Primary succession trajectories on a barren plain, Mount St. Helens, Washington

Primary succession trajectories on a barren plain Mount St Helens

Washington

Roger del Moral Jason M Saura amp Jennifer N Emenegger

AbstractQuestions Have predictable relationships betweenenvironmental variables and vegetation developedin primary succession following a volcanic eruptionHas the rate of succession changed Have vegetationtrajectories converged or divergedLocation The Abraham Plain of Mount St HelensWashington USA (4611204200N 12210802700W ele-vation 1360m) was sterilized in 1980 by a blastscoured by lahars and buried by pumiceMethod We monitored 400 100m2 contiguous per-manent plots annually (1988ndash2008) and classifiedeach plot from every year into ten community types(CTs) We characterized the terrain by topographyand surface features Redundancy analysis assessedrelationships between vegetation and possible ex-planatory variables which included samplelocation We used detrended correspondence analy-sis (DCA) to assess successional rates and trendsResults Relationships between species compositionand explanatory variables were only significant after1996 when position and presence of rills becamesignificant By 2006 explained variation remainedlow (13) but significant Species accumulatedslowly restricted by stress and isolation Changesin mean DCA position slowed Composition shiftedfrom pioneer to persistent species and vegetationbecame more stable with time Species accumulatedfor two decades and then stabilized while cover hascontinued to increase Diversity increased and thendeclined slightly as dominance developed and pio-neer species became less commonConclusions We demonstrate weak but increasinglypredictable trends in species composition using en-vironmental variables The rate of succession slowedand trajectories formed a reticulate network oftransitions dominated by divergence Convergencewas not evident because vegetation responded dis-tinctively to minor topographic features thatallowed alternative stable communities to develop

Keywords Community assembly Convergence Di-vergence Permanent plots Rate of successionRedundancy analysis Restoration Stochastic as-sembly Succession network

NomenclatureUSDA NRCS (2009 Appendix S1)

Introduction

One goal of the study of succession is to assesshow vegetation develops in response to explanatoryenvironmental variables Studies of succession insevere habitats suggest that the control of successionshifts from stochastic events to predictable causes(Baasch et al 2009) but this shift is poorly under-stood and rarely documented directly Severalsuccession trajectories are probable during earlyprimary succession when competitive interactionsare weak

Early in primary succession landscape factorsthat influence dispersal may be the only predictorsof plant patterns (Prach amp Rehounkova 2006) Thiscan produce variable patterns in composition withrespect to topography (Felinks amp Wiegand 2008)Rarity of seedling establishment and extremeweather events can obscure predictable patternTherefore variability of early primary successionvegetation is expected (Robbins amp Matthews 2009)because similar sites that receive different colonistsoften follow multiple trajectories

While the response of a species to one environ-mental factor might be predictable vegetationpatterns are less tractable due to complex inter-dependent species responses to environmentalfactors Variables that are often related to speciespatterns include soil properties that favour speciesdifferentially favourable microsites and biotic in-teractions (del Moral 2009a) The effects of suchfactors should increase with time because as dis-persal limitations ease more biomass can producestronger interactions and relationships between to-pography or nurse plants can become more evident(but see Walker et al 2006)

A successional trajectory describes vegetationchange these were once thought inevitably to belinear convergent and predictable (Pickett et al

del Moral R (corresponding author moraluwedu)

Saura J M (saurauwedu) amp Emenegger J N (jne2

uwashingtonedu) Department of Biology Box 351800

University of Washington Seattle Washington USA

Journal of Vegetation Science 21 857ndash867 2010DOI 101111j1654-1103201001189xamp 2010 International Association for Vegetation Science

2009) Now ecologists recognize that vegetation candiverge or follow intricate temporal braids in re-sponse to stochastic processes contingencies andlandscape factors

The vegetation of most new surfaces on MountSt Helens has developed markedly since the massiveeruption of 1980 but on the east slope vegetationremains sparse Slow primary succession on thisbarren plain offers a superb chance to evaluaterelationships between environmental factors andvegetation at different stages of development Weask has a more predictable relationship betweenenvironmental variables and vegetation developedover time has the rate of succession changed asspecies accumulate and have trajectories convergedor diverged

Methods

Study site

Abraham Plain a barren nearly level site 4 kmeast of the cone of Mount St Helens is centered at4611204200N12210802700Wmean elevation 1360m Itreceived three catastrophic volcanic impacts in rapidsuccession on 18051980 The lateral blast removedall soil and melted ice fields to produce massive la-hars that scoured the site (Swanson amp Major 2005)Pumice then smothered the landscape Pumice rockshave decomposed to gravel and erosion has carvedgullies and rills to create protected microsites Iso-lation from surviving vegetation few animaldispersal vectors and winds that direct seeds awayfrom barren sites all restricted colonization Estab-lishment was constrained by drought stress andinfertility

