I

Reprinted from MADRONO, Volume 23, Number 7, July 1976(Published July t , 1976)

VEGETATION SURROUNDINGKINGS LAKE BOG, WASHINGTON

GnorcuBN K. LnsBoNrxDepartment of Botany,

University of British Columbia, Vancouver V6T 1W5

Rocnn lor, MonalDepartment of Botany, University of Washington, Seattle 98195

t

VEGETATION SURROUNDINGKINGS LAKE BOG, WASHINGTON

GnercHBN K. LoBBoNrxDepartment of Botany,

University of British Columbia, Vancouver V6T lW5

Rocpn DEL MoRALDepartment of Botany, University of Washington, Seattle 98195

This communication describes vegetation surrounding a small bog lakein lowland King County, Washington. We tested the hypothesis that theplants comprising vegetation surrounding this bog are arranged in zonesrather than continuously in response to environmental gradients.

Vegetation can be analyzed by two distinct approaches. Numericalclassification methods group samples into categories and emphasize simi-larities within and differences between catesories. Ordination methodsarrange samples sequentially based on environmental data or on composi-tional similarity of the vegetation itself or both. Both methods can con-tribute to the understanding of vegetation and the causes of vegetationorganization.

Gauch and Whittaker (1972) documented the growing realization ofmany ecologists that simple geometric ordinations are superior to moreelegant methods in several ways. They are mathematically straightfor-ward and the results are usually more reliable. Ordination approacheshave been successfully applied to many situations but seem best adaptedto those in which beta diversity is low and vegetation scale being investi-gated is intermediate (Gauch, 1973).

r9761 LEBEDNIK E

that stresses dominance ttpein an ordination. However, csiderable variation within aeight types. Since we obtainwe believe that this conclusgreater species richness.

Fitzgerald ( 1966) studiedBraun-Blanqu6t. She discendiscussions. Our methods sqpattern exists. They also retare arbitrary modes in an es

Coxcr,The vegetation of Kings L

and ordination methods. Thrvegetation forms distinct zocontinuous pattern. Combinrgradient except for one disalba and several other sedgposed portion of the floatinginvaded. Zones may be recqdient from lake to forest. budominant species. Qu1 slasqi'were recognized by Fitzgeralthat these form a nearly cont

The combined use of clastudy reveals a correlatiolproaches. The similarity betsimilarity projection analysismathematical approaches. Frwe recommend numerical cluse of discriminant analysismeans for objective reallocatof each group. Where the tthen ordination methods aresegments fo'r ease of considerbe done on the basis of numr

AcWe are indebted to A. R. I

nik, who each made valuablevided by the Department ofthe College of Arts and Scienhaeuser Company permittedtheir Snoqualmie Tree Farm

LEBEDNIK & DEL IVIORAL: VEGETATION 387

\regetation classification has a venerable history with many "tradi-

tions" (Shimwell, 1973; Mueller-Dombois and Ellenberg, 1974) thatrvill not be reviewed here. We favor methods that emphasize dominantsrather than differential or character species. We selected an agglomera-tive hierarchical method devised by Orloci (1969) based on comparisonsof information gained when two groups are merged and that produces adendrogram. This method is superior to non-hierarchical numerical clas-sification methods (Williams, 197l) .

All hierarchical classifications are plagued by the problem of termina-tion; analysis proceeds until all samples are placed into a single group.The investigator must intuitively determine when joining groups shouldcease. Del N{oral (1975) proposed a solution to this problem by usingstepwise discriminant analysis. Clustering terminates when as many ormore groups are recognized by the classification as are tentativelythought to occur in the field. Discriminant analysis assesses statisticaldifferences between groups, reallocates individual samples if necessary,and indicates the relative importance of species in constructing theclassification.

Kings Lake Bog lies in the western foothills of the Cascade Mountains(N47 o3 5'40" , W l2la 46'35") , at 280 m, well within the Tsuga hetero-phylla zone in the Puget Trough (Franklin and Dyrness, 1973). KingsLake is 1.1 ha and is surrounded by a 2l ha bog. It is 4.6 m deep off themat edge (Wolcott, 197 3) . The surrounding forest was logged and burnedin the 1930's by the Weyerhaeuser Company, upon whose land the lakeis located (Fitzgerald, 1966).

The area we sampled extends from the floating mat edge to a portionof the surrounding forest (flS. t). A lagg (Oswald, 1933) borders theeast and southeast lakeside; it is covered by fallen logs and abruptlyjoins higher ground. Spiraea douglasii occupies much of this area. Theremaining sides of the bog exhibit a more gradual change in forest andour sample was confined to this zone of transition.

