Content uploaded by David Zelený
Author content
All content in this area was uploaded by David Zelený
Content may be subject to copyright.
Environmental control of the vegetation pattern in deep river valleys
of the Bohemian Massif
Vliv faktorů prostředí na vegetaci hlubokých říčních údolí Českého masivu
David Z e l e n ý1,2 & Milan C h yt r ý1
1Department of Botany and Zoology, Masaryk University, Kotlářská 2, CZ-611 37 Brno,
Czech Republic, email: zeleny@sci.muni.cz, chytry@sci.muni.cz; 2Department of Botany,
Faculty of Biological Sciences, University of South Bohemia, Branišovská 31, CZ-370 05
České Budějovice, Czech Republic
Zelený D. & Chytrý M. (2007): Environmentalcontrol of vegetation pattern in deep river valleys of
the Bohemian Massif. – Preslia 79: 205–222.
The pattern of natural vegetation on non-calcareous soils in two deep river valleys of the Bohemian
Massif (Vltava and Dyje rivers, Czech Republic) was analyzed in order to determine the main topo-
graphic and soil variables affecting the composition of the vegetation. Vegetation data together with
topographic and soil variables were collected along transects down the slope from the upper edge to
the bottom of the valley. The distribution of vegetationtypes within the valleys was described using
cluster analysis and non-metric multidimensional scaling (NMDS). Effects of topographic and soil
variables were compared using a set of canonical correspondence analyses (CCAs) with explana-
tory variable selection based on the Akaike information criterion (AIC). In order to describe the
non-linear interaction between the two topographic variables, elevation and aspect, a new method
(moving window CCA) was introduced. This method assessed the explanatory power of aspect at
various elevations above the valley bottom. Results show that main vegetation coenoclines are cor-
related with two complex environmental gradients: the moisture-nutrient-soil reaction and light-
temperature-continentality gradients. Soil variables are slightly better predictors of vegetation com-
position than topographic variables. Altogether, these variables explain 18.8–21.6% of the total
inertia. Although soil development depends on topography, the variation jointly explained by both
groups of variables is only 3.9–5.2%, indicating that each of these two groups of variables influ-
ences vegetation pattern in a different way. Variables selected by the most parsimonious modelfor
the Vltava valley are aspect, soil pH, soil type fluvisoland soil depth. For the Dyje valley the same
variables as in Vltava valley were selected except for soil depth, which was replaced by soil type
cambisol. Aspect has a strong effect on vegetation on the middle slopes but not on the lower slopes
of the valleys. The results of all analyses are similar between the two valleys, suggesting that similar
patterns may also occur in other deep river valleys of mid-altitudes of the Bohemian Massif.
K e y w o r d s : canonical correspondence analysis, cluster analysis, deep river valleys, non-metric
multidimensional scaling, moving window CCA, vegetation-environment relationships
Introduction
In the gently undulating landscape of the Bohemian Massif, which occupies a large part of
the Czech Republic and adjacent areas of Germany and Austria, deep river valleys are
a distinct topographic feature. Compared to other valley types, these are narrow, V-shaped
valleys with steep slopes, large meanders and a narrow, discontinuously developed
floodplain. They are sharply incised in the flat or hilly landscapes, predominantly formed
of granite or gneiss bedrocks. These valleys are mainly found at middle altitudes between
200 and 700 m. All river valleys of this type are of Quaternary age, when the uplift of the
Preslia 79: 205–222, 2007 205
Bohemian Massif increased the erosion power of rivers and caused the deepening of previ-
ously shallow and broad valleys into deep and narrow ones (Kopecký 1996).
Botanical diversity of deeply incised river valleys in the Bohemian Massif has the same
general characteristics as other river corridors, such as high species richness (Gould &
Walker 1999), linear plant migration (Naiman et al. 1993, Burkart 2001, Mouw et al.
2003) and sensitivity to alien plant invasions (Pyšek & Prach 1993, Planty-Tabacchi et al.
1996). Additionally, these valleys possess some other characteristics, in particular a high
beta diversity of the plant communities on hillsides, caused by high topographic, geologi-
cal and mesoclimatic diversity (Chytrý & Tichý 1998). There are sharp environmental gra-
dients within relatively small areas, some of which are large enough to include both ex-
treme as well as intermediate values of environmental factors (e.g. different moisture in
the floodplain and on the south-facing upper parts of the valley slopes, or soil pH on out-
crops of siliceous and calcareous bedrocks). Consequently, deep river valleys represent lo-
cal biodiversity hotspots in an otherwise rather uniform landscape in the middle altitudes
of the Bohemian Massif.
The high biotic diversity in the Bohemian Massif deep river valleys is coupled with lim-
ited human impact in some places. This feature also sharply contrasts with the adjacent
landscape, which is dominated by an intensively managed mosaic of arable fields and sec-
ondary forest plantations. Because of the limited accessibility due to the steep valley
slopes, there are complete zonations of near-natural vegetation types, predominantly for-
ests, in several sections of the valleys.
Strong topographic gradients in the river valleys affect the variation in several environ-
mental variables, which directly affect plant growth, e.g. moisture, nutrient availability
and pH. Thus, topography strongly co-varies with vegetation pattern, and at some scales,
topographic variables can be robust predictors of vegetation patterns. Such relationships
are clear in river valleys with broad floodplains (e.g. Sagers & Lyon 1997, Gould &
Walker 1999, van Coller et al. 2000, Goebel et al. 2006), but may be accentuated in deep
river valleys due to the complex topography of the slopes adjacent to the floodplain.
Specific features of the vegetation pattern of these deep river valleys were summarized
under the heading “river phenomenon” in the descriptions provided by Czech vegetation sci-
entists in the 1960s (Blažková 1964, Jeník & Slavíková 1964). The “river phenomenon”
concept describes how topographic and mesoclimatic features of the deeply incised river
valleys of the Bohemian Massif affect their vegetation diversity. Of the abiotic factors, this
concept stresses the sharp contrast between the deep river valleys and adjacentgently undu-
lating landscape, the pronounced effect of exposed rocks occurring on steep slopes on the
vegetation, the contrast between the sunny and warm south-facing slopes and shaded and
cold north-facing slopes, high diversity of various extreme habitats situated next to each
other and specific mesoclimatic conditions causing temperature inversions. Of the vegeta-
tion features, the “river phenomenon” concept emphasizes (1) the high biodiversity in deep
river valleys, (2) non-random distribution of vegetation types and species richness within the
valleys, (3) concentration of relict species, resulting from the fact, that deep river valleys
probably served as a Pleistocene refuges for plant and animal species, and (4) migration of
plants and animals along the rivers, connecting mountains with lowlands. Although there are
several published local descriptions of plant communities in the Bohemian Massif deep river
valleys (Blažková 1964, Türk 1994, Chytrý & Vicherek 1995, 1996, 2003, Kolbek et al.
