comments on thegeology

References

Abbe C (1898) A general report on the physiography of Maryland JohnsHopkins University (Johns Hopkins University Press Baltimore MD)

Abrams MD (1992) Fire and the development of oak forests Bioscience 42346ndash353 doi1023071311781

Abrams MD (1998) The red maple paradox Bioscience 48 355ndash364doi1023071313374

AbramsMD (2006) Ecological and ecophysiological attributes and responsesto fire in eastern oak forests In lsquoFire in eastern oak forests deliveringscience to landmanagersrsquo (EdMBDickinson) pp 74ndash89 (USDA ForestService General Technical Report NRS-P-1)

Alexander EB (2009) Serpentine geoecology of the eastern and southeasternmargins of North America Northeastern Naturalist 16 223ndash252doi1016560450160518

Alexander EB Ellis CC Burke R (2007) A chronosequence of soils andvegetation on serpentine terraces in the Klamath Mountains USA SoilScience 172 565ndash576 doi101097ss0b013e31804fa22d

Arabas KB (2000) Spatial and temporal relationships among fire frequencyvegetation and soil depth in an easternNorthAmerican serpentine barrenThe Journal of the Torrey Botanical Society 127 51ndash65doi1023073088747

Archer S Schimel DS Holland EA (1995) Mechanisms of shrublandexpansion land use climate or CO2 Climatic Change 29 91ndash99doi101007BF01091640

Axmann BD Knapp AK (1993) Water relations of Juniperus virginiana andAndropogon gerardii in an unburned tallgrass prairie watershed TheSouthwestern Naturalist 38 325ndash330 doi1023073671610

Barton AM Wallenstein MD (1997) Effects of invasion of Pinus virginianaon soil properties in serpentine barrens in southeastern PennsylvaniaThe Journal of the Torrey Botanical Society 124 297ndash305doi1023072997264

Bazzaz FA (1968) Succession on abandoned fields in the Shawnee Hillssouthern Illinois Ecology 49 924 doi1023071936544

Becker-Heidmann P Scharpenseel H-W (1992) Studies of soil organicmatterdynamics using natural carbon isotopes The Science of the TotalEnvironment 117ndash118 305ndash312 doi1010160048-9697(92)90097-C

Belcher JW Keddy PA Twolan-Strutt L (1995) Root and shoot competitionintensity along a soil depth gradient Journal of Ecology 83 673ndash682doi1023072261635

BouttonTWArcher SRMidwoodAJZitzer SFBolR (1998)d13Cvalues ofsoil organic carbon and their use in documenting vegetation change in asubtropical savanna ecosystem Geoderma 82 5ndash41doi101016S0016-7061(97)00095-5

Bouyoucos GJ (1936) Directions for making mechanical analyses of soilsby the hydrometer method Soil Science 42 225ndash230doi10109700010694-193609000-00007

Briggs JMGibsonDJ (1992)Effect offire on tree spatial patterns in a tallgrassprairie landscape Bulletin of the Torrey Botanical Club 119 300ndash307doi1023072996761

BrownML Brown RG (1984) lsquoHerbaceous plants of Marylandrsquo (Universityof Maryland College Park MD)

Burgess JL Lev S Swan CM Szlavecz K (2009) Geologic and edaphiccontrols on a serpentine forest community Northeastern Naturalist 16366ndash384 doi1016560450160527

Butaye J Jacquemyn H Honnay O Hermy M (2002) The species poolconcept applied to forests in a fragmented landscape dispersal limitationversus habitat limitation Journal of Vegetation Science 13 27ndash34doi101111j1654-11032002tb02020x

Callaway RM Davis FW (1993) Vegetation dynamics fire and the physical-environment in coastal central California Rid B-7010-2009 Ecology 741567ndash1578 doi1023071940084

Carroll W D Kapeluck P R Harper R A and Van Lear D H (2002)Background paper historical overview of the southern forest landscapeand associated resources In lsquoSouthern forest resource assessmentrsquo (EdsDNWear JGGreis) pp 583ndash605 (USDepartmentofAgricultureForestService Southern Research Station Asheville NC)

Cerling TE Quade J Wang Y Bowman JR (1989) Carbon isotopes in soilsand palaeosols as ecology and palaeoecology indicators Nature 341138ndash139 doi101038341138a0

