Aboveground and belowground effects of single-tree removals in new Zealand rain forest

Ecology,89(5),2008,pp.1232–1245Ó2008by the Ecological Society of AmericaABOVEGROUND AND BELOWGROUND EFFECTS OF SINGLE-TREE REMOVALS IN NEW ZEALAND RAIN FORESTD AVID A.W ARDLE,1,2,5S USAN K.W ISER,2R OBERT B.A LLEN,2J AMES E.D OHERTY,3K AREN I.B ONNER,2AND W ENDY M.W ILLIAMSON41Department of Forest Ecology and Management,Swedish University of Agricultural Sciences,SE90183,Umea˚,Sweden2Landcare Research,P.O.Box40,Lincoln7640,New Zealand3T¯u hoe Tuawhenua Trust,P.O.Box4,Murupara,Bay of Plenty,New Zealand4ESR,Christchurch Science Centre,P.O.Box29181,Christchurch8030,New ZealandAbstract.There has been considerable recent interest in how human-induced species loss affects community and ecosystem properties.These effects are particularly apparent when acommercially valuable species is harvested from an ecosystem,such as occurs through single-tree harvesting or selective logging of desired timber species in natural forests.In New Zealandmixed-species rain forests,single-tree harvesting of the emergent gymnosperm Dacrydiumcupressinum,or rimu,has been widespread.This harvesting has been contentious in partbecause of possible ecological impacts of Dacrydium removal on the remainder of the forest,but many of these effects remain unexplored.We identified an area where an unintended40-year‘‘removal experiment’’had been set up that involved selective extraction of individualDacrydium trees.We measured aboveground and belowground variables at set distances fromboth individual live trees and stumps of trees harvested40years ago.Live trees had effectsboth above and below ground by affecting diversity and cover of several components of thevegetation(usually negatively),promoting soil C sequestration,enhancing ratios of soil C:Pand N:P,and affecting community structure of soil microflora.These effects extended to8mfrom the tree base and were likely caused by poor-quality litter and humus produced by thetrees.Measurements for the stumps revealed strong legacy effects of prior presence of trees onsome properties(e.g.,cover by understory herbs and ferns,soil C sequestration,soil C:P andN:P ratios),but not others(e.g.,soil fungal biomass,soil N concentration).These resultssuggest that the legacy of prior presence of Dacrydium may remain for several decades orcenturies,and certainly well over40years.They also demonstrate that,while large Dacrydiumindividuals(and their removal)may have important effects in their immediate proximity,within a forest,these effects should only be important in localized patches containing highdensities of large trees.Finally,this study emphasizes that deliberate extraction of a particulartree species from a forest can exert influences both above and below ground if the removedspecies has a different functional role than that of the other plant species present.Key words:Dacrydium cupressinum;microbial community;New Zealand;removal experiment;rimu;selective logging;single-tree harvesting;species effects.I NTRODUCTIONThere has been considerable recent interest in predicting how human-induced losses of species and functional groups influence community and ecosystem properties(Hooper et al.2005).The ecological impacts of species loss can be especially apparent when particular plant or animal species are deliberately harvested from the ecosystem(e.g.,Bodmer et al. 1997,Cardillo et al.2005).Single-tree(or selective) harvesting in natural forests that contain several tree species represents such a situation;here,individuals of one or more tree species with desired timber properties are selectively removed from the forest with the remainder of theflora being left.In a mixed-species forest,different tree species may have vastly different effects on both aboveground and belowground proper-ties(Binkley and Giardina1998,Finzi et al.1998, Binkley and Menyialo2005),and in such forests, individual tree effects may be important drivers of spatial heterogeneity of both soil properties and understory vegetation(Boettcher and Kalisz1990, Saetre and Ba a th2000).It follows that selective removal of some tree species using logging,while leaving other tree