Vegetation sampling

Errors using chronosequence approaches canaccrue when multiple trajectories remain un-recognized Using permanent plots mitigates thepotential for such errors (Johnson amp Miyanishi2008) In 1988 we established 400 contiguous10m10m plots (a 1040 grid) We estimated spe-cies cover with this index 15o6 individuals 25 6to 20 individuals 35420 individuals or cover of025m2 to 05m2 45 cover 405m2 to 1m25541m2 to 2m2 6542m2 to 4m2 7544m2 to8m2 48m2 recorded directly (Wood amp del Moral1988) R del Moral determined cover from the ver-tical projection of the canopy of each speciesannually between 1988 and 2008

Explanatory variables

We estimated percentage cover of topographic(smooth rill and gully) and surface (rock pumiceand sand) features in each plot Rills are narrowwith gentle slopes and have not exposed the originalsurface Gullies are more than 1-m wide with steepslopes and have reached the original surface Cate-gories were as follows For smooth 05o20cover 15 21ndash50 cover 25 51ndash90 cover35490 cover For rills 05 none 15 1ndash5cover 25 6ndash10 cover 35410 cover For gul-ly 05 none 15 1ndash10 25 11ndash20 cover35420 cover For rocks 05 none 15 1ndash2cover 25 3ndash5 cover 3545 cover For pu-mice 05o80 cover 15 81ndash90 cover 25 91ndash97 cover 35497 cover For sand 05 0ndash3cover 15 4ndash6 cover 25 7ndash10 cover35410 cover Position was defined by x (northndashsouth) and y (eastndashwest) grid coordinates

Analyses

We calculated species richness plot cover indexSimpsonrsquos dominance [D5 1Spi

2] and Shannondiversity (H05 [Spi log pi]) where pi is the pro-portion of the cover index represented by the ithspecies from the cover index of each species (MjMSoftware Design Gleneden Beach OR US) Weclassified all plots in each year into community types(CTs) with flexible sorting (an effective space-con-serving group-linkage method that employsb5 025 to limit chaining) using the Euclideandistance between samples (MjM Software Design)The 6464 samples formed a dendrogram that we di-vided into 10 CTs recognized when members sharedat least 50 of the information Large differences inspecies richness distort dendrograms (Clifford1976) so divisions occurred at different fusion le-vels We used multiple response permutationprocedures (MRPP MjM Software Design) to as-sess validity of the CTs MRPP calculates aweighted mean within-group distance and a T-valuethat describes the distinctiveness of the groups A(0 to 10) by describing group homogeneitycompared to a random classification We formed sixhabitat types (HTs) from the topographic and sur-face variables using the same classification approachas for species

Recognizing that succession has encompassedonly early primary succession we divided the CTsinto four stages For convenience these are pioneerearly mid- and late seral The categories were based

858 del Moral Roger et al

on the dispersal growth form and longevity ofdominant species (Appendix S1)

After 1990 annual species turnover was low solinear methods could be applied (Legendre amp An-derson 1999) We used redundancy analysis (RDALeps amp Smilauer 2003 CANOCO DLO-Agri-cultural Mathematics Group Wageningen NL) toexplore the relationship between species composi-tion (index values) and explanatory variables in1990 1993 1996 1999 2003 2006 and 2008 Speciescomposition in each of these years was regressedwith explanatory variables to create fitted speciesscores We applied principal components analysis tothese values to produce canonical plot scores RDAthen formed linear equations to predict the positionof plots We assessed variation by comparing theresult to a null model (2000 random trials) The re-gression coefficient of each variable with each RDAaxis estimated how well variables predicted plot po-sition determined by a t-value (df5 400 samples ndash 8variables ndash 15 391) The surplus species variationsuggested the importance of unmeasured variablessampling error and stochastic effects

Spatial pattern of CTs was determined by com-paring the inter-plot distances of a given CT to thedistances between N random plots and the nearestplot of that CT Adjacent plots were 0 distance lar-ger spacings were calculated by Euclidean distancein grid units The observed mean between-plot dis-tance was compared by t-test to that determinedfrom the hypothesis of random distribution

We used Statistix 9 (Analytical Software Talla-hassee FL US) to conduct statistical analysesComparisons of cover index values among species inthe CTs and among DCA scores of CTs were madewith one-way ANOVA or repeat measures ANO-VA as appropriate each followed by a Bonferronitest of differences We used w2 association tests torelate community patterns to surface patterns Weused Axum 7 (Mathsoft Insightful CorporationSeattle WA US) to produce graphs and generaterandom numbers

Results

Vegetation

Succession on Abraham Plain has been leisurelyBy 2009 vegetation remained sparse (Fig S1) andpercentage cover was one-quarter of that observedon a similar grid on the north side of Mount StHelens with three-quarters as many species (delMoral 2009b)