Fitzgerald (1966) showed that the microclimate of the open bog wasmore extreme than that of the surrounding forest. Temperature extremesvaried annually from 43" C to - 17o C on the bog and from 21" to6" C in the forest. Diurnal variation, due to more insolation and intensereradiation, was much greater on the bog than in the forest. For example,in one 24 hr period in July, 1965, temperature varied 20o C on the bogcompared to 60 C in the forest. The bog experiences many more days ofsub-freezing temperature than the surrounding forest. The water tablefluctuates such that during early spring, standing water may be foundabove the moss layer throughout the bog while in late summer the matis very dry. By these climatic criteria, the bog habitat is more severe thanthe forest.

MrrHonsVegetation sampli.ng. We sampled vegetation along a 216 m by 10 cm

line intercept transect from the mat edge into forest. Within alternate

MADRONO lVol. 23 r97 6l LEBEDNIK T

T,qnr,n 4. Mrew Cown PrnRnvrsno Gnoups Itnnrrrno nl I

Taxon

Ledum groenlandicum 1.0Kalmia occidentalis +.9Vaccinium oxycoccus fSJRhynchospora alba 8OJSphagnum subsecundum 0.0Pleurozium schreberi 0.0Cladonia rangiferina 3S.OTsuga heterophylla 0.0Gaultheria shallon O.0Pteridium aquilinum 0.0

var. pubescensHylocomnium splendens O-0Eurhynchiumpraelongum 0.0Cornus canadensis 0.0Maianthemum dilatatum 0.0

larity projection ordination '

approach recognizes the comThe first axis is defined t

97 (group F) and the secorand 67 (group H). The ininated vegetation and the sThe resultant ordination is dnized communities are circuldiversity) along the first pnis distortion of the type re;Bray-Curtis polar ordinatioris closely related. Samples trtend to be distant from it sohump results because Percerlogical relationships between

Ordination suggests that,spora dominated communitlspora community is uniqueshrubs. Communities B, C, afication hierarchy, are likewicover of Ledum and Kalnnium) and G (Pteri,dium / H'cept for sample 93 ) and careach other and from E (GGaultheria/ Pleurozi.um) . Hc

T

oo

l Fl o

!

o

o3

oot o

|-el

EF,I'TIIN

ffi.$-'jr.-"$

LAKE

S E D G E Z O N E

SHRUB ZONE

TRANSITION ZONE

LAGG

o I1ETERS 5oo

Frc. 1. Vegetation zone map of Kings Lake Bog showing location of accesstrail and transect. Forest is unshaded.

I m segments, we recorded leaf intercept at each centimeter. The bogcontains hummocks between the forest and the floating mat. In order toreduce sampling errors inherent in such heterogeneous communities,cover was expressed as the percentage of centimeter occurrences permeter. We did this because we were not interested in microtopographicpatterns, which we observed on a scale of a few centimeters. Samplingerror is fairly high for rare species but lower for common species. A pre-liminary study compared 1 m2 quadrats with line intercept values and

t 9 7 6 1 LEBEDNIK & DEL MORAL: VEGETATION 389

gave similar results (means and standard deviations). The analyses re-ported here are confined to l4 more common species.

Quantitativc analysis. Percent cover for the 14 dominant species wasused in our analyses. Orloci's (1969) mutual information method wasimplemented with a computer program modified from Goldstein andGrigal (1971). The matrix is inspected for local minima and the corre-sponding subsets are united to form new groups with minimum informa-tion gain. The results are displayed by a dendrogram that shows themutual information at which each subset is joined.

Once tentative groups are identified tesLs of significance of group dif-ferences and tests of group membership are accomplished with stepwisemultiple discriminant analysis (SMDA). N{athematical details are foundin del Moral (1975), who followed Dixon (1970). Given a partit ioned setof data, SN{DA determines the order in which variables (species) areuseful to distinguish between groups. At each successive step an F-testfor differences between groups is made. Individual samples are evaluatedto determine the class to which they belong. This permits an a posteriori,objective, reallocation of the samples. The phytosociological relationshipsamong the groups are determined by means of canonical rotation of thediscriminant axes. What results is a set of mutually orthogonal (inde-pendent) axes in which variation is maximized on the first axis. Samplesare plotted according to the first two canonical axes, thus providing anoptimum two-dimensional picture of the relationships between groups.

\Ve also r-rsed similarity projection ordination (Gauch and Whittaker,197 2 ) to determine vegetational relationships. This is a multidimensionalextension of the polar ordination method of Bray and Curtis (1957). Toeffect an ordination, we selected two samples that contrasted strongly.Sev:ral trials, bas:d on the similarity matrix, were used. We selected aninternal association of 90 percent arbitrarily since a) published valuesvary from 80 to 100 percent and b) Gauch and Whittaker (1972) con-cluded that internal association has little effect on rank order in ordina-tions. All other samples are projected orthogonally onto the axis definedby the end stands. A second axis was defined by stands chosen with thefollowing criteria: stands from the center of the first axis, stands quitedissimilar to each other, and stands distinct from the first axis. Addi-tional dimensions did not improve the ordination.