species,can have important effects on the remainder of the ecosystem(Pollmann2002,Verburg and van Eijk-Bos2003,Villela et al.2006).What is less clear is how long these effects persist following harvesting.The vast majority of experiments investigating eco-logical effects of human-induced loss of biodiversity have involved randomly assembled communities(Bal-vanera et al.2006,Cardinale et al.2006).However,in reality,human activity does not result in species beingManuscript received20September2007;accepted25 September2007.Corresponding Editor:J.Belnap.5E-mail:david.wardle@svek.slu.se1232lost at random from the community(Wardle1999, Solan et al.2004,Zavaleta and Hulvey2004),and some taxa with specific attributes or traits may be lost from a community more rapidly than others(Duncan and Young2000,Petchey and Gaston2002).The ecological impacts of nonrandom species losses in real communi-ties are therefore arguably often better investigated by ‘‘removal experiments’’in which the effects of biodiver-sity loss are studied following the experimental removal of various components of those communities(e.g., Wardle et al.1999,Symstad and Tilman2001,Wardle and Zackrisson2005,Polley et al.2006;for a review see Dıaz et al.2003).It has been recognized that forests from which particular tree species have been removed for timber(e.g.,through selective logging)may serve as long-term replicated‘‘removal experiments’’for investi-gating the ecological impacts of nonrandom loss of species through human activity(Dıaz et al.2003), although few studies have specifically utilized such experiments in this way.In New Zealand,about25%of the land area is covered by indigenous forest,and this includes signifi-cant areas of mixed podocarp(gymnosperms in the Podocarpaceae)–hardwood forest.One of the most widespread and conspicuous tree species in the lowland or lower montane mixed forests of New Zealand is Dacrydium cupressinum(Podocarpaceae),or rimu, which commonly reaches.35m(and up to60m)in height,and often grows as a long-lived emergent tree above the forest canopy(Franklin1968,Six Dijkstra et al.1985,Norton et al.1988;see Plate1).This species hashighly desirable timber properties and,as such,was widely logged throughout New Zealand until the1980s. There are some areas where single-tree harvesting of Dacrydium has been performed from the1950s to the present day,with coexisting tree species(notably hardwood)being left relatively intact.This selective harvesting has been a contentious political issue in New Zealand for over three decades(James and Norton 2002),in part because of possible effects of Dacrydium removal on the composition and functioning of the remainder of the forest.These effects remain unex-plored.In the present study,we identified an area in which an unintended‘‘removal experiment’’had been performed;40years ago in a mixed podocarp–hardwood forest from which Dacrydium had been selectively logged.Specifically,we sought to answer this question:do long-lived large trees of Dacrydium(a gymnosperm)have a legacy(on the order of decades) above and below ground when the residual canopy is dominated by hardwood(angiosperm)species?To do this,we assessed aboveground and belowground prop-erties and aboveground–belowground relationships at set distances from both live trees and from stumps of Dacrydium trees that had been removed40years ago. Further,we sought to better understand the long-term influence of selective logging on individual tree effects, and in a more general sense to generate insights about the ecological impacts of human-induced losses of key biotic components from natural forest ecosystems.M ETHODSStudy system and tree selectionThe study area is in the central North Island of New Zealand;part of the area is in the land administered by the T¯u hoe Tuawhenua Trust,and the remainder is in Te Urewera National Park.It is located at388370S,1768520 E,the mean annual rainfall is1600–2400mm,equally distributed throughout the year,and the mean January and July air temperatures are148C and48C,respectively. The soil present is Urewera steepland soil,which is a type of podzolic orthic pumice soil known in the USDA soil taxonomy as a Vitrand or