1997, 1999–2003), there are no quantitative studies that test the predictions of the “river phe-
206 Preslia 79: 205–222, 2007
nomenon” concept (Blažková 1964, Jeník & Slavíková 1964) and summarize the general
features of vegetation patterns and their driving environmental factors.
The aim of this study is to produce a quantitative description of the vegetation-environ-
ment relationships in deep river valleys of the Bohemian Massif, focusing on the patterns oc-
curring in a cross-section of the valleys. To avoid the effect of local idiosyncrasies on the re-
sults, we studied two valleys, Vltava and Dyje, differing markedly in both climatic and
floristic characteristics. Specifically, we analyzed correlations between vegetation pattern
and topographic variables, measured soil factors and species indicator values, in order to re-
veal the most important factors determining the pattern of vegetation in deep river valleys.
Methods
Study sites
One study site is a part of the Vltava river valley in S Bohemia, N of Český Krumlov
(Fig. 1). The section of valley studied is situated between Zlatá Koruna (48°51' N, 14°22'
E) and Boršov nad Vltavou (48°55' N, 14°26' E), with an altitudinal range of 400–540 m
a.s.l. and maximum valley depth of around 100 m. Climate in this area is moderately
warm, with mean January temperatures –3 to –1 °C and mean July temperatures 16–17 °C.
Average annual precipitation is 550–600 mm (Tolasz 2007). Phytogeographically this
area belongs to the Hercynian floristic region with some components of the forest flora of
the Alps and continental thermophilous flora of Central Bohemia. Bedrock types include
mainly acidic gneiss and granulite, with patchy occurrence of crystalline limestone (mar-
ble), serpentine and amphibolite (Chábera 1985).
Zelený & Chytrý: Environmental control of vegetation in deep river valleys 207
Fig. 1. – Location of the studied sections of the Vltava and Dyje valleys in the Czech Republic.
The other site is located in the Dyje (in German Thaya) river valley in the
Podyjí/Thayatal National Park on the border between the Czech Republic and Austria
(Fig. 1). The studied section of this river valley is between the towns of Vranov nad Dyjí
(48°54' N, 15°49' E) and Znojmo (48°52' N, 16°03' E) on the Czech and Austrian sides of
the national border, respectively. Altitudinal range is 220–536 m a.s.l. and maximum val-
ley depth almost 200 m. Climate in this area is generally warmer and more continental
than at the previous site, with mean January temperatures ranging between –3 and –2 °C,
mean July temperatures 18–19 °C and mean annual precipitation 550–600 mm (Tolasz
2007). This site is located close to the boundary of the Hercynian and Pannonian floristic
regions (Chytrý et al. 1999) and therefore has a significant proportion of thermophilous
and continental species. Geological characteristics are similar to those of the previous site.
Predominant bedrocks include acidic gneiss and granite, with some restricted occurrences
of crystalline limestone (Batík 1992).
Data sampling
Fieldwork was conducted in 1992–1993 by M. C. (Dyje valley) and 2001–2003 by D. Z.
(Vltava valley). The standardized sampling protocol (Chytrý 1995) was applied in both
valleys. Vegetation was sampled along transects from the upper edge to the bottom of the
valley in places where there was no artificialor human-disturbed vegetation. Transect sites
were selected to include the maximum diversity of habitat types occurring in the valleys.
Along these transects, vegetation and environmental data were collected in plots of
10 × 15 m (with longer axis situated along the isohypse) placed equidistantly every 30 m in
the Vltava valley and 40 m in the Dyje valley, which reflects the greater depth of the Dyje
valley. In each plot, all the vascular plants were recorded, plus an estimate of the cover
based on the nine-degree Braun-Blanquet scale (Westhoff & van der Maarel 1978). No-
menclature of plant taxa follows Kubát et al. (2002).
Various topographic and soil factors were either measured directly, estimated or calcu-
lated (Table 1). Of them, heat index (Parker 1988) measures relative differences in the solar
energy arriving at the different sites. It is calculated from the slope and aspect, using the for-
mula heat index = cos (aspect – 202.5°) × tg (slope), where 202.5° represents the warmest
SSW aspect. Although in theory solar irradiance in the northern hemisphere peaks at solar
noon and a 180° aspect, delayed ground heating is responsible for the fact that the highestdi-
urnal heat load is experienced on SW–SSW facing slopes (Geiger 1966). Aspect, due to its
circular nature, was not used per se, but calculated as the deviance of the measured plot as-
pect from 22.5° (NNE), thus reaching the highest value of 180° on SSW slopes.
Within each plot, five measurements of soil depth were made using a metal gouge auger
with an operational length of 70 cm and a diameter of 1.5 cm; the five values were aver-
aged and used as an estimate of soil depth (note that the actual soil depth may be underesti-
mated when the auger is used in stony soils). Due to a strongly skewed distribution, this
variable was log-transformed before further analyses. At each plot, five soil samples from
the A-horizon (depth 0–10 cm after litter removal) were collected from different places,
mixed together and used to measure soil pH in water solution (dried samples were placed
in distilled water for 24 hours; weight ratio of soil/water = 0.4). For each plot, soil types ac-
cording to ISSS-ISRIC-FAO (1998) were recorded, using a simplified categorization of
the following four broadly conceived classes: fluvisol – soils directly affected by a river
208 Preslia 79: 205–222, 2007
water regime, with fluvic soil material (inspected using the auger); skeletic – skeletic and
hyperskeletic leptosols on steep scree slopes, containing various proportions of gravel or
coarse stones; cambisol – deeper and matured cambisols on slight slopes; lithic – shallow
and undeveloped lithic leptosols on and near to rocky outcrops. As most plots were on
acidic bedrock, data from transects containing plots on calcareous soils were removed
from the data set (3 plots in the Vltava and 22 in the Dyje valley). These plots, representing
vegetation types sharply different from those on acidic soils, might produce an undesirable
outlier effect. The data set used for the analyses included 94 plots situated along 26
transects in the Vltava valley and 82 plots from 14 transects in the Dyje valley.
Classification and indirect ordination
To identify the main vegetation types, plots were classified by cluster analysis, performed
separately on the data sets from each valley. Several pilot analyses with various combina-
tion of clustering methods and distance measures were calculated. For presentation, the
relative Euclidean (chord) distance and Ward’s clustering algorithm based on square-root
transformed percentage cover data were used, because they best reflected the pattern of
vegetation differentiation as judged by expert knowledge. The resulting classifications
were projected onto an ordination diagram using non-metric multidimensional scaling
(NMDS; Minchin 1987) performed on a matrix of Bray-Curtis dissimilarities between
relevés, together with passively projected Ellenberg indicator values (EIV; Ellenberg et al.
1992) calculated as non-weighted averages of the values for all species in merged vegeta-
tion layers. NMDS was calculated using the advanced algorithm proposed by Minchin
(1987). It includes several random calculations in order to search for a robust global solu-
tion and post-analysis rotation of NMDS axes based on principal components analysis so
that the variance of points is maximized on the first dimension (for more details see
Oksanen et al. 2006). Polarity of axes in resulting diagrams was adjusted in order to unify
Zelený & Chytrý: Environmental control of vegetation in deep river valleys 209
Table 1. – Explanatory variables used in the study.
