The flora of German cities is naturally species rich
Ingolf Kühn,1* Roland Brandl2 and Stefan Klotz1
1Department of Community Ecology, Centre for Environmental Research Leipzig-Halle,
Theodor-Lieser-Str. 4, 06120 Halle and 2Department of Animal Ecology,
University of Marburg, Karl-von-Frisch-Straße, 35032 Marburg, Germany
Previous studies on various scales and for various European regions and North America have
shown that cities harbour more plant species than the surrounding landscape. It has been
argued that the greater number of plant species is usually caused by a high number of alien
plants promoted by human influence. We analysed native and naturalized vascular plant species
distribution data from a comprehensive German database comparing city and non-city grid
cells of 10 minutes latitude ×6 minutes longitude (c. 130 km2). The number of city grid cells
(n=68) and non-city grid cells (n=1856) differed by two orders of magnitude and species
richness was highly autocorrelated. We therefore used resampling techniques. We resampled the
species richness of 68 randomly selected grid cells 9999 times. This showed that not only
naturalized alien but also native plant species richness was significantly higher in city grid cells.
To relate environmental variables to species richness, we used 10,000 analyses of covariance
of 68 city grid cells and 68 randomly selected non-city grid cells. We demonstrated that a
large proportion of the higher native plant species richness could be explained by the number of
geological types per grid cell (i.e. a measure of natural geological diversity). Additionally, we
showed by resampling the number of geological types per grid cell that cities are not randomly
distributed but are in fact in areas of high geological diversity. Hence, we conclude that city
areas are preferentially located in pre-existing biodiversity hotspots and argue that they are
species rich not because of but in spite of urbanization.
Keywords: alien plants, environmental correlates, environmental heterogeneity, native plants,
resampling methods, species richness, urbanization.
Across the globe and on various scales, strikingly similar patterns of correlation between
human population density and species richness have been recorded. On the regional scale,
Walters (1970) was the first to point out that cities harbour more spontaneous (not
cultivated) plant species than the surrounding landscape. This pattern was confirmed on
several scales and for various European regions (Haeupler, 1975; Klotz, 1990; Pysˇek, 1993;
Kowarik, 1995; Araújo, 2003; but see Roy et al., 1999) and North America (Dobson
et al., 2001; McKinney, 2002a; Hope et al., 2003).
* Author to whom all correspondence should be addressed. e-mail: firstname.lastname@example.org
Consult the copyright statement on the inside front cover for non-commercial copying policies.
Evolutionary Ecology Research, 2004, 6: 749–764
© 2004 Ingolf Kühn
This pattern is even more striking considering how changes and developments in human
land use alter ecosystems. Such land-use changes induce changes in a variety of factors
(e.g. Sukopp and Werner, 1983; Gilbert, 1989; Wittig, 1991; Collins et al., 2000; Pickett
et al., 2001), including increases in temperature (‘heat effect’ in the city) and rainfall
(Landsberg, 1981; Oke, 1982), pollution (Douglas, 1983), impervious surfaces, habitat
fragmentation and disturbance (Kowarik, 1995; Trepl, 1995). However, fragments of
semi-natural vegetation and agricultural land within or surrounding the cities contribute
to higher biodiversity.
It is often argued that the majority of alien species establish in cities by means of culti-
vation (Barthlott et al., 1999; Kent et al., 1999). Thus, the higher species richness in cities
could mainly be ascribed to alien species (e.g. McKinney and Lockwood, 2001). Plants that
are adapted to stress or human land use and occur preferably within cities were termed
‘urban specialists’ by Hill et al. (2002). Such species could increase urban plant species
richness (Pysˇek, 1998) further. Similarly, disturbance or perturbation may lead to increased
plant species richness at intermediate levels (Sukopp and Trepl, 1987; Kowarik, 1990, 1991,
1995; Pysˇek, 1993; see Huston, 1994, for a general framework). Additionally, because
botanical research institutes tend to be more often in big cities, some have argued that the
species richness of cities is a sampling artefact (Barthlott et al., 1999). Further general issues
are the effect of area size on species richness (e.g. Rosenzweig, 1995) and the influence of
spatial autocorrelation on statistical analyses (e.g. Fortin et al., 1989; Lichstein et al., 2002).
These can cause severe problems but have largely been neglected by previous analyses.
Previous studies focused primarily on the increased richness of alien plant species due to
human influence. However, richness patterns of plant species correlate, for example, with
productivity (e.g. Waide et al., 1999; Mittelbach et al., 2001), landscape heterogeneity
(e.g. Wohlgemuth, 1998; Deutschewitz et al., 2003; Kühn et al., 2003), climate
(e.g. Wohlgemuth, 1998) and other ‘natural’ factors, which, in turn, may covary with human
population density or settlements. As natural factors are important determinants for plant
species richness and as naturalized aliens account usually for just a small fraction of
urban plant species richness in Germany (e.g. Kowarik, 1995; Pysˇek, 1998), we focus on the
proportions of natural and man-made factors as correlates for native and naturalized alien
plant species richness. From previous studies we know that natural factors such as
geological diversity, soil diversity and landscape diversity are positively correlated with
native and alien plant species richness (Deutschewitz et al., 2003; Kühn et al., 2003).
Thus, we hypothesize that these natural factors are not only the most important correlates
of plant species richness but are also relevant factors for the location of cities.
