Correspondence analysis of functional groups in a riparian landscape

Jonathan Lyon1,2 and Cynthia L. Sagers1,*1Department of Biological Sciences, University of Arkansas, Fayetteville, Arkansas 72701, USA; 2Currentaddress: Biology Department, University of Wisconsin – Whitewater, 800 West Main Street, Whitewater,WI 53190-1790, USA; *Author for correspondence (e-mail: csagers@comp.uark.edu; phone: (501) 575-6349;fax: (501) 575-4010)

Received 14 September 1999; accepted in revised form 29 March 2001

Key words: CCA, DCA, Missouri, Ordination, Ozark National Scenic Riverways, Riparian forest, Vegetationanalysis

Abstract

We used multivariate analysis and ordinations to characterize the composition and distribution of woody vegeta-tion within the Ozark National Scenic Riverways (ONSR), Missouri, USA. The objectives of the study were to:1) evaluate patterns of woody species distributions along existing environmental gradients; 2) determine if dif-ferent classes of woody plants (i.e., dominant overstory trees, all trees, understory trees, and shrubs) respondedsimilarly to the same suite of environmental variables; and 3) determine if discrete ecotonal and/or ecoclinalvegetation patterns were present across the landscape. Woody vegetation was sampled from 94 plots along 35transects positioned at right angles to the river channel. Sample plots were analyzed with Detrended Correspon-dence Analysis (DCA), Canonical Correspondence Analysis (CCA), and TWINSPAN. Overall, woody vegetationwas correlated with several environmental gradients, including elevation of the plot above the river, soil pH, soilmoisture, and soil particle size. Responses to secondary gradients differed among the four classes of plants ana-lyzed, however. CCA biplots of understory trees indicated that patterns of those species were strongly correlatedwith slope through the plot and sand content of soil. CCA biplots of shrubs showed that CCA axes were moststrongly correlated with soil organic matter content, soil moisture, and silt content. Further, there was limitedevidence for discrete assemblages of woody species, with the exception of streamside vegetation. Instead, mix-ing of woody species was observed across a broad transition zone. Because there is little correspondence be-tween vegetation layers, our results demonstrate including plant classes other than a subset of canopy dominanttrees can provide additional resolution in characterizing vegetation responses along complex environmental gra-dients.

Introduction

Riparian plant communities perform an array of im-portant ecosystem functions, including stream bankstabilization (Osborne and Kovacic 1993), thermalregulation of streams (Gray and Eddington 1969), fil-tering and retention of nutrients (Vought et al. 1994),maintenance of ecosystem stability (Wiens et al.1985), provision of important animal and wildlifehabitat (Sparks 1995) and migration corridors (Sim-berloff and Cox 1987), and supply of organic matterto aquatic consumers (Cummins et al. 1989). Manyriparian forests also support diverse flora (Nilsson et

al. 1989; Bratton et al. 1994; Planty-Tabacchi et al.1996). Based on these attributes, there has been grow-ing interest in characterizing the composition and spa-tial boundaries of riparian forests as well as ascertain-ing the links between riparian vegetation and under-lying environmental gradients (Naiman and Décamps1990; Hansen and Di Castri 1992; Bendix 1994; Nils-son et al. 1994). These approaches have provided es-sential information in the formulation of effectiveprotection, management, and restoration efforts in ri-parian forests (Naiman and Décamps 1990; Hansenand Di Castri 1992; Hupp 1992; Bendix 1994; Nils-son et al. 1994).

171Plant Ecology 164: 171–183, 2002.© 2002 Kluwer Academic Publishers. Printed in the Netherlands.

In the United States, the need for proper charac-terization of riparian vegetation is important in lightof the mandates of the National Forest ManagementAct of 1976 (Federal Register 47(190), 219.26,219.27(g), 1982) and recent efforts to pursue ecosys-tem management, both which require the use of ef-fective quantitative approaches to ensure that man-agement practices maintain the integrity and biodiver-sity of forest systems (Thomas 1996). Most efforts tocharacterize riparian vegetation (and forest vegetationin general) have centered on the use of dominant orcommercially important canopy species. Yet, for theseclassifications to be useful in an ecosystem manage-ment context, their ecological validity needs to be as-certained (Bailey 1984; Jensen et al. 1991; Sharitz etal. 1992; Slocombe 1993; Bailey et al. 1994). Manystudies of riparian vegetation have employed multi-variate statistical techniques to characterize vegeta-tion patterns (Rochow 1972; Robertson et al. 1984;Collins et al. 1981; Hardin et al. 1989; Hupp 1992;Nilsson et al. 1994), but fewer studies have attemptedto incorporate other vegetation layers in evaluatingvegetation patterns and underlying environmentalgradients (Nilsson et al. (1989); Dollar et al. (1992);Ware et al. (1992), but see Sagers and Lyon (1997)).

One approach to addressing the complexity of for-est classification is functional group analysis. Plantspecies can be classified into specific guilds or func-tional groups based on a variety of characteristics, in-cluding morphology (Raunkiaer 1934), physiology(Mueller-Dombois and Ellenberg 1974; Smith et al.1993), reproductive, ruderal or competitive status(Grime 1979) or location in a successional sere (Ba-zzaz 1979). Each functional group potentially willpartition the environmental gradient(s) differently(Austin 1985, 1990; Smith and Huston 1989). Thus,as resource levels change spatially and/or temporally,the growth and distribution of different functionalgroups is predicted to change also. The use of verti-cally stratified growth forms or vegetation layers(e.g., overstory trees, understory trees, shrubs) as ameans of separating functional groups has been dem-onstrated as an approach that integrates many of thepreviously described characteristics and has a soundecological and physiological basis (Noble et al. 1988;Chapin 1993; Grime 1993; Smith et al. 1993; Körner1994). However, there is limited information on thedifferential responses to environmental gradientsamong vegetation layers (Lippmaa 1939; McCuneand Antos 1981; Hardin and Wistendahl 1983; Dunnand Stearns 1987; Gilliam et al. 1995).

We carried out the study within the Ozark NationalScenic Riverways (ONSR), a riparian corridor alongthe Current and Jacks Fork Rivers in southeasternMissouri, USA. Little specific information is avail-able concerning the riparian forests of the Ozarks(Ware et al. 1992); the vast majority of woody veg-etation studies in the Ozarks have focused on domi-nant canopy species in upland vegetation (Zimmer-man and Wagner 1979; Nigh et al. 1985; Pallardy etal. 1988; Cutter and Guyette 1994) or on specificplant assemblages in riparian areas (Witherspoon1971; Autry 1988; McKenney et al. 1995). The spe-cific objectives of the study were to: 1) evaluate pat-terns of woody species distributions along existingenvironmental gradients; 2) determine if differentfunctional groups (dominant canopy trees, overstorytrees, understory trees, shrubs) exhibited differentialresponses to the same suite of environmental varia-bles; and 3) determine if discrete shifts in vegetationcould be detected across an elevational gradient fromstreamside to ridgetop.