The composition of each of the 10 CTs was dis-tinctive (MRPP A5 0278 T5 1772 Po00001Table 1) Each pair-wise comparison by MRPPshowed that CTs differed significantly Pioneer CTsvanished by 2001 (Fig 1a) they had lower richness(Fig 1b) and cover indices (Fig 1c) than did persis-tent CTs H0 (Fig 1d) increased with time and washigher in persistent CTs Each plot supported amean of 495 106 CTs between 1988 and 2008We characterize CT composition below The num-ber of plots within which the CT ever occurred is inparentheses

There were two pioneer community types CT-A (Anaphalis-Chamerion-Hypochaeris n5 397 of400 plots) had sparse cover of a few pioneer specieseach with long-distance wind dispersal herbaceouslife form and short life span (Appendix S1) As coverincreased plots developed into other CTs CT-B(Lupinus lepidus n5 11) established first in sevenplots near surviving vegetation and was character-ized by L lepidus with sparse Anaphalis and Carexmertensii We recognized three early seral CTs thatalways developed from other CTs and had morespecies and cover than did pioneer CTs These in-cluded a few species that are more persistent Plotscharacterized as CT-C (Anaphalis n5 107) usuallydeveloped from CT-A plots CT-D (Anaphalis-Cis-tanthe-Agrostis) n5 319) was the main nexusthrough which pioneer CTs developed It had lowerrichness and cover than did persistent CTs CT-E(Anaphalis-Chamerion n5 208) had more Hy-pochaeris Luetkea andRacomitrium than did CT-D

Transitions to persistent CTs involved increas-ing species richness and cover and greaterdominance by longer-lived often woody speciesWe characterized two persistent CTs as mid-seralMost common species have less effective dispersalthan those in pioneer CTs The richness of CT-F(Penstemon-Agrostis-Cistanthe n5 338) peakedwhen it contained both pioneer and seral speciesH0

declined as the vegetation matured CT-G (Pen-stemon-Agrostis-Juncus n5 75) retained Anaphalisand developed more of the persistent species (egLuetkea and Racomitrium) than did CT-F Threelate-seral CTs may represent divergent trajectoriestowards mature vegetation CT-H (Penstemon-Agrostis-Salix n5 266) also had Juncus Cistantheand Luetkea Common species are long-lived andseveral are shrubs CT-H had more species highercover and greater diversity than other CTs It hadchanged little since 2004 CT-I (Penstemon-Agrostis-Cistanthe n5 125) also supported Juncus and mos-ses Richness cover and H0 were intermediateamong persistent CTs CT-J (Penstemon-Agrostis-

Primary succession trajectories 859

Cistanthe-Juncus n5 145) also had substantialLuetkea and Racomitrium Pioneer species were un-common and taller woody species (Abies spp andSalix) absent Richness cover and diversity were allleast of the persistent CTs

Species patterns

Many species increased in cover (Fig 2 Fig S2)during the study but cover of pioneer species de-clined after the mid-1990s In 2008 1093 speciesplot were wind-dispersed while only 011 speciesplot were animal-dispersed (Appendix S1) Pioneerspecies have effective long-distance dispersal me-chanisms and while seral species are also wind-

dispersed their dispersal abilities are more limited(Fuller amp del Moral 2003) Penstemon and Agrostispallens expanded to occupy all plots while Juncusparryi and Cistanthe umbellata occurred in mostLuetkea increased steadily but by 2008 it appearedto be stable Salix increased gradually and spor-adicallyRacomitrium occurred in four plots in 1988while Polytrichum did not occur until 1995 Carexmertensii and C microptera were widely distributedand occasionally abundant both occur on halfthe grid Saxifraga once common declined in2008 but this may be associated with dry summerconditions

Relative cover clearly shows the contrast be-tween pioneer and persistent species (Fig 3 Fig S3)

Table 1 Structure and mean cover index in 10 community types (CTs) Numbers in parentheses after species are occurrences(plotsyears) Values in bold indicate dominant species (high cover index) in that CT Each structural variable differedsignificantly among CTs (one-way ANOVA Po0001 values with same superscript fall within the same group as determinedby Bonferroni comparisons) Species cover index means also differed significantly (ANOVA Po0001) but patterns weretoo complex to display grouping patterns CTs and species were each arranged in order of their DCA scores

Structure CT-A CT-B CT-C CT-D CT-E CT-F CT-G CT-H CT-I CT-J

Richness 37a 31a 53b 95d 89c 109e 125f 127f 111e 91cd

Cover (sum indexscores)

52a 38a 70a 159c 136b 242d 321g 298f 281f 238e

Dominance 0516a 0575a 0734b 0851cde 0840c 0871def 0890ef 0890ef 0877def 0843cd

H0 1001a 0999a 1512b 2078d 2008cd 2204e 2347f 2361f 2227e 2001c

Number of plotsyears 1558 21 108 877 444 1289 219 1334 421 495First year observed 1988 1988 1988 1990 1989 1992 1996 1994 1997 1997Last year observed 2000 1991 1998 2001 2000 2008 2008 2008 2008 2008Hypochaeris radicata

