Agriculture, Ecosystems and Environment 182 (2014) 2536

Contents lists available at ScienceDirect

Agriculture, Ecosystems and Environment

journal homepage: www.elsevier.com/locate/agee

Species richness in dry grassland patches of eastern Austria:

A multi-taxon study on the role of local, landscape and habitat quality
variables
Klaus Peter Zulka a,b, , Max Abensperg-Traun a,c , Norbert Milasowszky a , Georg Bieringer d ,
Barbara-Amina Gereben-Krenn a , Werner Holzinger e , Gerald Hlzler f,g ,
Wolfgang Rabitsch a,b , Alexander Reischtz h , Pascal Querner a,i , Norbert Sauberer j,k ,
Ingrid Schmitzberger j,l , Wolfgang Willner j,k , Thomas Wrbka j , Harald Zechmeister j
a

Department of Integrative Zoology, University of Vienna, Althanstrae 14, 1090 Vienna, Austria
Environment Agency Austria, Spittelauer Lnde 5, 1090 Vienna, Austria
Federal Ministry of Agriculture, Forestry, Environment and Water Management, Division for Nature Conservation and Species Protection, Stubenbastei 5,
1010 Vienna, Austria
d
Technisches Bro fr Biologie, Umlauffgasse 29, 2544 Leobersdorf, Austria
e
koteam, Institute for Animal Ecology and Landscape Planning, Bergmanngasse 22, 8010 Graz, Austria
f
Institut fr angewandte Biologie und Umweltbildung, Argentinierstrae 54/21, 1040 Vienna, Austria
g
Institute for Wildlife Biology and Game Management, University of Natural Resources and Life Sciences, Gregor-Mendel-Strae 33, 1180 Vienna, Austria
h
Untere Augartenstrae 7/2/24, 1020 Vienna, Austria
i
Institute of Zoology, University of Natural Resources and Life Sciences, Gregor-Mendel-Strae 33, 1180 Vienna, Austria
j
Department of Conservation Biology, Vegetation Ecology and Landscape Ecology, University of Vienna, Rennweg 14, 1030 Vienna, Austria
k
VINCA, Vienna Institute for Nature Conservation and Analyses, Giessergasse 6/7, 1090 Vienna, Austria
l
coopNATURA, Bro fr kologie und Naturschutz, Kremstalstrae 77, 3500 Krems an der Donau, Austria
b
c

a r t i c l e

i n f o

Article history:
Received 31 December 2012
Received in revised form
19 November 2013
Accepted 20 November 2013
Available online 15 December 2013
Keywords:
Fragmentation
Habitat quality
Extinction debt
Connectivity
Patch
Isolation

a b s t r a c t
According to island biogeography theory, the species richness of patches is determined by their size
and spatial isolation, while in conservation practice, it is patch quality that determines protection and
guides management. We analysed whether size, isolation or habitat quality are most important for the
species richness in a set of 50 dry grassland fragments in agricultural landscapes of eastern Austria.
We studied two plant taxa (vascular plants, bryophytes) and 11 invertebrate taxa (gastropods, spiders,
springtails, grasshoppers, true bugs, leafhoppers and planthoppers, ground beetles, rove beetles, butteries and burnets, ants and wild bees). The species richness of three categories was analysed: (1) dry
grassland specialist species, (2) all grassland species and (3) all species. We used regression and hierarchical partitioning techniques to determine the relationship between species richness and environmental
variables describing patch size and shape, patch quality, landscape conguration and landscape quality.
The area-isolation paradigm was only applicable for dry grassland specialists, which comprised 12% of all
species. Richness of all grassland species was determined mostly by landscape heterogeneity parameters.
Total species richness was highly inuenced by spillover from adjacent biotopes, and was signicantly
determined by the percentage of arable land bordering the patches. When analysing all taxa together,
species richness of dry grassland specialists was signicantly related to historical patch size but not to
current patch size, indicating an extinction debt. At the landscape scale, the variable short-grass area was
a better predictor than the less specic variable area of extensively used landscape elements. Distance
to mainland was a good predictor for specialists of mobile animal taxa. Plant specialists showed a pronounced dependence on quality measures at the patch scale and at the landscape scale, whereas animal
specialists were inuenced by patch size, patch quality, landscape quality and isolation measures. None

Corresponding author at: Environment Agency Austria, Spittelauer Lnde 5, 1090 Vienna, Austria. Tel.: +43 1 31304 3391.
E-mail addresses: klaus.peter.zulka@univie.ac.at, peter.zulka@umweltbundesamt.at (K.P. Zulka).
0167-8809 2013 The Authors. Published by Elsevier B.V. Open access under CC BY-NC-ND license.
http://dx.doi.org/10.1016/j.agee.2013.11.016

26

K.P. Zulka et al. / Agriculture, Ecosystems and Environment 182 (2014) 2536

of the taxa beneted from linear structures in the surroundings. In conclusion, high patch quality and a
network of high-quality areas in the surrounding landscape should be the best conservation strategy to
ensure conservation of dry grassland specialists. This goal does not conict with the specic demands of
single taxa.
2013 The Authors. Published by Elsevier B.V. Open access under CC BY-NC-ND license.