MATERIALS AND METHODS
Species numbers were calculated from the database on German flora (FLORKART, see
http://www.floraweb.de), maintained by the German Centre for Phytodiversity at the
Federal Agency for Nature Conservation (Bundesamt für Naturschutz, BfN). FLORKART
is the central database that includes all provincial or regional mapping schemes throughout
Germany. The data were mainly assembled by thousands of volunteers who recorded the
flora of their respective grid cells. This work was initiated by a mapping scheme for Central
Europe (Ehrendorfer and Hamann, 1965; Ellenberg et al., 1968), which unfortunately could
Kühn et al.750
not be realized in the desired manner. Nevertheless, a multitude of more regional mapping
schemes led to distribution atlases for West Germany (Haeupler and Schönfelder, 1989) and
East Germany (Benkert et al., 1996). Both atlases, many publications on local or regional
floras from the nineteenth century and the first half of the twentieth century, several very
recent mapping schemes and corrections were incorporated into the recent version of
FLORKART. We choose to use FLORKART for our analysis because it contains more
than 14 million records, of which 2.6 million are unique for each species in a 6⬘×10⬘ grid
cell in the most recent time period, making it one of the most comprehensive European
databases on plant species distribution. Therefore, we analysed this database. In our version
of FLORKART, the time of recording is referenced as three time periods: before 1950, 1950
until 1979 and from 1980 (until 2001). We used the records from 1950 onwards. Localities
of plant species are referenced within this database according to the German 1:25,000
Ordnance Survey Maps (10 minutes latitude ×6 minutes longitude, c. 130 km2, hereafter
called ‘grid cells’). Species numbers were the sums of all occurrences of both natives
and naturalized aliens (thus omitting casual and cultivated occurrences, i.e. in gardens,
arboretums, managed road verges, etc.).
Mapping intensity in Germany is very heterogeneous. Hence, only sufficiently well
mapped grid cells were used for analysis. Mapping intensity was evaluated by designating
50 ‘control’ species. These control species were the 45 most frequent species mentioned
by Krause (1998) and five additional species. The latter five species were generalists and
preferably inconspicuous or regarded by many volunteers as difficult to identify. By this, we
wanted to minimize potential observer bias towards obvious, easily identifiable species. All
50 control species had to be present to include a grid cell in the analysis. This control
reduced the number of grid cells from 2995 to 1928. These control species were omitted
from further analyses to avoid circular reasoning. However, as the number of control
species is a constant in all grid cells under analysis, this does not change the general
outcome. We also carried out preliminary tests to establish if we could work with less than
50 or up to 100 species. Using less than 50 species (e.g. 30 or 40) included too many grid cells
not sufficiently well mapped (since we knew at least some regions that were poorly mapped).
Having more than 50 species (e.g. 80 or 100) led partially to the exclusion of well mapped
regions as the additional set of species included less common ones.
Despite this control mechanism, it was impossible to use data from smaller grid cells.
Though there are some regions that have been extremely well mapped in the past on the
scale of a quarter ordnance survey map (5⬘×3⬘) or smaller, such as Saxony (Hardtke and
Ihl, 2000), Thuringia (Korsch et al., 2002), Westphalia (Haeupler et al., 2003) and
the Dessau region (see Deutschewitz et al., 2003), the chosen grid cell size for this study was
the only one with a reasonably good coverage and quality for the whole of Germany.
We distinguished between native species (native to Germany) and alien species (species
not native to Germany). Alien species were divided into pre-1500 aliens (so-called
archaeophytes, promoted by agriculture from the Neolithic prior to the discovery of the
Americas) and post-1500 aliens (so-called neophytes, introduced due to long-distance
transport subsequent to the discovery of the Americas). We designated the immigration
status according to BiolFlor (Kühn and Klotz, 2002). This distinction is established in
Central European botanical research (e.g. Schroeder, 1969; Sykora, 1990; Pysˇek, 1998;
Hill et al., 2002; Pysˇek et al., 2003), as both groups of aliens differ markedly (e.g. Pysˇek
et al., 2002a,b), for example in ecology (the first are mainly species of arable fields, the latter
occupy a wide variety of habitats), mode of introduction (the former immigrated across
The flora of German cities 751
short and medium distances without the assistance of man but into habitats provided by
man, the latter arrived often by human mediated long-distance transport) and thus in
evolutionary history. The combination of the two databases FLORKART and BiolFlor left
3150 species for analyses, of which 2411 are native, 239 are (presumed) pre-1500 aliens
and 500 are post-1500 aliens. Furthermore, we used red list species (n=797 within
FLORKART) as another species group according to the German ‘Red List’ of threatened
and endangered plants (Korneck et al., 1996).
City grid cells are those containing the centres of cities with more than 100,000
inhabitants (Statistisches Bundesamt, 2001). The centres of the German cities could easily
be recognized on the map as the old town (often of medieval origin) with usually a cathedral
and a marketplace in its centre and its often somewhat circular patterns (surrounded
by ancient city walls, ring walls, circular roads, etc). The city of Kassel could not
unambiguously be assigned to a grid cell (the city centre is located in the corners of four grid
cells), and thus the corresponding grid cells were omitted from further analyses.
Rural grid cells were defined as those with less than 5% cover of urban land use according
to the Corine Land Cover maps provided by the Federal Statistical Office of Germany
(Statistisches Bundesamt, 1997). This resulted in 1924 grid cells for Germany, 68 of which
contained city centres and 1099 of which were defined as rural. We calculated the number
of geological patches and geological types according to the Geological Survey Map
(Bundesanstalt für Geowissenschaften und Rohstoffe, 1993) per grid cell. The 241
geological types were aggregated into five substrate classes: lime, sand, loess, clay and
others. The number of natural soil types (n=69) and six natural landscape types (coasts,
valleys, plains, loess landscape, low mountains, high mountains) were calculated per
grid cell according to the 1: 1,000,000 German soil survey map (Bundesanstalt für
Geowissenschaften und Rohstoffe, 1995).
For information on land cover, we used Corine Land Cover maps (Statistisches
Bundesamt, 1997). These land cover types are hierarchically aggregated. The lowest (third)
level contains 34 land cover type classes for Germany (represented by a three-digit code
in the CLC legend, URL: http://reports.eea.eu.int/COR0-part2/en/) and will be called ‘level
3 CLC classes’ hereafter. The highest (first) level aggregates the lower level CLC type classes
into five ‘level 1 CLC classes’ (artificial surfaces, agricultural areas, forests and semi-natural
areas, wetlands and water bodies). From this information, we calculated the number of level
3 CLC patches, the number of level 3 CLC classes and the number of level 1 CLC classes per
The Federal Agency for Nature Conservation provided the respective data in grid cell
format, which were transformed from the digital maps mentioned above on geology, soil
and land cover with polygon-topology.