Methods

Study area

The Ozark National Scenic Riverways (ONSR) occu-pies some 26,306 ha along a narrow corridor enclos-ing a 161 km stretch of the Current River and 55 kmof the Jacks Fork River in south-central Missouri(Figure 1). The ONSR occupies portions of Dent, Sh-annon, Carter, and Texas counties in Missouri, USA,and is located on the Salem Plateau of the Ozark Pla-teaus physiographic province (Fenneman 1938). Mostof the ONSR area has a karst drainage system that hasdeveloped in the carbonate rocks in the region (Vine-yard and Feder 1974) and as much as 60% of the tworivers’ flow is from karst springs (Jacobson andPrimm 1994). The Ozark Plateau has been a continu-ous land area since the end of the Paleozoic (Branson1944; Steyermark 1959; Vineyard 1969) and becausethe region has never been glaciated, it has been openfor plant migration since the Tertiary.

The Current and Jacks Fork River systems drain apart of the Salem Plateau in south central Missouriand flow through an open valley bounded by bedrockwalls with an average height of 60 m (McKenney etal. 1995). The Jacks Fork watershed drains some1046 km2 in the Salem Plateau. The Current Riverwatershed is substantially larger (9560 km2) with

172

only a small proportion of that watershed (26,306 ha)protected within the boundaries of the ONSR. TheCurrent and Jacks Fork Rivers experience periodicflooding. Typical yearly floods range from 2–3 mabove baseflow, a 25 to 50 year flood reaches 4–6 mabove baseflow (Jacobson and Primm 1994), and themaximum flood stage of 9.4 m above baseflow wasrecorded along the Jacks Fork in November, 1904.The Current and Jacks Fork systems flow south intothe White River drainage of Arkansas and into theMississippi River. Much of the riparian landscape ofthe ONSR has been highly disturbed following Euro-pean settlement around 1820 (Jacobson and Primm1994). The forests in the ONSR experienced indis-criminate, and widespread clearcutting from 1890 to1920. These anthropogenic disturbances have alteredvegetation cover, forest density, and fire regimes. Theexisting secondary forests have been broadly classi-fied as oak-pine and oak-hickory (Braun 1950; Eyre1980), but specific assemblages range from wet bot-tomland to mesic mid-slope to more xeric upland.

Vegetation sampling

Between June and August 1994, and in August 1995,we established elevational transects at 35 sites alongthe Current and Jacks Fork Rivers. Sites were sepa-rated by approximately 4.8 river km. Transects beganat the river’s edge and continued upland until the can-opy of the study plot was dominated by oak (Quercusspp.) and hickory (Carya spp.), species characteristicof ridgetop communities in these forest systems.

Rectangular plots were spaced at 15 m intervals alongthe transect. Rectangular plots, rather than square orcircular plots, were used because rectangular plotsmore adequately sample the existing diversity (Bor-mann 1953). Each plot was 10 × 20 m, with the longside of the plot parallel to the river’s edge. A total of135 - 10 × 20 m plots were sampled during the sur-vey, 94 of these plots contained woody vegetation >1 cm dbh. These 94 plots are included in the analysespresented here.

We categorized plants as trees, shrubs or herbsbased on stem diameter, height, and the presence ofwoody tissue. We segregated the vegetation into threea priori defined functional groups: overstory trees,understory trees, and shrubs. All tree species (over-story and understory) > 1 cm in diameter at 1.3 m inheight (dbh) and all shrub species were measured.Each tree within the plot was identified to species andits diameter was recorded. Shrub species were sam-pled in four 3-m2 circular subplots within each 10 ×20 m plot, and their occurrence and cover classeswithin a subplot were recorded. Cover classes fol-lowed Daubenmire (1959).

Physical attributes

Each study plot was characterized by three physicalfactors: slope, aspect, and height above river. We col-lected soil samples at a depth of 10 cm from threelocations chosen haphazardly within each 10 × 20 mplot. Details of soil analyses are given in Sagers andLyon (1997). In summary, each bulk soil sample wasair-dried and passed through a 2 mm sieve to sepa-rate fine and crude soil fractions. The total sampleweight, and the weight of the smaller size fractionwere recorded to calculate the percentage of totalsample < 2 mm (fines). All subsequent analyses wereperformed on the fine fraction. Soil pH was measuredfollowing the methods of McLean (1982). Soil tex-ture was measured using a method modified slightlyfrom Bouyoucos (1951). Hydrometer readings weretaken 0 s, 40 s and 120 min after mixing with a stan-dard hydrometer (ASTM no. 152 H with Bouyoucosscale in g/l) to characterize sand, silt and clay, respec-tively. Hydrometer readings were corrected for devi-ations from normal room temperature. Container ca-pacity (cont cap), the water content of a saturated soilafter is has been allowed to drain, was determinedfollowing the methods of Cassel and Nielsen (1986).Container capacity was calculated as the differencebetween the post-, and pre-wetting weights divided by

Figure 1. Map of the Ozark National Scenic Riverways (ONSR)depicting the location in the State of Missouri and the ONSRboundaries.

173

the post-wetting weight of the sample. Organic mat-ter content (OM) was determined by loss on ignitionfollowing Lim and Jackson (1982). Loss on ignitionincludes structural water, organic matter, and somesoluble volatile salts.

Ordinations and classification

Both detrended correspondence analysis (DCA) andcanonical correspondence analysis (CCA) were con-ducted on plant species-environmental variable matri-ces using the programs PC-ORD (McCune and Mef-ford 1995) and CANOCO 3.10 (ter Braak 1990). Im-portance values (IV’s) were calculated for all tree spe-cies for each sampling plot as the sum of relativedensity (% of stems) + relative dominance (% of basalarea). Importance value, therefore, reflects numericaland physical dominance in the study plot and wouldreach a maximum value (200%) in a monotypic plot.Tree species IV’s were used in all the ordination andclassification methods described. The CCA procedureinvolved linear combination of variables for sitescores, no transformation of species abundance ma-trices, and the use of a Monte Carlo permutation testto test the significance of the first axis eigenvalue (terBraak 1990). To determine if different functionalgroups exhibited differential responses to the samesuite of environmental variables, separate CCA ordi-nations were performed on four a priori defined func-tional groups: all trees, dominant overstory trees, un-derstory trees, and shrubs. (Appendix 1) contains alist of all species encountered and their classificationinto functional groups.

Classification of woody plant assemblages wasalso conducted with the program TWINSPAN (Hill1979). TWINSPAN classifications were partiallymodified to help clarify the validity of the indicatorspecies detected (Dale 1995). The soil and environ-mental variables described in the Methods sectionwere transformed, when necessary, to meet the as-sumptions of normality. Correlations between envi-ronmental variables and ordination axes scores wererun using Minitab 8.2 (Minitab 1991). Significance isreported at the alpha = 0.05 level, unless noted other-wise in the text.