1. Introduction
Large grassland areas once were a prominent landscape feature in many temperate regions, particularly in central and eastern
Europe. They resulted from a long history of low-intensity farming systems until the middle of the 20th century (Poschlod and
WallisDeVries, 2002; WallisDeVries et al., 2002). Change in land
use after the Second World War, in particular agricultural intensication and afforestation, led to large-scale fragmentation of
these grasslands. In most regions, semi-natural grasslands have
been reduced to small, isolated habitat patches, where soils are
too poor or slopes are too steep for a more intensive land use.
In such landscapes, semi-natural grassland fragments have been
increasingly appreciated as important biodiversity sources (Bignal
and McCracken, 1996; Hodgson et al., 2005), providing not only
valuable ecosystem services for agriculture such as pollination
and biological control, but also harbouring a number of highly
threatened species that are dependent on high temperatures and
insolation rates. Owing to their small size, grassland fragments
are increasingly threatened by pesticide and fertiliser spillover
from adjacent arable elds. Cessation of grazing and fertilisation may trigger shrub encroachment and eutrophication, leading
to an overall deterioration of habitat quality and the loss of
original open short-grass conditions. Stochastic extinction events
affecting patch organisms may no longer be counterbalanced by
re-immigration from other grassland sites owing to dispersal
impediments imposed by an increasingly hostile and intensively
farmed agricultural matrix and large distances between grassland
patches.
Ecology and conservation biology have developed a comprehensive framework to predict the conditions for species persistence
in small isolated habitat patches. According to the theories of
island biogeography (MacArthur and Wilson, 1967) and metapopulation dynamics (Hanski, 1999), species will survive in isolated
fragments if the patch area is sufciently large to support a
viable population and/or if the landscape structure allows reimmigration into patches once a population becomes extinct.
Consequently, in principle, populations in small isolated grassland
fragments could be managed by (1) improving the local conditions, e.g. by enlarging the patch or (2) improving the landscape
connectivity, e.g. by providing stepping stones or migration corridors.
The question whether local or landscape factors are more important in determining the fate of species in habitat fragments has
been studied extensively over the past decades. Nonetheless, the
results have remained inconclusive even for single taxa. For example, buttery species richness was found to increase with patch
size in several studies (Brckmann et al., 2010; Krauss et al., 2003;
Steffan-Dewenter and Tscharntke, 2000; Wettstein and Schmid,
1999), but showed no signicant relationship in others (Pyry et al.,
2009). Landscape factors (such as patch isolation) were found to
be sometimes highly important for vascular plant species richness (Piessens et al., 2004) and sometimes insignicant (Dauber
et al., 2003; Herrera and Laterra, 2011). Reviewing the evidence on
landscape variables and invertebrate species richness in patches,
Mazerolle and Villard (1999) concluded that most studies failed to
detect a signicant relationship. However, more recently several
studies found an important landscape effect on invertebrates (e.g.
Dauber et al., 2003; Marini et al., 2008). Apparently, the constantly
growing body of scientic evidence on the relevance of local and
landscape factors for species in fragmented habitats has not yet

produced a coherent picture, let alone led to guidelines for conservation managers on how to manage fragmented landscapes to
maximise the benets for patch-dependent species.
Apart from the dichotomy between local and landscape factors, a second dichotomy, that between size and quality measures,
presents an even greater complication for an efcient conservation
management of fragmented grasslands. The basic model underpinning island biogeography theory is a binary island-ocean contrast,
which distinguishes only between habitat and non-habitat for a
particular organism group. Consequently, fragmentation theory is
couched in size measures, such as the size of the local habitat
patch or the total area of stepping stones, typically without considering variation in habitat conditions (Hanski, 1998). By contrast,
practical conservation usually attempts to improve local habitat
conditions, e.g. by implementing grazing management, mowing
or shrub removal, normally without any consideration of landscape geometry, conguration, habitat isolation patterns or patch
network structure. Although evidence shows that a theory-driven
network approach and local quality management could complement each other (Thomas et al., 2001), it remains to be analysed
which size or quality variables are the most inuential. While
some studies have incorporated habitat quality measures into their
set of patch-level variables (e.g. Abensperg-Traun et al., 1996;
Collinge et al., 2003), landscape-level habitat quality measures have
been rarely tested in fragmentation studies, despite accumulating
evidence that matrix quality may signicantly modulate species
persistence in the patches (e.g. Bender and Fahrig, 2005; Ricketts,
2001).
Many of the problems and unsolved questions in fragmentation
research appear to result from focusing on single organism groups,
often butteries or vascular plants. On the one hand, the level of
a single group is too narrow to derive comprehensive landscape
management guidelines, since other taxa might show quite different responses to landscape fragmentation. On the other hand,
the level of single taxa is usually too broad for a detailed analysis
of the relationship between fragmentation response and particular biological or ecological traits, since taxa are often comprised of
species with a wide variety of life cycle and dispersal strategies. As
shown for range shifts in response to global warming in fragmented
habitats, within-taxon variation in habitat preference, dispersal
capacity, longevity and body size is so high that generalisations
at the taxon level are usually of limited value (Hickling et al., 2006).
Even within the butteries, a group supercially appearing to be
homogeneous with regard to dispersal capacity and habitat use,
ckinger et al. (2009) detected disparate responses towards habitat fragmentation between sedentary and mobile species. Studies
addressing taxon species richness will thus combine species with
different response patterns in unknown proportions, and are thus
likely to lead to inconsistent results (Debinski and Holt, 2000).
To address these problems, we based our study not on a single
selected taxon, but on a large set of plant and animal taxa with
pronounced differences in dispersal power, habitat requirements,
life cycles and food chain positions. This means that responses to
fragmentation are potentially quite diverse. While it is impossible
to investigate all taxa living in a landscape, this broad selection
should provide a more balanced and representative picture on
fragmentation responses. Any conservation measures altering the
landscape conguration or the ecological conditions in grassland
patches will affect all species living there, not just a single taxon
addressed in a particular study. Therefore, the rst step was to analyse the net response of wholesale biodiversity. In a second step,