Rnsur,rs eNu DrscussrtxQaalitative Description. Prior to selection of transect location, a gen-

eral survey was made. Fitzgerald (1966) divided this bog subjectivelyinto five zones. We added the lagg area and classified samples of vegeta-tion into one of the six zones and recorded species composition. Table 1provides a qualitative list of species encountered in each zone. Nomen-clature of vascular plants follows Hitchcock and Cronquist (1973), mossnomenclature follows Lawton (1971), and lichen nomenclature followsHa le (1969 ) .

390 MADRONO [Vo l .23

Teers 1. OccunnnNcns ol Ar-r- Tex,r Ercourmnno rrv Srx ZoNns Anouro KrrcsBoc. F - forest, T: transition zone, S: shrub zone, Se = sedge zone, L = lagg,P : picneer zone. Species used in quantitative analysis are marked *.

r97 6l LEBED\IX &

Tlsr,E, 3. F-vsurs Bf,TT-rsdom: 14, 87. \;alues esceedingthe same group u'ith P < O-0f .

A46.8856r.44863.06160.62678.2853 8 .05159.262

B6.66s C

10.438 6.S;;20.405 2+.16-i54.639 7:.O;i27 .907 3l,85S25.696 27.9r9

drogram tends to confrrm thtA joins B, C, and D at a hi1related and D is more closehof E and H is demonstratedtween H and D and closer togram.

Discriminant analysis indfirst canonical axis accountssecond accounts f.or 21.2 Pthe discriminant axes. Figurrcal distribution of sample-cerning the relationships amto be closely related, E and Itinct and A is very distinct-Group A is in the sedge zonrsome rotation of these asestransect that is a gradient otent (Lebednik, unpublishec

Key to Vegetation Clustamation to produce a key to tthose species highest in the rthe composition of the eighlsubjectively. The followingthe field:

a. Rhynchospora alba doriaa. Rhyncospora alba rare a

shrubs commonb. Kalmia occi.dentulis d

c. Plewozi.um schrebe,cc. P leur o ziurn s c hr e b a

d. Sphagnum subserdd.Sphagnum sabse,

terina common

Taxon FVegetation zone

T S S e L

*Tsuga heterophyllaThuja plicataPicea sitchensisPseudoi.suga menziesiiAcer circinatum

*Ledum groenlandicum*Kalmia occidentalis*Vaccinium oxycoccusVaccinium alaskenseVaccinium ovalifoliumVaccinium parvifoliumSpiraea douglasii

*Gaultheria shallonMenziesia ferrugineaRubus spectabilisSambucus racemosaSymphoricarpos occidentalisMaianthemurnr dilatatum

*Cornus canadensisD - , - ^ I ^ , , - i A ^ * ^! J r u 1 4 u r r r r r u r 4

Circaea alpinaMontia sibiricaGalium aparine var. echinospermumActaea rubraCorallorhiza sp.Listera sp.Lycopodium annotinum

*Pteridium aquilinum var. pubescensTrientalis arcticaAgrostis oregonensisPotentilla palustrisNuphar polysepalumMenyanthes trifoliataDrosera rotundifoliaCarex cusickiiCarex vesicaria

Juncus ensifolius

Juncus effusus var. compactus*Rhynchospora albaCicuta douglasiiLysichitum amrcricanumPolysticum munitum

*Cladonia rangiferinaUsnea longissima

*Sphagnum subsecundum*Pleurozium schreberi

Rhytidiadelphus triquetrusAulacomnium androgynePolytrichum juniperinum

xXXXx

Xx

XX

X

xXx

xXXxXXXXxXxX

xXXx

xXxx

xx

XXXX

xXxxx

X

X

xXXX

XXX

xxX

XxX

X

XXxXxxxXxXx

X

xxXXXxXx

XxxX

x xxxXxXX

x

x

x

xx xX Xx x

xxX

x

X

xXx

XXx

X

797 6 l LEBEDNIK & DEL MORAL: VEGETATION

Tirsr,r 1. Continued.

*Hylocomnium splendens*Eurhynchium praelongum var. stokesii

Isothecium stoloniferumDicranum sp.Bryum sp.

Total numbers

Each vegetation zone is recognized by both dominants and physlog-nomy. Aspect zonation around the lake is clear. At the edge of the sphag-num mat is a peat-sedge zone in which dominance is shared by Rhyncho-s p or a al b a, V ac c i,nium oxy c o c cu s, and C lad onia r an gi, t' erina. A shrub zonedominated by Ledum. groenlandicum, Kalnoia occidentalis, and stuntedTsuga keteropltyllaborders the first zone. This mounded zone is hetero-geneous with mounds and intermounds having different species composi-tion. Intermounds are dominated by Sphagnum subsecundum and Pleur-oziunt schreberi. The forest proper rises abruptly from the edge of t'.:eshrub zone.