Transitions between vegetation types along an el-evational transect were determined using both a com-bined dataset of all tree species and by segregatingthe data into vegetation layers (i.e., dominant over-story trees, understory trees, shrubs). Data were com-piled from all transects into a single elevation (height

above river) transect. Differential DCA profiles werederived from each vegetation layer using DCA scores(Hill 1979; Hobbs 1986). DCA graphical profileswere created by plotting sample plot scores from thefirst axis of the respective DCA ordinations versus theelevation of the plot. In a DCA graphical profile, thesteeper the profile slope, the more abrupt the changein the composition of the vegetation. Concordance inDCA scores for each pair-wise combination of plantclasses was evaluated with Pearson product-momentcorrelations (Sagers and Lyon 1997).

Results

Soil pH, soil organic matter content, soil particle size(fines), and percent clay content were highly variable,indicating the presence of broad soil chemical andphysical gradients (Table 1). This variability, in part,reflects the diversity in soil parent materials and geo-morphology within the ONSR (Jacobson and Primm1994). Thus, the sampling regime was effective insurveying across several environmental gradients inthe riparian landscape. The corresponding range inplot locations relative to river elevation (0.1 – 40.0m) indicates that the vegetation sampling cut across atopographical gradient that included both flood proneand flood immune areas.

Ordinations of all tree species

To ascertain overall patterns of plant species distribu-tions based on measured physical and environmentalgradients, both detrended correspondence analysis(DCA) and canonical correspondence analysis (CCA)ordinations were performed on a matrix containingIV’s for all tree species (n = 65 species) on 94 plots.In all CCA ordinations performed, the Monte Carlopermutation test indicated that the eigenvalues for thefirst axes were all significant (P < 0.05). Table 2 pro-vides a comparison of eigenvalues and environmentalvariable correlations with the first three axes of theDCA and CCA ordinations. The correlations of eachenvironmental variable with the DCA and CCA axeswere very similar (except for a reversal of sign onaxis 1) for the first two axes. The similarity in envi-ronmental correlations on the first axis of both theDCA (unconstrained) and CCA (constrained) ordina-tions, suggests that the environmental/soil variablesmeasured were good indicators (either directly or ascovariates) of key underlying environmental gradi-

174

ents that exist within the study area (Jongman et al.1995).

The biplot of the CCA ordination is shown in Fig-ure 2. The eigenvalues for the first two axes (�1 =0.553 and �2 = 0.322, respectively) indicate separa-tion along the measured gradients. The dominant en-vironmental variables correlated with the first axiswere elevation (height above river) and pH (Table 2).Fines (soil particle size) showed the highest correla-tion with the second axis (Table 2). Separation ofvegetation plots located on gravel bar and directlyadjacent to the river channel is represented by pointsin the upper left quadrant of the CCA biplot (low el-evation, high pH) (Figure 2). CCA ordination of alltree species (results not presented) indicated that bot-tomland tree species such as Platanus occidentalis,Salix nigra, and Ulmus pumila were dominant in thisregion of the biplot. Some separation of plots support-ing upland oak assemblages was also noted in the up-per right quadrant of the biplot (high elevation, lowpH). Upland species such as Quercus alba, Q. velu-tina, Q. marilandica, and Carya texana were noted inCCA species ordinations in this quadrant. Distinctseparation and grouping of other plots and/or specieswere not as pronounced, indicating a continuum ofwoody vegetation moving left to right beneath thecentroid of the ordination. Overall, the CCA biplot

depicted in Figure 2 indicates a transition in woody

Table 1. Correlation coefficients between 10 environmental variables measured in this study. The means, medians, and ranges of variables(where appropriate) are also listed at the bottom of the table. Significant correlations between variables are noted by a * (= P < 0.05).

Variable† Slope Aspect HAR pH Fines Cont cap Sand Silt Clay OM

Slope

Aspect −0.079

HAR 0.386* −0.060

pH −0.217 −0.088 −0.473*

Fines −0.071 0.176 −0.012 −0.215

Cont cap 0.382* −0.111 0.436* −0.150 −0.150

Sand −0.165 0.018 −0.123 0.456* −0.247 −0.126

Silt 0.118 −0.031 0.142 −0.380* 0.219 0.268 −0.877*

Clay 0.383* 0.066 0.129 −0.257 −0.093 0.134 −0.553* 0.362*

OM 0.385* −0.094 0.363* −0.006 −0.003 0.880* −0.260 0.414* 0.233

units deg deg m pH % % % % % % LOI

Mean 20.7 – 8.75 6.07 60.2 33.9 33.8 8.1 58.1 8.3

Median 16.0 – 4.35 6.19 57.0 33.6 27.8 5.6 66.7 6.8

Range

low 0.0 – 0.1 3.54 1.2 20.2 0.0 0.0 0.0 0.6

high 65.0 – 40.0 7.40 99.9 52.5 97.2 36.1 99.4 43.1

† Variables are defined as follows: slope = slope (°) through the vegetation plot; aspect = aspect of plot (°); HAR = height above river ofvegetation plot (m); pH = soil pH of top 10 cm of soil; Fines = % of total sample < 2 mm dia; Cont cap = container capacity of soil samples(%); Sand = % sand in soil; Silt = % silt in soil; Clay = % clay in soil; OM = organic matter content (%) in top 10 cm of soil determined byLOI (loss on ignition).

Table 2. Comparison of eigenvalues, correlations, and species-en-vironment correlations between environmental variables and DCAand CCA ordination axes. All CCA correlations listed below are‘intraset correlations’ as described by ter Braak (1986)

Based on All Trees > 1 cm dbh

DCA AXES CCA AXES

1 2 3 1 2 3

Eigenvalue 0.761 0.454 0.374 0.553 0.322 0.193

Variables*

Slope −0.398 0.190 0.079 0.465 0.199 −0.271

Aspect 0.046 0.060 −0.140 −0.063 −0.128 0.434

HAR −0.685 0.289 −0.183 0.789 0.338 0.034

pH 0.612 −0.193 0.090 −0.757 −0.243 −0.327

Fines −0.258 −0.024 0.053 0.391 −0.801 0.160

Cont cap −0.388 0.133 −0.027 0.461 −0.003 −0.569

Sand 0.341 −0.052 0.096 −0.450 0.118 −0.217

Silt −0.317 0.073 −0.110 0.436 −0.175 0.217

Clay −0.251 0.108 −0.032 0.278 0.145 −0.115

OM −0.334 0.118 0.001 0.423 −0.200 −0.528

spp-Envt† 0.921 0.795 0.608 0.880 0.754 0.738

* Descriptions of environmental variables and how they were de-termined can be found in the Methods section and at the bottom ofTable 1.† Spp-Envt correlations refer to Pearson correlations between sam-ple scores that are linear combinations of environmental variablesand sample scores that are based on species data.