Chao A (1984) Nonparametric estimation of the number of classes in apopulation Scandinavian Journal of Statistics 11 265ndash270

ChaoA (1987) Estimating the population size for capturendashrecapture data withunequal catchability Biometrics 43 783ndash791 doi1023072531532

Chazdon RL Colwell RK Denslow JS Guariguata MR (1998) Statisticalmethods for estimating species richness of woody regeneration in primaryand secondary rain forests of Northeastern Costa Rica In lsquoForest

Conifer encroachment on ultramafics Australian Journal of Botany M

xerothermic to mesic andor nutrient-demanding speciesFigure 9 details a conceptual model for pedologic and ecologicchanges at serpentine sites of theMid-Atlantic with successionaltrajectories converging on a species redistribution and loss oftypical serpentine species such as Cerastium arvense varvillosissimum Andropogon gerardii and the xeric oaks(Q marilandica and Q stellata)

Although the canopy is still oak or pine dominated thetrajectory points to a maplendashbeechndashred oak-dominated forestThe Mid-Atlantic has already lost much of its originalgrasslandndashoak woodlandndashforest mosaic This state transitionmay lead to the loss of a unique ecosystem and thustaxonomic homogenisation with the regional flora Thesechanges result from anthropic influences through changes inland use and fire frequency the loss of Castanea dentata(American chestnut) and changing herbivore populationsNearly a century ago these areas gave way to a mosaic ofsmall woodlands typically characterised by xeric oaksinterspersed with shallow rooting low nutrient-tolerantconifers Released from suppressive disturbance regimes(herbivory and fire for example) Pinus virginiana andJuniperus virginiana quickly became the dominant speciesHowever our data suggest that the landscape is nowundergoing another distinct change as the once-open canopystructure has becomemore closed Edaphic properties such as soildepth chemistry and moisture that once drove communityorganisation across environmental stress gradients are nowbeing modulated by species no longer filtered by theserpentine syndrome This modulation may allow for therecruitment and establishment of drought-sensitive speciesbeyond their typical tolerance limits (Warman and Moles2008) This unique ecosystem which took centuries tomillennia to evolve appears to be fading and may soonevanesce into the past

Sources of funding

Our work was funded by the National Science Foundation(DEB-0423476 and DEB-1027188) and grants from theR Balk and D Elliott Field Fund (Johns Hopkins University)

Conflicts of interest

No conflicts of interest

Acknowledgements

We thank Drs Richard Back Naomi Levin Bruce Marsh and Ben Passey foruse of their elemental and isotopic facilities We also thank Dr David Veblenand two anonymous reviewers for numerous constructive comments on thegeology

References

Abbe C (1898) A general report on the physiography of Maryland JohnsHopkins University (Johns Hopkins University Press Baltimore MD)

Abrams MD (1992) Fire and the development of oak forests Bioscience 42346ndash353 doi1023071311781

Abrams MD (1998) The red maple paradox Bioscience 48 355ndash364doi1023071313374

AbramsMD (2006) Ecological and ecophysiological attributes and responsesto fire in eastern oak forests In lsquoFire in eastern oak forests deliveringscience to landmanagersrsquo (EdMBDickinson) pp 74ndash89 (USDA ForestService General Technical Report NRS-P-1)

Alexander EB (2009) Serpentine geoecology of the eastern and southeasternmargins of North America Northeastern Naturalist 16 223ndash252doi1016560450160518

Alexander EB Ellis CC Burke R (2007) A chronosequence of soils andvegetation on serpentine terraces in the Klamath Mountains USA SoilScience 172 565ndash576 doi101097ss0b013e31804fa22d

Arabas KB (2000) Spatial and temporal relationships among fire frequencyvegetation and soil depth in an easternNorthAmerican serpentine barrenThe Journal of the Torrey Botanical Society 127 51ndash65doi1023073088747

Archer S Schimel DS Holland EA (1995) Mechanisms of shrublandexpansion land use climate or CO2 Climatic Change 29 91ndash99doi101007BF01091640

Axmann BD Knapp AK (1993) Water relations of Juniperus virginiana andAndropogon gerardii in an unburned tallgrass prairie watershed TheSouthwestern Naturalist 38 325ndash330 doi1023073671610