Closer inspection reveals additional distinct areas. Between the sedgezone and the lake lies a very narrow band consisting of tlvo phases andtermed the pioneer zone: Ledum groenlandicum, Kalmia occidentalis,and Potentilla palustri,s form a matrix extending into the lake. \Vithinthis matrix may be found Sphagnum subsecundum and Drosera roiund.i-Jolia.

\\rithin tire sedge-peat zonet considerable seasonal variation in aspectdominance occurs. R.ttynchospora dies back to an underground rhizomeannuaily. Cladonia is sporadic. While numerous, Vaccinium ox1,6,66,5x4slras low cover. All of these are obscured by Rhynchosloro during springand summer.

The shrub zone maintains its appearance throughout the year sincethe dominants are evergreen and form a dense intertwined layer. The matlayer beneath the shrubs consists of Vaccinium oxycoccus, Spltagnumsub s e cundum, C ladonia ran gi! erina, and Pleur ozi.um s chr eb eri. Punctu-ating this mat are sporadic Lysi,chitum am.eri,cqnum individuals in de-pressions. Pools of various sizes also occur and are dominatedby Iuncusefiusus, J. ensilolius var. compactus, Carer cusicki,i, and C. aersicaria,all of which also occur in the sedge zone. (These pools were not sampledby the intercept transect.)

The forest shrub transition zone is an interdigitation of forest andshrub zone species. Pteridi,um aquili.num var. pubescens and Gaultheriashallon extend farther into the shrub zone than other forest species. tr/ac-ci.nium alaskense and then V. ovalifoli.uln are encountered as the hem-locks gain stature. Cladonia rangilerina and, V. ofiyclccus are the firstbog species to disappear. Sphagnum subsecundum and Pleurozium schre-beri. are replaced gradually by Hylocomniurn splendens, Eurynchiunx

xxxxX33

MADRONO [\;ol. 23

praelongum, and Rhytidiadelphus triquestrus. Within the forest, the ter-rain is level and species common to Tsuga heterophytla forests occur(Tab le 1 ) .

The lagg is a special type of vegetation that does not extend com-pleiely around the bog. It is richer in species than all but the forest zone.Spilaea douglasit,, Lysichitum ameri.canurn, and Symphoricarpos occi-d^nta!is are among the common species.

\\rhile most species on the bog display regular patterns with respect tothe major environmental gradients associated with the degree of "boggi-

n?ss'), some orcur sporadically in response to the presence of water pools.Lysichitwm. americanum is sporadic throughout the bog from mat

edge to forest margin. Turesson ( l916) believed that Lysichiturn was re-lictual, maintaining itself by shading and mechanically destroying Sphag-num, the main invader. Each individual Lysichi,tum is rooted in a deephollow, the edges of which may be lined with forest moss species suchas Aulac omnium andr o gyne, Rhy ti,diadelp hus tri.quetrus, or P olytri;c humjwniperinum. Turesson's hypothesis does not explain the presence of Ly-sichitutn in the hanging mat.

I{any individuals of Lysichitum are found in the lagg, which is similarto its usual swampy habitat. Pools may be formed in the mat by variousdisturbances, including wind throw of adjacent trees and trampling byman or beast. These pools may be invaded and enlarged by Lysichitum.Therefore, we believe that Lysicltitum invades disturbed bogs. Its densityis being monitored annually to test this hypothesis.

Other species occur in atypical patterns in this bog. Most patterns aredue to the presence of the lagg or to downed timber creating less boggyconditions and variable seed beds. Species such as l{uphar polysepalumand Menyanthes triloliatd occur in pools in the mat even though theseare usually considered lake margin species. Where fallen logs occur, forestunderstory species invade to the edge of the lake. Thus Vaccini,um parvi-folium, Cornus canadensis, and Maianthemum dilatatum, occlr on rot-ting logs at lake margin. These species often occur on fallen logs in theforest and such a distribution is not surprising.

Zones viewed on the basis of dominant species are identified in thefield. However, our fundamental question is not whether zones can bedistinguished subjectively. Several generations of ecologists have dem-onstrated that this is possible and useful. We wished to determine ifsuch subjective zones maintain their discreteness in the face of an objec-tive analysis.

Quanti,tatile Descripti,on. The location of the transect with respect toa vegetation map of Kings Lake Bog, which emphasizes zonation, isshown in Figure 1.