Topographic variables (quantitative and ordinal variables):
Elevation relative elevation above the valley bottom (range 0–1, 0 for the valley bottom, 1 for the upper
edge)
Aspect aspect, expressed as deviation of plot aspect from 22.5° (NNE); it reaches the highest value
for the supposedly warmest SSW aspect
Slope slope inclination (°); observed range: Vltava 0–88°, Dyje 5–77°
Heat index heat index = cos (aspect – 202.5°) × tg slope
Surface SL landform shape in the downslope direction (three-degree ordinal scale: –1 concave, 0 flat, 1
convex)
Surface ISO landform shape along an isohypse (three-degree ordinal scale: –1 concave, 0 flat, 1 convex)
Soil variables and soil types (quantitative and categorical variables):
pH active soil pH measured in water solution
Soil depth soil depth, expressed as log [soil depth (cm)]
Fluvisol fluvisols (water-influenced soils formed from alluvial deposits)
Skeletic skeletic and hyperskeletic leptosols (stony soils on scree accumulations)
Cambisol cambisols (well-developed zonal soils)
Lithic lithic leptosols (shallow soils near rock outcrops)
the directions of EIV and signs of correlations with axes in both valleys. Clusters obtained
for each valley were ordered along the moisture gradient (according to cluster median EIV
for moisture) from the driest (Cluster 1) to wettest (Cluster 5) to ensure that in both valleys
the clusters with the same numbers represent analogous vegetation types. Interpretation of
particular clusters in terms of vegetation types was based on expert judgement, supported
by the list of diagnostic, constant and dominant species identified for each cluster (not
shown; diagnostic species were determined using the phi coefficient of association, cor-
rected for even group sizes according to Tichý & Chytrý 2006).
Correlation among explanatory variables and Ellenberg indicator values
Correlation matrix of all explanatory variables and EIV was calculated, using Spearman
rank coefficients for all variables except the relationships between binary variables (soil
types); these were calculated using contingency tables, with significance derived from
Pearson’s chi-square test with Yates’s continuity correction (Sokal & Rohlf 1995). Corre-
lations between explanatory variables were based on data merged from both valleys, while
correlations of explanatory variables and EIV were made separately for each valley in or-
der to detect local differences in observed patterns.
Variation partitioning between topographic and soil variables
Relationships between vegetation composition and environmental variables were ana-
lyzed using canonical correspondence analysis (CCA; ter Braak 1986), a method modified
to handle unimodal species responses. In the first step, models based only on topographic
(model V.topo for Vltava and D.topo for Dyje valley, respectively) and only on soil vari-
ables (models V.soil and D.soil) were developed for each valley in order to assess the
amount of variation explained by each of these two types of explanatory variables. Models
were built using a stepwise algorithm, combining forward and backward selection of ex-
planatory variables. Evaluation of models’ parsimony was based on the Akaike informa-
tion criterion (AIC; Akaike 1973) as implemented in the R package Vegan (Oksanen et al.
2006). Conditional and shared effects of selected topographic and soil variables were cal-
culated by partial CCA, using topographic variables as explanatory variables and soil vari-
ables as covariables (models V.topo.cond and D.topo.cond) and vice versa (V.soil.cond and
D.soil.cond). In order to quantify the amount of variation explained by the model, the ratio
of the sum of the constrained eigenvalues to total inertia was used. Like Økland (1999),
this ratio was not interpreted as the proportion of the explained variation, but as the frac-
tion of the total inertia explained by the model. A stepwise algorithm was used also to
build a parsimonious model that combined both topographic and soil variables.
Moving window CCA: quantifying interaction between aspect and elevation
Preliminary analyses indicated that species composition mainly varies along two gradi-
ents, directly influenced by the topographic position in the valley – relative elevation
above the valley bottom and aspect. However, aspect may have a different effect on vegeta-
tion in deeper, shaded parts of the valley, where it plays a less important role than in the up-
per parts, where the contrast in irradiation between north-facing and south-facing slopes is
much more pronounced. To test this hypothesis, we proposed a method inspired by the
210 Preslia 79: 205–222, 2007
moving window regression analysis (e.g. Walker et al. 2003, Palmer 2006), which was
originally designed to detect changes in vegetation composition along transects. However,
our analysis did not employ linear regression, but CCA with one explanatory variable,
which analyzed the changes in explanatory power of this variable along a gradient of an-
other variable. We call this method “moving window CCA”. In our case, the method was
used to quantify changes in the explanatory power of aspect when moving from the bottom
to the upper edge of the valley. Plots were sorted by their relative elevation above the val-
ley bottom (from 0 to 1) and a virtual moving window was set at the beginning of this se-
ries. The window then moved by steps of constant length toward the opposite end of the
relative elevation interval (elevation). In each step, CCA analysis of the plots included in
the window, with aspect as an explanatory variable, was calculated to quantify the amount
of variation explained by aspect at particular elevations above the valley bottom, measured
by the fraction of total inertia explained by the first axis of CCA. The size of the window
and hence the gradient length was kept constant in all steps, which resulted in different
numbers of plots being included in the window in particular steps. However, to make the
analyses of all steps comparable (in the sense of variation explained by aspect), it was es-
sential to keep constant the number of plots in each analysis. This was done by random se-
lection (without replacement) of a constant number of plots in each particular step within
the virtual window. This random selection was repeated 20 times and averaged fractions of
total inertia together with confidence intervals were plotted against the relative elevation.
Generally, the shape of the analyzed relationship depends on the gradient length (or size of
the window) analyzed in each step, which corresponds to the scale of the studied relation-
ship. To make the results comparable, this parameter was kept the same in both valleys.
After several pre-analysis runs, the size of the window was set to 0.35 units of relative ele-
vation, the number of steps of the window towards the end of the elevation gradient to 20
and number of plots randomly selected per window and used in CCA in a particular step to
17 in both valleys. To visualize the trend, the averages of the explained variation were
smoothed by a curve fitted using a general additive model with three degrees of freedom
(Hastie & Tibshirani 1990).
Software
TURBOVEG 2 database program (Hennekens & Schaminée 2001) was used for storing
vegetation data, JUICE 6.3 (Tichý 2002) for data editing and calculation of Ellenberg indi-
cator values and PC-ORD 4 (McCune & Mefford 1999) for processing cluster analysis.
The calculation routine for moving window CCA analysis was written in R language and
run in R software (R Development Core Team 2005) with Vegan package (Oksanen et al.
2006). R software was used also for calculating and drawing NMDS and CCA ordinations.