Barton AM Wallenstein MD (1997) Effects of invasion of Pinus virginianaon soil properties in serpentine barrens in southeastern PennsylvaniaThe Journal of the Torrey Botanical Society 124 297ndash305doi1023072997264

Bazzaz FA (1968) Succession on abandoned fields in the Shawnee Hillssouthern Illinois Ecology 49 924 doi1023071936544

Becker-Heidmann P Scharpenseel H-W (1992) Studies of soil organicmatterdynamics using natural carbon isotopes The Science of the TotalEnvironment 117ndash118 305ndash312 doi1010160048-9697(92)90097-C

Belcher JW Keddy PA Twolan-Strutt L (1995) Root and shoot competitionintensity along a soil depth gradient Journal of Ecology 83 673ndash682doi1023072261635

BouttonTWArcher SRMidwoodAJZitzer SFBolR (1998)d13Cvalues ofsoil organic carbon and their use in documenting vegetation change in asubtropical savanna ecosystem Geoderma 82 5ndash41doi101016S0016-7061(97)00095-5

Bouyoucos GJ (1936) Directions for making mechanical analyses of soilsby the hydrometer method Soil Science 42 225ndash230doi10109700010694-193609000-00007

Briggs JMGibsonDJ (1992)Effect offire on tree spatial patterns in a tallgrassprairie landscape Bulletin of the Torrey Botanical Club 119 300ndash307doi1023072996761

BrownML Brown RG (1984) lsquoHerbaceous plants of Marylandrsquo (Universityof Maryland College Park MD)

Burgess JL Lev S Swan CM Szlavecz K (2009) Geologic and edaphiccontrols on a serpentine forest community Northeastern Naturalist 16366ndash384 doi1016560450160527

Butaye J Jacquemyn H Honnay O Hermy M (2002) The species poolconcept applied to forests in a fragmented landscape dispersal limitationversus habitat limitation Journal of Vegetation Science 13 27ndash34doi101111j1654-11032002tb02020x

Callaway RM Davis FW (1993) Vegetation dynamics fire and the physical-environment in coastal central California Rid B-7010-2009 Ecology 741567ndash1578 doi1023071940084

Carroll W D Kapeluck P R Harper R A and Van Lear D H (2002)Background paper historical overview of the southern forest landscapeand associated resources In lsquoSouthern forest resource assessmentrsquo (EdsDNWear JGGreis) pp 583ndash605 (USDepartmentofAgricultureForestService Southern Research Station Asheville NC)

Cerling TE Quade J Wang Y Bowman JR (1989) Carbon isotopes in soilsand palaeosols as ecology and palaeoecology indicators Nature 341138ndash139 doi101038341138a0

Chao A (1984) Nonparametric estimation of the number of classes in apopulation Scandinavian Journal of Statistics 11 265ndash270

ChaoA (1987) Estimating the population size for capturendashrecapture data withunequal catchability Biometrics 43 783ndash791 doi1023072531532

Chazdon RL Colwell RK Denslow JS Guariguata MR (1998) Statisticalmethods for estimating species richness of woody regeneration in primaryand secondary rain forests of Northeastern Costa Rica In lsquoForest

Conifer encroachment on ultramafics Australian Journal of Botany M

biodiversity research monitoring and modeling conceptual backgroundandoldworld case studiesrsquo (EdsFDallmeier JAComiskey)pp 285ndash309(Center for International Forestry Research Paris)

Clements FE (1916) lsquoPlant succession an analysis of the development ofvegetation No 242rsquo (Carnegie Institution of Washington WashingtonDC)

Cumming JR Kelly CN (2007) Pinus virginiana invasion influences soilsand arbuscular mycorrhizae of a serpentine grassland The Journal of theTorrey Botanical Society 134 63ndash73doi1031591095-5674(2007)134[63PVIISA]20CO2

Curtis JT (1959) lsquoThe vegetation of Wisconsin an ordination of plantcommunitiesrsquo (University of Wisconsin Press Madison WI)

Deines P (1980) The carbon isotopic composition of diamonds relationshipto diamond shape color occurrence and vapor compositionGeochimicaet CosmochimicaActa 44 943ndash961 doi1010160016-7037(80)90284-7

DeNiro MJ Epstein S (1977) Mechanism of carbon isotope fractionationassociated with lipid synthesis Science 197 261ndash263doi101126science327543