175

vegetation from bottomland to upland species that isinfluenced primarily by height above river and pH onthe first axis and by the proportion of fines or stoni-ness on the second axis.

Functional group analysis

The relative importance of the environmental varia-bles differed among plant classes. A comparison of‘intraset correlations’ (ter Braak 1986) with the firstthree CCA ordination axes of the three functionalgroups is given in Table 3. The biplot of the CCA or-dination that included only the 14 dominant tree spe-cies (species with the highest overall basal areassummed across plots) is shown in Figure 3. Similarto the CCA ordination for all tree species, dominanttrees species showed strong correlations between thefirst axis and pH and HAR (Figure 3, Table 3). Com-parisons of all three CCA axes of this analysis re-vealed additional consistencies with the analysis of alltree species (Table 2). Nine environmental variables(HAR, pH, cont cap, OM, slope, silt, sand, fines, as-pect) were largely correlated with species distribu-tions whether all trees, or only the dominant tree spe-cies, were included. In contrast, soil particle size(fines) was important in the ordination of all tree spe-cies (Figure 2), but were not strongly correlated in theanalysis of dominant tree species (Figure 3). Further-more, container capacity and organic matter content

were less important in the analysis of all tree speciesthan in the analysis of dominant tree species. Plotscontaining higher elevation oak and hickory specieswere found exclusively in the upper right quadrant ofthe CCA biplot (analyses not presented). Speciescharacteristic of low elevation plots were noted in theupper left quadrant, and a varied assortment of wet-mesic to mesic species were noted in the lower quad-rants, including Acer saccharum, Acer negundo, Ul-mus rubra, Juglans cinerea, and Quercus rubra.

Understory tree species (n = 15 species) showedstrong correlations between the first CCA axis andHAR and pH (Figure 4, Table 3), similar to the over-all and dominant overstory functional group ordina-tions. However, distinct differences from the otherfunctional group ordinations were evident. Strongcorrelations between percent sand and CCA axes 2and 3, and between CCA axis 2 and slope through theplot were noted (Figure 4, Table 3). The correlationof axis 2 with fines also was much weaker among theunderstory trees compared to the overstory and over-all CCA ordinations. CCA ordination of understoryspecies (results not presented) indicated that Cornusflorida was the only species in the upper right quad-rant of the CCA biplot. Bottomland species such asSambucus canadensis and Asimina triloba were cen-tered in the upper left quadrant. The species positivelycorrelated with percent sand included Rhus aromat-ica, R. glabra and Staphylea trifolia; species posi-tively correlated with increased slope included Sassa-fras albidum. These results indicate that understorytrees exhibited a differential response to the same un-derlying environmental gradients compared to theother functional groups, thus reflecting a differentialenvironmental response between overstory and under-story tree species.

The final functional group investigated was theshrubs (n = 23 species). The CCA biplot for shrubs isshown in Figure 5. The relationships between shrubsand the measured environmental variables differedmarkedly from the other functional groups. CCA axis1 was strongly correlated with organic matter contentand container capacity (Table 3). Axis 2 showed thestrongest correlations with percent silt and percentsand (Table 3). Correlations with pH and HAR werestill evident, but not as pronounced (Table 3). Over-all, responses of shrubs to environmental gradientsdiffered substantially from the other functional groupsanalyzed; shrub plots were more strongly correlatedwith container capacity, OM, and silt than the othervegetation layers.

Figure 2. CCA ordination of 94 plots and 65 tree species (> 1cm dbh) with 10 environmental variables. Biplot vectors shown re-present the major explanatory environmental variables (see Table 1for codes). Longer vector lines represent stronger ‘intraset correla-tions’ (after ter Braak 1986).

176

Classification of woody riparian vegetation

TWINSPAN was used to classify the overall woodyvegetation data (Hill 1979) to determine if discreteassemblages of all trees and shrubs were evident inthe ONSR, The analysis was stopped at level 5 andthe subsequent TWINSPAN results were used to aid

in the classification of the woody vegetation. Whilethe TWINSPAN results indicated 16 associations, wecollapsed these groups to form a total of 12 associa-tions based on the high degree of species similarityin some of the groups. To graphically depict these as-sociations on the CCA ordinations, the centroids ofeach association (and their standard deviations) werecalculated from CCA scores on the first two axes, andwere then plotted (Figure 6). Three distinct groupings

Table 3. A comparison of CCA ordination results between dominant trees and three functional type groupings: overstory trees, understorytrees, and shrubs (see Appendix 1 for categorization of dominants, overstory, understory, and shrub species). Eigenvalues, and both environ-mental variable and species-environment correlations with CCA ordination axes, are shown for comparison.

DOMINANT TREES UNDERSTORY SHRUBS

AXIS AXIS AXIS

1 2 3 1 2 3 1 2 3

Eigen-values 0.511 0.243 0.170 0.510 0.256 0.175 0.907 0.863 0.631

Variables*

Slope −0.553 0.031 −0.295 0.393 −0.540 −0.130 0.500 0.012 0.276

Aspect 0.331 0.046 0.423 −0.079 0.155 0.088 0.129 0.020 −0.007

HAR −0.701 0.250 0.078 0.888 −0.026 0.329 0.555 −0.038 0.236

pH 0.631 −0.545 −0.147 −0.677 −0.148 0.224 −0.170 0.595 −0.409

Fines −0.001 −0.251 0.414 0.194 0.027 0.011 0.518 −0.077 −0.054

Cont cap −0.699 −0.265 −0.061 0.123 −0.101 0.334 0.759 −0.159 0.160

Sand 0.474 0.211 −0.498 −0.224 −0.565 0.593 −0.098 0.641 −0.484

Silt −0.487 −0.249 0.614 0.024 0.415 0.056 0.236 −0.749 0.166

Clay −0.347 −0.213 0.098 0.324 −0.157 −0.224 −0.056 −0.283 0.379

OM −0.641 −0.362 0.073 0.208 −0.001 0.031 0.881 0.011 0.045

Spp-Envt† 0.814 0.699 0.542 0.832 0.711 0.583 0.956 0.950 0.819

* Descriptions of environmental variables and how they were determined can be found in the Methods section and at the bottom of Table 2.† Spp-Envt correlations refer to Pearson correlations between sample scores that are linear combinations of environmental variables andsample scores that are based on species data.

Figure 3. CCA ordination of 93 plots and the 14 most dominanttree species found throughout the ONSR with 10 environmentalvariables. Biplot vectors shown represent the major explanatoryenvironmental variables (see Table 1 for codes).