To relate the native and alien species number of rural and city grid cells to natural and
man-made factors, we used an analysis of covariance (ANCOVA; Crawley, 2002; Quinn and
Keough, 2002). The number of city grid cells and non-city grid cells differed by two orders
of magnitude, there was strong autocorrelation (see Fortin et al., 1989) between a focus grid
cell and the adjacent ones (native plant species of our analysis show a highly significant
Moran’s I>0.04 [P<0.001] up to a Euclidean distance of <22 grid cells) and not all grid
cells with a specific combination of abiotic features (potentially ‘suitable’ for cities) were
Kühn et al.752
actually cities. Hence we used a combination of parametric and resampling methods. We
used the 68 city grid cells and randomly sampled 68 non-city grid cells to perform an
ANCOVA. This was repeated 10,000 times with random resampling of non-city grid cells
each time. We used the species number of natives, pre-1500 aliens and post-1500 aliens as
dependent variables in three analyses. The categorical predictor was ‘city’ versus ‘non-city’
grid cells, while the metric predictors were the number of geological patches, geological
types, natural substrates, natural soil types and natural landscape types, level 3 Corine Land
Cover patches, level 3 Corine Land Cover classes and level 1 Corine Land Cover classes per
grid cell. We modelled the main effects as well as interactions between each of the metric
variables and the categorical variable. We estimated the two-tailed error probability from
the 95% confidence interval of the 10,000 sums of squares per variable of the ANCOVA. To
get a minimum adequate model, we stepwise deleted the least non-significant variable(s)
after one cycle of 10,000 resamples.
We partitioned the variation among natural factors (variables derived from geology
natural soil types and natural landscape types), land use factors (variables derived from
Corine Land Cover) and a city effect (city grid cell, ‘yes/no’) following the method proposed
by Legendre and Legendre (1998). Variation could be partitioned among those fractions
that (i) corresponded exclusively to one (group) of predictor(s), (ii) were joint contributions
of each (group) of predictor(s) in conjunction with the other(s) and (iii) remained
unexplained (see also Quinn and Keough, 2002). Given that there are only two predictor
variables, it is possible to compute three linear models using both predictor variables:
(model i) variable 1 as predictor and variable 2 as co-variable, (model ii) variable 2 as
predictor and variable 1 as co-variable, (model iii) variables 1 and 2 as predictors. The
R2 coefficient of model iii yields the complete explained variation. Similarly, the R2
coefficient of model i (or model ii) is the variation exclusively explained by variable 1 (or 2).
Thus the fraction of joint contributions R2
model iii −(R2
model i +R2
To assess the significance of differences between species richness and the number of
geological types in city grid cells and randomly selected (rural) grid cells, we used the means
of 68 cells resampled 9999 times (‘simple resampling’) and compared this distribution to the
mean of the city grid cells.
We present the results from analyses where we performed both the resampling ANCOVA
and the simple resampling (to compare mean species richness) without replacement. For the
simple resampling, we included the city grid cells into the universe for random sampling.
This analysis was also performed with different resampling techniques (i.e. with and
without replacement, including and excluding city grid cells into the universe for random
selection) and definitions of city grid cells (e.g. >30% urban land use as the definition for
‘city grid cells’). The results were not influenced by these differences.
To regress species richness of red list species versus the number of natives plus pre-1500
aliens (both of which are pooled together in the red list of Korneck et al., 1996), we used a
major axis regression (Legendre and Legendre, 1998). This method was more appropriate
for our data than an ordinary least square regression for two reasons. First, the numbers
of red list species and of native +naturalized pre-1500 alien plant species were in the
same dimension and sampled with the same error structure. Secondly, the error probability
of the major axis regression was assessed after 1000 permutations. Therefore, we did not
infer an error probability from an inflated number of degrees of freedom due to spatial
autocorrelation (see Dutilleul, 1993).
All analyses were performed using R (http://www.r-project.org/).
The flora of German cities 753
First, we checked if the use of control species resulted in a higher proportion of up-to-date
records compared with those grid cells that were omitted due to a lack of control species.
On average, 82% (standard deviation 17.5%) of the data of city grid cells used in this
analysis were from after 1980. Similarly, an average of 80% (standard deviation 23.5%)
of the non-city data were from after 1980. There were no significant differences among
the groups of natives and aliens, or between city and non-city cells. In contrast, the average
proportion of the data from 1980 onwards of those cells that were omitted was 40%
(standard deviation 36%). This differed significantly from the former groups (ANOVA,
A brief overview of mean, quartiles and ranges of plant species richness in city or
non-city grid cells is given by Fig. 1. The average numbers of native species were roughly five
to six times higher than those of pre-1500 aliens and eight to twelve times higher than those
of post-1500 aliens. While the average species number outside city grid cells was always
lower than within city grid cells, their ranges were higher, as was their sample size (68 city
grid cells, 1856 non-city grid cells).
The simple resampling analysis showed that the number of pre-1500 aliens, post-1500
aliens and native plant species were significantly higher in city grid cells than expected by
random or than in rural grid cells (Fig. 2). The mean species number differed significantly
between city grid cells and both randomly sampled grid cells and randomly sampled rural
Fig. 1. Box-and-whisker plots of the species richness of native, pre-1500 alien and post-1500 alien
plant species. White bars represent city grid cells (n=68), grey bars represent non-city grid cells
(n=1856). The figure shows median species richness (lines), 25% to 75% quartiles (boxes) and ranges
(whiskers). Open circles are shown if extreme values are more than 1.5 times the interquartile range of
Kühn et al.754
grid cells for all groups (all P<0.001) (Table 1). Additionally, those species threatened or
endangered on the national scale were more abundant in city grid cells than in randomly
selected grid cells (P=0.03) or in randomly selected rural grid cells (P=0.001; see Table 1).