Dendrogram. We clustered 108 samples by MINFO (Goldstein andGrigal, l97 l). The dendrogram (fig. 2) confirms the validity of thezonation concept when dealing with dominant vegelation. Sample num-bers across the bottom of Figure 2 are the sequence in which samples

l -

r97 6l LEBED\IK &

1 0 -

Samde

Frc. 2. Dendrogram produceused in subsequent analyses. Tlsimilarity required to produce Itransect.

were encountered from the Iat the 5 X 103 level of mlgroups, corresponding to theand the forest, are recognizersults in the eight groups labenoted in dendrograms emerp

c.9oE0

s(E=3

tr.9GEo

G

=E

t e 7 6 l

1 0 -

LEBBDNIK & DEL MORAL: VEGETATION

Sample Number

Frc. 2. Dendrogram produced by MINFO. Letters indicate the eight groupsused in subsequent analyses. The dotted line (5 X 103) indicates the level of

;lTj*:i* required to produce four zones. Sample numbers are sequence in the

were encountered from the Lake to the forest. rf clustering is terminatedat the 5 X 103 level of mutual information (dotted line), then fourgroups, corresponding to the sedge zone, the shrub zone, the transition,and the forest, are recognized. Termination at 2 X 103 (dashed line) re-sults in the eight groups labeled in Figure 2. Some features not ordinarilynoted in dendrograms emerge from a consideration of this one. In the

l9.1 MADRONTO lVc l . 23

dendrogram, sample similarity is indicated by the level of first joining.The degree of sample s,milarity appears highest in the sedge zone, leastin the transition, and intermediate in the others. Thus, the dendrogramproduces classes but simultaneously differential relationships betweensamples imply a gradational pattern.

Discri,minant analysis. The eight subsets were investigated by step-wise discriminant anaiysis to determine whether any samples should bereallocated, to determine whether any pair of clusters should be merged,to determine the relationship of the groups to each other, and to deter-mine the relative importance of species. In Table 2 we list the order inwhicir species were used to distinguish among the classes. The two mostirnportant species distinguished between forest and sedge zones, the thirdis primarily a shrub zone species, and the fourth is a transition zones)er ies.

.After 14 stens, the eight groups are all highly statistically significantlyd i f ferent (P < 0.01, d. f . : 14,87) . Six samples are real located by th isanalvsis, four transferring samples from C to B. The final F-table (Table3) indicates that clusters B, C, and D are closely related. These threeare united by Ledum dominance, but D bas high cover oI Pleuroziumschreberi, rvhile B and C are distinguished on the basis of Vacciniunoox\coccus. Cluster A can be recognized by dominanceby Rhyncltosporaand Cladonia. Clusters F, G, H, and E belong to the forest and transi-tion. Clusters E and H are related on the basis of. Pleurozium and Ledum.Cluster F is disiinguished by high concentrations of. Hylocomnium andPteri,di.um, while G is a mixture dominated by Pteri,di.um. Cluster Hlacks Tsuga and is characterized,by Pteridium, Pleurozium, and Ltdum,while E is dominated bv Gaultheria. Ledum. and Pleurozium. The den-

Tlnr,E 2. RaNr Ononn or Spncns DnrnnlrrNnn sy DrscnnlrNlxr Axitr,vsts.F-value indicates the relative information and degree of significance rvith 7 and 87degrees of freedom.

Number Species F-valuleP of Larger

F-value

I

23156789

101 1

l . )

l 4

Hylocomnium splendensRhynchospora albaKalmia occidentalisPleuroziurn schreberiCladonia rangiferinaGaultheria shallonLedum groenlandicumPteridium aquilinumSphagnum subsecundumEurhynchium praelongumMaianthemum dilatatumVaccinium oxycoccusCornus canadensisTsuga heterophylla

130.51 1 1 . 9o.) . /24 .7r3.57 2 . 7r 5 . 78 . 16 .94.42 .32 .3l . J

1 .0

0.0050.0050.0050.0050.0050.0050.0050.0050.0050.0050.050.05

n.s.

1e7 6l LEBEDNIK & DEL MORAL: VEGETATION 395

Tleln 3. F-vrrr,uns BETWEEN Prrrns or,Gnoups (see fig. 2). Degrees of free-dom: 14, 87. Values exceeding F * 2.28 occur between two sets of samples fromthe same group with P < 0.01.

46 .88561 .44863 .06160.62678 .2 8538 .05159.262

B6.663

10.43820.40.554.63 927.90725.696

Co . 5 / I

24.1637 2 .07 33 2 .86827 .9 t9

D17.4837 2 .59734.340z J . z J +

E49 .95529.83672.389

f

24.937 G55.365 79.449

drogram tends to confirm the impressions produced by the F-table. Here,A joins B, C, and D at a high level of information. B and C are closelyrelated and D is more closely related to these than to others. The affrnityof E and H is demonstrated as is that of F and G. E is intermediate be-tween H and D and closer to H by F-test as is confirmed by the dendro-gram.