K.P. Zulka et al. / Agriculture, Ecosystems and Environment 182 (2014) 2536

we sought to determine whether the general pattern obtained for

all taxa analysed together would also apply for animal and plants
analysed separately and for individual taxa.
We tested a number of variables that were not only related
to patch and landscape element size and conguration, but also
to habitat quality. In addition to patch quality characteristics, we
explored quality attributes of landscape elements that could potentially function as corridors or stepping stones. The aim of the study
was thus to relate the species richness of patches to a 2 2 predictor matrix reecting the local-landscape dichotomy and the
size-quality dichotomy. As the concepts of island biogeography and
metapopulation ecology are only applicable to species with habitat
preferences that are consistent with the island-ocean paradigm, we
classied species according to their dependence on dry grassland
patches into categories; in particular, we distinguished between
species completely restricted to those patches and species preferring dry grassland patches but occurring also in other non-arable
landscape elements. Our main questions were: (1) Which percentage of species conforms to the island-ocean paradigm? (2) How
important are local, landscape and quality variables for the species
richness of dry grassland specialist species, all grassland species,
all species and individual taxa? (3) To what extent does the pattern
observed for specialist species richness of all taxa agree with the
pattern observed for single taxa?
2. Materials and methods
2.1. Study area
The study was conducted in the Pannonian part of eastern
Austria, i.e. in the lowlands that extend from the eastern Alps
to the Carpathians (Fig. 1). Until the Second World War, these
landscapes were characterised by large, extensively used grassland areas (Bieringer and Grinschgl, 2001; Kohler et al., 1994).
With the exception of some nature reserves (Paar et al., 1993)
and military training areas (Bieringer et al., 2001), most of the
large grassland areas have disappeared. However, within the landscape now largely dominated by intensive agriculture, a number
of small grassland patches have persisted (Holzner et al., 1986;
Paar et al., 1994). Agricultural use of these patches gradually ceased
during the 1960s and 1970s, but has partly been re-established during the last decades, encouraged by agri-environmental schemes.
Fifty dry semi-natural grassland patches were selected in the
study area between the outskirts of Vienna, the eastern slopes
of the Alps, the Seewinkel region east of Lake Neusiedl, and the
river Danube in the north. The climate is subcontinental with
610660 mm annual precipitation, average temperatures between
9.4 C and 9.8 C and an annual temperature amplitude of about
20 C between the temperature average of the coldest and the
warmest month (Walter and Lieth, 1967). Altitudes range from 117
to 290 m a.s.l.
2.2. Site selection
From a set of about 90 known dry grassland fragments (Holzner
et al., 1986; Paar et al., 1994, unpubl. data), we selected 50 study
patches. Selection criteria were (1) the dominant plant community
belonging to the class Festuco-Brometea Braun-Blanquet et Txen
ex So 1947 (Mucina and Kolbek, 1993; Willner et al., 2013), (2)
a patch size between 0.05 and 10 ha and (3) Tertiary or Quaternary geology (rock grasslands along the periphery of the region
were excluded). To avoid pseudoreplication due to spatial proximity, only one patch of a group of patches within a distance of patch
centroids less than 1000 m was retained in the set of study sites.
Preliminary correlation analyses were calculated to identify groups
of patches that showed a high degree of autocorrelation. To ensure