Results
Vegetation types and their ecological relationships
Differentiation of the vegetation types in the river valleys is illustrated in Fig. 2, which
combines the results of cluster analysis and NMDS ordination with passively projected
Ellenberg indicator values (see Table 2 for explanation of vegetation types). Number of
Zelený & Chytrý: Environmental control of vegetation in deep river valleys 211
212 Preslia 79: 205–222, 2007
Fig. 2. – Non-metric multi-dimensional scaling (NMDS) ordination diagrams of vegetation plots from the Vltava
and Dyje valleys with projected cluster membership (1–5; see Table 2 for cluster descriptions). Each spider con-
nects individual plots with the averagescore for plots belonging to the same cluster. Ellenberg indicator values for
LIGHT, TEMPerature, CONTinentality, MOISTure, soil REACTion and NUTRients are passively projected
onto these ordination diagrams.
(A) Vltava (B) Dyje
Fig. 3. – “Iris diagrams” showing distribution of particular vegetation types (clusters) in idealized space of the
deep river valleys. Diagrams combine aspect and relative elevation above the valley bottom in the following way:
central circle represents valley bottom, outer margin represents upper edge of the valley and the direction from the
centre represents the direction, in which the slopes face (see the scheme in the middle). Point types refer to the
vegetation types described in Table 2.
Zelený & Chytrý: Environmental control of vegetation in deep river valleys 213
Table 2. – Brief description of clusters derived from cluster analysis of the Vltava and Dyje vegetation plot data,
including number of plots in each cluster, average values ± S.D. of selected environmental variables (slope, soil
pH, soil depth) and two most frequently occurring soil types (see Table 1 for abbreviations).
Cluster Vegetation characteristics No. of
plots
Slope
(°)
pH Soil
depth
(cm)
Two most frequent
soil types
Vltava
1 thermophilous oak forests
(Quercus petraea, Q. robur)
6 35±8 4.4±0.5 22±8 cambisol/lithic
2 acidophilous pine and oak forests
(Pinus sylvestris, Quercus petraea)
24 41±18 3.8±0.2 16±9 cambisol/lithic
3 ravine and oak-hornbeam forests
(Acer, Tilia, Quercus petraea)
29 39±10 4.2±0.4 23±8 cambisol/skeletic
4 fir forests (Abies alba) 18 35±8 4.1±0.3 32±10 cambisol/skeletic
5 alluvial alder forests (Alnus glutinosa) 17 26±25 4.6±0.4 33±17 fluvisol/skeletic
Dyje
1 thermophilous oak forests (Quercus
petraea)
14 36±8 4.5±0.4 21±9 cambisol/lithic
2 acidophilous pine and oak forests
(Pinus sylvestris, Quercus petraea)
17 39±13 4.1±0.2 22±14 cambisol/lithic
3 ravine and oak-hornbeam forests (Acer,
Tilia, Carpinus betulus, Quercus petraea)
25 35±12 5.0±0.7 28±16 cambisol/skeletic
4 beech forests (Fagus sylvatica) 19 33±6 5.0±0.5 40±9 cambisol/skeletic
5 alluvial alder forests (Alnus glutinosa) 7 10±6 5.2±0.6 65±27 fluvisol/skeletic
Fig. 4. – Joint NMDS ordination of all plots from the Vltava and Dyje valleys. Each spider connects individual
plots with the average score for plots from each valley. Ellenberg indicator values are passively projected onto this
ordination diagram (for abbreviations see Fig. 2).
clusters was arbitrarily set to five in both valleys. There are corresponding patterns in both
valleys, with major vegetation types similarly scattered in the ordination diagrams. The
first axes show strong correlations with EIV for moisture, nutrients and soil reaction,
whereas the second axes correlate with light, temperature and continentality (although not
so clearly in the Dyje valley). Central position in both ordination diagrams is occupied by
ravine and oak-hornbeam forests (Cluster 3). The most dry, nutrient-poor, light and warm
habitats are occupied by thermophilous oak forests (Cluster 1). More acidic and cooler
habitats support acidophilous pine and oak forests (Cluster 2). The opposite part of the or-
dination diagrams, with wet and nutrient-rich habitats, is occupied by alluvial forests
(Cluster 5). The vegetation in the coolest and most shady habitats, on the north-facing
slopes, is in Cluster 4 and occupies similar habitats in both valleys, but with different spe-
cies composition: in the Vltava valley, this cluster includes ravine forest dominated by fir
with the tall forb Lunaria rediviva dominating the herb layer, whereas in the Dyje valley it
is represented by beech forests. The spatial pattern of the distribution of particular vegeta-
tion types in idealized space of river valley is presented in “iris diagrams” (Fig. 3). The dif-
ference between the vegetation in the two valleys is shown in Fig. 4, with the Dyje valley
being generally warmer.
Correlations among explanatory variables and Ellenberg indicator values
Distribution of soil types strongly depends on topographic features (Table 3): fluvisols and
(hyper)skeletic leptosols are found in the lower and bottom parts of the valleys, while
lithic leptosols and cambisols are confined to the middle and upper slopes. Lithic leptosols
are shallow soils with low pH and are restricted to steep, upward convex and sun-exposed
slopes. Fluvisols in the floodplain are deeper and less acid and together with hyperskeletic
leptosols of stony screes occupy concave landforms. Elevation is strongly associated with
soil reaction, with base-rich soils in the lower parts of the valleys. Slope is also negatively
correlated with pH, with more acidic soils found on steeper slopes.
Correlations between Ellenberg indicator values and explanatory variables were (in
contrast to the correlation of explanatory variables with one another) calculated for sepa-
rate data sets from each river valley (Table 3). Even though the results are generally consis-
tent between valleys, they show some regional differences. EIVs are closely associated
with topography: sites with warmer aspects and higher heat index values are positively
correlated with EIVs for light, temperature and continentality, and negatively correlated
with EIVs for moisture and nutrients. The bottom of the Vltava valley is cold and shaded
(in terms of EIV) and in both valleys the bottom is more wet, basic and nutrient-rich. Con-
vex topography and slope are negatively correlated with moisture, soil reaction and nutri-
ent availability. Soil variables also correlate with several EIVs: fluvisols are wet, basic and
nutrient-rich; lithic leptosols are dry, acidic and nutrient-poor; hyperskeletic leptosols are
wetter and richer in nutrients. Soil depth in both valleys is negatively correlated with EIVs
for light and temperature, and positively with moisture, soil reaction and nutrients. In both
valleys measured soil pH is strongly positively correlated with EIVs for soil reaction,
moisture and nutrients; only in the Dyje valley is pH negatively correlated with EIVs for
light and temperature.
214 Preslia 79: 205–222, 2007
Effect of topographic and soil variables on vegetation
Table 4 shows the results of direct ordination analyses, processed separately for data from
each river valley and each set of topographic and soil explanatory variables. The most par-
simonious model (based on AIC), including only topographic variables, explains 10.3%
of total inertia in the Vltava valley (V.topo) and 12.0% in the Dyje valley (D.topo), while
models including only soil variables explain slightly more – 12.5% in the Vltava valley
(V.soil) and 14.8% in the Dyje valley (D.soil). Partial CCA revealed conditional and shared
effects of these models (Fig. 5). Full models, including all topographic and soil variables
selected by previous topographic and soil models, explain 18.8% in the Vltava (V.full) and
21.6% in the Dyje valley (D.full). However, these models are not parsimonious, as mea-
sured by the AIC criterion. Parsimonious models including both topographic and soil vari-
ables (V.parsim and D.parsim, respectively) include only four out of the seven explanatory
variables included in the full models and explain 13.8% in the Vltava valley (V.parsim) and
16.8% in the Dyje valley (D.parsim).