Discriminant analysis includes a determination of canonical axes. Thefirst canonical axis accounts for 41.7 percent of the variation while thesecond accounts for 21.2 percent; both axes have high correlation withthe discriminant axes. Figure 3 is a two dimensional plot of the canoni-cal distribution of samples. This plot reconfirms our conclusions con-cerning the relationships among the eight groups. B, C, and D are shownto be closely related, E and H are relatively close, F and G are more dis-tinct and A is very distinct. A, F, and G are all loosely defined groups.Group A is in the sedge zone and F is in the forest zone, suggesting thatsome rotation of these axes represents an approximation of the originaltransect that is a gradient of nutrient availability and soil oxygen con-tent (Lebednik, unpublished).

Key to Vegetation Clusters. The discriminant analysis provides infor,mation to produce a key to the vegetation types recognized. Our key usesthose species highest in the rank order of information. In Table 4 we listthe composition of the eight groups according to the zonation perceivedsubjectively. The following key may be used to identify these types inthe field:

a. Rhynchospora alba dominant (Sedge Zone)aa. Rhyncospora alba rare or absent, Ericaceous

shrubs common

Type A

b. Kalmia occidentalis dominant (Shrub Zone)c. Pleuroziu?n schreberi, over 25% cover Type Dcc. Pleuryziunx schreberi rare or absent

d. Sphagnuno sabsecundum common Type Cdd. Sphagnum subsecundunx absent, Cladonia rangi,-

Jeri,na common Type B

Similarity Projection Ordination. The dendrogram suggests greatercontinuity in the data than is manifest by classification approaches. Simi-

u

& o G

G n- G q

u

F

-FrFfFF

396 MADRONO

bb. Kalwia occidentalis rare or absente. Pleurozi.um schreberi, over 50/a, or if less, then

Le dum gr o e nl andi culn common ( Transi tion zone )I. Pteridium aquili,nuno common .If.. Pteridium aquilinum rare

ee. Both Pleurozi,um schreberi. and Ledum grlenlandi.culn,rare (Forest Zone)

g. Gaultheria shallon and Hylocomnium splend.enscommon

gg. Pteridium aquilinum dominant, others mixed

[Vol. 23

Type HType E

Type FType G

19761 LEBED\IK I

gave similar results (meansported here are confined to l

Quanti.t ati,a e analy si s. P sused in our analyses. Orlocimplemented with a comptGr igal ( 1971 ) . The matr i rsponding subsets are unitedtion gain. The results are rmutual information at *'hicl

Once tentative groups :u€ferences and tests of groupmultiple discriminant anall-sin del Moral ( 1975), who folof data, SMDA determinesuseful to distinguish betrreefor differences between groqto determine the class to whiobjective, reallocation of theamong the groups are detendiscriminant axes. What rependent) axes in which rariare plotted according to throptimum two-dimensional p

We also used similarity p1972) to determine vegetatirextension of the polar ordinieffect an ordination, we selSeveral trials, based on theinternal association of 90 pvary from 80 to 100 percencluded that internal urssocialtions. All other samples areby the end stands. A secondfollowing criteria: stands fidissimilar to each other, antional dimensions did not in

RnsrQualit ati.a e D e s cription. )

eral survey was made, Fitzinto five zones. We added tltion into one of the six zonrprovides a qualitative list oclature of vascular plants fonomenclature follows LawtrHale (1969).

E

Lt r - q E c -

D- h^ t "t*. ' . . t

h'Hn' 'Bffii u'. *J T'F;"

n H Htr0X

ctr-JctrCJ

z.Iz.ctr.(-J

Gz.a(-JtrJa

Frc. 3.analvsis.

FIRSTCanonical distribution of

Letters refer to groups in

0 3 6 9CRNONICFL FXISthe eight groups determined by discriminant

FiEure 2.

r97 6l LEBEDNIK & DEL MORAL: VEGETATION

TlsrE 4. MllN Covnn Prncnwucn on Dourr.lNr Spncns lon rnr ErcsrRovrsnn Gnoups Innwru.mo rw MINFO.

ClusterTransect

MeanTaxon

397

Ledum groenlandicumKalmia occidentalis\iaccinium oxycoccusRhynchospora albaSphagnum subsecundumPleurozium schreberiCladonia rangilerinaTsuga heterophyllaGaultheria shallonPteridium aquilinum

var. pubescensHylocomnium splendensEurhynchium praelongumCornus canadensisMaianthemum dilatatum

L O 9 9 . 24.9 85.0

15 .2 5 .080.3 0.00.0 0.00.0 4.9

38.0 78.00.o 23.30.0 0.00.0 0.0

0.0 0.00.0 0.00.0 0.00.0 0.0

97 .6 99.290.3 84.665.6 50 .40.1 0.0

58.2 8.60.0 7 5.6

7 2 . 8 5 . 03 . 6 1 5 . 00 .0 3 .20.0 0.2

0.0 0.00.0 0.00.0 0.00.0 0.0

67.3 4 .26.2 0.90.0 0.00.0 0.02 .3 5 .0

78.8 2.50.8 0.00.8 0.9

5 .2 22 .8

41.8 53 .05.6 34.20.0 18.60.0 7.4

28.9 16.565.3 32 .80.0 13.30.0 5.8

49.5 30.677.6 23 .6

0 .8 16 .80 .0 2 .o0.3 7.40.0 0.3

4.00.00.00 .01 .0

14.00.0

I J . J

14.064.O

5.8 92 .5 22 .O0.0 0.3 21.53 .8 3.8 4.00.0 0.3 3.0

larity projection ordination was used to determine whether a continuumapproach recognizes the community types reported above.