Figure 4. CCA ordination of 81 plots and 15 understory tree spe-cies (> 1 cm dbh) with 10 environmental variables. Biplot vectorsshown represent the major explanatory environmental variables(see Table 1 for codes).

177

of the 12 associations can be ascertained. The asso-ciation among low elevation plots (group 12) has amean height above river of 3.5 m (thus clearly proneto flooding) and is the only group that clearly sepa-rated out on the CCA plot. A cluster of upland groupsdominated by upland oak (Quercus) and hickory(Carya) species is indicated by the cluster of groups1, 2, 3 and 4 on the high elevation, low pH side ofCCA axis 1. The third cluster (groups 5–11) are rela-tively close in ordination space and many associationsexhibit a high degree of overlap.

Transitions among species assemblages

We used results of the DCA analysis to distinguish ifthere were specific ecotonal transitions along the to-pographic gradient from the river’s edge to the up-lands. The results of these analyses are shown in Fig-ure 7, a DCA profile of vegetation plots plottedagainst elevation (height above river). The steeperslopes represent more abrupt changes in the compo-sition of vegetation along the elevational gradient. Allplots reveral abrupt and numerous turnovers in treespecies composition, especially at the lower eleva-tions (Fig. 7A). At higher elevations, plant speciesassemblages appear more homogeneous, althoughsome variation is still evident. Similar patterns areobserved for both overstory trees and the sub-groupof canopy dominants. The understory trees (Fig. 7B)also showed rapid turnover at the lower elevations buta distinct change at approximately 7 m. Shrubs, on

the other hand, showed little change in compositionacross the gradient (Fig. 7C).

Plot positions in the DCA did not correspond wellamong plant classes. In all cases, sampling plot ordi-nation scores of DCA axis 1 were significantly corre-lated (P < 0.01). DCA scores of all trees and domi-nant overstory trees were highly correlated (r = 0.94),but scores of all trees corresponded less well withunderstory tree (r = 0.78) and shrub (r = 0.36) classes.Scores of dominant overstory trees showed similarweak correlations with the understory and shrubclasses (r = 0.65, and 0.57, respectively).

Three broad vegetation categories or zones are de-picted in Figure 7: flooded, transition, and upland.These zones correspond to the three clusters indicatedin Figure 6. The flooded zone refers to woody plantassemblages that experience various levels of flood-ing. The upper position (4.6 m) reflects the meanflooding level for a 10-year flood (Jacobson andPrimm 1994). The upland oak zone refers to a suiteof relatively homogeneous assemblages of upland oak(Quercus) and hickory (Carya) species. The uplandoak zone reflects more xeric conditions with low soilpH. The transition zone refers to those plots and veg-etation assemblages that bridge between the often-flooded forests and the flood immune upland oaks.The flooded area represents the spatial zone with thehighest level of species turnover, whereas the transi-tion zone appears to bounded by abrupt changes inspecies composition.

Figure 5. CCA ordination of 41 plots and 25 woody shrub species(cover estimates) with 10 environmental variables. Biplot vectorsshown represent the major explanatory environmental variables(see Table 1 for codes.)

Figure 6. CCA ordination of 94 plants and 70 woody species (> 1cm dbh) with 10 environmental variables. Centroids (_1 SE) of the12 vegetation associations determined with TWINSPAN areshown.

178

Discussion

Woody vegetation in the ONSR is responding directlyor indirectly to several important environmental var-iables/gradients, including elevation (height aboveriver), soil pH, soil moisture (container capacity),sand content, and soil particle size (fines). Responsesof woody vegetation to these gradients have been ob-served in other studies in the Ozarks (Ware et al.1992; Sagers and Lyon 1997; Lyon and Sagers 1998).Topographic and soil pH gradients have been corre-lated with vegetation type in a study of larger trees (>10.16 cm dbh) in the ONSR (Ware et al. 1992), inupland sugar maple stands (Nigh et al. 1985), and insimilar riparian-upland landscapes (Pallardy et al.1988; Dollar et al. 1992). Strong correlations betweenvegetation assemblages and soil water holding capac-ity have also been noted in Missouri forests (Zimmer-

man and Wagner 1979; Robertson et al. 1984; Nighet al. 1985; Ware et al. 1992). However, all of the re-gional studies mentioned focused on the distributionof canopy species (i.e., either large trees or a sub-group of canopy dominants). A more detailed ap-proach, incorporating all trees and shrubs, revealssome of the specific interactions between environ-mental gradients and woody vegetation. Further, theseclasses of woody vegetation can be included in thesample with limited additional costs of field time ortaxonomic expertise.

That vegetation patterns are often correlated withpatterns of resource variation and resource gradientshas been well established in vegetation science (Glea-son (1926); Whittaker (1956, 1967, 1978), Austin1980, Smith and Huston (1989)). Different plant spe-cies or groups of species may have different resource-use strategies, physiologies, and competitive abilities,and thus may be segregated into different functionalgroups. O’Neill et al. (1986) noted that many vegeta-tion studies avoid complexity by over-emphasizing asingle type of observation set, typically a subset ofcanopy tree species (Barnes et al. 1982). In this study,an analysis of canopy dominants corresponded wellwith an analysis of all tree species, but failed to pre-dict the distribution of understory trees and shrubs, orto characterize transitions among forest types. Thegeneral discordance among plant classes in this studyemphasizes the importance of including additionaldata in characterizing the structure of complex forestsystems (Lippmaa 1939; McCune and Antos 1981;Dunn and Stearns 1987). Including different func-tional groups in multivariate assessments, classifica-tion, and ordinations can provide more detailed andmore comprehensive information on the distributionsof “vegetation types” and the response of woody plantspecies to underlying gradients. Furthermore, vegeta-tion layer/functional group analyses can be used in amanagement context to identify important landscapecharacteristics, including zones of richness, key habi-tats, and vegetation layer interactions.

Many efforts to characterize forest vegetation ingeneral, or riparian forest vegetation specifically,have focused on the presence of dominant and/orcommercially important canopy species in the devel-opment of classification systems. These classificationschemes are then used as a template for forest man-agement. However, the accuracy of these classifica-tions is questionable in terms of their inclusiveness ofvegetation layers and their ecological validity (Bailey1984). The arguable goal of forest classification/plant

Figure 7. DCA score profiles of vegetation plots plotted against aheight gradient (m). Plot (A) shows the profile for both overstorytree and canopy dominants; plot (B) shows the profile for under-story trees; and plot (C) depicts the profile for shrubs. The DCAvalues plotted are based on a five-point moving weighed averageof DCA axis 1 scores to reduce variation. Distinctions between theflooded zone, transition zone, and upland zone are shown on allplots.