Zelený & Chytrý: Environmental control of vegetation in deep river valleys 215
Table 3. – Correlations between all explanatory variables and correlations between explanatory variables and
Ellenberg indicator values (the latter calculated separately for the Vltava and Dyje valley). Spearman rank coefficients
are used for all correlations except for soil type categories (fluvisol, skeletic, cambisol and lithic), for which the chi-
square test is used. Positive correlation: +++ P < 0.001, ++ P < 0.01, + P < 0.05; negative correlation: –––P<0.001,
– – P < 0.01, – P < 0.05. Meaning of P: for correlation of categorical soil variables – significance of Pearson’s chi-
squared test with Yates’ continuity correction derived from a contingency table; for the others – significance of
Spearman rank coefficient. For definitions of environmental variables, see Table 1.
Elevation x
Slope n.s. x
Aspect n.s. n.s. x
Heat index n.s. n.s. +++ x
Surface SL +++ +++ n.s. n.s. x
Surface ISO ++ n.s. n.s. n.s. +++ x
Soil depth – – – – – – – – – – – – n.s. x
pH – – – – – n.s. n.s. – n.s. +++ x
Fluvisol – – – – – – n.s. n.s. – – – n.s. +++ +++ x
Skeletic – – – n.s. n.s. n.s. – – – –-- n.s. + + x
Cambisol +++ n.s. n.s. n.s. n.s. n.s. n.s. – – – – – – x
Lithic +++ +++ +++ +++ +++ n.s. – – – – – – – – – – – n.s. x
Elevation Slope Aspect Heat
index
Surface
SL
Surface
ISO
Soil
depth
pH Fluvisol Skeletic Cambisol Lithic
Vltava
Light n.s. n.s. +++ +++ n.s. + – – – n.s. n.s. – – n.s. +
Temperature n.s. n.s. +++ +++ n.s. n.s. – n.s. n.s. n.s. n.s. +
Continentality n.s. n.s. +++ +++ n.s. n.s. – – n.s. n.s. n.s. n.s. n.s.
Moisture – – – – – – – – – – – – – – – n.s. +++ +++ +++ +++ – – – –
Reaction – – – – n.s. n.s. – – – n.s. +++ +++ +++ + n.s. – – –
Nutrients – – – – – – – – – – – – – n.s. +++ +++ +++ +++ n.s. – – –
Dyje
Light +++ n.s. +++ +++ n.s. n.s. – – – – – – n.s. n.s. n.s. +++
Temperature + n.s. +++ +++ n.s. n.s. – – – – – n.s. – n.s. ++
Continentality – – – n.s. +++ +++ n.s. n.s. n.s. ++ + n.s. n.s. n.s.
Moisture – – – – – – – – – – – – – – +++ +++ +++ ++ – – – – – –
Reaction – – – – – – + n.s. – n.s. +++ +++ +++ n.s. n.s. – –
Nutrients – – – – – – n.s. – – – – n.s. +++ +++ +++ ++ – – – – –
Relationship between vegetation, elevation and aspect: moving window CCA
Although elevation and aspect are not correlated, moving window CCA revealed that the
explanatory power of aspect changes at different elevations above the valley bottom
(Fig. 6). Explanatory power of aspect is lowest near the valley bottom, reaches the maxi-
mum half way up the side of the valley and decreases again near the top. Fractions of total
inertia explained by aspect in particular steps range between 7–13% in the Vltava and
9.5–13.0% in the Dyje valley.
Discussion
Indirect ordination revealed that in both river valleys, the main gradients in vegetation
composition are similar – the first NMDS axis represents a complex nutrient-moisture-soil
reaction gradient and the second a light-temperature-continentality gradient. Distribution
of vegetation types determined by cluster analysis along these gradients also displays sim-
ilar patterns in both valleys, both in ordination space (Fig. 2) and the idealized spatial
model of a deep valley (Fig. 3). This degree of similarity between the two river valleys
accords with the results of previous phytosociological studies from deep river valleys (e.g.
Blažková 1964, Türk 1994, Chytrý 1995, Chytrý & Vicherek 1995, 1996, Kolbek et al.
1997, Kolbek 1999–2003) and suggests that the patterns described in this study are
general for river valleys of the Bohemian Massif, and not a result of local coincidences of
vegetation and environment. Vegetation patterns in the Vltava and Dyje valleys are similar
in spite of the fact that the former is situated in a cooler and wetter macroclimatic region
than the latter (Fig. 4).
This analysis shows that soil and topographic variables are both good predictors of vegeta-
tion pattern in river valleys, but the former explains slightly more variation. While topo-
graphic variables can be derived from high resolution digital elevation maps, soil variables
need detailed field inspection, which is more time and budget demanding. If money or time
are limiting factors, topographic variables itself, such as elevation above the valley bottom,
aspect and slope (or landform shape), can be still considered as good predictors of vegeta-
tion pattern (Tichý 1999a). These variables have no direct effect on plants, but exert a strong
control on the distribution of resources and conditions necessary for plant growth, such as
216 Preslia 79: 205–222, 2007
Fig. 5. – Venn diagrams showing conditional and shared effects of the groups of topographic and soil variables as
fractions of the total inertia.
moisture availability, nutrients or temperature (Pabst & Spies 1998). Aspect and elevation
determine mesoclimatic conditions such as incoming solar radiation (Austin et al. 1984) or
formation of temperature inversions in river valleys (Quitt 1996, Chytrý & Tichý 1998,
Tichý 1999b). Slope is closely related to disturbance, caused by falling rocks, soil creep, sur-
face erosion etc. (Rejmánek et al. 2004). The down slopeincrease in soil pH revealed in this
study is probably also connected with down slope mass and nutrient migration, induced both
by groundwater flow (Campbell 1973, Zinko et al. 2006) and superficial erosion (Cox et al.
2002), causing increased leaching of the upper slopes, followed by transport and accumula-
tion of soluble base cations in the lower parts of the valley (Silver et al. 1994, Chen et al.
1997). Surface erosion is perhaps also responsible for the negative correlation between soil
pH and slope, as steep slopes on acidic bedrock support the development of shallow soils
with an acidic reaction. Apart from this, nutrient accumulation in the lower parts of the val-
leys is connected with flooding (in the case of fluvisols) or more intensive microbial activity
in the highly skeletic soils of ravine forests on the lower slopes (Ellenberg 1996).