The first axis is defined b5r samples 32 (a member of group C) and97 (group F) and the second axis is defined by samples 10 (group A)and 67 (group H). The first axis represents shifts within shrub dom-inated vegetation and the second in non-woody dominated vegetation.The resultant ordination is shown in Figure 4, in which previously recog-nized communities are circumscribed. The rate of change of species (betadiversity) along the first projection axis is relatively large so that thereis distortion of the type reported by Gauch (1973) in his anal5rsis ofBray-Curtis polar ordination, to which Similarity projection Ordinationis closely related. Samples toward the middle of the first projection axistend to be distant from it so that there is a ,,hump,, in the analysis. Thishump results because Percent Distance is not a linear estimator of eco-logical relationships between the samples.

Ordination suggests that, with the exception of type A, the Rhyncho-spora dominated community, the data are continuous. The Rhyncho-spora community is unique by virtue of the virtual absence of erectshrubs. Communities B, C, and D, which are closely related in the classi-fication hierarchy, are likewise clustered in the ordination and linked bycover of Ledum and Kalmi,a. Communities F (Gaultheria/Hylocom-nium) and G (P t eridium / H ylo c om.nium-Eur hync hium) ar e distinct ( ex-cept for sample 93) and can correctly, if arbitrarily, be separated fromeach other and from E (Gaultheria/Pleurozi,um) and H (pteridi,um-Gaultheria/Pleurozium). However, communities E and H. which are as

398 MADRONO

I6-- 65 56 59

45 Utro u,

qts t - E 6 i H

lVo ] .23 19761 LEBEDNIK I

Vegetation classifi cationtions" (Shimwell, 1973: \ lwill not be reviewed here. Trather than differential or ctive hierarchical method derof information gained whendendrogram. This method issifi cation methods (Williaml

All hierarchical classificattion: analysis proceeds untiThe investigator must intuilcease. Del Moral (1975) pstepwise discriminant analvmore groups are recogniz€thought to occur in the fieldifferences between groups,and indicates the relativeclassification.

Kings Lake Bog lies in thr(N47 035'�40", W l2lo 46'�35phylla zone in the Puget TrLake is 1.1 ha and is surrourmat edge (Wolcott, 1973).7in the 1930's by the Weyerlis located (Fitzgerald, 19661

The area we sampled ert(of the surrounding forest (:east and southeast lakesidejoins higher ground. Spiroetremaining sides of the bog rour sample was confined to I

Fitzgerald (1966) showecmore extreme than that of tlvaried annually from 43o (60 C in the forest. Diurnal rreradiation, was much great(in one 24 hr period in July,compared to 60 C in the fonsub-freezing temperature tlfluctuates such that duringabove the moss layer throulis very dry. By these climatirthe forest.

V e g et at i,on s ampli.n g. W e tline intercept transect from

74"6b831 1

5D

7770

El3A.#

7#57

76

zo

oUlo

zO6

59

#'

40 50 50 70FTRST PROJECTION RXIS

Frc. 4. Similarity projection ordination using samples 32 and,97 as end pointsior the first axis and samples 10 and 67 for the second. Samples representing thecight grcups revealed by MINFO (A-H, see fig. 2) are circumscribed.

related to each other as are F and G, are scrambled in the middle of a,xisone at the upper end of axis two. These samples occur in a transitionzone, show little tendency to cluster, are well separated from other sam-ples, and are intermixed. The ordination reflects the classification betterat the Iour group level than at the eight group level.

Although similarity projection and discriminant analysis are entirelydifferent analytical tools, they produce similar results (cf. figs. 3 and 4).All groups occupy similar relative positions though spacing is distorteddue to different mathematical assumptions. Community type G is plottedlower on the second canonical axis in relation to F than in similarityprojection. In both, the integradation of E and H is evident though it ismore pronounced in the similarity projection.

Comparison ol Classification and Ordination. The two approaches ofthis study reveal that while the vegetation can be treated as four vege-tation zones, the continuum approach suggests the gradient nature of thevegetation. The vegetation may be classified numerically by a method

40

1976) LEBEDNIK & DEL MORAL: VEGETATION 399

that stresses dominance types, and the resultant classes can be recognizedin an ordination. However, canonical analysis of these types reveals con-siderable variation within a type and continuity between seven of theeight types. Since we obLained this result with only 14 common species,we believe that this conclusion applies even more strongly to cases ofgreater species richness.