27

homogenous coverage of the variable space, only one patch of such

groups was retained in the nal set of studied patches. We standardised the sampling effort across all 50 habitat patches. Within
each patch, we established two nested permanent sampling plots of
5 m 5 m and 20 m 20 m around the patch centroid, henceforth
referred to as 25-m2 plot and 400-m2 plot.
2.3. Sampling of organism groups
We sampled vascular plants within the 25-m2 plot. All species
were recorded, and their cover and abundance was estimated using
a modied Braun-Blanquet scale (Rabotnov, 1984). Additionally,
to determine the variable patch heterogeneity (PHET; see Section
2.4), plant species richness within the 400-m2 plot was estimated
by recording all additional species occurring in four rectangles of
5 m 7.5 m adjacent to the central quadrat. The plots were visited
three times during the year (early spring, early and late summer). A
comprehensive account on the plant communities recorded in the
patches was provided in Willner et al. (2004).
Bryophytes were collected once in April 2001, a time of the
year known to be optimal for bryophyte growth in the investigated
area (Zechmeister, 1998). Thirty minutes were spent recording all
bryophytes within the 25-m2 plot. Specimens were collected and
identied in the laboratory.
We sampled gastropods (Gastropoda), spiders (Araneae),
surface-active springtails (Collembola), grasshoppers (Orthoptera),
ground beetles (Carabidae), rove beetles (subfamily Staphylininae)
and ants (Formicidae) by pitfall trapping. Three pitfalls (glass jars of
4.5 cm opening diameter, converging on top, half-lled with ethylene glycol as preservative and some droplets of a detergent) were
placed in a triangle around the patch centroid, with two pitfalls in
the corners of the 25-m2 plot and one in the middle of the opposite side. Pitfalls were continuously exposed from 12 April to 9
November 2001 and emptied every three weeks. No roof was used
to avoid microclimate modication. Trap loss was negligible. In the
laboratory, we pooled the contents of the three traps per site, sorted
them by organism groups and stored the material in ethanol for
identication. Owing to the large numbers, surface-active springtails were only used from the rst trapping period (12 April to 4
May 2001). Grasshoppers were additionally assessed by two eld
surveys. The period of the rst eld survey was from 30 June to
10 July 2001. From the centre of each grassland patch, all singing
species were recorded for 5 min with an ultrasound detector (Mini3-Bat, Ultra Sound Advice). All patches were visited between 12:00
and 18:00. The second eld survey was conducted from 24 to 27
August between 08:30 and 16:00. Circles of 20 m diameter around
the centre of each patch were observed for 10 min. Species were
identied by acoustic or morphological characters, as appropriate.
True bugs, planthoppers and leafhoppers (Heteroptera, Fulgoromorpha and Cicadomorpha) were sampled with a sweeping net
(20 strokes along each length in the periphery of the 25-m2 plot).
Additional specimens were obtained by visually collecting specimens on food plants within the 400-m2 plot for half an hour per
site and from pitfall trapping. Field work was done during twoweek periods in May, July and September between 9:00 and 18:00
on clear days without precipitation.
True butteries (Papilionoidea), skippers (Hesperiidae) and burnets (Zygaenidae), were surveyed three times between May and
August. The 400-m2 plot was paced in loops lasting 15 min per sampling period. Surveys were conducted between 10:00 and 17:00
during weather conditions suitable for buttery recording (Pollard,
1977), which was controlled by air-temperature (Testotherm testo)
and wind-velocity measurements (ATP Messtechnik) as well as
cloudiness estimation. Individuals were caught with a sweep net
(38 cm in diameter), identied to species level and released. Species
requiring genital identication were killed and transferred to the

28

K.P. Zulka et al. / Agriculture, Ecosystems and Environment 182 (2014) 2536

Fig. 1. Study area in eastern Austria and spatial arrangement of the dry grassland patches in the region.

laboratory. The species pairs Zygaena purpuralis and Z. minos, Colias

hyale and C. alfacariensis, Plebeius idas and P. argyrognomon, as well
as Melitaea aurelia and M. britomartis were not separated, referred
to as species complexes, and counted as one species each.
Wild bees (Hymenoptera: Apoidea) were collected by sweepnetting, using a sweep-net of 30 cm diameter. Every patch was
visited between 14 April and 14 June and between 14 July and
14 September. In every 25-m2 plot, three 1 m 1 m squares were
haphazardly positioned. Within these squares, all owers were
sampled for 5 min. About 40% of the individuals were identied in
the eld, the remainder collected for later laboratory identication.
Elutriation and subsequent dry sieving of soil samples complemented pitfall trapping results for gastropods. Between 8 and 10
August, three soil samples of 10 cm2 area and 7 cm depth were
taken in each patch along the sides of the 25-m2 plot. Samples
were pooled, homogenised and elutriated. The remaining soil was
dried and sifted through ve sieves of different mesh width. Sieve
remainders were inspected under a stereomicroscope for live and
fresh dead gastropods.

2.4. Environmental variables

We recorded a series of variables classied into local vs. landscape, and quantity vs. quality variables (Table 1). Current patch
area (AR00) was obtained by GIS analysis of aerial photographs
taken near the year 2000. On these photographs, a 1 km 1 km
quadrat around each patch centroid (henceforth referred to a
1-km2 quadrat) was delineated and digitised, using the program
ArcInfo (ESRI Inc.). The variables AR00, patch fractal dimension PFD
(the ruggedness of the patch border) and shape index SHAPE (the
deviation of the patch shape from a quadrat) were obtained from
this digitisation using the program ArcView Patch Analyst (Elkie
et al., 1999) and Fragstats (McGarigal and Marks, 1995). Historical
patch area (AR50) was delineated on aerial photographs taken in
1950. We used SigmaScanPro 5.0 (Systat Software Inc.) for the area
calculation.