Vitricryands(Hewitt 1998).The underlying parent material is greywacke(a fine-grained sandstone)overlain by rhyolitic ash.The dominant(canopy-forming)hardwood tree species is Beilschmiedia tawa,with Dacrydium,Prumnopitys ferru-ginea,and Metrosideros robusta as emergents.Many of the Dacrydium trees that occur are.500years old and have a diameter at breast height(dbh,about1.4m high) of.1m.This species produces very poor litter that decomposes slowly,especially in comparison with Beilschmiedia(data in Wardle et al.2002).The understory species consist of smaller hardwoods(Meli-P LATE 1.An emergent Dacrydium cupressinum tree and associated forest in Te Urewera National Park,North Island, New Zealand.Photo credit:S.K.Wiser.May20081233RAIN FOREST SINGLE-TREE REMOVAL EFFECTScytus ramiflorus,Pseudowintera spp.),tree ferns(Dick-sonia squarrosa and Cyathea smithii),ground ferns (Blechnum discolor,Asplenium bulbiferum,Leptopteris hymenophylloides),and other herbaceous plants(Micro-laena avenacea,Uncinia uncinata).The area specifically considered in our study is a strip of land;1km long3300m wide.On one side of this strip(1km3150m)all Dacrydium trees of millable size (dbh.30cm)were logged between1956and1970 (mean time since harvest¼40years).On the other side of the strip(1km3150m),no logging was ever performed and all Dacrydium trees remain intact.In the absence of logging,the density of Dacrydium stems with a dbh.30cm is14.4stems/ha.At these densities, Dacrydium trees(or stumps)were usually far enough apart to be considered independent.Total tree basal area(dbh.30cm)is43.2m2/ha,and Dacrydium stems comprise35%of this basal area(R.Scott,unpublished data).Of the total basal area,3%consists of stems,10 cm dbh,42%of stems10–30cm dbh,28%of stems30–50dbh,13%of stems50–80cm dbh,and14%of stems .80cm dbh.During19–27November2004,we randomly selected 20live Dacrydium trees and20stumps of Dacrydium individuals that had been logged40years earlier.We selected only live trees with dbh.80cm and only stumps that had comparable diameters at ground level to the live trees we used.We treated each live tree or tree stump as a discrete replicate,which we maintain is defensible,given the narrow dimensions of the land used relative to its length.Plot demarcation and vegetation measurementsAll vegetation measurements were performed over19–27November2004.For each live tree or stump,the diameter at ground level was recorded,and for each live tree,the dbh was also noted.Three plots were then set up at increasing distances from the base of each tree or stump,so that each plot was24m2.Thefirst plot was arc shaped and extended from0to4m from the base of the tree or tree stump.The left-and right-hand boundaries of the plot were both segments of radial lines extending from the center of the stump or tree,with the angle between the two lines being calculated so that the plot area was24m2.This angle varied slightly according to the diameter of tree or stump base,but was on average818.The second plot was also arc shaped and extended from4to8m from the base of the tree or stump,and with its left-and right-hand boundaries being determined in the same way as for thefirst plot,so as to yield a total plot area of24m2.Here,the angle between left and right boundaries was,on average,408. The third plot extended from12to16m from the base of the tree or stump,and had dimensions of436m;it was rectangular rather than arc shaped because the degree of curvature at that distance from the tree was minor relative to the dimensions of the plot.The direction of plots from the tree or stump base was chosen at random,except that bulldozer tracks made during tree harvesting were avoided.On each plot,vascular species present were recorded in six height tiers(.25m,12–25m,5–12m,2–5m,0.3–2m,and,0.3m).Foliar cover over the plot area for each species in each tier was estimated using a modified Braun-Blanquet cover scale(1¼few plants,cover,1%; 2¼1–5%cover;3¼6–25%;4¼26–50%;5¼51–75%;6¼.75%;Allen,1992).Species were classified intofive groups,namely woody canopy species(primarily Beilschmiedia),tree ferns,understory woody plants, understory ferns,and understory herbs.Total cover for each group was calculated by summing the midpoints of the cover class recorded for each species in that group across all the tiers in which it occurred,as described by Wiser et al.