Fitzgerald (1966) studied this bog by means of relev6s in the style ofBraun-Blanqu6t. She discerned the four zones we have retained in ourdiscussions. Our methods suggest that greater complexity of communitypattern exists. They also recognize that the classes, with one exception,are arbitrary modes in an essentially continuous pattern.

CoNcr,usroNS AND SUMMARyThe vegetation of Kings Lake Bog has been described by classification

and ordination methods. The question posed by this study was whethervegetation forms distinct zones around this bog or whether it forms acontinuous pattern. Combined analyses suggest that vegetation forms agradient except for one distinct type, distinguished by Rhynchosporaalba and several other sedge species. This type occurs on the most ex-posed portion of the floating mat where woody ericacous shrubs have notinvaded. Zones may be recognized in the remaining portions of the gra-dient from lake to forest, but these zones reflect only the distribution ofdominant species. Our classification constructed more communities thanwere reco64nized by Fitzgerald (1966) and discriminant analysis suggeststhat these form a nearly continuous sequence.

The combined use of classification and ordination methods in thisstudy reveals a correlation between philosophically contrasting ap-proaches. The similarity between canonical distribution of samples andsimilarity projection analysis is striking when one considers the disparatemathematical approaches. For mapping and synthesis of vegetation data,we recommend numerical classification methods such as MINFO. Theuse of discriminant analysis to refine resultant classifications provides ameans for objective reallocation of samples and for assessing the validityof each group. Where the behavior of individual species is important,then ordination methods are superior. Ordinations may be divided intosegments for ease of consideration, but we propose that such partitioningbe done on the basis of numerical classification methods.

Acr<wowr,nocMENTSWe are indebted to A. R. Kruckeberg, M. F. Denton, and P. A. Lebed-

nik, who each made valuable contributions to the paper. Funds were pro-vided by the Department of Botany, University of Washingtorl, and bythe College of Arts and Sciences, University of Washington. The Weyer-haeuser Company permitted this study to be carried out on a portion oftheir Snoqualmie Tree Farm.

400 MADRONTO lvcl. 23

Lrrpnrrrunrr CrrnnBnlv, J. R. and J. T. Cuntq, 1957. An ordinat ion of the upland forest communi-

t ies of southern Wisconsin. Ecol . Monogr. 27:325-349.nnr. Mon.+l, R. 1975. Vegetation ciustering by means of Isodata: Revision by

multiple discriminant analysis. Vegetatio 29:779-190.Drxox, W. J. (ed.). 1970. BMD Biornedical computer programs. Univ. Calif.

Press, Berkeley.Frrzcnnaro, B. J. 1966. The microenvironment in a Pacific Northwest bog and its

implications for establishment of conifer seedlings. M.S. Thesis, Univ. Wash-ington, Seattle.

FnlNxr-rx, J. F. and C. T. Dvnxrss. 1973. Natural vegetation of Oregon andlVashington. Pac. N. W. For. Range Exp. Sta. Gen. Tech. Rep. pNW8,Portland, Ore.

GATTCH, H. G., Jn. 1973. The relationship between sample similarity and ecologicaldistance. Ecology 54:678-622.

and R. H. Wnrrr.lrnp.. 1972. Comparison of ordination techniques.Ecology 53:868-875.

Gor,lsrnrN, R. A. and D. F. Gnrcal. 1971. Computer programs for the ordinationand classification of ecosystems. ORNL-IBP-71-10, Oak Ridse NationalLabo ra to r y .

Her.E, M. E. 1969. How to know the l ichens. Wm. C. Brown Co., Dubuque, Iowa.HrrcHcocx, C. L. and A. CnoNgursr. 1973. Flora of the pacific Northwest. Univ.

Washington Press, Seattle.Lawron, B. 1971. Moss flora of the Pacific Northwest. The Hattori Botanical

Laboratory, Miyasaki, Japan.Munrr,en-Dorrnors, D. and H. Elr,r,Nsrnc. 1974. Aims and methods of veeetation

ecology. John Wiley & Sons, New York.onrocr, L. 1969. Information analysis of structure in biological collections. Nature

2 2 3 : 2 2 8 .Oswar.o, H. 1933. Vegetation of the Pacific Coast bogs of North America. Acta

Phytogeogr. Suec. V.surrrwn'r., D. 1971. The description and classification of vegetation. univ. wash-

ington Press, Seatt le.Tunnnsor, G. 1916. Lysichiton camtschatense (L.) schott, and its behavior in

sphagnum bogs. Amer. J. Bot . 3:189-209.wrr.rrarrs, w. T. r971. Principles of clustering. Ann. Rev. Ecol. system. 2:3o3-326.Wolcorr, E. E. 1973. Lakes of Washington I. Water supply Bull. #14, Dept.

Ecology, Olympia, Washington.