To quantify the habitat quality of the patch, we measured aboveground standing phytomass in the patches in April and August
(PHYT4 and PHYT8) using a disc pasture meter (Bransby and Tainton,
1977; Drgeloh, 2002). A 20 cm 20 cm disc of 100 g weight was
released from 1 m height and its nal position above the ground
was measured. An average of fteen measures, made within the
400-m2 plot, was used in the analyses. The proportion of these
measures below 10 cm was used to calculate short-grass area 2000
(SH00) by multiplying area with this fraction (Table 1) and thus to
enable a direct comparison between geometrical patch measures
and quality-weighted patch measures. To account for a possible
extinction debt, a historical version of this variable was similarly
calculated by multiplying the current proportion of short-grass area
with historical patch area (SH50). The variable shrub encroachment
(SHRUB), quantifying the proportion of the patch dominated by
shrubs, was obtained from the GIS analysis of aerial photographs
and ground truthing.
We determined soil temperature sums (TEMP) during a 25-day
period in August and September using the polarimetric sugar inversion method of Pallmann et al. (1940). The velocity of sucrose
hydrolysis into glucose and fructose is temperature-dependent; all
of the sugars are optically active in solution, from the rotation angle
ratio before and after exposition, the sugar concentration ratio and
the temperature sums during exposition can be determined. In
each plot, two plastic tubes containing 20 ml of sugar solution were
buried in 5 cm depth. During transport, the samples were kept in a
cold box. Rotation angles were measured immediately before the
start and shortly after the end of the exposition period with a circular polarimeter (Atago Polax-D). We used rotation angles instead
of mean temperatures in the analyses.
Another modier of the patch quality is soil composition, both
directly by inuencing the habitat of soil-inhabiting and surfaceactive organisms, and indirectly by modifying microclimate. To
analyse soil composition, we took three soil samples from the 25m2 plot boundary in August. Soil cylinders of 10 cm2 surface area
and 10 cm depth were homogenised and dried at 105 C. Dry sieving separated gravel content (GRAV, >2 mm) from the remaining

K.P. Zulka et al. / Agriculture, Ecosystems and Environment 182 (2014) 2536

29

Table 1
Independent environmental variables, their abbreviations, variable transformations used in the analyses and gradient lengths. Further explanations of the variables can be
found in the text.
Variable

Abbreviation

Unit

Transformation

Local size and shape measures

Patch area 2000
Patch area 1950
Patch fractal dimension
Shape index

AR00
AR50
PFD
SHAPE

ha
ha

ln
ln

Local quality measures

Above-ground standing phytomass April
Above-ground standing phytomass August
Short-grass area 2000
Short-grass area 1950
Shrub encroachment
Temperature
Clay content
Silt content
Sand content
Gravel content
Organic material content
Agricultural edge
Shading edge
Patch heterogeneity

PHYT4
PHYT8
SH00
SH50
SHRUB
TEMP
CLAY
SILT
SAND
GRAV
ORGM
AGRIC
SHADE
PHET

cm
cm
ha
ha
%
%
%
%
%
%
%
%

angular
angular
angular
angular
angular
angular
angular

Landscape scale: size and distance measures

Area of extensively used elements
Area of linear elements
Distance to mainland

EXTEN
LINEA
MAINL

ha
ha
m

Landscape scale: quality measures

Area of extensive grassland
Area of fallow land
Area of short-grass dry grassland
Area of short-grass linear elements
Landscape heterogeneity

EXTGR
FALLO
SHGRA
SHLIN
LHET

ha
ha
ha
ha

ne soil (<2 mm). The ne soil components were further separated

by sedimentation analysis. Percentages of sand fractions (SAND),
silt (SILT) and clay (CLAY) fractions refer to ne soil; the percentage of gravel content (GRAV) refers to total soil weight. The content
of organic material (ORGM) was determined by burning ne soil
at 500 C and assessing the weight loss (Austrian Federal Agency
for Soil Analysis). Patch heterogeneity PHET was measured as the
increase in vascular plant species richness between the 25-m2
quadrat and the 20 m 20 m cross (cf. Section 2.3).
To determine edge and landscape effects on patch species richness, spatial elements (Forman, 1995, p. 39) were delineated within
the 1-km2 quadrat around the patch on the aerial photograph. By
ground-truthing, one of 62 pre-dened land use types out of 10 predened land cover classes was assigned to each landscape element.
Disturbance effects of an agricultural matrix (variable agricultural
edge, AGRIC) were determined by calculating the percentage of
the edge bordering intensively farmed agricultural land use types
(agricultural elds, vineyards, intensively used meadows). Shading
effects (SHADE) were quantied by calculating the percentage of
woodland types (forests, single trees, shrubs and hedges) bordering
the patch.
To dene a quantitative isolation measure, the distance between
patch centroid and the nearest large dry grassland area (>15 ha,
termed as mainland) was measured on a topographical map (scale
1:50,000; MAINL). Additionally, a series of landscape composition
variables was determined. Within the 1-km2 quadrat around the
patch, a subset of land use types that were assumed to facilitate re-colonisation and to enhance landscape connectivity of dry
grassland patches were classied in closer detail. Spatial elements
containing grassland of various kind, verges, margins and other
semi-natural biotope types dominated by herbaceous vegetation
were considered, provided that they were only slightly affected by
human inuence (-hemerobic or less intensive, cf. Zechmeister
and Moser, 2001). Attributes describing vegetation height,

