DISCRIMINANT ANALYSIS OF COVE

FORESTSOF THE CUMBERLANDPLATEAU

OF TENNESSEE

Paul A. Schmalzer and C. Ross HinkleGraduate Program in Ecology

and

H. R. DeSelmBotany Department and Graduate Program in Ecology

408 lOth StreetThe University of TennesseeKnoxville, Tennessee 37916

ABSTRACT

Coves and gorges of the Cumberland Plateau in Tennessee areoccupied by a complex array of forest communities. Vegetationand environmental data from 178 .04 ha plots in five cove or gorgesystems were analyzed in this study. Numerical classificationanalysis was used to establish fourteen plant communities.Multiple discriminant analysis using species importance valuesindicated that these types were distinct at the .05 significancelevel. Discriminant analysis using selected environmental vari-ables indicated that several of the type were not distinct fromeach other on this basis. This analysis indicated that factorsrelated to slope position which may indicate a soil moisture axiswere important in segregating the communities as were factors suchas pH which are related to soil nutrient status.

63

INTRODUCTION

Coves and gorges of the Cumberland Plateau in Tennessee areoccupied by a complex mosaic of plant communities. Many of thesecommunities were termed "mixed mesophytic" by Braun (1950) anddiffer considerably from the oak dominated forests of the uplandsof the Cumberland Plateau.

Understanding of the vegetation-environment relationships inthese coves is complicated by the species richness of these areasand the complex environmental gradients which occur there.Discriminant analysis has been used in this study as a means ofexamining the distribution of community types along theseenvirnmental gradients.

Management of cove areas that are held in the public domainas state and national parks, recreation areas_ natural areas, andwildlife management areas as well as those that are used for timberproduction also require a better understandinq of the naturalvegetation of coves and gorges of the Cumberland Plateau.

THE STUDY AREA

The Cumberland Plateau is one of the major physiographicregions of Tennessee, extending from 84o30 ' to 85o10' longitude atthe Tennessee-Kentucky border and from 85o20 ' to 86o10' longitudeat the southern border of the state. It is bordered by the GreatValley to the east, the Cumberland Mountains to the northeast,and the eastern Highland Rim to the west. Part or all of 21 countiesare located on the Plateau.

The climate of the Plateau is considered humid and mesothermal.Precipitation averages from 127 cm to 140 cm per year and thus ishigher than that of the Highland Rim to the west and the Great Valleyto the east of the Plateau (Dickson 1960).

Physiographically, the Cumberland Plateau is the southernextension of the Appalachian Plateau from which it is somewhat

arbitrari]y designated. In Tennessee, the Plateau is underlain byresistant Pennsylvanian sandstones of the Pottsville series whichare responsible for maintaing the "submaturely dissected" flat torolling surface of the Plateau. Where these sandstones have beenbreached, erosion of the less resistant Mississippian limestoneshave produced steep-sided gorges (Fenneman 1938). The easternescarpment of the Plateau is a nearly linear ridge extending fromHarriman southwest to Chattanooga and about 300 m higher than theGreat Va]ley. The western escarDment, in contrast, is highlydissected by streams flowing west onto the Eastern Highland Rim.

Soils of the coves and gorges of the Plateau are complex andhave not been mapped in detail° The soils of the upper slopes ofthese coves are generally derived from sandstone colluvium overshale or sandstone bedrock_ are sandy in texture, and are acid.Soils of the middle and lower slopes below the level of theMississippian strata are derived from sandstone and limestonecolluvium and limestone residuum, are less acid and have a greaterclayfractiono

Braun (1950) included the Cumberland Plateau of Tennesseewithin the Mixed Mesophytic Forest Region because of the presenceof mixed mesophytic forest communities in Plateau coves and gorges_She recognized, however that "oak or oak-hickory forest prevailsover the large area of submaturely dissected Plateau in Tennessee..."(Braun 1950_ p_ 114).

Caplenor (1965) described hemlock, mixed mesophytic, oak-hickoryand chestnut oak communities in the gorges of Fall Creek Falls State

Park (B]edsoe and Van Buren Counties)° Safley (1970) described avariety of communities in the gorges of the Big South Fork of the iCumberland River. These included types dominated by tulip, hemlock,northern red oak, white oak, chestnut oak, shortleaf and Virginiapines_ Floristic studies have been made by Sherman (1958) and Clark i(1966) in Plateau coves° Recent vegetation studies have been

completed by Ninkle (1978) and Schmalzer (1978).

METHODS i!

Old growth, uneven_age forests in coves and gorges of the i!Cumberland Plateau of Tennessee were selected for sampling. Plots _.from five such coves are included in the present study. These are iDick Cove near Sewanee, Franklin County, Doe Creek Cove,White County, a section of the Obed River Valley, Cumberland County,Flint Fork Cove in Pickett State Forest, Pickett County, and LittlePiney Creek gorge, Rhea County. Circular plots of 0.04 ha weredistributed in a stratified random manner by slope position. Canopyspecies, greater than 12o5 cm dbh, were recorded in 5 cm diameterclasses by taxon. Saplings, 2.5 cm to 12_5 cm dbh were recorded bytaxon. Nomenclature follows Radford et al. (1968).

A soil pit was dug near the center of each plot and samples ofthe A and B horizons were collected. Thickness of each horizon,depth to bedrock (to 91 cm), and stone content of each horizon were 'irecorded. !i

Slope angle was measured in four directions from plot center;aspect was also measured. Evidence of past disturbance such as firescars or cut stumps was noted.