The resampling analysis of covariance showed for native plant species richness no
significant effect of grid cell type (city or non-city). There was a significant effect of the
number of different geological types and a marginally significant effect of land use on the
number of native species (Table 2). The species number of aliens was significantly influenced
by all three variables. For all three species groups, the analysis showed no interaction
between city and geological, soil or land use variables, respectively. The number of red list
Fig. 2. Histograms of mean species richness per grid cell in Germany for (a) native plant species, (b)
pre-1500 alien plant species and (c) post-1500 alien plant species of 68 grid cells randomly sampled
9999 times. Hatched bars represent the frequency of randomly sampled grid cells (n=1924), white
bars represent the frequency of randomly sampled rural grid cells (n=1099). The labels of the x-axis
provide the number of the midpoint of species richness classes. The interval range for native plant
species is 5, for pre-1500 alien species it is 2 and for post-1500 alien species it is 2.5. The mean species
number of city grid cells (n=68) is indicated by the black triangle. City grid cells contain centres from
cities with more than 100,000 inhabitants; rural grid cells contain less than 5% urban land cover.
Table 1. Average richness of different groups of plant species per grid cell in Germany
Species richness of native plant species 580.7 536.1*** 516.3***
Species richness of pre-1500 alien plant species 116.7 92.4*** 85.3***
Species richness of post-1500 alien plant species 85.7 46.0*** 37.1***
Species richness of threatened or endangered
42.6 38.7* 36.7**
Note: Species richness in city grid cells was the average of 68 city grid cells with more than 100,000 inhabitants. The
number of random grid cells equalled the number of city grid cells and were 9999 times randomly selected for (i) all
German grid cells or (ii) only for rural grid cells (<5% cover of urban land use). *One-tailed error probability
(compared with species number in city grid cells) <0.05. **One-tailed error probability <0.01. ***One-tailed error
The flora of German cities 755
species was influenced by almost the same variables as that of native species, namely the
number of geological patches and the number of land use types but not by a city effect
(I. Kühn et al., unpublished).
To test the assumption that the high number of red list species in cities was not caused by
urbanization, we tested whether the species richness of red list plant species was a constant
property of the species richness of natives +naturalized pre-1500 aliens. A major axis
regression of the log-transformed species number yielded a slope of 3.8 (P=0.001) and
unity was outside of the confidence interval [3.52, 4.04]. (The slope of the ordinary least
square regression was 1.5 [P<0.001] and differed significantly from 1 [t-test, P<0.001].)
The results were similar whether city grid cells were included or not.
Partitioning the variation among the variables revealed that geology was especially
important for natives, whereas land use and in particular the city effects were increasingly
important for aliens (Fig. 3). Joint contributions or effects were those where, for example,
geology might influence land use or where land use is inseparable from city effects.
Calculating how much each of these factors contributed to the total (native and alien)
additional species in cities showed that geology was the most important single factor (30%),
while land use and city/non-city each accounted for 12% of the additional species; 46% was
attributed to combined effects.
We hypothesized that natural factors influencing plant species richness covary
with urbanization – that is, they were more abundant in city grid cells and not randomly
distributed. To test this hypothesis, we resampled (i) randomly selected and (ii) randomly
selected rural grid cells in Germany and compared these to city grid cells. This showed
that the single most important factor influencing plant species richness was significantly
higher in city areas (number of geological types =7.4) than on average (number of
geological types =6.6, P=0.007) or than in rural grid cells (number of geological
types =6.2, P<0.001) (Fig. 4). Of course, there are less level 3 CLC classes outside city grid
cells (10.0) than in city grid cells (15.1), but as this is trivial and not the focus of our analysis
it was not analysed further.
Table 2. Error probabilities of the minimum adequate model of a resampling analysis of covariance
on the species number of natives and aliens between city and non-city grids in Germany
Native species Pre-1500 aliens Post-1500 aliens
Number of geological
238521.0 <0.001 12290.1 <0.001 12523.5 0.001
Number of level 3 land
78489.6 0.040 18082.4 <0.001 44013.4 <0.001
City/non-city 4866.2 0.981 4401.1 0.035 16787.1 <0.001
Residuals 905685.3 48178.9 88065.0
Note: Categorical factors were all 68 ‘city’ and 68 randomly selected ‘non-city’ grid cells, resampled 10,000 times.
Metric predictors of the full model were the number of geological patches, number of geological types, number
of natural geological substrates, number of natural soil types, number of natural landscape types, number of
level 3 Corine Land Cover (CLC) patches, number of level 3 CLC classes and number of level 1 CLC classes per
Kühn et al.756
The results of this study clearly showed that not only naturalized alien plant species but also
native plant species contributed to higher plant species richness in urban areas. Thus, the
high species richness of cities was not only due to cultivated or introduced alien plants, as
suggested by several authors (e.g. Barthlott et al., 1999; Kent et al., 1999; McKinney,
Fig. 3. Partitioning the variance of variables correlating with plant species richness in Germany. Slices
represent the average proportions of explained variance. These are calculated as the averages of R2
values of 10,000 analyses of covariance, each with 68 city grid cells and 68 randomly sampled grid
cells from Germany. The pie charts show the independent contribution to the explained variance of
geology (i.e. number of geological types), land use (number of level 3 Corine Land Cover classes) and
city (city or non-city grid cell); joint contributions are those that could only be explained by two or
three of the previous components together. Note that circle sizes are proportional to the explained
Fig. 4. Histogram of the mean number of geological types of 68 grid cells randomly sampled 9999
times. Hatched bars represent the frequency of randomly sampled grid cells (n=1924), white bars
represent the frequency of randomly sampled rural grid cells (n=1099). The labels of the x-axis
provide the number of the midpoint of species richness classes with ranges of 0.2. The mean number
of geological types in city grid cells (n=68) is indicated by the black triangle. City grid cells contain
centres from cities with more than 100,000 inhabitants; rural grid cells contain less than 5% urban
The flora of German cities 757
2002a). The number of geological types was the most important environmental correlate for
native and for total plant species richness. However, geological diversity itself was probably
not directly responsible for plant species richness, but within a climatic zone, geology defines
resource availability through other factors, for example soil chemistry, (micro) relief or
habitats. All this adds to environmental heterogeneity caused by geological richness,
which contributes to species richness (e.g. Wohlgemuth, 1998; Levine and D’Antonio, 1999;
Lonsdale, 1999; Stohlgren et al., 1999; Davis et al., 2000; Stadler et al., 2000; Deutschewitz
et al., 2003).