Diacuteaz S Cabido M Casanover F (1998) Plant functional traits andenvironmental filters at a regional scale Journal of Vegetation Science9 113ndash122 doi1023073237229

Ehleringer JR Buchmann N Flanagan LB (2000) Carbon isotope ratios inbelowground carbon cycle processes Ecological Applications 10412ndash422 doi1018901051-0761(2000)010[0412CIRIBC]20CO2

Fernandez-GoingBMHarrison SPAnacker BL SaffordHD (2013)Climateinteracts with soil to produce beta diversity in Californian plantcommunities Ecology 94 2007ndash2018 doi10189012-20111

Foster D Swanson F Aber J Burke I BrokawN Tilman D Knapp A (2003)The importance of land-use legacies to ecology and conservationBioscience 53 77ndash88doi1016410006-3568(2003)053[0077TIOLUL]20CO2

Fox DL Honey JG Martin RA Pelaez-Campomanes P (2012) Pedogeniccarbonate stable isotope record of environmental change during theNeogene in the southern Great Plains southwest Kansas USA carbonisotopes and the evolution of C4-dominated grasslands GeologicalSociety of America Bulletin 124 444ndash462 doi101130B304011

Friedli H Siegenthaler U Rauber D Oeschger H (1987) Measurements ofconcentration 13C12C and 18O16O ratios of tropospheric carbon dioxideover Switzerland Tellus 39 80doi101111j1600-08891987tb00272x

Gates AE (1992) Domainal failure of serpentinite in shear zones state-linemafic complex Pennsylvania USA Journal of Structural Geology 1419ndash28 doi1010160191-8141(92)90141-I

Gleason HA (1926) The individualistic concept of the plant associationBulletin of the Torrey Botanical Club 53 7ndash26 doi1023072479933

Goumltzenberger L de Bello F Braringthen KA Davison J Dubuis A Guisan ALepš J LindborgRMooraM PaumlrtelM (2012) Ecological assembly rulesin plant communities ndash approaches patterns and prospects BiologicalReviews of the Cambridge Philosophical Society 87 111ndash127doi101111j1469-185X201100187x

Grime JP (1979) Primary strategies in plants Transactions of the BotanicalSociety of Edinburgh 43 151ndash160

Haas JP Heske EJ (2005) Experimental study of the effects of mammalianacorn predators on red oak acorn survival and germination Journal ofMammalogy 86 1015ndash1021doi1016441545-1542(2005)86[1015ESOTEO]20CO2

Harris T Rajakaruna N (2009) Adiantum viridimontanum Aspidotis densaMinuartiamarcescens and Symphyotrichum rhiannon additionalserpentine endemics from eastern North America NortheasternNaturalist 16 111ndash120 doi1016560450160509

Harrison SP Rajakaruna N (2011) lsquoSerpentine The evolution and ecology ofa model systemrsquo (University of California Press Berkeley CA)

HilgartnerWBNejakoMCaseyR (2009)A200-year paleoecological recordof Pinus virginiana trace metals sedimentation and mining disturbance

in a Maryland serpentine barren 1 The Journal of the Torrey BotanicalSociety 136 257ndash271 doi10315908-RA-0881

Holthuijzen AM Sharik TL (1984) Seed longevity and mechanisms ofregeneration of eastern red cedar (Juniperus virginiana L) Bulletin ofthe Torrey Botanical Club 111 153ndash158 doi1023072996014

Holthuijzen AM Sharik TL (1985) Colonization of abandoned pastures byeastern red cedar (Juniperus virginiana L) Canadian Journal ofResearch 15 1065ndash1068 doi101139x85-173

HuggettRJ (1995) lsquoGeoecologyAnEvaluationApproachrsquo (RoutledgeNewYork NY)

Huggett RJ (1998) Soil chronosequences soil development and soilevolution a critical review Catena 32 155ndash172doi101016S0341-8162(98)00053-8

Huston M (1979) A general hypothesis of species diversity AmericanNaturalist 113 81ndash101 doi101086283366

Inouye RS Huntly NJ Tilman D Tester JR Stillwell M Zinnel KC (1987)Old-field succession on a Minnesota sand plain Ecology 68 12ndash26doi1023071938801

JefferisWW (1880)Menaccanite and talc fromMarylandProceedings of theAcademy of Natural Sciences of Philadelphia 32 292