65 iiii;i¸

The pH of air dry soil sampleswas determined in a 1"I soil"water slurry (Jackson, 1968)o Soil texture standards were determined

using a modified Day (1956) hydrometer method. Soil texture of theother samples were determined by the "feel" method (Soil SurveyStaff 1951)o Water holding capacity' of the soil was calculatedusing the methods of Longwe11 et a]o (1963)o Topographic quadranglemaps were used to determine distance from the top and bottom of theslope, gorge width, and vertical distance of the plot below the

surface of the Plateau. _

Absolute and relative densities and basal areas for the canopy ispecies in each plot were calculated° An importance value wascalculated as the sum of relative density and relative basal area

for each species (IV = RD + RBA = 200). i!;

Cluster analys'is (Orloci 1967) was used to place plots intogroups (community types) of similar species composition°Discriminant analysis (Nie et at. 19759 Dixon 1975) was used to iexamine the separation of types using vegetational and environmentalcharacteristics.

THEORY AND APPL!CATION OF DISCRIMINANT ANALYSIS

Discriminant analysis is a multivariate statistical proceduredesigned to distinguish between two or more groups Of observations !on the basis of a set of discriminanting variables. The analysisderives linear combinations of the discriminating variables whichprovide maximum separation between the groups (Pielou i977) o Theselinear combinations are termed discriminant functions and have the

form D1 = diIZ 1 + d12Z2 ÷ ° o ° + d.lnnZ i

where Di is the score on discriminant function i o the d's areweighting coefficients, and the Z's are the standardized values ofthe discriminating variables (Nie et al. 1975)o

The first discriminant funciton is inclined such that it

represents the dimension along which the maximum group differentation !;occurs The second function is idependent of the first and isinclined along the largest group differences not represented by thefirst function (Tatsuoka !971).

Two quantities measure the relative importance of a givendiscriminant function. The canonical correlation coefficient is

a measure of thefunctibns ability to distinguish between the groups.The second measure is a relative percentage computed by dividingthe eigenvalue of a given function by the sum of the eigenvaluesof the functions derived;, this sum is a measure of the total varianceof the discriminating variables. The ration of the eigenvalue ofone function to the total is the a measure of how much of thistotal variance is explained by that function (Nie et al., 1975).

Discriminant analysis may be used for two basic purposes,analysis and classification. One analytical application is theinterpretation of the discriminant function coefficients. Ignoringsign_ the magnitude of these standardized discriminant coefficientsrepresents the relative contribution of that variable to thediscriminant function° Considering the sign, if the coefficient

_ is negative then increases in the variable Zi will decreasevalue of the discriminant score Di and conversely for positivecoefficients (Nie et al. 1975). If the discriminant function isplotted as an axis, then a negative coefficient of a variable willbe associated with a movement in the negative direction along thisaxis and a positive coefficient with movement in the positivedirection°

Plots of the group centroids (means) along the discriminantfunctions or axes allow a ready comparison of the degree ofseparation of these groups along the plotted functions. Significancetests(assuming multivariate normality) may be used to examine thetotal separation of the groups in multivariate space (Nie et al.1975)o

Classification is also a valuable feature of discriminantanalysis. Once the discriminant functions have been derived,classification functions for each group may be found. Theseclassification functions are then employed to predict the probabilityof group membership either for cases not included in the originalanalysis or for the original cases. The second option may be usedas a check on the effectiveness of the classification procedure.If much misclassification occurs, this may be an indication ofpoor discrimination by the original variables (Nie et al. 1975).

Bias is introduced into the classification analysis becauseplots are placed into groups, functions are derived based on thesegroups, and the functions are used to classify the same plots.The "jacknife" procedure is intended to reduce this bias in thattwo sets of classification functions are derived. The first by thestandard procedure while in the second, functions are adjusted sothat the probability of a plot belonging to a group is calculatedexcluding that particular plot. This bias is of particular concernif the classification success of the standard procedure is nothigh (Dixon 1975).

67!ii!

Discriminant analysis has been used to define nicherelationships in various animals (Anderson and Shugart 1974) and r rubr_recently in sympatric herb species (Mann 1977). Norris and Barkham A° sacchal(1970) examined the environmental relationships of the ground flora cu]usoof English beechwoods using discriminant analysis. Kessel and ._tu]a letWhittaker (1976) found discriminant analysis to be generally . ]uteaunsatisfactory as an ordination technique but useful in . nigraclassification. Kercher and Goldstein (1977) determined, using Carp_mus _discriminant analysis, that "core groups" (community types) defined Carya cor_in the vegetation of the Walker Branch watershed, Oak Ridge, C. glabraTennessee were well classified by environmental variables and Co ovalisthus appeared to be environmentally distinct, ii'__C. ovata

iI C. foment(

Hinkle (1978) used discriminant analysis to test the vegetational _iCarya spoand environmental distinctness of upland and ravine communities of _i_ Cercis calthe Cun_erland Plateau in Tennessee and also in identifying which _ C]adrasti_-

_Comus f_(environmental variables were most important in producing thisseparation This technique has also been applied to individual !i Fagus gra_

" _:!Fraxinuscoves (Hinkle et al. 1978, Schmalzer 1978). iiI I]ex opac_

A plant com_nunity, defined by a number of plots for which ci!Jug]ans nimeasures of species importance are available may be conceptualized _Ka]mia lalas existing in a vegetation hyperspace in which the importance of !! L_quldamb_n species is represented by n axes (Goodall 1963). Each plot is a i Liriodend_point in the space defined by coordinants of species importance Magnoliavalues. Community types may be considered clusters of these points, i _. macrop_The relative degree of separation between these types should be Nyssa syl_revealed by discriminant analysis ,iOstrya vil