A possible source of error might have been the use of our 50 control species. Two of the
authors (I.K. and S.K.) have in-depth experience in floristic mapping and were involved
in several schemes incorporated in FLORKART. Therefore, some experience in plant
distribution, plant habitat preferences, mapping procedures and mapping accuracy exists
that, unfortunately, could not be quantified in any way that we are aware without either
employing circular reasoning or calibrating to a benchmark, which does not yet exist. As
far as the distribution maps in Haeupler and Schönfelder (1989) and Benkert et al. (1996),
the experience of the authors as well as the species’ ecological behaviour (very broad
generalists) are concerned, it is reasonable to assume that all of the 50 control species occur
in every 6⬘×10⬘ grid cell. However, without using the 50 control species, our randomly
resampled grid cells would have included many more insufficiently mapped grid cells, largely
from more rural areas (for mainly stochastic and not methodological reasons). This would
have resulted in an even lower plant species richness in randomly selected or rural grid cells
while having only a minor effect on the species number per city grid cell.
An inconsistency might be the different scales for plant species distribution (1: 25,000),
geology and soil (1: 1,000,000) and land cover (1: 100,000). Unfortunately, there is no
information available for the complete area of Germany on geology, soil and land cover at a
scale of 1: 25,000. Scaling up the grid cells of plant species distribution would have resulted
in levelling off differences between city and non-city grid cells. Recognizing the appropriate
scale of an analysis is important (e.g. Allen and Hoekstra, 1991, 1992; Scheiner et al., 2000).
Scheiner et al. (2000) distinguished three different components of scale: (1) the grain, which
is the level of resolution at which the sampling took place; (2) the focus, being the level at
which the analysis takes place, which can be the same size or larger than the grain; and (3)
the extent, which is the span of the geographic area of the samples. Although we employed
different scales for plant species distribution and environmental variables, we used the same
focus and the same extent; differences were only in the grain. Using a relatively coarse grain
for environmental variables may result in underestimating the number of different patches,
and less in underestimating the number of different classes which were defined very broadly.
Therefore, we believe that the use of a smaller grain would probably lead to reduced noise in
the analysis but would not influence the general outcome of our study.
We showed that anthropogenic activities, namely land use heterogeneity, is least
important for native plant species richness and most important for the post-1500 aliens.
This is to be expected, since native plant species were once independent of humans (during
evolution and colonization of pre-human habitats) while post-1500 aliens have been
largely dependent on humans in their new range. Indeed, they would not have crossed
biogeographic barriers into foreign areas without human facilitation.
We showed that cities are not randomly distributed among areas of different geological
heterogeneity. There are good reasons why big cities are more often in areas of high
geological heterogeneity, namely to provide enough means for founding a settlement and
Kühn et al.758
facilitating development into big cities. Analyses from prehistoric settlements in ancient
Greece (Davidson, 1972) and Roman settlements in Britain (Branigan, 1972) demonstrated
the importance of geomorphology for the placement of settlements and traffic routes.
Theoretical considerations (Blouet, 1972) and empirical evidence (Southall, 2000) showed
that the availability of numerous things, such as drinking water (i.e. rivers), arable soils,
defence (e.g. outlooks) and mineral resources, promoted the growth of cities. All these
factors add to geological richness compared to average areas. Actually, most German cities
are situated on large, usually navigable, rivers. As the number of geological patches was the
most important correlate for native plant species richness in our analysis, we reason that
areas where cities were developed were already species rich before the arrival of man. This
means that, at least for native species, today’s city areas were species rich independent of the
presence of a city.
Unlike natives, the groups of alien species showed a significant city effect (i.e. higher
richness within versus outside urban areas). This might be due to variables not included in
our model (for example, location of transportation corridors). Another argument is that
centres of floristic research (such as universities or research centres) and hobby botanists
were more often in larger cities than in rural areas and thus the observed pattern might
simply be a sampling artifact (Barthlott et al., 1999). On balance, we think that this result
is unlikely to be a sampling artifact, for the following reasons. While we omitted all grid
cells that were poorly mapped, there might still be a ‘positive’ bias for areas mapped extra-
ordinarily well, but there is little reason to expect that alien plant species would be sampled
differently from natives, which showed no city effect.
We believe that the clear pattern of enhanced plant species diversity seen for German
flora at the national scale is likely to be similar across large parts of temperate Europe; for
example, areas of high natural heterogeneity with high native plant species richness are also
areas of high population density. However, in other climatic zones (e.g. Hope et al., 2003),
areas with another history of settlement and land use, at other spatial scales or with
different definitions of ‘urbanization’ (e.g. Roy et al., 1999), the pattern might be different.
Therefore, we need to conduct similar analyses especially from other biogeographic regions.
Our data do not allow us to infer directly whether the high number of red list species in
city grid cells is due to or in spite of urbanization. While Dobson et al. (2001) could only
show a correlation between population density and proportions of imperilled plants in
California, analyses from Britain clearly showed a decline of rare species with increasing
population density (Thompson and Jones, 1999). However, our results may be caused by
similar underlying processes, as almost the same factors correlate with native plant species
richness and with those of the red list plant species.
We demonstrated that species richness of red list species is a constant property of
native +naturalized pre-1500 alien plant species richness and that the proportion of red
list plant species increases more than proportionately. Hence we reason that species-rich
grid cells have a considerably larger proportion of rare (and endangered) species than
species-poor grid cells. As city grid cells are relatively species rich, the number of red list
species is quite high as well. Thus it can also be reasoned that endangered plant species
survived in small (semi-) natural areas or those that were not intensely utilized and
which are frequently found in and around city areas and are themselves promoted by
geological diversity. However, as underlying mechanisms for plant species richness
may be co-correlated (Loreau et al., 2001), cause and effect are easily confounded
(e.g. Araújo, 2003).