(3011)0444 0000 0491 1197 1380 0497 0365 0412 0214 0040

Anaphalis margaritacea(5784)

1557 0810 1713 3304 3459 2370 2717 2147 1337 0802

Chamerionangustifolium (4712)

0514 0191 0593 1911 1803 1442 0991 1077 0691 0513

Hieracium albiflorum(3907)

0336 0143 0398 0904 0930 0929 0753 0813 0411 0293

Lupinus lepidus (248) 0007 1667 0028 0051 0220 0076 0383 0021 0088 0036Lupinus latifolius (122) o0001 0095 0000 o0001 0047 0020 0932 0037 0000 0000Abies sp (736) 0106 0000 0120 0123 0124 0087 0032 0352 1062 0020Luetkea pectinata

(4075)0182 0095 0157 0691 0916 1389 2078 2524 1981 1432

Carex mertensii (2691) 0128 0428 0046 0379 0373 0680 0827 1249 1057 0649Saxifraga ferruginea

(1596)0045 0095 0083 0276 0253 0632 1082 1079 0456 0247

Racomitrium canescens(2498)

0072 0000 0019 0243 0943 0421 2183 1080 1064 0990

Cistanthe umbellata(4867)

0123 0048 1370 2222 0523 3199 3603 3469 4033 4853

Agrostis pallens (5060) 0294 0000 0528 1331 0738 2855 3457 3271 3924 4301

Eriogonum pyrolifolium(653)

0016 0000 0000 0071 0057 0120 0338 0281 0591 0059

Juncus parryi (4421) 0116 0191 0250 1008 0550 2066 4064 2633 3565 3465

Penstemon cardwellii(4489)

0195 0000 0278 0520 0484 3165 4489 3591 4197 4216

Salix sitchensis (1858) 0037 0000 0000 0140 0188 0216 1237 3144 0271 0073Carex microptera

(1723)0035 0095 0065 0161 0091 0412 0763 0752 0644 1372

Polytrichumjuniperinum (1133)

o0001 0000 0000 0047 0093 0182 0182 1601 2183 0168

860 del Moral Roger et al

Species such as Penstemon Agrostis Juncus parryiCarex spp and Luetkea increased across the CTswhen listed by their first appearance These speciesare long-lived with strong vegetative growth

Fig 1 Structure of CTs from 1988 to 2008 (a) Number of plots (b) Mean species richness (c) Mean cover index (d) Meandiversity index (H0)

Fig 2 Changes in mean cover index for abundant species

Fig 3 Changes in relative cover (proportion of total cov-er in the sample) of common species in the four stages ofearly primary succession The first four species are persis-tent the last two are pioneer species (defined in AppendixS1)

Primary succession trajectories 861

In contrast pioneer species soon peaked and thendeclined

Habitat relationships

We could use only position topography andsurface conditions as explanatory variables in RDAExplained variation was low but increased four-fold from 1990 to 2006 (Table 2) In 1996 we ob-tained the first significant result when the y-axis(possibly related to distance from survivors) and rillfraction became significant By 2006 smooth sur-face also contributed but topography did notinfluence the vegetation

We recognized six habitat types (Table 3) A fivepersistent CT (Table S1) by six HT w2 test was sig-nificant (w2 5 935 Po00001 df 5 20) CT-F(Penstemon-Agrostis-Cistanthe) was more commonin rills CT-G (Penstemon-Agrostis-Juncus) wasmore common in gullies CT-H (Penstemon-Agros-tis-Salix) tended to avoid smooth plots and wasmore common in plots with rills or gullies and CT-J(Penstemon-Agrostis-Cistanthe-Juncus) tended tooccur in smooth plots but avoided rills or gullies

The spatial pattern displayed by the CTs re-flected some environmental sorting (Fig 4) Nearestneighbour distances compared to the randommodel suggested that CT-F formed scattered clus-ters (mean observed distance5 142 mean randomdistance5 224 t5 225 P5 003) Some plots ofCT-G were confined to the eastern edge while oth-ers were in the southwest corner but overall it wasnot clustered (observed5 190 random5 243t5 190 P5 006) CT-H was clustered (ob-served5 015 random5 058 t5 483 Po00001)as was CT-I (observed5 052 random5 145t5 448 Po00001) concentrated on the east halfof the grid CT-J was clustered (observed5 119random5 264 t5 383 P5 0002) and con-centrated on the eastern part of the grid

Trajectories

We explored vegetation trajectories in severalways In 1988 richness the mean cover index(Fig 5) H0 and dominance (Fig S4) were leastRichness increased until 1999 after which time itvaried around 11 species per plot The cover indexcontinued to increase while H0 and dominance sta-bilized by the mid-1990s

Succession began with colonization by highlyvagile pioneers such as Anaphalis Chamerion Hier-acium and Hypochaeris Plots changed into otherCTs at variable rates While we found transitionsamong all CTs some trajectories were more typicalthan were others (Fig 6) The most likely trajec-