Min

Mean

Max

0.04
0.05
1.28
1.15

1.29
6.69
1.41
1.57

9.73
60.41
1.60
2.61

5.3
5.9
0.01
0.01
0%
11.3
1%
0%
10%
0%
1%
0%
0%
5

9.8
12.3
0.82
4.32
7%
19.3
17%
28%
55%
21%
8%
19%
21%
17.14

15.8
26.7
9.73
60.02
30%
32.6
44%
73%
99%
72%
24%
95%
100%
40

ln
ln
ln

1.64
0.24
250

10.22
1.03
4610

ln
ln
ln
ln

0.00
0.97
0.00
0.00
13

1.39
10.61
0.51
0.10
22.34

ln
ln
angular

35.54
2.67
12,250
27.18
31.02
6.49
0.99
37

vegetation density, substrate conditions, shape (in particular

linearity), resource availability (dryness, nutrients), richness in
structural elements (owers, stones), and shrub density were
assigned to each spatial element or sub-element. Landscape elements along with their attribute vector were entered into a
database. Queries against this database provided land use type
selections of increasing quality. The variable area of extensively
used elements EXTEN provided a landscape measure without any
quality considerations; in this variable, essentially all landscape
elements with no or extensive agricultural use were combined
(i.e. set-asides, verges, extensively used grasslands, vineyards).
The variables area of fallow land FALLO and area of extensive
grassland EXTGR reduced this selection to fallows and extensively
used grasslands, respectively. The variable area of short-grass dry
grassland SHGRA restricted the selection of landscape elements to
natural high-quality grasslands with a short sward. Similarly, we
distinguished between linear landscape elements without quality
restriction (LINEA, typically eld margins) and high-quality linear
elements with short-grass vegetation (SHLIN). Finally, to obtain a
measure for landscape heterogeneity, the number of land use types
in the 1-km2 quadrat around the patch was counted (LHET).
2.5. Species richness variables
We used the species richness values obtained by the various
trapping methods as dependent variables. Since only the centre
of the patch was sampled, the numbers should not be regarded
as indicative of the whole-patch species richness, but rather as
species density values. We classied the species into the following categories of patch dependence: (1) Highly patch-dependent
dry grassland specialist species. Owing to their habitat requirements, persistence of these species in the agricultural matrix is not
possible. Species in this category usually require dry and warm
conditions. (2) Species with a preference for grassland habitats.

30

K.P. Zulka et al. / Agriculture, Ecosystems and Environment 182 (2014) 2536

Table 2
Cumulative species richness of all 50 dry grassland patches. Species were classied according to their patch dependence: 1 = dry grassland specialists; 2 = other
grassland species; 3 = other species.
Organism group
Mosses and Liverworts (Bryophyta)
Vascular plants (Trachaeophyta)
Gastropods (Gastropoda)
Spiders (Araneae)
Springtails (Collembola)
Grasshoppers (Orthoptera)
True bugs (Heteroptera)
Leafhoppers and planthoppers
(Fulgoromorpha and
Cicadomorpha)
Ground beetles (Carabidae)
Rove beetles (Staphylininae)
Butteries and burnets (Rhopalocera
and Zygaenidae)
Ants (Formicidae)
Wild bees (Apoidea)
Total

8
60
2
24
9
5
16
14

21
119
11
60
12
18
96
41

43
164
17
102
65
15
151
63

Total
72
343
30
186
86
38
263
118

24
7
10

42
16
45

47
30
8

113
53
63

3
1

23
96

20
35

46
132

183

600

760

1543

These species prefer the patches over the agricultural matrix, yet
are not restricted to the patch habitats; they may live and reproduce in other elements of the agricultural landscape as well, e.g.
in road verges or set-asides. (3) Other species. In the analyses,
we tested species richness of the category 1 species (dry grassland specialists), numbers of group 1 and 2 species combined (all
grassland species) and all categories together (all species). Species
were assigned using habitat preference data obtained from an earlier project (Sauberer et al., 2004), literature data, own collection
data and personal experience.
2.6. Statistical analysis
In the rst step, we calculated a Pearson correlation matrix
between the 26 independent variables (Table 1, ESM 1). In the
second step, we performed single regressions between the independent variable and the three species richness measures (category
1, categories 1 + 2, all species). In test runs, we modelled species
richness using generalised linear modelling (McCullagh and Nelder,
1983) with log link and Poisson errors, but owing to overdispersion, quasi-Poisson error structures had to be assumed (software R
3.0.1, package glm, R Core Team, 2013). However, GLM with quasiPoisson errors led to error probabilities that were numerically very
similar to error probabilities obtained with ordinary least squares
regression. Consequently, for sake of simplicity and better interpretability, the latter regression modelling approach was applied
for single regression analyses (Table 3). Variables were transformed
as necessary (Tables 1 and 3). Owing to the large number of tests
on the same data body, we calculated the positive false discovery
rate using the R package qvalue 1.36.0 (Dabney et al., 2013) and set
the probability threshold to q = 0.05 (Table 3, ESM 2). The positive
false discovery rate approach developed by Storey et al. (2004) is
a far more powerful approach to signicance correction in multiple testing compared to Bonferroni-type adjustments and allows
for the initial screening of a large number of variables (Roback and
Askins, 2005).
Based on the correlation and single regression results, we subjected species richness values of category-1 species (dry grassland
specialists) to multiple regression analyses. We selected nine variables representing the local vs. landscape and quantity vs. quality
categories with high explanatory power in single regressions and
a low degree of mutual correlation. Since the main goal was to
identify variables with a high direct inuence on species richness
(cf. Mac Nally, 2000), we used hierarchical partitioning for the