Zelený & Chytrý: Environmental control of vegetation in deep river valleys 217
Fig. 6. – Joint effect of aspect and relative elevation above the valley bottom on species composition, analyzed us-
ing moving window CCA for the Vltava and Dyje river valleys. Horizontal axis is relative elevation above the val-
ley bottom. The bottom parts of the diagrams show distributionof relative elevation (the points are slightly jittered
along the vertical line to visualize overlapping values). The middle parts demonstrate the position of each moving
window along the gradient of elevation (triangle shows the mean of all the values in a particular window, horizon-
tal bar shows the range of relative elevation in a particular window). The upper part shows the explainedvariation
(the fraction of total inertia explained by aspect), calculated by CCA for a given step; for each step of the moving
window, mean (triangle) and confidence interval (vertical bar) of the explainedvariation, calculated using the ran-
dom sub-samples, are shown; the positions of triangles along the horizontal axis correspondin the middle and up-
per part of the diagram.
Table 4. – CCA models with various combinationsof explanatory variables and covariables. Total inertia: Vltava
= 7.144, Dyje = 7.898. See Table 1 for variable definitions. AIC = value of (generalized) Akaike informationcri-
terion; Σeig. = sum of all canonical eigenvalues; % expl. = fraction of total inertia explained by the model. All
models (excluding conditional effect models,which have not been tested) give significant results of Monte Carlo
permutation test (P < 0.001, 1000 permutations). Model abbreviations: V.topo, D.topo – explanatory variables in-
cluding topographic factors only for Vltava and Dyje valleys, respectively; V.soil, D.soil – explanatory variables
including soil factors only; V.topo.cond, D.topo.cond – conditional effects of topographic variables with soil vari-
ables as covariables; V.soil.cond, D.soil.cond – conditional effects of soil variables with topographic variables as
covariables; V.full, D.full – combines topographic and soil variables from V.topo, D.topo and V.soil, D.soil;
V.parsim, D.parsim – the most parsimonious models including both topographic and soil variables.
Explanatory variables Covariables AIC Σeig. % expl.
Vltava
V.topo elevation + aspect + surface SL – 447.00 0.733 10.3
V.soil skeletic + fluvisol + soil depth + pH – 446.59 0.896 12.5
V.topo.cond elevation + aspect + surface SL skeletic + fluvisol + soil depth
+ pH
0.452 6.3
V.soil.cond skeletic + fluvisol + soil depth + pH elevation + aspect + surface SL 0.615 8.6
V.full elevation + aspect + surface SL +
skeletic + fluvisol + soil depth + pH
– 1.347 18.8
V.parsim aspect + fluvisol + soil depth + pH – 445.21 0.987 13.8
Dyje
D.topo elevation + aspect + slope – 384.42 0.945 12.0
D.soil cambisol + fluvisol + soil depth + pH – 383.78 1.166 14.8
D.topo.cond elevation + aspect + slope cambisol + fluvisol +
soil depth + pH
0.537 6.8
D.soil.cond cambisol + fluvisol + soil depth + pH elevation + aspect + slope 0.758 9.6
D.full elevation + aspect + slope + cambisol
+ fluvisol + soil depth + pH
– 1.703 21.6
D.parsim aspect + cambisol + fluvisol + pH – 381.83 1.324 16.8
Due to the complex topography of the valleys, the effects of some topographic vari-
ables on vegetation pattern are not easy to identify. In particular, the non-linear interaction
between the elevation above the valley bottom and aspect can mask the effect of the latter
when standard procedures of constrained ordination are used. The new method of moving
window CCA, proposed here, proved successful in disentangling the complex effect of
these two variables on vegetation (Fig. 6). It clearly showed that near the valley bottom,
where the valley is rather narrow and shaded by the adjacent slopes in many places, aspect
does not explain much of the variation in vegetation. Moving up the valley sides, the im-
portance of aspect as a determinant of species composition increases, because of the more
pronounced contrast between the dry and warm south-facing and more shaded, wetter and
cooler north-facing slopes. At the upper edges of the valley, the importance of aspect de-
creases again, as the difference in insolation of south-facing and north-facing slopes di-
minishes due to the less steep topography.
218 Preslia 79: 205–222, 2007
Despite the strong correlations between several topographic and soil variables (Table 3),
the shared fraction of variation in species composition explained by both topographic and
soil variables is relatively low (Fig. 5). It means that soil variables explain a different part of
the variability in vegetation than topography. Therefore, recording several simple soil vari-
ables, such as pH, soil depth and soil type, can significantly improve the explanatory power
of vegetation-environment models, even in a landscape with strong topographic contrasts.
Conclusions
The similarity of the vegetation patterns between the two river valleys studied and their
correspondence with the earlier phytosociological studies indicate that the patterns re-
vealed in the present study are reasonably robust and can be generalized for most deep
river valleys on non-calcareous soils at middle altitudes of the Bohemian Massif. Main
topographic factors driving vegetation pattern are elevation above the valley bottom, as-
pect (being more important half way up the valley sides) and slope. Soil variables such as
measured pH and soil type (mainly fluvisols vs the others) may significantly improve veg-
etation-environment models for these valleys. The vegetation pattern of the valleys can be
briefly summarized as follows:
(1) Floodplain forests, mostly dominated by Alnus glutinosa, occur on the valley bot-
tom on deep and moist fluvisols, which are rich in nutrients and have a relatively high pH.
(2) On lower valley slopes, there is usually a small difference in the vegetation on
south-facing and north-facing slopes. Here the main factor is slope, which determines
whether cambisols (on less steep slopes) or skeletic leptosols (on steeper slopes) develop,
with the former supporting oak-hornbeam forests and the latter ravine forests of Acer,
Tilia and Carpinus betulus (Carpinus being locally absent in the Vltava valley).
(3) Half way up the valley sides and to a lesser extent further up, there is a striking con-
trast between the vegetation on southern and northern slopes. The warmest south-facing
slopes support thermophilous oak forests with Quercus petraea, which are better devel-
oped in the warmer and more continental Dyje valley. In contrast, north-facing slopes sup-
port forests dominated by fir (Vltava) or beech (Dyje) on relatively deep and nutrient-rich
cambisols.
(4) Other habitats in deep river valleys are covered with acidophilous oak and/or pine
forests on more or less shallow lithic leptosols of various aspects and pure pine stands re-
stricted to extreme habitats on rocky outcrops.
Acknowledgements
The helpful comments of two anonymous referees are appreciated. Tony Dixon kindly improved our English.
This study was supported by the long-term research plan MSM 0021622416.