Table 2 Summary of redundancy analysis results usingposition topography and surface variables on the gridThe alternate plots sampled in 2007 and 2008 were ana-lysed separately Percentage variation5 how much ofspecies variation is explained by variables F-values5 sig-nificance level of Axis 1 trace5 percentage variance onAxis 1 Correlations 405

Years Percentagevariation

F-value

Trace Correlatedvariables

1990 34 707 157 ndash1993 65 128 340 ndash1996 79 178 420 y rill1999 109 331 598 y rill2003 112 338 616 y rill2006 125 390 698 y rill flat2008 122 335 654 y rill flat

Table 3 Mean value of characteristics of six habitat types (HT) HTs differed significantly for each terrain characteristic(Po00001 one-way ANOVA followed by Bonferroni tests for differences among means Values with same superscript fallwithin the same group)

Terrain HT-1 HT-2 HT-3 HT-4 HT-5 HT-6 F-value(n5 128) (n5 28) (n5 37) (n5 64) (n5 61) (n5 82)

Smooth 300a 286a 289a 192bc 198b 180c 1911Gully 005b 036b 030b 203a 026b 222a 2085Rill 015d 029cd 057c 061c 238a 117b 1141Rock 000d 100bc 132b 080c 062c 216a 1466Sand 000b 000b 019b 002b 010b 090a 306Pumice 300a 300a 200c 261b 257b 128d 1470

Fig 4 Distribution of CTs on the grid in 2008

862 del Moral Roger et al

tories proceeded from pioneer plots dominated byAnaphalis to early seral types that had accumulatedCistanthe Agrostis and Luetkea These developedfurther into mid-seral communities when Penstemonbecame dominant and Juncus became widespreador they developed directly into late seral commu-nities dominated by Penstemon and Agrostis In2008 some mid-seral CTs persisted but most plotswere in a late-seral community that had changedlittle since 2005 For each annual interval 1988 to2008 a plot could persist in a CT progress to a moredeveloped one or regress to an earlier stage Therewere 301 retrogressions and 1014 progressions The

five extant CTs persisted significantly longer thandid the five vanished CTs (Table 4)

We clarified temporal vegetation changes usingDCA (Fig 7) Total variance (l) was 159 (DCA-15 155 DCA-25 74) Turnover on DCA-1was 53 half-changes on DCA-2 it was 39 Pioneerspecies had high values on DCA-1 species with lowDCA-1 values appeared later and were persistentMean DCA-1 position of the plots changed sig-nificantly between years (r2 5 0637 Po00001Spearman nonparametric correlation r5 079Po00001) DCA-1 scores by year changed rapidlythrough 2001 (ANOVA Po00001) then slowedDCA-2 changes ceased after 1997 but ANOVA was

Fig 5 Annual changes of mean species richness and meancover index Differences (Po005) among measures de-termined by a repeat measures ANOVA followed by theBonferroni test Values with the same associated letter fallwithin the same group of values

Fig 6 Trajectories showing the number of transitionsbetween CTs Numbers in parentheses after persistent CTsare the number of plots in 2008 Many arrows transitionnear the mid-point to differentiate between transitions toand from a given CT

Table 4 Number of continuous years (standard devia-tion) for which each CT persisted Differences amongmeans are significant (one-way ANOVA Po00001 fol-lowed by Bonferroni tests for differences among meansSuperscripts indicate membership in same group) Mini-mum run is one in each case

CT Years SD Max

A 329b 135 7B 200ab 115 4C 107a 026 2D 238b 142 7E 181ab 117 7F 398c 274 10G 391c 172 6H 558d 256 10I 419cd 167 8J 378c 185 7

Fig 7 Detrended correspondence analysis of plots by CTand by year Differences (Po005) among measures de-termined by ANOVA followed by the Bonferroni testFor each set values with the same associated letter fallwithin the same group of values for DCA-1 (no data for1998 and 2002 alternate plots were sampled in 2007 and2008) The inset is a separate DCA of plots in 2008

Primary succession trajectories 863

significant (r2 5 0206 Po00001 Spearman non-parametric correlation r5 025 Po00001) Thevariation in DCA-1 scores within years declined (thelinear correlation of years vs standard deviationwas Po00001 t5 533 r2 5 0637)

The mean DCA position of the 10 CTs did notreveal a single trajectory CT-C (Anaphalis) wasmore closely associated with pioneer CTs but wasnever the first CT on a plot CT-G (Penstemon-Agrostis-Cistanthe) was aligned with late-seral CTsWe also conducted a DCA of 2008 plots (Fig 7inset) b-diversity was low with l5 097 (DCA-15