JennyH (1941) lsquoFactors of soil formation a systemof quantitative pedologyrsquo(McGraw-Hill Book Co New York)

Jenny H (1980) lsquoThe soil resource Origin and behaviorrsquo (Springer-VerlagNew York)

Johnson EA Miyanishi K (2008) Testing the assumptions ofchronosequences in succession Ecology Letters 11 419ndash431doi101111j1461-0248200801173x

Jones CC del Moral RD (2009) Dispersal and establishment both limitcolonization during primary succession on a glacier foreland PlantEcology 204 217ndash230 doi101007s11258-009-9586-3

Keddy PA (1992) Assembly and response rules two goals for predictivecommunity ecology Journal of Vegetation Science 3 157ndash164doi1023073235676

Keever C (1950) Causes of succession on old fields of the PiedmontNorth Carolina Ecological Monographs 20 229ndash250doi1023071948582

Kindt R Coe R (2005) lsquoTree diversity analysis a manual and software forcommon statistical methods for ecological and biodiversity studiesrsquo(World Agroforestry Centre (ICRAF) Nairobi Kenya)

Kruckeberg AR (1984) lsquoCalifornia serpentines flora vegetation geologysoils and management problemsrsquo (University of California PressBerkely CA)

Krull ES Skjemstad JO BurrowsWH Bray SGWynn JG Bol R SpouncerL Harms B (2005) Recent vegetation changes in central QueenslandAustralia evidence from C13C14 analyses of soil organic matterGeoderma 126 241ndash259 doi101016jgeoderma200409012

Latham RE (1993) Global patterns of tree species richness in moist forestsenergy-diversity theory does not account for variation in species richnessOikos 67 325ndash333 doi1023073545479

Lawton JH (1999) Are there general laws in ecology Oikos 84 177ndash192doi1023073546712

Lorimer CG (1980) Age structure and disturbance history of a southernAppalachian virgin forestEcology61 1169ndash1184 doi1023071936836

Lortie CJ Brooker RW Choler P Kikvidze Z Michalet R Pugnaire FICallaway RM (2004) Rethinking plant community theory Oikos 107433ndash438 doi101111j0030-1299200413250x

Ludwig JA Reynolds JF (1988) lsquoStatistical ecology a primer onmethods andcomputingrsquo (J Wiley amp Sons New York)

Mantel N (1967) Ranking procedures for arbitrarily restricted observationBiometrics 23 65ndash78 doi1023072528282

Marye WB (1955a) The great Maryland barrens I Maryland HistoricalMagazine 50 11ndash23

Marye WB (1955b) The great Maryland barrens II Maryland HistoricalMagazine 50 120ndash142

N Australian Journal of Botany J Burgess et al

Marye WB (1955c) The great Maryland barrens III Maryland HistoricalMagazine 50 234ndash253

Maryland State Climatologist Office (2014) lsquoState Climatologists ReportrsquoMaryland State Climatologist Office Available at wwwatmosumdeduclimate [Verified August 2014]

McCuneBGrace JB (2002) lsquoAnalysis of ecological communities 28rsquo (MjMSoftware Design Gleneden Beach OR)

McCune B Keon D (2002) Equations for potential annual direct incidentradiation and heat load Journal of Vegetation Science 13 603ndash606doi101111j1654-11032002tb02087x

McEwan RW Dyer JM Pederson N (2010) Multiple interacting ecosystemdrivers towardan encompassinghypothesis ofoak forest dynamics acrosseastern North America Ecography 34 244ndash256doi101111j1600-0587201006390x

MisraKCKellerFB (1978)Ultramaficbodies in the southernAppalachians areview American Journal of Science 278 389ndash418doi102475ajs2784389

Mittwede SK Stoddard EF (1989) lsquoUltramafic rocks of the AppalachianPiedmontrsquo (Geological Society of America Boulder CO)

Muller PD Candela PA Wylie AG (1989) Liberty complex polygeneticmelange in the centralMarylandPiedmont lsquoLiberty complexpolygeneticmelange in the central Maryland Piedmontrsquo (Eds JW Horton N Rast)pp 113ndash134 (Geological Society of America Boulder CO)

Noss RF (2013) Natural history of a forgotten American grassland InlsquoForgotten grasslands of the Southrsquo pp 1ndash32 (Island PressCenter forResource Economics Washington DC)