" i:I(7xyd_endrulSimilarly,an environmentalhyperspacemay be definedwhere i_Pinus str(

the z axes are the z environmental variables measured for the plots. !i_P, virginiThe degree of separation between the communities based on environmental i P]atanus (variables may also be revealed by discriminant analysis. _iPrunus se_

ii Quercus al

RESULTS iliO.coccin__:Q. muehler:.iQ, prinus

Classification i Qo rubra

Fourteen community types (Table I) were derived from the cluster i Q. ve]utirRhododendranalysis of 178 plots (Figure 1). These were distinct at the level _ Robinia pswhere 35% of the variance is within-group and 65% is between group. _ Sa#safrasEnvironmental characteristics of these types were also summarized(Table 2).. The types noted here are described more fully in ! Ti]ia an_rT. heteroGSchmalzer (1978)and are similar to those described by Hinkle (1978). i Tsuga cana

i Ulmus amer

:i',ii_)

68

,i_iiiiii!!iiiiii!!i!iil!! !i _TAB mE I IMPORTAN mE VALUES OF CANOPY S PEel iES IN THE COVE COM_U_ I TI ES _!!i! _ i_]

<ii<ii_!i i_ili!iill

CO CO_WO WP-CO NRO WO-NRO TUL SM-NRO

r rubrum 5,1 5,6 14,7 2.8 6,4 1.4o saccharum I, 3 0,4 i, I 17,0 7,7 6.9 44. I

....Aescu_uso_tandra io5 O,8 5.4_etula ]enta 0o5 9,2 1.2

]ute,a I.6nigra 6.3

rp_nus carolinianaCarya cordiformis 5.2 2.9 4.7Coglabra 849 7,4 3,0 2.6 15.0 8.3 6.0Coova]is 2.4 1.0 1.2 8.4C. ovata 3.8 5.5 3.7 26.6 9.6 4.6 10.8

Co t o_mntosa 4.6 18,7 5.8 10,9 7.3 6.2_l /a Spo 0,5 1.6

Cercis canadensis 1.2 0.6]adrastis ]urea 1'7_ florida 2.1 3.7 2.4 1.1 0.5 1 1

Fa_us grandifolia 2.9 4.2 7.7 2.0Fraxinus a_mricana 1.3 1.6 1.5 2.9 4.6 12.6

]ex opacaug]ansnigra 3.6 0.2 1 2 2 2

mia ]atifo]iaiquidambar styraci f]uai riodendron tulipi fera 5.4 9.6 11.6 2.1 17.4 80.8 19.3gno]ia ac_inata 0,9 1.1 4.1 4.5 I. 7 1.0m_crophy]]a 8.1 1.0sa sylvatica I0.7 3.5 4. I 3.0 4.2 0.6 14.6

virginiaria O.3endrumarboreum 6.1 4.6 8.2 1.2 2.5 4.9 1.1

inusstrobus 1.3 46.3 0.9 9.1virginiana 2.3 3.1 0.7 0.8

P]atanus occidenta]is

r_us serotina 1.4 0,2 1.6Q uercusalba 0.6 32.8 4.3 4.7 69,4 2.5 2.8

. coccinea 8.2 2,7 2.4 0.7mueh]enbergii 7.5prinus 120.9 56.6 42.4 14.2 8.4 6.8 6.5rubra 8.3 22.2 3.5 93.3 20.9 12.9 32.9velutina 3.8 14.2 2.7 1.7

Jendron maximum

inia pseudoacacia 0.8 0.4 0.4 1.8 3,2_%safras albidum 1.4 1.1 0.3 0.6 0.9ilia americana 8,3 8.3o heterophylla 0.8 11.1 1,8ugacanadensis 28.1 1.2 6.2

s americana 0.4 2.2

1 CO-chestnut oak, CO-WO-chestnut oak-white oak, WP-CO-white pine-chestnutoak, NRO-northern red oak, WO-NRO-white oak-northern red oak, TUL-tulip, SM-NRO-sugar maple-northern red oak

i

7O

FIGUREI. Dendrogram of cluster analysis of cove vegetation.Communities are defined at the 35% dispersion level, I =white oak-northern red oak, 2= sugar maple-shagbarkhickory-white oak, 3= sugar maple-northern red. oak, 4=white basswood-sugar maple-buckeye, 5= white ash-sugarmaple, 6= northern red oak, 7= beech-tulip, 8= tulip,9= hemlock-tulip, 10= river birch, II= beech, 12=chestnut oak-white oak, 13= white pine-chestnut oak,14= chestnut oak.

1TA

BLE

2S

ELE

CTE

DE

NV

IRO

NM

EN

TAL

CH

AR

AC

TER

ISTI

CS

OF

THE

CO

VE

CO

MM

UN

ITY

TYP

ES

CO

-W

P-

WO

-S

M-

SM

-W

A-

BA

S-

HE

M-

BE

-V

ari

able

CO

WO

CO

NR

ON

RO

TU

LN

RO

SH

AG

SM

SM

TU

LT

UL

BE

RB

Slo

peP

osi

tion

2U

UU

,MU

,ML,

ML,

FM

,LM

,UM

M,L

GL,

ML,

MF

Dis

tanc

efr

omth

eto

po

fth

e

slo

p(m

)72

.491

.779

.114

0.9

210.

926

0.1

220.

318

4.9

164.

222

i.220

8.7

398.

736

6.3

292.