(2002).As a consequence,the total cover of any given group can exceed100%.Belowground measurementsWhen the vegetation was measured,soil samples were collected midway between the innermost and outermost edges of each of the vegetation plots,at2,6,and14m from the base of the tree or tree stump.For each plot, three intact cores of soil with surface litter removed (each5cm diameter35cm depth)were taken,1m apart from each other.At this depth the soil is primarily humus with some incorporation of mineral soil,with an organic matter content of67%on average.The three samples were bulked for each plot and the wet mass determined;the dry mass(dried at808C for24h)was then determined for a20-g subsample for determination of soil bulk density(dry mass per unit volume).The soil sample from each plot was sievedfield-moist to pass through a4-mm mesh and a subsample was used for determination of the following soil chemical proper-ties:soil pH,total soil C using a Leco FP-2000Analyzer (Laboratory Equipment Corporation,St.Joseph,Mich-igan,USA,),total N(Kjeldahl method),P(as described by Jackson1958),mineral N(ammonium and nitrate) using automated colorimetric procedures(Technicon Instruments1977)following extraction with2mol/L KCl,and bicarbonate extractable C using the Olsen method(Jackson1958).Because the bulk density of each soil sample was known,total C,N,and P per unit area stored in the soil to5cm depth were also calculated. Microbial basal respiration(BR)and substrate-induced respiration(SIR),a relative measure of active microbial biomass,were determined for each sample as described by Wardle(1993).Briefly,a15-g(dry mass equivalent)subsample of soil was amended to150% gravimetric moisture content(dry mass basis),either by gradual drying or rewetting with afine mist,placing it in a130-mL airtight vessel,and incubating at228C. Evolution of CO2between1and4h was then determined by injecting1-mL subsamples of headspace gas into an infrared gas analyzer(ADC2250Gas Analyzer,ADC BioScientific,Hoddesdon,UK)used as a measure of BR.Measurement of SIR(Anderson andDAVID A.WARDLE ET AL.1234Ecology,Vol.89,No.5Domsch1978)was performed in the same way,but with amendment of the sample with150-mg anhydrous D(þ)-Glucose at the start of the incubation(Wardle1993). The microbial metabolic quotient(q CO2),the ratio of BR to SIR,was used as a measure of relative microbial-C-use efficiency;this ratio declines as efficiency increases (Anderson and Domsch1985).The ratio of SIR to organic C was used as an indicator of soil C quality because when soil C is more favorable for microbes,a higher proportion of it is assimilated in microbial tissue (Insam and Domsch1988).Microbial community composition for each sample was assessed by measuring the composition of microbial phospholipids fatty acids(PLFAs)using the method of Bligh and Dyer(1959),as modified by White et al. (1979)and Bardgett et al.(1996);different PLFAs are derived from different subsets of the soil microflora.For each soil,the abundance of each of the individual fatty acids extracted was expressed as relative nmoles per gram of dry soil using standard nomenclature(Tunlid et al.1989).The PLFAs that were used to represent bacteria were i-15:0,a-15:0,15:0,16.0,i-16.0,16:1x9,i-17:0,cyc-17:0,17:0,18:1x9c,18:1x9t,and cyc-19:0,and the PLFA used to represent fungi was18:2x6(Parekh and Bardgett2002).The PLFAs used for calculation of total branched PLFAs(indicative of gram-positive bacteria)were i-15:0,a-15:0,i-16:0,and i-17:0,while those used for determining total cyclic PLFAs were cyc-17:0and cyc-19:0.The ratio of total bacterial PLFAs to fungal plus bacterial PLFAs was used as an estimate of the relative importance of the bacterial and fungal energy channels(Bardgett et al.1996,Parekh and Bardgett2002).To assess plant litter decomposition,two nylon-mesh litter bags were positioned at each of three distances(2, 6,and14m)from the edge of each tree or stump,which was in the center of the plots used for vegetation measurement.Each litter bag measured10310cm with