179

community delineation in an ecosystem managementcontext should be to develop a classification schemethat integrates all existing vegetation, landscape con-siderations, and ecological principles while concomi-tantly providing applicable assessments for riparianforest management (Jensen et al. (1991); Sharitz et al.(1992); Bailey et al. (1994), Minshall, 1994). The in-tegration of functional group analysis, ecotonal anal-ysis, and multivariate ordination techniques discussedin the present paper is an effort toward a more com-prehensive approach to classification of riparian for-est systems.

Given the spatial complexity of woody species dis-tribution and composition within the ONSR, the de-lineation of distinct vegetation boundaries (e.g., ripar-ian versus mesic) remains problematic. The lack ofany consistent delineation between plant assemblageslimits the value of designating specific managementzones based on any single landscape attribute (e.g.,topography, soil type). Managing larger landscapeunits based on comprehensive vegetation analysesmay be a more effective management strategy in thisspatially complex landscape. Stratifying vegetationby functional groups or specific guilds in vegetationanalyses can allay ecosystem management concernsby providing managers with a broader view of veg-etation structure and composition. A landscape man-agement approach based on integrated vegetationanalyses is also more likely to buffer the impacts ofsuccessional processes (temporal complexity) that isaltering the composition of existing assemblages andthe abundance and distribution of specific species inthis highly dynamic landscape. Overall, there is aneed to acknowledge the spatial and temporal con-straints of any forest classification scheme, especiallyas it relates to the protection, enhancement and/orrestoration of specific plant species and assemblages.

Acknowledgements

We thank C. Crownover, A. Delp, H. Hubbard, B.Madison, D. Moore, J. Sharma and A. Spann for theirhelp with data collection. Drs. E.E. Dale, Jr. and E.Smith, and B. Hintertheur identified many haggardplant specimens. Dr. R.D. Evans provided advice onexperimental design. The Arkansas Water ResourcesCenter and the Department of Biological Sciences,University of Arkansas provided logistical support.

The Department of the Interior/National Park ServiceSubagreement No. 9 to Cooperative Agreement No.CA 7150-4-0001 funded this work.

Appendix 1

Table A1. Latin binomials, functional type (Func Type) designa-tion, and categorization as a dominant (Dom) of all 88 woody spe-cies encountered in this study. Nomenclature follows after Steyer-mark (1968).

Latin binomial Func Type* Dom†

Acer negundo L. O DAcer rubrum L. OAcer saccharinum L. OAcer saccharum Marsh. O DAesculus glabra Wild. OAgrimonia pubescens Wallr. SAgrimonia rostellata Wallr. SAmelanchier arborea (Michx. F.) Fern. SAmpeliopsis arborea (L.) Koehne SAsimina triloba (L.) Dunal UBetula nigra L. OBumelia lanuginosa (Michx.) Pers. UCampsis radicans (L.) Seem. SCarpinus caroliniana Walt. UCarya cordiformis (Wang) K. Koch O DCarya glabra (Mill.) OCarya illinoensis (Wang) K. Koch OCarya lacinosa (Michx. F.) Loud. OCarya ovata (Mill.) K. Koch OCarya texana Buckl. OCarya tomentosa (Poir.) Nutt. OCatalpa speciosa Warder OCeltis laevigata Wild. OCeltis occidentalis L. O DCeltis tenuifolia Nutt. SCercis canadensis L. UCornus drummondii Meyer UCornus florida L. UCorylus americana Walt. SCotinus obovatus Raf. SCrataegus viridis L. ODiospyros virginiana L. OFagus grandiflora Ehrh. OFraxinus americana L. O DFraxinus pennsylvanica Marsh. OFraxinus quadrangulata Michx. OGleditsia tricanthos L. OHamamelis virginiana L. UIlex decidua Walt. UJuglans cinerea L. OJuglans nigra L. O

180

Table A1. Continued.

Latin binomial Func Type* Dom†

Juniperus virginiana L. OJusticia americana (L.) Vahl SLindera benzoin (L.) Blume UMoru rubra L. ONyssa sylvatica Marsh. OOstrya virginiana (P. Mill.) K. Koch UPhiladelphus hirsutus Nutt. SPhiladelphus pubescens Loisel. SPhysocarpus opulifolius (L.) Raf. SPinus echinata Miller OPlatanus occidentalis L. O DPopulus deltoides Marsh. OPrunus serotina Ehrh. OQuercus alba L. O DQuercus bicolor Wild. OQuercus falcata Michx. OQuercus imbricaria Michx. OQuercus lyrata Walt. OQuercus marcrocarpa Michx. O DQuercus marilandica Muench. OQuercus muhlenbergii Engelm. O DQuercus rubra L. O DQuercus shumardii Buckl. OQuercus stellata Wang. OQuercus velutina Lam. O DRhamnus caroliniana Walt. URhus aromatica Ait. SRhus glabra L. URosa sp SSalix nigra Marsh. OSambucus canadensis L. SSassafras albidum (Nutt.) Nees UStaphylea trifolia L. UTilia americana L. OToxidendron radicans (L.) Kuntze SUlmus alata Michx. O DUlmus americana L. O DUlmus pumila L. OUlmus rubra Muhl. O DViburnum prunifolium L. UViburnum rufidulum Raf. SVitis aestivitis Michx. SVitis cinerea Engelm. Ex Millard SVitis riparia Michx. SVitis rupestris Scheele SVitis vulpina L. SZanthophyllum americanum P. Mill. S

* Func Type refers to plant functional type: O = overstory tree, U= understory tree, S = woody shrub† Dom (D) refers to the 14 most dominant tree species (based onsummed basal area across all 94 plots)

References

Austin M.P. 1985. Continuum concept, ordination methods, andniche theory. Annual Review of Ecology and Systematics 16:39–61.

Austin M.P. 1990. Community theory and competition in vegeta-tion. In: Grace J.B. and Tilman D. (eds), Perspectives on plantcompetition. Academic Press, New York, New York, pp. 215–238.

Autry D.D. 1988. Plant communities on riparian limestone bluffsin Ozark National Scenic Riverways. PhD Dissertation, South-ern Illinois University at Carbondale, Carbondale.

Bailey R.G. 1984. Testing an ecosystem regionalization. Journal ofEnvironmental Management 19: 239–248.

Bailey R.G., Jensen M.E., Cleland D.T. and Bourgeron P.S. 1994.Design and use of ecological mapping units. In: Jensen M.E.and Bourgeron P.S. (eds), Ecosystem management: principlesand applications. USDA Forest Service, Pacific Northwest Re-search Station, Portland, Oregon, pp. 95–106.

Baker W.L. 1989. Macro- and micro-scale influences on riparianvegetation in western Colorado. Annals of the Association ofAmerican Geographers 79: 65–78.

Bazzaz F.A. 1979. The physiological ecology of plant succession.Annual Review of Ecology and Systematics 10: 351–371.

Bendix J. 1994. Scale, direction, and pattern in riparian vegetation-environment relationships. Annals of the Association of Ameri-can Geographers 84: 652–665.