133 DCA-25 84 ca 18 HC along DCA-1)No trajectory was evident from this analysis

Discussion

There are so few long-term longitudinal studiesof early primary succession on stressful sites be-cause initially so little happens The vegetation ofthe Abraham Plain remains sparse 30 years after theeruption It has developed by interwoven succes-sional trajectories that include plots that divergeconverge regress and progress to form a braidedsuccession Neither habitat factors nor competitionappears to have provided strong filtering effects onspecies composition although unmeasured soilvariables (eg moisture and nitrogen content) andfactors related to stochastic colonization and estab-lishment (cf del Moral amp Ellis 2004) may explainsome of the remaining unexplained variation Theimmigration seed density was low (Wood amp delMoral 2000) many suitable sites appeared vacant inearly years (del Moral amp Bliss 1993) and severalspecies occupied similar sites Chance may governsuch patterns We propose that when environmentalvariables have only slight effects on species compo-sition consistent successional trajectories areunlikely to emerge Below we discuss questionsabout predictability rate and pattern of successionon Abraham Plain that are the focus of this study

Environmentndashvegetation connection

Predictable relationships between environ-mental variables and vegetation had developed by2008 but they remained low Priority effects sec-ondary disturbances elk grazing and seasonalfluctuations may affect these trajectories in the fu-ture rendering it unlikely that explained variationwould become the dominant factor before forestsdevelop

We did not use soil factors because plants es-tablish at a smaller scale than the plot size and it isimpractical to measure such factors as moisture ineven a few plots Although soils were young andhomogeneous it is likely that inclusion of soil fac-tors (eg pH and moisture) would have improvedexplanatory power based on similar studies (delMoral 2009a) explained variance might doubleTopographic surface and position factors alone diddemonstrate increasing explained variance of spe-cies pattern over time but this trend was lesspronounced than in other studies On the northslope of Mount St Helens in 1992 only spatial fac-tors explained any variation (15) but after 10 yrenvironmental factors dominated (28 del Moralamp Lacher 2005) Baasch et al (2009) produced oneof the very few primary succession studies to showthat relationships between vegetation and ex-planatory variables increased over time They toosuggested that the impact of stochastic processesdeclined over time

The spatial distribution of CTs suggested someenvironmental influence Rills provide havens forpioneers (eg Anaphalis Hieracium and Hy-pochaeris) Gullies offer protection from wind andaccumulate snow which allowed species such asAnaphalis and Salix to colonize and persist Pen-stemon and Juncus were common in more exposedsites where cover and diversity were least In suchsites species that are less stress tolerant (eg SalixPolytrichum Saxifraga and Juncus mertensianus)were uncommon while stress-tolerant species suchas Cistanthe Juncus parryi Agrostis pallens andCarex microptera reached their maximum values

Rates

Habitats with limited fertility or moisture typi-cally have low succession rates (Donnegan ampRebertus 1999 Anderson 2007) and succession onAbraham Plain has been no exception Mean annualDCA scores were in a single group of values Be-tween 1988 and 1999 these scores formed sevengroups Surrounding habitats on Mount St Helensdeveloped more rapidly On an exposed lahar on thesouth slope richness averaged 16 species and thecover index averaged 73 (del Moral 2009b) prior to2008 In contrast mean richness on the AbrahamPlain in 2008 was 108 and the mean cover index was314 The proximity to forest more fertile soils andabundance of the N-fixing Lupinus may all con-tribute to accelerated primary succession at suchsites relative to Abraham Plain

864 del Moral Roger et al

Trajectories

Succession trajectories have been discussedthoroughly (cf Leps amp Rejmanek 1991) When thevegetation of initially homogeneous sites developslocally distinctive characteristics divergence occurs(Bossuyt et al 2003) Divergent trajectories occur onmany surfaces including landslides glacier fore-lands wet slack dune vegetation floodplains andmine tailings (Walker amp del Moral 2003) Lanta ampLeps (2009) showed experimentally that differentialdispersal could promote divergence when there arepriority effects Divergence may be more likely instressful habitats when different species arrive firstand where climatic factors are highly variable(MacDougall et al 2008) Convergence describestrajectories that become more similar over timeConvergence has been assumed to result when cli-matic factors dominate and there is evidence thatwhen dispersal effects are small and competitive ef-fects are strong convergence to a single communitydoes occur (van Oijen et al 2005 Anthelme et al2007) Convergence is also more likely when thereare few alternative stages as on glacial forelands(Hodkinson et al 2003 but see Robbins ampMathews2009)

Here we show directly that early succession un-der stressful conditions can include trajectories thatdiverge converge regress and progress to form areticulate succession Our study joins a growingbody of work that demonstrates the frequency ofmultiple trajectories in early primary succession (egdel Moral 2007 Prach amp Hobbs 2008 Walker amp delMoral 2009) Forecasting trajectories and identify-ing factors that limit their development is crucial tounderstanding community assembly mechanismsOn the Abraham Plain persistent species graduallyreplaced readily dispersed pioneer species Subtlehabitat variation annual weather fluctuation andlandscape effects combined to produce communitiesrelated through a network of development and re-sponsible to topographic variation