mesh size of1mm and contained3g(dry-mass basis)of air-dried fresh litter of Beilschmiedia tawa(N and P concentrations¼1.34%and0.067%,respectively).Litter bags were positioned at the interface of the humus and litter layers during19–27November2004,and all bags were harvested approximately one year later during7–8 December2005.Following harvest,the litter in each bag was cleaned and the remaining dry mass(dried at808C for24h)was determined.Data analysisAll response variables,except for those summarizing community structure,were analyzed using split-plot analysis of variance(ANOVA)with each tree or stump serving as the main plot and with the three distances from each tree or stump serving as the subplots.For the soil abiotic variables,split-plot multivariate ANOVA (MANOVA)was also used for the entire data set because of the large number of variables measured (other multivariate methods were used for the plant and microbial community data).In all cases,the effects tested for were tree vs.stump(S),distance from tree or stump(D),and the S3D interaction.When any of these effects was significant at P,0.05,differences among means were compared using the least significant difference test at P,0.05.Variables were transformed as necessary to satisfy assumptions of all parametric data analyses performed in this study.For analysis of both the vegetation and PLFA community structure data we used the constrained-ordination split-plot analogue as implemented in the software package CANOCO(Microcomputer Power, Ithaca,New York,USA;Baar and ter Braak1996,ter Braak and S milauer2002).This was chosen because it is the only ordination approach available for performing the analogue of a split-plot ANOVA.To do this,we used redundancy analysis,a linear ordination method, because a preliminary analysis by the nonlinear de-trended correspondence analysis method resulted in short gradient lengths(i.e.,,3.0standard deviations;ter Braak and S milauer2002).We then used Monte Carlo permutation tests with plots permuted as appropriate to test for the statistical significance of S,D,and S3D effects on vegetation and PLFA community structure,as described by ter Braak and S milauer(2002).To examine the relationship between vegetation composition and PLFA composition across plots,we analyzed both data sets using the unconstrained linear ordination technique,principal components analysis (PCA),and determined Spearman rank correlation coefficients between vegetation,and PLFA ordination plot scores for the two main axes.This was done using only the stump data,only the live tree data,and all stump and live tree data combined.R ESULTSAboveground responsesFor the20live Dacrydium trees,dbh was136.6632.3 cm(mean6SD).The mean diameter at ground level of these trees was208.2650.1cm,while that of the20 stumps was240.3652.2cm.The live trees and tree stumps were therefore of comparable dimensions,as the cut stumps remain largely intact and have diminished little since being harvested.The cover of understory herbs,understory ferns,and tree ferns significantly increased with increasing distance from trees and stumps,according to ANOVA(Table1,Fig.1);the same pattern occurred for understory woody plants in relation to live trees,but this effect was not statistically significant,according to ANOVA.Meanwhile Beilsch-miedia cover was least at the greatest distance from trees and stumps.Further,the cover of both tree ferns and understory woody plants was greater for stumps than for live trees(Fig.1).There was no interactive effect between distance from trees or stumps,and trees vs. stumps,for the cover of any plant group(Table1). When the overall species composition of the vegeta-tion was analyzed,both distance from tree or stump andMay20081235RAIN FOREST SINGLE-TREE REMOVAL EFFECTStree vs.stump had significant effects (Table 1,Fig.2).There was no interactive effect,however,between these two factors.The ordination revealed that some plant species were particularly important in separating the plots from each other (Fig.2).Thus,the tree fern Cyathea smithii had increased cover in plots that were farthest from the live trees and stumps.The understory woody plant Pseudowintera colorata