Souhrn
Cílem této studie je kvantitativní popis faktorů prostředí, které zásadním způsobem ovlivňují druhové složení
a prostorové rozmístění vegetace v hlubokých říčních údolích Českého masivu s vyvinutými projevytzv. „říčního
fenoménu“. Problematika byla studována ve dvou klimaticky odlišných územích: údolí Vltavy v jižních Čechách
a údolí Dyje na jižní Moravě. Data o vegetaci a proměnných prostředí byla sbírána na transektech vedených po
Zelený & Chytrý: Environmental control of vegetation in deep river valleys 219
spádnici údolních svahů z horní hrany údolí k bázi svahu. Vegetační data byla analyzována kombinací shlukové
analýzy a nepřímé ordinace (nemetrického mnohorozměrného škálování, NMDS). Vliv geomorfologických
a půdních proměnných na vegetaci byl porovnáván sérií kanonických korespondenčníchanalýz (CCA) s metodou
postupného výběru vysvětlujících proměnných založenouna Akaikeho informačním kritériu (AIC). Pro analýzu
vlivu nelineárních interakcí mezi dvěmi proměnnými prostředí na vegetaci byla navržena nová metoda nazvaná
„moving window CCA“. V této studii metoda popsala, jak se mění vysvětlující síla jedné proměnné (orientace
svahu) se změnou druhé proměnné (výšky nad řekou). Hlavní směry variability ve vegetaci jsou v hlubokých říč-
ních údolích korelovány s dvěma komplexními gradienty proměnných prostředí: vlhkost-živiny-půdní reakce
a světlo-teplota-kontinentalita. Přímá ordinační analýza ukázala, že půdní faktory lépe korelují s druhovým slože-
ním vegetace než geomorfologie terénu, přičemž dohromady obě tyto skupiny proměnných vysvětlily
18.8–21.6 % celkové variability v druhovém složení vegetace. Ačkoliv některé půdní a geomorfologické proměn-
né těsně korelují, množství variability vysvětlené sdíleným vlivem obou skupin není příliš vysoké (3.9–5.2 %),
což znamená, že každá skupina proměnných ovlivňuje vegetaci poněkud jiným způsobem. Nejvíce parsimonní
model CCA pro údolí Vltavy vysvětluje druhové složení vegetace pomocí následujících faktorů: orientace svahu,
půdní pH, přítomnost fluvizemě a hloubka půdy; pro údolí Dyje vypadá model podobně, jen faktor hloubka půdy
je nahrazen přítomností kambizemě. „Moving windows CCA“ ukázala, že orientace svahu má na vegetaci vliv
nejvíce ve střední části údolního svahu a nejméněpři bázi svahu.Výsledky všech analýzukazují výraznou shodu
ve vztazích mezi vegetací a prostředím v obou říčních údolích, což naznačuje možnosti zobecnění popsaných
vztahů i na další hluboká říční údolí Českého masivu.
References
Akaike H. (1973): Information theory as an extension of the maximum likelihood principle. – In: Petrov B. N. & Csaki
F. (eds), Second international symposium on information theory, p. 267–281, Akadémiai Kiadó, Budapest.
Austin M. P., Cunningham R. B. & Fleming P. N. (1984): New approaches to direct gradient analysis using envi-
ronmental scalars and statistical curve-fitting procedures. – Vegetatio 55: 11–27.
Batík P. (1992): Geologická mapa Národního parku Podyjí [Geological map of the Podyjí National Park]. –
Český geologický ústav, Praha.
Blažková D. (1964): Rozčlenění vegetace na údolních svazích v oblasti Orlické nádrže [Pattern of distribution of
plant communities in the area of Orlík reservoir]. – In: Jeník J. (ed.), Vegetační problémy při budování
vodních děl [Vegetation issues related to the construction of water reservoirs], p. 21–37, NČSAV, Praha.
Burkart M. (2001): River corridor plants (Stromtalpflanzen) in Central European lowland: a review of poorly un-
derstood plant distribution pattern. – Global Ecol. Biogeogr. 10: 449–468.
Campbell I. B. (1973): Pattern of variation in steepland soils: variation on a single slope. – New Zeal. J. Sci. 16:
413–434.
Chábera S. (1985): Neživá příroda. Jihočeská vlastivěda, přírodní vědy [Abiotic nature. Nature history of South
Bohemia]. – Jihočeské nakladatelství, České Budějovice.
Chen Z. S., Hsieh C. F., Jiang F. Y., Hsieh T. H. & Sun I. F. (1997): Relations of soil properties to topography and
vegetation in a subtropical rain forest in southern Taiwan. – Plant Ecol. 132: 229–241.
Chytrý M. (1995): Are species with similar ranges confined to similar habitats in a lanscape? – Preslia 67: 25–40.
Chytrý M., Grulich V., Tichý L. & Kouřil M. (1999): Phytogeographical boundary between the Pannonicum and
Hercynicum: a multivariate analysis of landscapein the Podyjí/Thayatal National Park, Czech Republic/Aus-
tria. – Preslia 71: 1–19.
Chytrý M. & Tichý L. (1998): Phenological mappingin a topographically complex landscape by combining field
survey with an irradiation model. – Appl. Veg. Sci. 1: 225–232.
Chytrý M. & Vicherek J. (1995): Lesní vegetace národního parku Podyjí/Thayatal [Forest vegetation of the
Podyjí/Thayatal National Park]. – Academia, Praha.
Chytrý M. & VicherekJ. (1996): Přirozená a polopřirozená vegetace údolířek Oslavy, Jihlavy a Rokytné [Natural
and semi-natural vegetation of the Oslava, Jihlava and Rokytná river valleys]. – Přírod. Sborn. Západomor.
Muz. Třebíč 22: 1–125.
Chytrý M. & Vicherek J. (2003): Travinná, keříčková a křovinná vegetace Národního parku Podyjí/Thayatal
[Grassland, heathland, and scrub vegetation of the Podyjí/Thayatal National Park]. – Thayensia 5: 11–84.
Cox S. B., Willig M. R. & Scatena F. N. (2002): Variation in nutrient characteristics of surface soils from the
Luquillo Experimental Forest of Puerto Rico: A multivariate perspective. – Plant Soil 247: 189–198.
Ellenberg H. (1996): Vegetation Mitteleuropas mit den Alpen in ökologischer, dynamischer und historischer
Sicht. Ed. 5. – E. Ulmer, Stuttgart.
220 Preslia 79: 205–222, 2007
Ellenberg H., Weber H. E., Düll R., Wirth V., Werner W. & Paulißen D. (1992): Zeigerwerte von Pflanzen in
Mitteleuropa. Ed. 2. – Scripta Geobot. 18: 1–258.
Geiger R. (1966): The climate near the ground. – Harvard Univ. Press, Cambridge.
Goebel P. C., Pregitzer K. S. & Palik B. J. (2006): Landscape hierarchies influence riparian ground-flora commu-
nities in Wisconsin, USA. – For. Ecol. Manage. 230: 43–54.
Gould W. & Walker M. D. (1999): Plant communitiesand landscape diversity along a Canadian Arctic river. – J.
Veg. Sci. 10: 537–548.
Hastie T. J. & Tibshirani R. J. (1990): Generalized additive models. – Chapman & Hall, London.
Hennekens S. M. & Schaminée J. H. J. (2001): TURBOVEG, a comprehensive data base management system for
vegetation data. – J. Veg. Sci. 12: 589–591.
ISSS-ISRIC-FAO (1998): Worldreference bas e for soil resources. – World Soil Res ources Reports 84, FAO, Roma.