Nowacki GJ Abrams MD (2008) The demise of fire and lsquomesophicationrsquo offorests in the eastern United States Bioscience 58 123doi101641B580207

Nuzzo VA (1986) Extent and status of Midwest oak savanna presettlementand 1985 Natural Areas Journal 6 6ndash36

Oosting HJ (1942) An ecological analysis of the plant communities ofPiedmont North Carolina American Midland Naturalist 28 1ndash126doi1023072420696

Ormsbee P Bazzaz FA Boggess WR (1976) Physiological ecology ofJuniperus virginiana in oldfields Oecologia 23 75ndash82doi101007BF00351216

Owen W (2002) The history of native plant communities in the south Thesouthern forest resource assessment technical report US Forest Servicegeneral technical report SRS-53 pp 47ndash62 (USDA Forest ServiceSouthern Research Station Asheville NC)

Oze C Fendorf S Bird DK (2004) Chromium geochemistry in serpentinizedultramafic rocks and serpentine soils from the Franciscan complex ofCalifornia American Journal of Science 304 67ndash101doi102475ajs304167

Peterson DW Reich PB (2001) Prescribed fire in oak savanna fire frequencyeffects on stand structure and dynamics Ecological Applications 11914ndash927 doi1018901051-0761(2001)011[0914PFIOSF]20CO2

Pickett ST White PS (1985) lsquoThe ecology of natural disturbance and patchdynamicsrsquo (Academic Press Orlando FL)

Potts PJ (1987) lsquoA handbook of silicate rock analysisrsquo (Blackie and SonsGlasgow)

Proctor J Nagy L (1992) Ultramafic rocks and their vegetation an overviewIn lsquoThe vegetation of ultramafic (serpentine) soilsrsquo (Eds AJM Baker JProctor RD Reeves) pp 469ndash494 (Intercept Andover UK)

Rajakaruna N Boyd RS (2009) Advances in serpentine geoecology aretrospective Northeastern Naturalist 16 1ndash7doi1016560450160501

Rajakaruna N Harris TB Alexander EB (2009) Serpentine geoecology ofeastern North America a review Rhodora 111 21ndash108doi10311907-231

Robertson GP Hutson MA Evans FC Tiedje JM (1988) Spatial variabilityin a successional plant community patterns of nitrogen availabilityEcology 69 1517ndash1524 doi1023071941649

SantRuCkova H Bird MI Lloyd J (2000) Microbial processes and carbon-isotope fractionation in tropical and temperate grassland soilsFunctionalEcology 14 108ndash114 doi101046j1365-2435200000402x

ShachakMBoekenBGroner EKadmonRLubinYMeronENersquoEmanGPerevolotskyA ShkedyYUngarED (2008)Woody species as landscapemodulators and their effect on biodiversity patterns Bioscience 58209ndash221 doi101641B580307

Shannon CE Weaver W (1949) lsquoThe mathematical theory ofcommunicationrsquo (University of Illinois Press Urbana IL)

Shea KL Stange EE (1998) Effects of deer browsing fabric mats and treeshelters on Quercus rubra seedlings Restoration Ecology 6 29ndash34doi101046j1526-100x199800614x

Shreve F Chrysler MA Blodgett FH Besley FW (1910) lsquoThe plant life ofMarylandrsquoSpecial publication vol III (JohnsHopkins Press BaltimoreMD)

Simpson EH (1949) Measurement of diversity Nature 163 688doi101038163688a0

Smith RC Barnes JH (2008) Geology of the Goat Hill serpentine barrensBaltimore Mafic Complex Pennsylvania Journal of the PennsylvaniaAcademy of Science 82 19ndash30

Speer JH (2010) lsquoFundamentals of tree-ring researchrsquo (University of ArizonaPress Tuscan AZ)

Stokes MA (1996) lsquoAn introduction to tree-ring datingrsquo (University ofArizona Press)

Stokes MA Smiley TL (1968) lsquoAn introduction to tree-ring datingrsquo(University of Chicago Press Chicago IL)

Thiet RK Boerner REJ (2007) Spatial patterns of ectomycorrhizal fungalinoculum in arbuscular mycorrhizal barrens communities implicationsfor controlling invasion by Pinus virginiana Mycorrhiza 17 507ndash517doi101007s00572-007-0123-8