8Tr

ans

f_rm

eda

sp

ectJ

0.9

11

.03

1.0

51

.33

1.0

61

.22

1.1

01

.35

1.5

51

.42

0.5

50

.98

0.7

00

.0S

]ope

ang]

e

(5)

33

.04

1.3

39

.42

3.4

31

.63

0.9

36

.22

4.3

32

.23

8.5

22

.82

7.9

25

.62

.6S

and

inth

e

A(%

)43.3

37.9

41.1

39.3

37.8

42.9

35.4

27.9

21.8

39.6

43.8

37.6

46.3

34.6

San

din

the

B(%

)4

1.8

35

.64

1.0

31

.33

4.5

43

.83

5.9

25

.71

5.5

35

.25

0.1

38

.84

4.2

37

.8°

Cla

yin

the

A12.4

17,4

10.3

16

.41

7.0

13

.3t5

.31

8.8

19.8

12.9

9.2

12,2

10.3

15.4

Cla

yin

the

B(%

)14.9

22.9

12.6

25.7

22.9

16.9

20.4

26.6

33.0

20.1

10.4

15.3

15.8

17.4

pH

of

the

A4

.74

.74

.45

.34

.85

.15

.75

.55

.95

.54

.65

.24

.84

.1

pH

of

the

B4

.84

.64

.64

.94

.85

.15

.55

_I

5.6

5.4

4.7

5.0

4.9

4.2

Tota

lav

ail-

ab

lew

ate

r7

.13

7.1

28

.26

9.0

19

.10

8.8

59

.00

6.3

27

.47

8.9

31

0o

96

9.0

11

0.5

12

0°8

4

Bed

rock

typ

ess,

ss,

sh

,sh

,ss,

ss,

Is,

I_,

Is,

Isss,

ss,

Is,

ss

sh

sh

ss

Issh

Isss

sh

sh

sh

Isss

1/V

alue

sw

ithin

the

tabl

ear

em

eans

unle

ssot

herw

ise

spec

ified

.2/

U-u

pper

,M

-mid

dle,

L-l

ower

,G

-gor

ge,

F-fl

oodp

lain

.3/

Tra

nsfo

rmat

ion

isaf

ter

Bee

rset

al.

(196

6).

4/ss

-san

dsto

ne,

sh_s

hale

,!s

-lim

esto

ne.

72

The chestnut oak type was common on upper slop positions.Chestnut oak was reproducing well based on presence in the saplinglayer and the type is apparently time stable° Chestnut oak

communities have previously been repomted from Plateau coves(Caplenor 1965) as well as from the Cumberland Mountains (Braun19509 Hinkle 1975) and in the Great Valley (Martin 1971).

The chestnut oak-white oak type occurred in similar sitesto the previous type. The major canopy species were reproducingand the type is apparently stable° Safley (1970) reproted a whiteoak_chestnut oak type from the Big South Fork area similar to thistype.

The white pine-chestnut oak type occurred primarily in onegorge (Little Piney Creek, Rhea County). The major canopy specieswere reproducing but the successional status of this community isuncertain. White pine is a constituent of certain upland forestsof the Cumberland Plateau (Smith 1977), but it is not common inPlateau gorges.

The northern red oak type is apparently successional withlittle reproduction by northern red oak although sugar maplereproduction was heavy. Cabrera (1969) reported a similar typefrom the Cumberland Mountains while Martin (1966) and Hinkle (1975)reported less mesic northern red oak types also from the CumberlandMountains.

The white oak-northern red oak type was widespread; white oakwas reproducing well in this type but northern red oak was not.White oak communities have previously been reported from Plateaucovers (Caplenor 1965, Safley 1970). White oak is important butwith different associate species on the Plateau uplands (Hinkle1978) and also dominates many of the forests of the adjacent GreatValley (Martin 1971).

The tulip type was apparently successional. The importance oftulip in certain secondary Plateau cove forests has previouslybeen noted by Braun (1950), Caplenor (1965), and Carpenter (1976).

The sugar maple-northern red oak type occurred on the westernescarpment of the Cumberland Plateau. Northern red oak and tulipwere not reproducing well in this type. The type is somewhatsimilar to conTnunities described by McCarthy (1976) from FentressCounty and by Cabrera (1969) in the Cumberland Mountains.

74

Discriminant Analysis

Discriminant analysis of the types using species importancevalues indicated that all l the types were different at the o05significance lever (Table 3). Classification success for thesetypes was also high (Table 4)°

Discriminant analysis using selected environmental variablesindicated environmental similarity between a number of the types(Table 5)° The chestnut oak_ white pine-chestnut oak_ hemlocktulip_ and river birch types were distinct environmentally fromall others at the 0°05 significance level° A number of the othertypes were not significantly different suggesting environmentaloverlap° The,!white oak-northern red oak type overlapped with thechestnut oakowhite oak, northern red oak_ and tulip types. Thenorthern red oak type also overlapped with the sugar maple-shagbarkhickory and the white ash-sugar maple types. Some overlap alsooccurred a_mng the sugar maple types. The white ash-sugar mapletype was not environmentally distinct from the other sugar mapletypes° The sugar maple-northern red oak type also overlappedsoiTmwhat with the sugar maple-shagbark hickory and the whitebasswood-sugar maple types. The two beech types also wereenvironmentally similar.

Examination o{ the classification matrix (Table 6) indicatedthat the,_overall classification success was low (37.6%). Certain

types including river birch, hemlock-tulip, beech, and chestnutoak were better classified than this, while the tulip and beechtulip types were poorly classified°

The coefficients of the discriminant functions (Table 7) andthe plot of the centroids of the groups along the first twodiscriminant functions (Figure 2) may be interpreted together inconsidering the environ_ntal separation of these groups. On thefirst discriminant function, which accounts for 28.5% of the variance,distance from the head of the gorge_ gorge width, slope position,and distance from the top of the slope were most important. The signsof these coefficients were all positive and increases in theassociated variables are associated with types to the right on thefirst discriminant function. These factors are correlated to somedegree and in general related to lower slope position andincreasingly mesic conditions; this is_however a complex axis inwhich other environmental factors also enter.