The first colonists were species with excellentdispersal but isolation allowed for only a meagerseed rain they established pioneer communities thatsoon developed into other CTs characterized byspecies with greater persistence (cf Ejrnaes et al2006) Several alternative transitions were possiblefor most CTs suggesting that stochastic processesinitially dominated transitions Gradually persis-tent species invaded and attained dominance Minortopographic variation affected composition some-what resulting in some successional transitions thatare more common than others Of annual transi-

tions between successional stages 51 wereregressions and 170 were progressions DCA re-vealed no single trend among the CTs The two mid-seral community types were central to the transitionpatterns while the three late-seral CTs occurred indifferent directions from this centre

It is unlikely that complete convergence will oc-cur Even when trees dominate understoreyvegetation is likely to retain variation initiated dur-ing early succession and maintained by topographicvariation and priority effects The understorey offorests on the south side of Mount St Helens de-monstrates such variation (del Moral amp Ellis 2004)There is no evidence yet for a single target commu-nity

Strong assembly rules are more likely to pro-duce convergent trajectories than weak ones Whilestrong assembly rules may exist in mature vegeta-tion (see Navas amp Violle 2009) they appear weak onAbraham Plain and are more effective for functionaltypes than species One rule may be that long-livedwind-dispersed species accumulate at the expense ofshort-lived pioneers A second may be that grami-noids with rhizomatous growth (eg Agrostispallens) and prostrate shrubs (eg Penstemon) arecomplementary and can coexist indefinitely A thirdmay be that fleshy fruit-producing shrubs (eg Ru-bus spp Vaccinium spp) cannot establish undercurrent conditions Finally although seeds of con-ifers and Populus balsamifera are commonconditions do not yet favour their developmentThese rules are weak and exceptions are commonThe spontaneously developing vegetation of theAbraham Plain is constrained by isolation from po-tential colonists as well as a habitat that challengesplant growth