Throop HL Lajtha K KramerM (2013) Density fractionation and 13C revealchanges in soil carbon following woody encroachment in a desertecosystem Biogeochemistry 112 409ndash422doi101007s10533-012-9735-y

Tieszen LL (1991) Natural variations in the carbon isotope values of plantsimplications for archaeology ecology and paleoecology Journal ofArchaeological Science 18 227ndash248doi1010160305-4403(91)90063-U

Tieszen LL Archer S (1990) Isotopic assessment of vegetation changes ingrassland and woodland systems lsquoIsotopic assessment of vegetationchanges in grassland and woodland systemsrsquo (CB Osmond LFPitelka GM Hidy) pp 293ndash321 (Springer-Verlag New York)

Tilman GD (1984) Plant dominance along an experimental nutrient gradientEcology 65 1445ndash1453 doi1023071939125

Tilman D (1987) Secondary succession and the pattern of plant dominancealong experimental nitrogen gradients Ecological Monographs 57189ndash214 doi1023072937080

TrolierMWhite JWCTans PPMasarieKAGemery PA (1996)Monitoringthe isotopic composition of atmospheric CO2 measurements from theNOAAGlobalAir SamplingNetwork Journal ofGeophysicalResearchAtmospheres (1984ndash2012) 101 25897ndash25916

Tyndall RW Farr P (1989) Vegetation structure and flora of a serpentinepinendashcedar savanna in Maryland Castanea 54 191ndash199

Tyndall RW (1992) Historical considerations of conifer expansion inMaryland serpentine barrens Castanea 57 123ndash131

Tyndall RW (2005) Twelve years of herbaceous vegetation change in oaksavanna habitat on a Maryland serpentine barren after Virginia pineremoval Castanea 70 287ndash297doi1021790008-7475(2005)070[0287TYOHVC]20CO2

Tyndall RW Farr P (1990) Vegetation and flora of the pilot serpentine area inMaryland Castanea 55 259ndash265

Van Auken OV (2000) Shrub invasions of North American semiaridgrasslands Annual Review of Ecology and Systematics 31 197ndash215doi101146annurevecolsys311197

Conifer encroachment on ultramafics Australian Journal of Botany O

Van Auken OW (2009) Causes and consequences of woody plantencroachment into western North American grasslands Journal ofEnvironmental Management 90 2931ndash2942doi101016jjenvman200904023

WaltersRSYawneyHW(2004) lsquoSilvicsmanual vol 2 hardwoodsrsquo (UnitedStates Department of Agriculture Washington DC)

Warman L Moles AT (2008) Alternative stable states in Australiarsquos WetTropics a theoretical framework for the field data and a field-case for thetheory Landscape Ecology 24 1ndash13 doi101007s10980-008-9285-9

Weiher E Keddy PA (1999) Relative abundance and evenness patterns alongdiversity and biomass gradients Oikos 87 355ndash361doi1023073546751

White J (1979) The plant as a metapopulationAnnual Review of Ecology andSystematics 10 109ndash145 doi101146annureves10110179000545

Whittaker RH (1960) Vegetation of the Siskiyou Mountains Oregon andCalifornia Ecological Monographs 30 279ndash338 doi1023071943563

Williams GH Darton NH (1892) lsquoGeology of Baltimore and its vicinitycrystalline rocksrsquo (John Murphy and Company Baltimore MD)

Williams BM Houseman GR (2013) Experimental evidence that soilheterogeneity enhances plant diversity during community assemblyJournal of Plant Ecology 7 461ndash469

Wright JP Fridley JD (2010) Biogeographic synthesis of secondarysuccession rates in eastern North America Journal of Biogeography37 1584ndash1596

Wynn JG Bird MI (2008) Environmental controls on the stable carbonisotopic composition of soil organic carbon implications for modellingthe distribution of C3 and C4 plants Australia Tellus Series B Chemicaland Physical Meteorology 60 604ndash621doi101111j1600-0889200800361x

Zavala MA Espelta JM Retana J (2000) Constraints and trade-offs inMediterranean plant communities the case of holm oakndashAleppo pineforests Botanical Review 66 119ndash149 doi101007BF02857785

P Australian Journal of Botany J Burgess et al

wwwpublishcsiroaujournalsajb