!i!!TABLE 3. MATRIX OF CLASSIFICATION SUCCESSFROMDISCRIMINATANALYSIS OF COMMUNITY TYPES USING SPECIES IMPORTANCE VALUES1

ACTUAL CO- WP- WO- SM- WA- BAS- HEM-BEE-TYPE CO WO CO NRO NROTUL NRO SHAG SM SM TUL TUL BEE RB •

CO 100.0

CO-WO 100.0

WP-CO 100.0

NRO 100.0

WO-NRO 3.6 92.9 3.6>

TUL 100.0 =

SM-NRO 91.7 8.3 _-

SM-SHAG 100.0

WA-SM I00.0

BAS-SM 7.7 923 c

HEM-TUL 100.0

BE-TUL 100.0

I

BE 100.0 u

RB I00,_ cF-

D

1Values within the table are percentages of the plots of the community types.

TAB

LE4.

MA

TRIX

OF

F-V

ALU

ES

FRO

MD

ISC

R!M

iNA

NT

AN

ALY

SIS

OF

CO

MI#

JNIT

YTY

PE

SU

SIN

GS

PE

CIE

SIM

PO

RTA

NC

EV

ALU

ES

i

COMMUNITY

CO-

WP-

WO-

SM-

SM-

WA-

BAS-

HEM-

BE-_

TYPE

CO

WO

CO

NRO

NRO

TUL

NRO

SHAG

SM

SM

TUL

TUL

BE

RB

CO

6.5

320.4

313.9

427,1

719.9

618o93

27°4

823°6

325o13

40°3

322°6

447°6

450°5

3

CO

-WO

15.3

16.

017.

0810

,16

9o41

13.6

817

o45

13o8

227

,56

11o1

330

o17

42,4

7

WP

-CO

12.5

818

.58

13.2

616

.38

19.7

218

o94

17o8

216

.53

15o7

625

°33

36,5

8

NR

O5.

827.

973.

905.

8611

.16

7°46

17,8

66.

4516

o01

31,4

1

WO

-NR

O9,

659.

8310

.93

i7.6

713

,97

27.2

59°

0830

.21

44°9

4

TU

L8,

0014

.68

15.0

09,

8418

o69

7o25

25,0

535

°20

SM

-NR

O4.1

98,4

47.3

221,7

68,5

828°6

541o15

SM

-SH

AG

10,8

79.

4825

.87

11.1

331

.96

44,2

6

WA

-SM

10.6

922,8

714o43

27.2

137,5

1

BA

S-S

M19.9

810.8

528,O

039.8

9

HE

M-T

UL

21

.33

35

.24

41

o79

BE-T

UL

9,1

13

0°1

5

BE

37.9

7

1The

over

all

Fof

14.0

3(d

.f.40

3,16

56)

issi

gnifi

cant

at_

=.0

01;

for

valu

esw

ithin

the

tabl

e,th

eF

-val

ueat

m=.

O01

(d.f.

31,1

34)

is2.

76

TAB

LE5.

MA

TRIX

OF

F-V

ALU

ES

BE

TWE

EN

CO

VE

CO

MM

UN

ITY

TYP

ES

FRO

MD

ISC

RIM

INAN

TAN

ALYS

lSSE

LEC

TED

ENVI

RO

NM

ENTA

LVA

RIA

BLES

1

CO

MM

UN

ITY

CO

-W

P-

WO

-S

M-

SM

-W

A-

BA

S-

HE

M-

BE

-T

YP

ES

CO

WO

CO

NR

ON

RO

TU

LN

RO

SH

AG

SM

SM

TU

LT

UL

BE

ER

B_.

....

....

....

....

._

L_

....

....,

m

CO

3.4

44.4

01.9

04.9

63.7

54.2

93.7

24.1

54.8

17.4

04.6

55.5

96.7

7

CO

-WO

2.3

32.2

81.3

9"

2.5

03.4

83.6

43.5

65.7

07.1

74.4

97.3

17.3

0

WP

-CO

2.4

52.5

83.2

44.2

05.1

84.2

94.6

62.4

75.5

25.8

96.4

2

NR

O1

.55

"2

.13

1.7

41

.26

"1

.51

"1

.85

4.4

42

.14

2.2

54

.91

WO

-NR

O1.6

1"

2.4

73.3

72.8

94.3

26.6

12.9

65.1

96.1

1

TU

L1.2

6"

2.5

92.1

91.7

94,7

0I.

62*

3.4

24.4

2

SM

-NR

O1

.54

*O

.98

*I.

59

*5

.39

i.2

6*

2.6

95

.39

SM

-SH

AG

1.1

3"

3.3

47.2

61.8

14.4

86.4

0

WA

-SM

1.3

9*

5.3

22

.16

3.2

45

.33

BA

S-S

M5.4

02.6

53.3

44.9

8

HE

M-T

UL

6.3

56.5

34o8

7

BE

-TU

L1.4

8*

3o8

3

BE

3o9

0_

__

,...

....

....

.,,

I/F(1

5,12

0)at

_of

.00i

=2.

51,

at_=

.05

=1.

67,

at_=

.!0=

1.49

.

*no

tsi

gnifi

cant

at_

=.0

5.