Bormann F.H. 1953. The statistical efficiency of sample plot sizeand shape in forest ecology. Ecology 34: 474–487.

Bouyoucos G.J. 1951. A recalibration of the hydrometer methodfor making mechanical analysis of soils. Agronomy Journal 43:434–438.

Branson E.B. 1944. The Geology of Missouri. University of Mis-souri Studies XIX, Columbia.

Bratton S.P., Hapeman J.R. and Mast A.R. 1994. The lower Sus-quehanna River Gorge and floodplain (U.S.A.) as a riparianrefugium for vernal, forest-floor herbs. Conservation Biology8: 1069–1077.

Braun E.L. 1950. Deciduous Forests of Eastern North America.Blakiston Company, Philadelphia.

Cassel D.K. and Nielsen D.R. 1986. Field capacity and availablewater capacity. In: Klute A. (ed.), Methods of soil analysis. Part1. Physical and mineralogical methods. Agronomy Monographno. 9. 2nd edn. American Society of Agronomy-Soil ScienceSociety of America, Madison, Wisconsin, pp. 198–223.

Chapin F.S. 1993. Functional role of growth forms in ecosystemand global processes. In: Ehleringer J.R. and Field C.B. (eds),Scaling physiological processes. Academic Press, New York,New York, pp. 287–312.

Collins S.L., Riser P.G. and Rice E.L. 1981. Ordination and clas-sification of mature bottomland forests in north central Okla-homa. Bulletin of the Torrey Botanical Club 108: 152–165.

Cummins K.W., Wilzbach M.A., Gates D.M., Perry J.B. and Talia-ferro W.B. 1989. Shredders and riparian vegetation. BioScience39: 24–30.

Cutter B.E. and Guyette R.P. 1994. Fire frequency on an oak-hickory ridgetop in the Missouri Ozarks. The American Mid-land Naturalist 132: 393–398.

Dale M.B. 1995. Evaluating classification strategies. Journal ofVegetation Science 6: 437–440.

181

Daubenmire R.F. 1959. A canopy-coverage method of vegetationanalysis. Northwest Scientist 33: 46–66.

Dollar K.E., Pallardy S.G. and Garrett H.G. 1992. Composition andenvironment of floodplain forests of northern Missouri. Cana-dian Journal of Forest Research 22: 1343–1350.

Dunn C.P. and Stearns F. 1987. Relationship of vegetation layersto soils in southeastern Wisconsin forested wetlands. TheAmerican Midland Naturalist 118: 366–374.

Eyre F.H. 1980. Forest Cover Types of The United States andCanada. Society of American Foresters, Washington.

Fenneman N.M. 1938. Physiography of The Eastern United States.McGraw-Hill, New York.

Gagnon D. and Bradfield G.E. 1987. Relationships among foreststrata and environment in southern coastal British Columbia.Canadian Journal of Forest Research 16: 1264–1271.

Gilliam F.S., Turrill N.L. and Adams M.B. 1995. Herbaceous-layerand overstory species in clear-cut and mature central Appala-chian hardwood forests. Ecological Applications 5: 947–955.

Gleason H.A. 1926. The individualistic concept of the plant asso-ciation. Bulletin of the Torrey Botanical Club 53: 1–20.

Gray J.R. and Eddington J.M. 1969. Effect of woodland clearanceon stream temperature. Journal of Fisheries Research Canada26: 399–403.

Grime J.P. 1979. Plant Strategies and Vegetation Processes. JohnWiley, New York.

Grime J.P. 1993. Vegetation functional classification systems asapproaches to predicting and quantifying global vegetationchange. In: Solomon A.M. and Shugart H.H. (eds), Vegetationdynamics and global change. Chapman and Hall, New York,New York, pp. 293–305.

Hansen A.J. and Di Castri F. 1992. Landscape Boundaries. Eco-logical Studies 92. Springer-Verlag, New York, New York.

Hardin E.D. and Wistendahl W.A. 1983. The effects of floodplaintrees on herbaceous vegetation patterns, microtopography andlitter. Bulletin of the Torrey Botanical Club 110: 23–30.

Hardin E.D., Lewis K.P. and Wistendahl W.A. 1989. Gradient anal-ysis of floodplain forests along three rivers in unglaciated Ohio.Bulletin of the Torrey Botanical Club 116: 258–264.

Hill M.O. 1979. TWINSPAN – a Fortran Program for ArrangingMultivariate Data in an Ordered Two-way Table by Classifica-tion of The Individuals and Attributes. Ecology and Systemat-ics, Cornell University, Ithaca, New York.

Hobbs E.R. 1986. Characterizing the boundary between Californiaannual grassland and Coastal sage scrub with differential pro-files. Vegetatio 65: 115–126.

Hupp C.R. 1992. Riparian vegetation recovery following streamchannelization: a geomorphic perspective. Ecology 73: 1209–1226.

Jacobson R.B. and Primm A.T. 1994. Historical land-use changesand potential effects on stream disturbances in the Ozark Pla-teaus, Missouri. United States Geological Service Report 94-333., Rolla, Missouri.

Jarvis P.G. 1995. Scaling processes and problems. Plant, Cell andEnvironment 18: 1079–1089.

Jensen M., McNicoll C. and Prather M. 1991. Application of eco-logical classification to environmental effects analysis. Journalof Environmental Quality 20: 24–30.

Johnston C.A., Pastor J. and Pinay G. 1992. Quantitative methodsfor studying landscape boundaries. In: Hansen A.J. and di Cas-tri F. (eds), Landscape Boundaries. Ecological Studies 92.Springer-Verlag, New York, New York, pp. 107–125.

Jongman R.H.G., ter Braak C.J.F. and van Tongeren O.F.R. 1995.Data Analysis in Community and Landscape Ecology. Cam-bridge University Press, Cambridge.

Körner C. 1994. Scaling from species to vegetation: the usefulnessof functional groups. In: Shulze E.D. and Mooney H.A. (eds),Biodiversity and ecosystem function. Springer-Verlag, NewYork, New York, pp. 117–140.

Lim C.H. and Jackson M.L. 1982. Dissolution for total elementalanalysis. In: Page A.L. (ed.), Methods of soils analysis. Part 2.Chemical and Microbiological Properties. Agronomy Series no.9. 2nd edn. American Society of Agronomy-Soil Science So-ciety of America, Madison, Wisconsin, pp. 1238–1252.

Lippmaa T. 1939. The unistratal concept of plant communities (theunions). The American Midland Naturalist 21: 111–143.

Lyon J. and Sagers C.L. 1998. Structure of herbaceous plant as-semblages in a forested riparian landscape. Plant Ecology 138:1–16.