analysis (Chevan and Sutherland, 1991). Hierarchical partitioning

calculates all possible regression subsets, averages model improvements across all hierarchies and thus permits a ranking of variables
by their independent effects. We calculated models for all taxa combined, for animals and plants and for individual taxa with at least
10 category-1 species (cf. Table 2) using the R package hier.part
(Walsh and Mac Nally, 2013). To identify non-redundant variables
in the set of the nine variables, we additionally performed multiple
regression analyses with backward elimination (PF to remove > 0.1),
starting with the full set of all nine variables. Pearson correlations,
single regressions and multiple regressions with backward elimination were calculated with the software SPSS 10.0.5 (SPSS Inc.,
now IBM Corp.).
3. Results
3.1. Catch results and species richness
We recorded a total of 1543 species in the 13 organism groups,
with vascular plants and true bugs being the most species-rich
groups. On average, only 12% of the species in the small grassland
fragments were dry grassland specialists (category 1), with proportions ranging from 1% in wild bees to 21% in ground beetles
(Table 2). About 51% of the species could be classied as grassland
species in a broader sense (categories 1 and 2 combined).
3.2. Correlations among environmental variables and single
regressions
Pearson correlation analysis of the independent variable data
set showed a generally low degree of correlation (ESM 1). However, some patch shape indices and measures of linear landscape
elements were substantially correlated.
In single regressions, AR00 was a poor predictor of the species
richness of dry grassland specialists; it performed even worse for all
grassland species (categories 1 and 2 combined) and the richness of
all species (Table 3). Using the historical patch size (AR50) instead of
the current patch size improved the explanatory power considerably, especially for dry grassland specialists (category 1). The sign
of the shape variables PFD and SHAPE was negative, but variance
explanation was limited and the models were insignicant.
By contrast, several variables describing the patch conditions
qualitatively were highly signicant predictors for dry grassland
specialist species, in particular PHYT4, PHYT8 and GRAV. The highest species richness of dry grassland specialists was found in open
patches on gravel with a short sward. The quality-weighted area
measures SH00 and SH50 had much higher explanatory power than
the geometrical area measures AR00 and AR50. None of the patch
quality measures was a signicant predictor for total grassland
species richness, but species richness of all species responded positively to the percentage of agricultural matrix around the patch
AGRIC and patch heterogeneity PHET (Table 3).
Species richness of specialists and grassland species increased
signicantly with the amount of extensively used landscape elements (EXTEN) in the 1-km2 quadrat around the patches. All three
measures of species richness decreased signicantly with increasing distance to the next large grassland area in the region (MAINL;
Table 3).
Among the qualitative landscape-scale variables, SHGRA was a
signicant predictor of all three species richness measures, though
variance explanation was about twice as high for specialist species
and grassland species richness than for the species richness of all
species. Specialist richness and grassland species richness were
also signicantly higher in complex landscapes with many biotope
types (LHET, Table 3). The explanatory power of other qualitative

K.P. Zulka et al. / Agriculture, Ecosystems and Environment 182 (2014) 2536

31

Table 3
Single regressions of the three species richness measures against patch and landscape variables (b = sign of the regression coefcient, r2 = goodness of t, P = error probability).
Variable

Category 1 (dry grassland specialists)

r2

Categories 1 and 2 (all grassland species)

All categories (all species)

r2

r2

Local quantity (size and shape) measures

+
0.020
AR00
+
0.122
AR50
PFD

0.057
SHAPE

0.074

0.331
0.013*
0.094
0.056

+
+

0.013
0.037
0.024
0.033

0.431
0.181
0.282
0.208

+
+

0.000
0.019
0.024
0.001

0.895
0.341
0.285
0.815

Local quality measures

PHYT4

PHYT8
+
SH00
SH50
+
SHRUB

+
TEMP

CLAY
SILT

SAND
+
GRAV
+
ORGM
+
AGRIC
+
+
SHADE
+
PHET

0.218
0.182
0.072
0.189
0.001
0.069
0.024
0.036
0.063
0.204
0.056
0.005
0.005
0.007

0.001*
0.002*
0.060
0.002*
0.870
0.066
0.284
0.184
0.080
0.001*
0.098
0.623
0.620
0.576

+
+
+
+

+
+
+
+
+
+
+

0.017
0.052
0.028
0.052
0.027
0.033
0.004
0.000
0.004
0.067
0.052
0.057
0.001
0.085

0.360
0.111
0.245
0.113
0.254
0.210
0.670
0.986
0.646
0.069
0.111
0.094
0.834
0.039

+
+

+
+

+
+

+
+
+
+

0.100
0.016
0.004
0.004
0.064
0.006
0.006
0.029
0.018
0.000
0.040
0.100
0.004
0.216

0.026
0.376
0.647
0.651
0.077
0.592
0.601
0.234
0.354
0.937
0.165
0.026*
0.655
0.001*

Landscape scale: quantity measures

+
0.109
EXTEN

0.029
LINEA

0.121
MAINL

0.019*
0.234
0.013*

0.098
0.031
0.147

0.027*
0.220
0.006*

0.040
0.058
0.108

0.165
0.092
0.020*

Landscape scale: quality measures

FALLO
+
+
EXTGR
+
SHGRA
SHLIN
+
+
LHET

0.097
0.945
0.001*
0.527
0.014*

+
+
+

0.019
0.017
0.199
0.003
0.146

0.336
0.367
0.001*
0.721
0.006*

+
+
+

0.005
0.031
0.104
0.022
0.033

0.614
0.224
0.023*
0.308
0.207

0.056
0.000
0.193
0.008
0.120

*An asterisk denotes a signicant test (P < 0.05) with a positive false discovery rate q < 0.05 (see ESM 2 for a full table of all q values and P values).