TA

BLE

6oC

LAS

SIF

ICA

TIO

NN

LAT

R!X

FR

OM

DIS

CR

IMIN

AN

TAN

ALY

SIS

OF

TAB

LE6o

CLA

SS

IFIC

ATI

ON

MA

TRIX

FRO

MD

ISC

RIM

INA

NT

AN

ALY

SIS

OF

CO

VE

CO

MM

UN

ITY

TYP

ES

US

ING

SE

LEC

TED

EN

VIR

ON

ME

NT

VA

RIA

BLE

S19

2

AC

TU

AL

CO-

WPo

WO-

SM

-S

M-W

A_B

AS-

HE

M-B

ET

YP

EC

OW

OC

ON

RO

NR

OT

UL

NR

OS

HA

GS

MS

MT

UL

TU

LB

ER

B

CO

55.6

16.7

5.6

11.0

5.6

5°6

CO

-QO

18.8

50.0

12.5

12.5

6.3

WP

-CO

12.5

12.5

37.5

12,5

25,0

_N

RO

14.2

28.6

!4.3

14,3

14.3

14,3

"_W

O-N

RO

7.1

21.4

7.1

25.0

10.7

3.6

10.7

7.1

3°6

3.6

TU

L1

6.7

25

.08

.32

5.0

8.3

8,3

8.3

0o

SM

-NR

O8.

38.

316

.725

.016

.78.

38_

3

SM

-SH

AG

20.0

6.7

6,7

26.7

13.3

13.3

WA

-SM

16.7

16,

750

.016

.7

BA

S-S

M7.

77.

77,

723

.130

.815

.47.

7

HE

M-T

UL

11

,11

1.1

66

.71

1_

1

BE

-TU

L16.7

8.3

8.3

8.3

8°3

33.3

16.7

BE

17

.65

.95

8.8

17

.6

RB

10

0.0

i/Val

ues

with

inta

ble

are

perc

enta

ges

ofth

epl

ots

ofa

type

.2/

Ove

rall

clas

sific

atio

nsu

cces

sw

as37

.6%

.

79

Table 7. Standaradized Coefficients of Environmental Variables on FirstTwo Discriminant Functions from Analysis of Cove Community Types

VARIABLE DF-1 DF-2

Transformedaspect -0.375 -0.050

Distant from the top of the slope 0.394 -0.092

Slopeposition 0.472 0.696

Slopeangle -0.156 0.109

Soil depth 0,122 -0.392

Bedrocktype -0.071 O.328

0 horizonthickness -0.056 0.255

A horizonthickness O.104 0.259

Stonein the A horizon 0.227 0.317

Sandin the A horizon 0.361 0 315 s

Sandin the B horizon -0.021 0.328

pHof the A (in water) 0.369 -0.518

pHof the B (in water) -0.138 0.120

Gorgewidth O.467 -0.307

Distance from the headof the gorge 0.566 0.158

Total available water 0.351 0.561

CANCORR O.774 O.743

PERCENT %8.5 23.5

8O

• HEM-TUL

_-,1WP-CO

_Eg _'UL_; _ BAS-SMWO-NRO

i 0

,l _o-wo _o _o _U-NROBE-rULSM-SHAG WA'SM 28.5%

-- d'--'--_-T_ -"- -- ' -_ 70 -0.20 O. 30 CJ. _}0 I, 30 _. 80 ,.. 30 2.80

'-z.zo -l. To -_.2o . DFI

FIGURE 2° Centroids of cove community types along the firsttwo discriminant functions from the analysis usingselected environmental variables.

81

The second axis (23.5% of the variance) is most highly relatedto slope position, pH of the A horizon, and total water. Gorge andfloodplain sites are at one end of this axis while cove slopes areat the other. The sign of the coefficient of pH of the A horizonis negative, indicating that increases in pH are associated withthe negative direction along this axis. This is confirmed by thecommunities on the negative end of this axis, white ash-sugar maple,sugar maple-shagbark hickory, and the others having higher pHvalues than the hemlock-tulip and river birch types on the positiveend of this function. Total available water also increases alongthis axis; the clayey soils of the sugar maple types have lowervalues for this variable than do the other types.

DISCUSSIONANDCONCLUSIONS

The community types defined by cluster analysis were foundto be vegetationally distinct by discriminant analysis° Thus i_appears that cluster analysis groups plots that are most similarin vegetational composition. Whether these types representrelatively discrete conTnunities or clusters within an overallcontinuum is not necessarily resolved by this analysis. They canhowever be successfully defined and classified.

Coefficients of the discriminant functions indicated thatslope position factors which are indicative of soil moistureconditions and soil pH which probably represents available soilnutrients are important in segregating the types. Hinkle (1978)showed that the amount of potassium in the A horizon and soil pHwere important in discriminanting between community types. Thusthe types appear to be distributed along a complex soil moisturenutrientaxis.

These community types were not so well separated byenvironmental variables. That separation could have a number ofcauses. The environmental variables used in this analysis mightnot be those most responsible for the separation of the types.Soil moisture and temperature, regimes were not directly measuredin this study but rather were inferred from more readily measuredparameters. BiologiQal interactions such as competition oralleopathy might be responsible for the separation of the typesbut there is no direct evidence to support this. Finally, it maybe that historical factors, past disturbance by logging, fire,and the removal of chestnut and the recovery ,from that disturbance,have resulted in different communities occupying similarenviroments. This seems to be a likely explanation for at leastsome of the lack of environmental discretenss between the types.

82

Several of the types which are not well distinguishedenvironmentally appear to be successional° The tulip and northernred oak types are in this category. Nithin the several types inwhich sugar maple is important, sugar maple-northern red oak,sugar maple-shagbark hickory-white oak, white ash_sugar maple,and white basswood-sugar maple-buckeye_ a considerable degree ofenvironmental overlap existed. These types generally occurredon middle to lower slopes of coves on the western escarpment ofthe Cumberland Plateau, often on northern aspects. Successionalrelationships bei_ween these types are unclear, but only thewhite basswood_sugar maple-northern red oak type appears to bein compositional equilibrium.