Jeník J. & Slavíková J. (1964): Střední Vltava a její přehrady z hlediska geobotanického [Middle tracts of the
Vltava river and its dams from geobotanical aspects]. – In: Jeník J. (ed.), Vegetační problémy při budování
vodních děl [Vegetation issues related to the construction of water reservoirs], p. 67–100, NČSAV, Praha.
Kolbek J. (ed.) (1999–2003): Vegetace Chráněné krajinné oblasti a Biosférické rezervace Křivoklátsko 1–3 [Veg-
etation of the Protected Landscape Area and Biosphere Reserve Křivoklátsko 1–3]. – Academia, Praha.
Kolbek J., Blažková D., Husová M., Moravec J., Neuhäuslová Z. & Sádlo J. (1997): Potential natural vegetation
of the Biosphere Reserve Křivoklátsko. – Academia, Praha.
Kopecký A. (1996): The part played by denudation in formingthe relief of the Bohemian Massif. – In: Kopecký
A., Loyda L. & Vyskočil P. (eds), Seismicity, neotectonics and recent dynamics with special regard to the Ter-
ritory of Czech Republic, VÚGTK, Zdiby, 42 (15): 55–75.
Kubát K., Hrouda L., Chrtek J. jun., Kaplan Z., Kirschner J. & Štěpánek J. (eds) (2002): Klíč ke Květeně České
Republiky [Key to the flora of the Czech Republic]. – Academia, Praha.
McCune B. & Mefford M. J. (1999): PC-ORD. Multivariate analysis of ecological data, version 4.0.– MjM Soft-
ware Design, Gleneden Beach.
Minchin P. R. (1987): An evaluation of the relative robustness of techniques for ecological ordination. – Vegetatio
69: 89–107.
Mouw J. E. B. & Alaback P. B. (2003): Putting floodplain hyperdiversity in a regional context: an assessment of
terrestrial-floodplain connectivity in a montane environment. – J. Biogeogr. 30: 87–103.
Naiman R. J., Décamps H. & Pollock M. (1993): The role of riparian corridors in maintaining regional
biodiversity. – Ecol. Appl. 3: 209–212.
Økland R. H. (1999): On the variation explained by ordination and constrained ordination axes. – J. Veg. Sci. 10:
131–136.
Oksanen J., Kindt R., Legendre P. & O’Hara R. B. (2006): Vegan: community ecology package version 1.6-10.–
URL: http://cc.oulu.fi/~jarioksa/ [accessed 7 Dec 2006].
Pabst R. J. & Spies T. A. (1998): Distribution of herbs and shrubs in relation to landform and canopy cover in ri-
parian forests of coastal Oregon. – Can. J. Bot. 76: 298–315.
Palmer M. W. (2006): Scale dependence of native and alien species richness in North American floras. – Preslia
78: 427–436.
Parker K. C. (1988): Environmental relationships and vegetation associates of columnar cacti in the northern
Sonoran Desert. – Vegetatio 78: 125–140.
Planty-TabacchiA. M., Tabacchi E., Naiman R. J., Deferrari C. & Décamps H. (1996): Invasibility of species-rich
communities in riparian zones. – Cons. Biol. 10: 598–607.
Pyšek P. & Prach K. (1993): Plant invasions and the role of riparian habitats: a comparison of fourspecies alien to
central Europe. – J. Biogeogr. 20: 413–420.
Quitt E. (1996): Změny mikroklimatu a topoklimatu způsobené výstavbou vodních nádrží Dalešice a Mohelno
[Changes in microclimate and topoclimate due to the construction of the Dalešice and Mohelno water reser-
voirs]. – Přírod. Sborn. Západomor. Muz. Třebíč 21: 1–26.
R Development Core Team (2005): R: a language and environment for statistical computing. – R Foundation for
Statistical Computing, Vienna, URL: http://www.r-project.org [accessed 7 Dec 2006].
Rejmánek M., Rejmánková E. & Holzner W. (2004): Species diversityof plant communities on calcareous screes:
the role of intermediate disturbance. – Preslia 76: 207–222.
Sagers C. L. & Lyon J. (1997): Gradientanalysis in a riparian landscape: contrast among forest layers. – For. Ecol.
Manage. 96: 13–26.
Silver W. L., Scatena F. N., Johnson A. H., Siccama T. G. & Sanchez M. J. (1994): Nutrient availability in
a montane wet tropical forest: spatial patterns and methodological considerations.– Plant Soil 164: 129–145.
Zelený & Chytrý: Environmental control of vegetation in deep river valleys 221
Sokal R. R. & Rohlf F. J. (1995): Biometry: the principles and practice of statistics in biological research. Ed. 3. –
W. H. Freeman and Co., New York.
ter Braak C. J. F. (1986): Canonical correspondence analysis: a new eigenvector technique for multivariate direct
gradient analysis. – Ecology 67: 1167–1179.
Tichý L. (1999a): Predictive modeling of the potential natural vegetation pattern in the Podyjí National Park,
Czech Republic. – Folia Geobot. 34: 243–252.
Tichý L. (1999b): Teplotní poměry a vegetace na lokalitě Sloní hřbet v Národním parku Podyjí [Temperature and
vegetation pattern at the Sloní hřbet Ridge, Podyjí National Park]. – Preslia 70 (1998): 349–363.
Tichý L. (2002): JUICE, software for vegetation classification. – J. Veg. Sci. 13: 451–453.
Tichý L. & Chytrý M. (2006): Statistical determination of diagnostic species for site groups of unequal size. – J.
Veg. Sci. 17: 809–818.
Tolasz R. (ed.) (2007): Atlas podnebí Česka [Climate atlas of Czechia]. – ČHMÚ, Univerzita Palackého
v Olomouci, Praha & Olomouc.
Türk W. (1994): Das “Höllental” im Frankenwald – Flora und Vegetation eines floristisch bemerkenswerten
Mittelgebirgstales. – Tuexenia 14: 17–52.
van Coller A. L., Rogers K. H. & Heritage G. L. (2000): Riparian vegetation-environment relationships:
complimentarity of gradients versus patch hierarchy approaches. – J. Veg. Sci. 11: 337–350.
Walker S., Wilson J. B., Steel J. B., Rapson G. L., Smith B., King W. Mc. G. & Cottam Y. H. (2003): Properties of
ecotones: evidence from five coastal ecotones. – J. Veg. Sci. 14: 579–590.
Westhoff V. & van der Maarel E. (1978): The Braun-Blanquet approach. – In: Whittaker R. H. (ed.), Classifica-
tion of plant communities, p. 289–399, W. Junk, The Hague.
Wiens J. A. (2002): Riverinelandscapes: taking landscape ecology into the water. – Freshw. Biol. 47: 501–515.
Zinko U., Dynesius M., Nilsson C. & Seibert J. (2006): The role of soil pH in linking groundwater flow and plant
species density in boreal forest landscapes. – Ecography 29: 515–524.
Received 6 February 2007
Revision received 13 May 2007
Accepted 18 May 2007
222 Preslia 79: 205–222, 2007