Acknowledgements We thank the US National Science

Foundation for funding (BSR8906544 DEB9406987

DEB0087040 and DEB0541972) and the Mount St He-

lens National Volcanic Monument for permissions

Valuable comments by J G Bishop M P Fleming Jan-

neke Hille Ris Lambers L R Walker J H Titus and two

anonymous reviewers improved the paper Field assis-

tance was provided through the years by S Anderson W

Arnesen S Bard A Coogan S Crawford K Dlugosch

A Eckert E Ellis C Fairbourn C Muerdter M Flem-

ing T Fletcher R Fuller A Grant E Jenkins M

Johns C Jones L McMullan P Moody R Robham L

Rozzell J Sandler R Sewell-Nesteruk M Spasojevic

M Tweiten S Wilson B Witte and C Wolfe This is

contribution No 60 to the Mount St Helens Succession

Project

Primary succession trajectories 865

References

Anderson KJ 2007 Temporal patterns in rates of

community change during succession American

Naturalist 169 780ndash793

Anthelme F Villaret JC amp Brun JJ 2007 Shrub

encroachment in the Alps gives rise to the

convergence of sub-alpine communities on a regional

scale Journal of Vegetation Science 18 355ndash362

Baasch A Tischew S amp Bruelheide H 2009 Insights

into succession processes using temporally repeated

habitat models results from a long-term study in a

post-mining landscape Journal of Vegetation Science

20 629ndash638

Bossuyt B Honnay O amp Hermy M 2003 An island

biogeographical view of the successional pathway in

wet dune slacks Journal of Vegetation Science 14 781ndash

788

Clifford HT 1976 Dendrograms and their

interpretation In Williams WT (ed) Pattern

analysis in agricultural science pp 96ndash109 Elsevier

Science Amsterdam NL

del Moral R 2007 Vegetation dynamics in space and

time an example from Mount St Helens Journal of

Vegetation Science 18 479ndash488

del Moral R 2009a Increasing deterministic control of

primary succession onMount St Helens Washington

Journal of Vegetation Science 20 1145ndash1154

del Moral R 2009b Mount St Helens permanent plots

and grids 1980 to present Available at http

protistbiologywashingtonedudelmoral Accessed 8

December 2009

del Moral R amp Bliss LC 1993 Mechanism of primary

succession insights resulting from the eruption of

Mount St Helens Advances in Ecological Research

24 1ndash66

del Moral R amp Ellis EE 2004 Gradients in hetero-

geneity and structure on lahars Mount St Helens

Washington USA Plant Ecology 175 273ndash286

del Moral R amp Lacher IL 2005 Vegetation patterns 25

years after the eruption of Mount St Helens

WashingtonAmerican Journal of Botany 92 1948ndash1956

Donnegan JA amp Rebertus AJ 1999 Rates and

mechanisms of subalpine forest succession along an

environmental gradient Ecology 80 1370ndash1384

Ejrnaes R Bruun HH amp Graae BJ 2006 Community

assembly in experimental grasslands suitable

environment or timely arrival Ecology 87 1225ndash1233

Felinks B ampWiegand T 2008 Exploring spatiotemporal

patterns in early stages of primary succession on

former lignite mining sites Journal of Vegetation

Science 19 267ndash276

Fuller RN amp del Moral R 2003 The role of refugia and

dispersal in primary succession on Mount St Helens

Washington Journal of Vegetation Science 14 637ndash

644

Hodkinson IA Coulson SJ amp Webb NR 2003

Community assembly along proglacial chrono-

sequences in the high Arctic vegetation and soil

development in north-west Svalbard Journal of

Ecology 91 651ndash663

Johnson EA amp Miyanishi K 2008 Testing the

assumptions of chronosequences in succession

Ecology Letters 11 419ndash431

Lanta V amp Leps J 2009 How does surrounding

vegetation affect the course of succession a five-year

container experiment Journal of Vegetation Science

20 686ndash694

Legendre P amp Anderson MJ 1999 Distance-based

redundancy analysis testing multispecies responses in

multi-factorial ecological experiments Ecological

Monographs 69 1ndash24

Leps J amp Rejmanek M 1991 Convergence or

divergence what should we expect from vegetation

succession Oikos 62 261ndash264

Leps J amp Smilauer P 2003 Multivariate analysis of

ecological data using CANOCO Cambridge

University Press Cambridge UK

MacDougall AS Wilson SD amp Bakker JD 2008

Climatic variability alters the outcome of longterm

community assembly Journal of Ecology 96 346ndash354

Navas ML amp Violle C 2009 Plant traits related to

competition how do they shape the functional

diversity of communities Community Ecology 10

131ndash137

Pickett STA Cadenasso ML amp Meiners SJ 2009

Ever since Clements from succession to vegetation

dynamics and understanding to intervention Applied

Vegetation Science 12 9ndash21

Prach K amp Hobbs RJ 2008 Spontaneous succession

versus technical reclamation in the restoration of

disturbed sites Restoration Ecology 16 363ndash366

Prach K amp Rehounkova K 2006 Vegetation succession

over broad geographical scales which factors

determine the patterns Preslia 78 469ndash480

Robbins JA amp Matthews JA 2009 Pioneer vegetation

on glacier forelands in southern Norway emerging

communities Journal of Vegetation Science 20 889ndash

902

Swanson FJ amp Major JJ 2005 Physical events

environments and geologicalndashecological interactions

at Mount St Helens March 1980ndash2004 In Dale

VH Swanson FJ amp Crisafulli CM (eds)

Ecological recovery after the 1980 eruption of Mount

St Helens pp 27ndash44 Springer New York NY US

USDA National Resource Conservation Service 2009

The Plants Data Base National Plants Data Center

Baton Rouge LA US Available at http

plantsusdagov Accessed October 15 2009

van Oijen D Hommel P den Ouden J amp de Waal R

2005 Effects of tree species composition on within-

forest distribution of understorey species Applied

Vegetation Science 8 155ndash166

Walker LR amp del Moral R 2003 Primary succession

and ecosystem rehabilitation Cambridge University

Press Cambridge UK

866 del Moral Roger et al

Walker LR amp del Moral R 2009 Transition dynamics

in succession implications for rates trajectories and

restoration In Hobbs RJ amp Suding KN (eds)

New models for ecosystem dynamics and restoration

pp 33ndash49 Island Press Washington DC US

Walker LR Bellingham PJ amp Peltzer DA 2006

Plant characteristics are poor predictors of microsite

colonization during the first two years of pri-

mary succession Journal of Vegetation Science 17

397ndash406

Wood DM amp del Moral R 1988 Colonizing plants on

the Pumice Plains Mount St Helens Washington

American Journal of Botany 75 1228ndash1237

Wood DM amp del Moral R 2000 Seed rain during early

primary succession onMount St Helens Washington

Madrono 47 1ndash9

Supporting Information

Additional supporting information may befound in the online version of this article

Fig S1 Abraham Plain Sept 1980 pumice de-posits over scoured bedrock looking north towardsthe study site

Fig S2 Abraham Plain July 2009 lookingnorthwest across the study site

Fig S3 Changes in mean cover index for lesscommon species

Fig S4 Relative cover (proportion of totalcover in the sample) of less common species in thefour stages of early primary succession The firstfour species are persistent the last two are pioneerspecies (defined in Appendix S1)

Fig S5 Annual changes of mean H0 and meanSimpsonrsquos dominance index Differences (Po005)among measures determined by a repeat measuresANOVA followed by the Bonferroni test Valueswith same associated letter fall within the samegroup of values

Table S1 Structure and species composition ofCTs found in 2008

Appendix S1 Characteristics of species en-countered on Abraham Plain grid

Please note Wiley-Blackwell is not responsiblefor the content or functionality of any supportingmaterials supplied by the authors Any queries(other than missing material) should be directed tothe corresponding author for the article

Received 9 December 2009

Accepted 14 March 2010

Co-ordinating Editor Dr Kerry Woods

Primary succession trajectories 867