and tree fern Dicksonia squarrosa were most closely associated with stumps,while the understory fern Polystichum richardii was associated most closely with live trees.The woody understory plant Melicytus ramiflorus and the understo-ry ferns Leptopteris hymenophylloides and Asplenium bulbiferum tended to be associated with live trees and with plots at greater distances from trees or stumps (12–16m vs.0–4m or 4–8m).Vegetation community composition in the ordination analysis was related to spatial variability in several soil properties,notably soil C concentration and ratios of C:N and N:P,which increased toward the stumps,and soil pH and bulk density,which increased away from them (Fig.2).Species richness of all vascular plants,and of understory herbs and understory ferns,increased significantly with increasing distance from trees or stumps (Table 1,Fig.1).Further,in the case of understory ferns,species richness was greater for live trees than for stumps (Table 1,Fig.1).Species richness of understory woody plants showed no significant relationship with stumps or with live trees.There was no interactive effect between distance from trees or stumps or trees vs.stumps for the species richness of any plant group (Table 1).Belowground responsesThe MANOVA performed over the full data set of abiotic soil properties revealed a highly significant overall interaction between trees vs.stumps and distance (Table 2).Univariate ANOVAs on these variables showed that distance and trees vs.stumps both significantly influenced soil variables,contributing to this overall interactive effect (Table 2).Concentrations of total soil C and N both declined at increasing distances from trees and stumps,and soil N concentra-tions were also greater for stumps than for trees (Table 2,Fig.3).In contrast,concentrations of total soil P and labile forms of N (i.e.,mineral N)and P (i.e.,Olsen P)were unaffected by trees or stumps.Mass ratios of C:N,C:P,and N:P were all significantly greater close to the stump or tree than farther away from it (Table 2,Fig.3).Bulk density of soil increased with increasing distance from the tree or stump,and the amounts of C and N sequestered in the soil per unit area (i.e.,soil content)declined.Soil pH increased at increasing distances from trees or stumps.The ratio of labile mineral N to total N and labile Olsen P to total P were unaffected by distance from tree or stump,although the ratio of mineral N to total N was marginally significantly greater (P ¼0.038)for live trees than for stumps.For the univariate ANOVAs,interactive effects between distance and stump vs.tree were significant only for mineral N and the soil C:N ratio,despite being highly significant for the MANOVA (Table 2).Two measures of soil biological activity,the soil BR and litter decomposition rate,declined significantly (though not strongly)with increasing distance from trees and stumps (Table 3,Fig.4).For BR this effect was greater for stumps than for live trees,though this effect was not statistically significant.Measures of the total microbial biomass (SIR and total PLFAs)were unre-sponsive to trees or stumps,but the ratio of SIR to organic C was significantly greater for stumps than for trees (Table 3,Fig.4).Further,there was evidence for effects of trees and stumps on microbial community composition as revealed using PLFA analysis.Fungal PLFA quantities were greater for live trees than for stumps (Fig.4)and declined with increasing distances from the tree (Table 3).T ABLE 1.Effect of stump vs.live tree,distance from stump or tree,and their interaction on vegetation biological properties as revealed by split-plot two-way analysis of variance (or redundancy analysis for the plant community compositional data).Response variableStump vs.tree (S)Distance from stump or tree (D)S 3D interaction F (df ¼1,38)P F (df ¼2,76)P F (df ¼2,76)P Multivariate analysesPlant community composition5.620.002 3.500.0020.830.824Univariate analysesBeilschmiedia tawa cover à0.210.652 5.320.0070.110.897Cover of tree ferns à11.510.00214.60,0.001 1.030.363Cover of understory herbs (excluding ferns)à0.070.788 3.840.0260.440.646Cover of understory ferns à0.730.3989.10,0.001 1.570.214Cover of understory woody plants à8.620.006 2.560.084 1.370.261Vascular plant species richness2.150.15112.29,0.0010.320.724Understory herb species richness (excluding ferns)0.210.6503.550.0340.580.564Understory fern species richness6.540.01512.71,0.001 1.630.202Understory woody plant species richness0.400.5282.530.0870.230.793Derived from the redundancy analysis presented in Fig.2.