Malanson G.P. and Butler D.R. 1991. Floristic variation amonggravel bars in a subalpine river, Montana, USA. Arctic and Al-pine Research 23: 273–278.

McCune B. and Antos J.A. 1981. Correlations between forest lay-ers in the Swan Valley, Montana. Ecology 62: 1196–1204.

McCune B. and Mefford M.J. 1995. PC-ORD. Multivariate Analy-sis of Ecological Data, Version 2.0. MjM Software Design,Gleneden Beach.

McKenney R., Jacobson R.B. and Wertheimer R.C. 1995. Woodyvegetation and channel morphogenesis in low-gradient, gravel-bed streams in the Ozark Plateaus, Missouri and Arkansas.Geomorphology 13: 175–198.

McLean E.O. 1982. Soil pH and lime requirement. In: Page A.L.(ed.), Methods of soils analysis. Part 2. Chemical and micro-biological properties. Agronomy Series No. 9. 2nd edn. Ameri-can Society of Agronomists, Madison, Wisconsin, pp. 97–119.

Minitab 1991. Minitab Reference Manual. Minitab Inc. Data TechIndustries, Valley Forge, Release 8.2.

Mueller-Dombois D. and Ellenberg H. 1974. Aims and Methods ofVegetation Ecology. Wiley, New York.

Naiman R.J. and Décamps H. 1990. The Ecology and Managementof Aquatic-terrestrial Ecotones. United Nations Education, Sci-entific and Cultural Organization, Paris.

Nigh T.A., Pallardy S.G. and Garrett H.E. 1985. Sugar maple-en-vironment relationships in the river hills and central OzarkMountains of Missouri. The American Midland Naturalist 98:469–476.

Nilsson C., Grelsson G., Johansson M. and Sperens U. 1989. Pat-terns of plant species richness along riverbanks. Ecology 70:77–84.

Nilsson C., Ekblad A., Dynesius M., Backe S., Gardfjell M., Car-lberg B. et al. 1994. A comparison of species richness and traitsof riparian plants between a main river channel and its tribu-taries. Journal of Ecology 82: 281–295.

Noble I.R., Moore A.D. and Strasser M.J. 1988. Predicting vegeta-tion dynamics based on structural and functional attributes. In:Proceedings International symposium on vegetation structure.,Utrecht.

O’Neill R.V., DeAngelis D.L., Waide J.B. and Allen T.F.H. 1986.A Hierarchical Concept of Ecosystems. Monographs in Popu-lation Biology 23. Princeton University Press, Princeton, NewJersey.

182

Osborne L.L. and Kovacic D.A. 1993. Riparian vegetated bufferstrips in water-quality restoration and stream management.Freshwater Biology 29: 243–258.

Pallardy S.G., Nigh T.A. and Garrett H.E. 1988. Changes in forestcomposition in central Missouri: 1968-1982. The AmericanMidland Naturalist 120: 380–390.

Planty-Tabacchi A., Tabacchi E., Naiman R.J., Deferrari C. andDécamps H. 1996. Invasibility of species-rich communities inriparian zones. Conservation Biology 10: 598–607.

Raunkiaer C. 1934. The Life-forms of Plants and Statistical PlantGeography. Oxford University Press, Oxford.

Robertson M.D., MacKenzie P.A. and Elliott L.F. 1984. Gradientanalysis and classification of the woody vegetation for foursites in southern Illinois and adjacent Missouri. Vegetatio 58:87–104.

Rochow J.J. 1972. A vegetational description of a mid-Missouriforest using gradient analysis techniques. The American Mid-land Naturalist 87: 377–398.

Sagers C.L. and Lyon J. 1997. Gradient analysis in a riparian land-scape: contrasts among forest layers. Forest Ecology and Man-agement 96: 13–26.

Sharitz R.R., Boring L.R., Van Lear D.H. and Pinder J.E. 1992. In-tegrating ecological concepts with natural resource manage-ment of southern forests. Ecological Applications 2: 226–237.

Simberloff D. and Cox J. 1987. Consequences and costs of conser-vation corridors. Conservation Biology 1: 63–71.

Slocombe D.S. 1993. Implementing ecosystem-based management.BioScience 43: 612–622.

Smith T. and Huston M. 1989. A theory of the spatial and temporaldynamics of plant communities. Vegetatio 83: 49–69.

Smith T.M., Shugart H.H., Woodward F.I. and Burton P.J. 1993.Plant functional groups. In: Solomon A.M. and Shugart H.H.(eds), Vegetation Dynamics and Global Change. Chapman &Hall, New York, New York, pp. 272–292.

Solomon A.M. and Shugart H.H. 1993. Vegetation Dynamics andGlobal Change. Chapman and Hall, New York.

Sparks R.E. 1995. Need for ecosystem management of large riversand their floodplains. BioScience 45: 168–182.

Steyermark J.A. 1959. Vegetational History of The Ozark Forest.University of Missouri Studies, Columbia, Missouri.

ter Braak C.J.F. 1990. CANOCO – a Fortran program for canoni-cal community ordination. Microcomputer Power, Ithaca, NewYork.

Thomas J.W. 1996. Forest service perspective on ecosystem man-agement. Ecological Applications 6: 703–705.

Vineyard J.D. 1969. Geological report on the Ozark National Sce-nic Riverways. National Park Service Report. Office of NaturalScience Studies, Rolla, Missouri.

Vineyard J.D. and Feder G.L. 1974. Springs of Missouri. In:Pflieger W.L. and Lipscomb R.G. (eds), Water Resources Re-port 29. Missouri Division of Geology and Land Survey, Rolla,Missouri, pp. 6–9.

Vought L.B.M., Dahl J., Pedersen C.L. and Lacoursiere J.O. 1994.Nutrient retention in riparian ecotones. Ambio 23: 342–348.

Ware S., Redfearn P.L., Pyrah G.L. and Weber W.R. 1992. Soil pH,topography, and forest vegetation in the central Ozarks. TheAmerican Midland Naturalist 128: 40–52.

Whittaker R.H. 1956. Vegetation of the Great Smoky Mountains.Ecological Monographs 26: 1–80.

Whittaker R.H. 1967. Gradient analysis of vegetation. BiologicalReview 42: 207–264.

Whittaker R.H. 1978. Ordination of Plant Communities. Junk, TheHague.

Wiens J.A., Crawford C.S. and Gosz J.R. 1985. Boundary dynam-ics: a conceptual framework for studying landscape ecosys-tems. Oikos 45: 421–427.

Witherspoon J.T. 1971. Plant succession on gravel bars along theCurrent and Jacks Fork Rivers in the Southcentral MissouriOzarks. M.S., Southwest Missouri State College, Springfield.

Zimmerman M. and Wagner W.L. 1979. A description of the woodyvegetation of oak-hickory forest in the Northern Ozark High-lands. Bulletin of the Torrey Botanical Club 106: 117–122.

183