The beech and beech-tulip types are environmentally similarto each other and may be successional ly related. They are distinct,however from the sugar maple dominated types.

Some community types were quite distinct environmentally.The hemlock-tulip type occurred in.distinct gorge sites. Thefloodplain sites of the river birch type were environmentallydistinct although this type is apparently successional. Thechestnut oak and white pine-chestnut oak types were distinct. Thechestnut oak-white oak overlapped with the white oak-northern redoak type but was otherwise distinct.

Analysis of individual coves (Schmalzer 1978) generallysupported these results. At the Obed River, Flint Fork, and Little _,Piney Creek sites factors relating to slope position were mostimportant in the analysis and the distribution of communities inthese sites appears to be related to the topographic moisturegradient. At these sites geology and soil parent material wererather uniform and the depth of colluvial material obscuredunderlying bedrock differences, such that major differences insoil texture and pH did not occur. At the Dick Cove and Doe Creeksites bedrock differences between the sandstone and shale of

upper slopes and the limestone on middle and lower slopes wereclearly reflected in soil texture and pH. The distribution ofthe community types here must be considered to be along a complexaxis which includes topographic moisture status, and soil pH.landtextural differences.

83 _i

S

These results are generally consistent with those of othertypes of gradient analysis of vegetation. Whittaker (1956) andGolden (1974) attributed the distribution of vegetation in the iGreat Smoky Mountains to topographic position and elevation onrelatively uniform parent material. Gradients of soil moistureand temperature related to slope position and aspect werefound to be important in determining species and community idistri'bution in Thompson Gorge in the North Carolina BlueRidge by Mowbray and Oosting (1968). Similarly, Cabrera (1969) Irelated variation in a mixed mesophytic forest on Ash LogMountain, Cambell County, Tennessee to aspect and topographic _form.

i

ACKNOWLEDGEMENTS

The Tennessee Heritage Program, State Department ofConservation provided support for field work for this project.The Graduate Program in Ecology of The University of Tennesseesupported other phases of this work. Use of the University ofTennessee Computing Center is also gratefully acknowledged.

ii!!_!i

84

LITERATURE CITED

Anderson, S° Ho and No H° Shugarto 1974. HABITAT SELECTION OFBREEDI_4G BIRDS IN AN EAST TENNESSEE DECIDUOUS FOREST°Ecology 55; 828.,837_

Beers_, To Wo_ P° Eo Dress° and L.o Co Wensel. 1966, ASPECTTRANSFORMATION IN SITE PRODUCTIVITY RESEARCH, J,Forestry 64(lC): 691-692o

Braun_ Eo Lo 1950o DECIDUOUSFORESTSOF EASTERN NORTHAMERICA° Blakiston, Philadelphia_ Penn. 596pp,

Buckner_ Eo and Wo McCrackeno 1978o YELLOW POPLARs A COMPONENTOF CLIMAX FORESTS? J o Forestry 76: 421-423.

Cabrera_ Ho 1969o PATTERNS OF SPECIES SEGREGATION AS RELATEDTO TOPOGRAPHICFORMAND ASPECT. Mo So Thesis. The Universityof Tennessee_ Knoxville. 111ppo

Caplenor_ Co Do 1965o THE VEGETATION OF THE GORGESOF FALLCREEK FALLS STATE PARK IN TENNESSEE, J. Tenno Acad. Sci.40: 27°49°

Carpemter_ S,, Bo 1976. STAND STRUCTUREOF A FOREST IN THECUMBERLAND PLATEAU OF EASTERN KENTUCKY FIFTY YEARS AFTERLOGGINGAND BURNING° Castanea 41: 325_337.

Clark, Ro C. 1966o THE VASCULAR FLORA OF THE FIERY GIZZARDGORGEIN SOUTHCENTRALTENNESSEE. Mo A. Thesis. Universityof North Carolina, Chapel Hill o 89ppo

Day, P,, R. i956. REPORTOF THE COMMITTEEON PHYSICAL ANALYSIS1954._1955o SOIL SCIENCE SOCIETY OF AMERICA. Soil Sci. Soc.Amero Proco 1956o

Dickson, Ro Ro 1960, CLIMATES OF THE STATES: TENNESSEE. U.S.Department of Commerce, Weather Bureau. Washington, D, C.16pp.

Dixon, W° Jo (ed.) 1975o BMDPBIOMEDICAL COMPUTERPROGRAMS,University of California Press_ Berkley. 792pp.

Fenneman, No M,, 1938. PHYSIOGRAPHYOF EASTERNUNITED STATES.McGrawoHill, New York° 712pp.

Golden, Mo So 1974o FOREST VEGETATION AND SITE RELATIONSHIPSIN THE CENTRAL PORTION OF THE GREAT SMOKY MOUNTAINSNATIONAL PARK. Pho Do Dissertation. The University ofTennesee, Knoxville. 275pp.

Goodall, Do Wo 1963. THE CONTINUUMAND THE INDIVIDUALISTICASSOCIATION. Vegetatio ]Is 297-316.

Hinkleo Co Ro 1975. A PRELIMINARY STUDY OF THE FLORA ANDVEGETATION OF CUMBERLANDGAP NATIONAL HISTORICAL PARK,MIDDLESBORO, KENTUCKY° M. So Thesis. The Universityof Tennessee, Knoxville. 236pp.

Hinkle, Co Ro 1978o THE CLASSIFICATION AND ANALYSIS OF FORESTCOMMUNITIES OCCURRING ON THE CUMBERLAND PLATEAU OF TENNESSEEINCLUDING FLAT TO ROLLING UPLANDAND RAVINE TOPOGRAPHY.Pho Do Dissertation. The University of Tennessee, Knoxville.