àData were log-transformed before analysis.DAVID A.WARDLE ET AL.1236Ecology,Vol.89,No.5The ratio of fungal to bacterial PLFAs showed a broadly similar response,though not as strongly.Total branched PLFAs also declined with increasing distance from trees or stumps,while cyclic PLFAs were unresponsive.Trees and stumps,however,had no detectable effect on the diversity of microbial PLFAs.Interactive effects between stump vs.tree and distance were significant only for the BR:SIR ratio (Table 3).The structure of the soil microbial community,determined through redundancy analysis of the PLFA data,showed significant overall responses to the distance from live trees or stumps,but not to trees vs.stumpsIG .1.Measures of plant cover and species richness at three distances from the center of live Dacrydium cupressinum trees and tree stumps.Within each panel,bars topped by the same lowercase letter do not differ significantly at P ¼0.05(least significant difference test)derived from split-plot ANOVA (Table 1).SEM is standard error of the mean pooled across distance and tree vs.stump.May 20081237RAIN FOREST SINGLE-TREE REMOVAL EFFECTS(Table 3,Fig.5).There was also a marginally nonsignif-icant interactive effect between these two factors (Table 3).Both bacterial PLFAs (e.g.,15:0,i16:0,16:0,cyc17:0)and the dominant fungal PLFA (18:2x 6)served as important discriminators among plots in the analysis (Fig.5).Microbial community composition in the ordination analysis was related to spatial variability in several soil properties,including C concentration,ratios of C:N and N:P,pH,and bulk density (Fig.5).Correlation analyses between PCA scores for plant community structure and PCA scores for microbial community structure (PLFA data)provided little evidence of relationships between the plant and micro-bial data across plots.Correlation coefficients amongpairs of PCA axes were never statistically significant when data for all plots or data for live-tree plots only were used (data not shown),and for stump plots the only significant relationship was between the second PCA axis for plants and the second PCA axis for microbes (r ¼0.293,P ¼0.023).D ISCUSSIONAboveground responsesOur study provided evidence that live Dacrydium trees have important effects on the rest of the vegetation;cover,and species richness of understory herbs and ferns,and cover of tree ferns were all lower closer totheF IG .2.Regression biplot derived from a redundancy analysis of plant species with respect to the model:stumps vs.trees;distance (0–4,4–8,12–16m).The biplot displays 8%of the variance in abundances (cover scores)and 43%of the variance in the fitted abundances.Eigenvalues of the first two axes are 0.051and 0.031,respectively;the sum of all canonical eigenvalues is 0.087.Only species with .20%frequency across all vegetation subplots are shown.C,N,and P are carbon,nitrogen,and phosphorus.Species codes are as follows:woody vascular plants,ARISER ¼Aristotelia serrata ,BEITAW ¼Beilschmiedia tawa ,BRAREP ¼Brachyglottis repanda ,CARSER ¼Carpodetus serratus ,COPTEF ¼Coprosma tenuifolia ,ELADEN ¼Elaeocarpus dentatus ,METPER ¼Metrosideros perforata ,MELRAM ¼Melicytus ramiflorus ,MYRAUS ¼Myrsine australis ,PARSON ¼Parsonsia spp .,PSEAXI ¼Pseudopanax axillaris ,PSECOL ¼Pseudowintera colorata ,PRUFER ¼Prumnopitys ferruginea ,and SCHDIG ¼Schefflera digitata ;understory herbs,COLHAS ¼Collospermum hastatum ,MICAVE ¼Microlaena avinacea ,and UNCUNC ¼Uncinia uncinata ;understory ferns,ASPBUL ¼Asplenium bulbiferum ,ASPFLA ¼Asplenium flaccidum ,ASPPOL ¼Asplenium polyodon ,BLEDIS ¼Blechnum discolor ;LEPHYM ¼Leptopteris hymenophylloides,MICPUS ¼Microsorum pustulatum ,and POLRIC ¼Polystichum richardii ;tree ferns,CYASMI ¼Cyathea smithii and DICSQU ¼Dicksonia squarrosa .DAVID A.WARDLE ET AL.1238Ecology,Vol.89,No.5。