The flora of German cities 759
Another reason for this pattern might be too long a time frame of the plant species
records. If the time frame was too long, species that never co-existed together might
co-occur together in the analysis. However, we used only records after 1950, when
city centres were already completely developed. Furthermore, there were on average less
than 20% of the records from before 1980 with no significant differences between
city or non-city grid cells. Thus, the proportions of species that could not be recorded later
from city and non-city grid cells were the same. Therefore, it is unlikely that the results
are derived from a higher number of pseudo co-occurrences in cities that were separated
It is well documented that urbanization is one of the major threats to biodiversity
(Wilson, 1988; Thompson and Jones, 1999; McKinney, 2002b; Liu et al., 2003). When
this takes place in areas of high biodiversity, the threat of species loss will increase.
Recent analyses on a global scale (Cincotta et al., 2000; Liu et al., 2003) have shown
that the human population and the number of households grow extraordinarily in
biodiversity hotspots. It appears that human habitat selection (a preference of urban
over rural habitats) is consistent with an evolutionarily stable strategy and that this pattern
of habitat use is associated with increased threats to the biodiversity of several animal
groups and higher plants (Morris and Kingston, 2002). For Germany, Korneck et al. (1998)
state that urbanization is one threat among others to biodiversity. Combining these
observations with our results, we reason that cities are still rich in native species despite
Additionally, cities form new habitats for completely human-managed cultivated plant
communities (e.g. Whitney and Adams, 1980; Wittig, 1991; Pickett et al., 2001; Hope et al.,
2003). These artificial (e.g. horticultural) plant assemblages, which are discussed as
functioning ecosystems and placed in conceptual frameworks (e.g. Savard et al., 2000;
Löfvenhaft et al., 2002), may add considerably to the biodiversity of cities. So besides the
effect on native as well as naturalized alien plant species richness demonstrated in this study,
the diversity of cities consists of these assemblages in gardens, parks, cemeteries or other
habitats completely maintained by humans.
In summary, our findings show that German cities, both currently and in the past, were
preferentially located in pre-existing biodiversity hot spots. We suspect that this pattern
is similar for other temperate cities in Europe and elsewhere under similar environmental
conditions and histories of settlement, and that urban areas are even more important
for the conservation of floristic biodiversity than previously realized. We conclude that
to maintain species diversity it is vital to increase conservation activities in human-
dominated ecosystems, particularly those in and around cities (cf. Balmford et al., 2001;
Rosenzweig, 2003; but see Dobson et al., 2001) and that a full understanding of the
importance of urban biodiversity be incorporated into urban development and planning
FLORKART was provided by the German Centre for Phytodiversity at the Federal Agency for
Nature Conservation. Furthermore, we thank H. Fink and R. May (Federal Agency for Nature
Conservation) for providing environmental data. Steve Higgins, Harald Auge and three anonymous
referees provided very valuable suggestions and comments on earlier drafts of the manuscript.
Steve Higgins and Sarah Gwillym improved our English.
Kühn et al.760
Allen, T.F.H. and Hoekstra, T.W. 1991. Role of heterogeneity in scaling ecological systems under
analysis. In Ecological Heterogeneity (J. Kolasa and S.T.A. Pickett, eds.), pp. 47–68. New York:
Allen, T.F.H. and Hoekstra, T.W. 1992. Toward a Unified Ecology. New York: Columbia University
Araújo, M.B. 2003. The coincidence of people and biodiversity in Europe. Global Ecol. Biogeogr.,
Balmford, A., Moore, J.L., Brooks, T. et al. 2001. Conservation conflicts across Africa. Science,
Barthlott, W., Biedinger, N., Braun, G. et al. 1999. Terminological and methodological aspects of the
mapping and analysis of the global biodiversity. Acta Bot. Fenn., 162: 103–110.
Benkert, D., Fukarek, F. and Korsch, H. 1996. Verbreitungsatlas der Farn- und Blütenpflanzen
Ostdeutschlands. Jena: Fischer.
Blouet, B.W. 1972. Factors influencing the evolution of settlement patterns. In Man, Settlement and
Urbanism (P.J. Ucko, R. Tringham and G.W. Dimbleby, eds.), pp. 3–15. Cambridge, MA:
Branigan, K. 1972. Verulamium and the Chiltern villas. In Man, Settlement and Urbanism (P.J. Ucko,
R. Tringham and G.W. Dimbleby, eds.), pp. 851–855. Cambridge, MA: Schenkman.
Bundesanstalt für Geowissenschaften und Rohstoffe. 1993. Geologische Karte der Bundesrepublik
Deutschland 1:1,000,000. Hannover: Bundesanstalt für Geowissenschaften und Rohstoffe
Bundesanstalt für Geowissenschaften und Rohstoffe. 1995. Bodenübersichtskarte 1: 1,000,000.
Hannover: Bundesanstalt für Geowissenschaften und Rohstoffe (Map).
Cincotta, R.P., Wisnewski, J. and Engelman, R. 2000. Human population in the biodiversity
hotspots. Nature, 404: 990–992.
Collins, J.P., Kinzig, A., Grimm, N.B. et al. 2000. A new urban ecology. Am. Sci., 88: 416–425.
Crawley, M.J. 2002. Statistical Computing: An Introduction to Data Analysis Using S-Plus.
Davidson, D.A. 1972. Terrain adjustment and prehsitoric communities. In Man, Settlement
and Urbanism (P.J. Ucko, R. Tringham and G.W. Dimbleby, eds.), pp. 17–22. Cambridge, MA:
Davis, M.A., Grime, J.P. and Thompson, K. 2000. Fluctuating resources in plant communities:
a general theory of invasibility. J. Ecol., 88: 528–534.
Deutschewitz, K., Lausch, A., Kühn, I. and Klotz, S. 2003. Native and alien plant species richness
in relation to spatial heterogeneity on a regional scale in Germany. Global Ecol. Biogeogr.,
Dobson, A.P., Rodriguez, J.P. and Roberts, W.M. 2001. Synoptic tinkering: integrating strategies for
large-scale conservation. Ecol. Appl., 11: 1019–1026.
Douglas, I. 1983. The Urban Environment. London: Edward Arnold.
Dutilleul, P. 1993. Modifying the t-test for assessing the correlation between two spatial processes.
Biometrics, 49: 305–314.
Ehrendorfer, F. and Hamann, U. 1965. Vorschläge zu einer floristischen Kartierung von
Mitteleuropa. Berichte der Deutschen Botanischen Gesellschaft, 78: 35–50.