85 _

Hinkle, Co R., P. A. Schmalzer, and H. R. DeSelmo 1978o THEVEGETATIONOF DICK COVE,SEWANEE,TENNESSEE. Abstract ofpresented paper in : The ASB Bulletin 25_51°

Jackson, M. L. 1958. SOIL CHEMICAL ANALYSIS. Prentice-Hall,Inc. Englewood Cliffs, New Jersey. 498ppo

Kercher, J. Ro and R. A. Goldstein. 1977. ANALYSIS OF AN EASTTENNESSEEOAK-HICKORYFORESTBY CANONICALCORRELATION OFSPECIESAND ENVIRONMENTALPARAMETERS. Vegetatio 35" 153-163. !i

Kessell, S. R. and R. H. Whittaker. 1976. COMPARISON OF THREEORDINATION TECHNIQUES. Vegetatio 32- 21-29.

Longwell, T. J., W. L. Parks, and M. E. Springer. 1963oMOISTURECHARACTERISTICSOF TENNESSEESOILS° University _of Tennessee Agr. Expt. Sta. Bull. 367. Knoxvilleo 46pp.

Mann, L. K. 1977. DISCRIMINANT ANALYSIS OF SOME EAST TENNESSEEFORESTHERBNICHES. M. S. Thesis° The University ofTennessee,Knoxville. 107pp.

Martin, W. H. III. 1966. SOMERELATIONSHIPSOF VEGETATION TOSOILANDSITE FACTORSONWILSONMOUNTAINMORGANCOUNTY

TENNESSEE.M. S. Thesis. The University of Tennessee, jiKnoxville. 134pp. _

Martin_ W. H° Ill. 1971. FORESTCOMMUNITIESOF THE GREAT VALLEYOF EASTTENNESSEEANDTHEIRRELATIONSHIPSTO SOIL ANDTOPOGRAPHICFACTORS.Ph. D. Dissertation. The University of !Tennessee, Knoxville. 366pp.

Mowbray, T. B. and H. j. Oosting. 1968. VEGETATIONGRADIENTSIN RELATION TO ENVIRONMENTAND PHENOLOGYIN A SOUTHERNBLUE RIDGE GORGE. Ecol. Monogr. 38: 309-344.

McCarthy, D. M. 1976. NUMERICALTECNIQUESFORCLASSIFYINGFORESTCOMMUNITIESIN THETENNESSEEVALLEY. Ph. D.

Dissertation. The University of Tennessee, Knoxville. 164pp.Nie_ N. H., C. H. Hull, J. G. Jenkins, K. Steinbrenner, and D. H.

Bent 1975 SPSSSTATISTICALPACKAGEFORTHE SOCIALSCIENCES. 2nd ed. McGraw-Hill, New York. 675pp.

Norris, J. Moand J. P. Barkham. 1970. A COMPARISONOF SOMECOTSWALDBEECHWOODSUSING MULTIPLE DISCRIMINANT ANALYSIS.J. Ecol. 58-603-619.

Oosting, H.J. 1942. AN ECOLOGICALANALYSISOF PLANT COMMUNITIESOF PIEDMONT,NORTHCAROLINA. Amer. Midland Nat. 28- 1-126. •

Orloci, L. 1967. AN AGGLOMERATIVEMETHODFOR CLASSIFICATIONOF PLANT COMMUNITIES. J. Ecol. 55: 193-206.

Pielou, E.C. 1977. MATHEMATICALECOLOGY. John Wiley and Sons,NewYork.385pp.

Radford, A. E., H. E. Ahles, and Co R. Bell. 1968. MANUAL OFTHEVASCULARFLORAOF THECAROLINAS. University of NorthCarolina Press, ChapelHill. 1183pp.

iii_,i' _ ii_

!_ i _iiiii_ii_i_i_ ii_ill

86

Safley_ J o Mo 1970o VEGETATION OF THE BIG SOUTH FORK,CUMBERLANDRIVERo KENTUCKYAND "TENNESSEE° Mo So Thesis.Yhe University of "Tennessee, Knoxvilleo 148pp.

Schmalzer_ Po Ao 1978o CLASSIFICATION AND ANALYSIS OF FORESTCOMMUNITIES iN SEVERAL COVES OF THE CUMBERLANDPLATEAU OFTENNESSEE° Mo So Thesis° The U_liversity of Tennessee,K_loxvi I I e,

Sherman_ Ho Lo 1958o THE VEGETATION AND FLORISTICS OF FIVEGORGESOF THE CUMBERLANDPLATEAU° Ms S. Thesis. TheUniversity of Tennessee, Knoxville° 103pp,

Smith_ Lo Ro 1977o .THE SWAMPAND MESIC FORESTSOF THECUMBERLANDPLATEAU IN TENNESSEE° M. So Thesis. TheUniversity of Tennessee, Knoxville° 121pp.

Soil Survey Staff° 1951o SOIL SURVEYMANUAL. U. S. Dept. Agr.Agricultural Research Administration Agro Handbook 18.Washington_ Do C. 503.pp,

Tatsuoka_ M,, Mo 1971o MULTIVARIATE ANALYSIS:TECHNIQUES FOREDUCATIONALAND PSYCHOLOGICALRESEARCH. John Wiley andSons, New York, 310pp°

Whittaker_ Ro Ho 1956. VEGETATION OF THE GREAT SMOKY MOUNTAINS,Ecol. Monogro 26: I_80o

Wolfe, C° Bo, Jro and Jo D, Pittilloo 1977. SOME ECOLOGICALFACTORSINFLUENCING THE DISTRIBUTION OF BETULA NIGRA L.IN WESTERNNORTHCAROLINA. Castanea 4_0--7--'-