Ellenberg, H., Haeupler, H. and Hamann, U. 1968. Arbeitsanleitung für die Kartierung der
Flora Mitteleuropas (Ausgabe für die Bundesrepublik Deutschland). Mitteilungen der Floristisch-
Soziologischen Arbeitsgemeinschaft N.F., 13: 284–296.
Fortin, M.J., Drapeau, P. and Legendre, P. 1989. Spatial autocorrelation and sampling design in
plant ecology. Vegetatio, 83: 209–222.
Gilbert, O.L. 1989. Ecology of Urban Habitats. London: Chapman & Hall.
The flora of German cities 761
Haeupler, H. 1975. Statistische Auswertungen von Punktrasterkarten der Gefäßpflanzenflora
Süd-Niedersachsens. Scripta Geobotanica, 8: 1–141.
Haeupler, H. and Schönfelder, P. 1989. Atlas der Farn- und Blütenpflanzen der Bundesrepublik
Deutschland. Stuttgart: Ulmer.
Haeupler, H., Jagel, A. and Schumacher, W. 2003. Verbreitungsatlas der Farn- und Blütenpflanzen
in Nordrhein-Westfalen. Recklinghausen: Landesanstalt für Ökologie, Bodenordnung und
Hardtke, H.-J. and Ihl, A. 2000. Atlas der Farn- und Samenpflanzen Sachsens. Dresden: Sächsisches
Landesamt für Umwelt und Geologie.
Hill, M.O., Roy, D.B. and Thompson, K. 2002. Hemeroby, urbanity and ruderality: bioindicators of
disturbance and human impact. J. Appl. Ecol., 39: 708–720.
Hope, D., Gries, C., Zhu, W.X. et al. 2003. Socioeconomics drive urban plant diversity. Proc. Natl.
Acad. Sci. USA, 100: 8788–8792.
Huston, M.A. 1994. Biological Diversity: The Coexistence of Species on Changing Landscapes.
Cambridge: Cambridge University Press.
Kent, M., Stevens, R.A. and Zhang, L. 1999. Urban plant ecology patterns and processes:
a case study of the flora of the City of Plymouth, Devon, UK. J. Biogeogr., 26: 1281–1298.
Klotz, S. 1990. Species/area and species/inhabitants relations in European cities. In Urban Ecology:
Plants and Plant Communities in Urban Environments (H. Sukopp and S. Hejný, eds.), pp. 99–103.
The Hague: SPB Academic Publishing.
Korneck, D., Schnittler, M. and Vollmer, I. 1996. Rote Liste der Farn- und Blütenpflanzen
(Pteridophyta et Spermatophyta) Deutschlands. In Rote Liste gefährdeter Pflanzen Deutschlands
(G. Ludwig and M. Schnittler, eds.), pp. 21–187. Bonn: Bundesamt für Naturschutz.
Korneck, D., Schnittler, M., Klingenstein, F. et al. 1998. Warum verarmt unsere Flora? Auswertung
der Roten Liste der Farn- und Blütenpflanzen Deutschlands. In Ursachen des Artenrückgangs von
Wildpflanzen und Möglichkeiten zur Erhaltung der Artenvielfalt (F. Klingenstein and G. Ludwig,
eds.), pp. 299–444. Bonn: Bundesamt für Naturschutz.
Korsch, H., Westhus, W. and Zündorf, H.-J. 2002. Verbreitungsatlas der Farn- und Blütenpflanzen
Thüringens. Jena: Weissdorn-Verlag.
Kowarik, I. 1990. Some responses of flora and vegetation to urbanization in central Europe. In
Urban Ecology: Plants and Plant Communities in Urban Environments (H. Sukopp and S. Hejný,
eds.), pp. 45–74. The Hague: SPB Academic Publishing.
Kowarik, I. 1991. The adaptation of urban flora to man-made perturbations. In Terrestrial and
Aquatic Ecosystems: Perturbation and Recovery (O. Ravera, ed.), pp. 176–184. London: Ellis
Kowarik, I. 1995. On the role of alien species in urban flora and vegetation. In Plant Invasions:
General Aspects and Special Problems (P. Pysˇek, K. Prach, M. Rejmánek and M. Wade, eds.),
pp. 85–103. Amsterdam: SPB Academic Publishing.
Krause, A. 1998. Floras Alltagskleid oder Deutschlands 100 häufigste Pflanzenarten. Natur und
Landschaft, 73: 486–491.
Kühn, I. and Klotz, S. 2002. Floristischer Status und gebietsfremde Arten. In BIOLFLOR – Eine
Datenbank zu biologisch-ökologischen Merkmalen der Gefäßpflanzen in Deutschland (S. Klotz, I.
Kühn and W. Durka, eds.), pp. 47–56. Bonn: Bundesamt für Naturschutz.
Kühn, I., May, R., Brandl, R. and Klotz, S. 2003. Plant distribution patterns in Germany: will
aliens match natives? Feddes Repertorium, 114: 559–573.
Landsberg, H. 1981. The Urban Climate. New York: Academic Press.
Legendre, P. and Legendre, L. 1998. Numerical Ecology. Amsterdam: Elsevier.
Levine, J.M. and D’Antonio, C.M. 1999. Elton revisited: a review of evidence linking diversity and
invasibility. Oikos, 87: 15–26.
Lichstein, J.W., Simons, T.R., Shriner, S.A. and Franzreb, K.E. 2002. Spatial autocorrelation and
autoregressive models in ecology. Ecol. Monogr., 72: 445–463.
Kühn et al.762
Liu, J., Daily, G.C., Ehrlich, P.R. and Luck, G.W. 2003. Effects of household dynamics on resource
consumption and biodiversity. Nature, 421: 530–533.
Löfvenhaft, K., Bjorn, C. and Ihse, M. 2002. Biotope patterns in urban areas: a conceptual
model integrating biodiversity issues in spatial planning. Landscape and Urban Planning,
Lonsdale, W.M. 1999. Global patterns of plant invasions and the concept of invasibility. Ecology, 80:
Loreau, M., Naeem, S., Inchausti, P. et al. 2001. Biodiversity and ecosystem functioning: current
knowledge and future challenges. Science, 294: 804–808.
McKinney, M.L. 2002a. Do human activities raise species richness? Contrasting patterns in United
States plants and fishes. Global Ecol. Biogeogr., 11: 343–348.
McKinney, M.L. 2002b. Urbanization, biodiversity, and conservation. Bioscience, 52: 883–890.
McKinney, M.L. and Lockwood, J.L. 2001. Biotic homogenization: a sequential and selective
process. In Biotic Homogenization (M.L. McKinney and J.L. Lockwood, eds.), pp. 1–17. New
York: Plenum Publishers.
Mittelbach, G.G., Steiner, C.F., Scheiner, S.M. et al. 2001. What is the observed relationship between
species richness and productivity? Ecology, 82: 2381–2396.
Morris, D.W. and Kingston, S.R. 2002. Predicting future threats to biodiversity from habitat
selection by humans. Evol. Ecol. Res., 4: 787–810.
Niemelä, J. 1999. Ecology and urban planning. Biodivers. Conserv., 8: 119–131.
Oke, T.R. 1982. The energetic basis of the urban heat island. Quart. J. R. Meteorol. Soc., 108: 1–24.
Pickett, S.T.A., Cadenasso, M.L., Grove, J.M. et al. 2001. Urban ecological systems: linking terres-
trial ecological, physical, and socioeconomic components of metropolitan areas. Annu. Rev. Ecol.
Syst., 32: 127–157.
Pysˇek, P. 1993. Factors affecting the diversity of flora and vegetation in central European
settlements. Vegetatio, 106: 89–100.
Pysˇek, P. 1998. Alien and native species in Central European urban floras: a quantitative
comparison. J. Biogeogr., 25: 155–163.
Pysˇek, P., Jarosˇik, V. and Kucera, T. 2002a. Patterns of invasion in temperate nature reserves. Biol.
Conserv., 104: 13–24.
Pysˇek, P., Sádlo, J. and Mandák, B. 2002b. Catalogue of alien plants of the Czech Republic. Preslia,
Pysˇek, P., Sádlo, J., Mandák, B. and Jarosˇik, V. 2003. Czech alien flora and the historical pattern of
its formation: what came first to Central Europe? Oecologia, 135: 122–130.
Quinn, G.P. and Keough, M.J. 2002. Experimental Design and Data Analysis for Biologists.
Cambridge: Cambridge University Press.
Rosenzweig, M.L. 1995. Species Diversity in Space and Time. Cambridge: Cambridge University
Rosenzweig, M.L. 2003. Win-Win Ecology. Oxford: Oxford University Press.
Roy, D.B., Hill, M.O. and Rothery, P. 1999. Effects of urban land cover on the local species pool in
Britain. Ecography, 22: 507–515.
Savard, J.P.L., Clergeau, P. and Mennechez, G. 2000. Biodiversity concepts and urban ecosystems.
Landscape and Urban Planning, 48: 131–142.
Scheiner, S.M., Cox, S.B., Willig, M. et al. 2000. Species richness, species–area curves and Simpson’s
paradox. Evol. Ecol. Res., 2: 791–802.
Schroeder, F.-G. 1969. Zur Klassifizierung der Anthropochoren. Vegetatio, 16: 225–238.
Southall, A. 2000. The City in Time and Space. Cambridge: Cambridge University Press.
Stadler, J., Trefflich, A., Klotz, S. and Brandl, R. 2000. Exotic plant species invade diversity hot
spots: the alien flora of northwestern Kenya. Ecography, 23: 169–176.
Statistisches Bundesamt. 1997. Daten zur Bodenbedeckung für die Bundesrepublik Deutschland
1: 100,000. Wiesbaden: Statistisches Bundesamt (Map).
The flora of German cities 763
Statistisches Bundesamt. 2001. Statistisches Jahrbuch für die Bundesrepublik Deutschland. Stuttgart:
Stohlgren, T.J., Binkley, D., Chong, G.W. et al. 1999. Exotic plant species invade hot spots of native
plant diversity. Ecol. Monogr., 69: 25–46.
Sukopp, H. and Trepl, L. 1987. Extinction and naturalization of plant species as related to ecosystem
structure and function. In Potentials and Limitations of Ecosystem Analysis (E.D. Schulze and
H. Zwölfer, eds.), pp. 245–276. Berlin: Springer.
Sukopp, H. and Werner, P. 1983. Urban environments and vegetation. In Man’s Impact on Vegetation
(W. Holzner, M.J.A. Werger and I. Ikusima, eds.), pp. 247–260. The Hague: Junk Publishers.
Sykora, K.V. 1990. History of the impact of man on the distribution of plant species. In Biological
Invasions in Europe and the Mediterranean Basin (F. di Castri, A.J. Hansen and G. Debussche,
eds.), pp. 37–50. Dordrecht: Kluwer Academic Publishers.
Thompson, K. and Jones, A. 1999. Human population density and prediction of local plant
extinction in Britain. Conserv. Biol., 13: 185–189.
Trepl, L. 1995. Towards a theory of urban biocoenoses: some hypotheses and research questions.
In Urban Ecology as the Basis of Urban Planning (H. Sukopp, M. Numata and A. Huber, eds.),
pp. 3–21. The Hague: SPB Academic Publishing.
Waide, R.B., Willig, M.R., Steiner, C.F. et al. 1999. The relationship between productivity and
species richness. Annu. Rev. Ecol. Syst., 30: 257–300.
Walters, S.M. 1970. The next twenty years. In The Flora of a Changing Britain (F. Perring, ed.),
pp. 136–141. Hampton: Classey.
Whitney, G.G. and Adams, S.D. 1980. Man as a maker of new plant communities. J. Appl. Ecol.,
Wilson, E.O. 1988. Biodiversity. Washington, DC: National Academy of Science.
Wittig, R. 1991. Ökologie der Großstadtflora. Stuttgart: Fischer.
Wohlgemuth, T. 1998. Modelling floristic species richness on a regional scale: a case study in
Switzerland. Biodivers. Conserv., 7: 159–177.
Kühn et al.764