landscape variables, such as fallow area (FALLO), extensive grassland area (EXTGR) or area of short-grass linear structures (SHLIN),
was low and insignicant (Table 3).
3.3. Hierarchical partitioning and backward elimination
For dry grassland specialists of all taxa combined, the patch variables PHYT4 and GRAV showed the highest independent effects,
followed by the landscape quality variable SHGRA (Fig. 2). Backward
elimination retained the quantitative landscape variables EXTEN
and LINEA as additional non-redundant variables in the nal model,
but their independent and joint contributions in hierarchical partitioning analysis were considerably lower. As in single regression,
specialist richness decreased with increasing area of linear elements (LINEA).
The pattern for specialist plant taxa was similar, but the direct
effects of quality variables at the patch scale (PHYT4, GRAV) were
more pronounced and landscape effects were lower than for all
taxa combined. By contrast, dry grassland specialist richness of animal taxa was more inuenced by landscape variables, in particular
SHGRA and MAINL. For animals, LINEA was a non-redundant variable with a negative sign. For animal species richness, patch quality
variables were less important and historical patch area was more
important compared to the all taxa pattern.
Except for vascular plants, variance explanation levels for single
taxon models were low. Historical patch size AR50 had a positive
effect on spiders (though eliminated as redundant), true bugs and
ground beetles, but had almost no inuence on leafhoppers and
planthoppers, butteries and burnets. The quality variable PHYT4,
which showed a negative direct effect on species richness of all
taxa combined, was also negative for spiders, true bugs and ground
beetles, but not for leafhoppers and planthoppers, butteries and

burnets. Among the landscape variables, the quality variable SHGRA

usually had higher explanatory power than the qualitative variable
EXTEN. MAINL was retained as non-redundant variable in the models for butteries/burnets and leafhoppers/grasshoppers; for these
taxa, its direct effect was amongst the highest of all variables.
In all models, the direct and indirect effects of SHAPE were
smaller than those of most all other variables and the variable
was typically agged as redundant (Fig. 2). LINEA had no direct
effect on plant species richness and a negative direct effect on
most animal taxa (though being eliminated as redundant in single
taxon analyses). The variable LHET was redundant in all models.
4. Discussion
4.1. Fragmentation
The habitat fragmentation paradigm emerged from island biogeography, metapopulation theory and landscape ecology. It has
been the governing explanation for biodiversity patterns in highly
modied agricultural landscapes over the past decades, although its
generality and value as an explanatory tool have been increasingly
questioned (Fahrig, 2003; Haila, 2002; Harrison and Bruna, 1999;
Lindenmayer and Fischer, 2007). Conceptually, the main problem
seems to be the conation of habitat, essentially a species property, and patch, essentially a human perception of a landscape
element (Lindenmayer and Fischer, 2007). In reality, species habitat requirements and patch area will never exactly match, even for
typical patch-dependent species. We made considerable efforts to
delineate those species across all organism groups for which the
species habitat perception and human perception of a grassland
patch corresponded sufciently, i.e. the dry grassland specialists of

32

K.P. Zulka et al. / Agriculture, Ecosystems and Environment 182 (2014) 2536

Fig. 2. Combination of hierarchical partitioning and backward elimination regression analyses for species richness numbers of dry grassland specialists. Independent (light
and dark grey columns) and joint variance explanation (white columns) averaged across all hierarchies of nine predictors of dry grassland species richness for all taxa
combined, for animal and plant taxa and for single taxa separately. Variable abbreviations and variable categorisations as size, shape and quality variables as in Table 1. The
sign indicates the direction of variable coefcients in the full nine-variable model. Variables excluded by backward elimination (PF to remove > 0.10) are displayed in light grey,
variables retained in the model are displayed in black. Negative white columns indicate suppression effects.

category 1. The proportion of species belonging to that category

was remarkably low, being just 12% of the total species richness in
the patches. For these species, it can be safely assumed that they
are restricted to the grassland patches, experience other habitats
as hostile and do not tolerate disturbances usually encountered in

arable elds. Results for these category-1 specialists are generally

in line with the predictions of the habitat fragmentation paradigm:
We observed an increase of plot species richness with patch area,
at least when measured as historical area or as current short-grass
area, with landscape-level stepping stone area, again measured

K.P. Zulka et al. / Agriculture, Ecosystems and Environment 182 (2014) 2536

preferably in area of high-quality habitat, and proximity to large

grassland areas.
For the species richness of all grassland species, the patchmatrix-corridor landscape model was not applicable, as reected
in the results of Table 3. Since these species are, by denition, not
restricted to the patches, it was not surprising that patch variables
were not found to have high explanatory power. By contrast, several
variables at the landscape scale were found to be good predictors
of grassland species richness, all of which can be interpreted as
enhancing landscape element diversity and complexity of agricultural landscapes. Richness of all grassland species was positively
related to the area of extensively used landscape elements EXTEN,
to the proximity of a large grassland MAINL, the area of shortgrass dry grassland SHGRA and the number of biotope types in the
surrounding 1-km quadrat LHET. Consequently, a landscape variegation model (McIntyre and Barrett, 1992; McIntyre and Hobbs,
1999) relating species richness patterns primarily to the diversity
of landscape elements, describes species richness of all grassland
species better than a binary island-ocean fragmentation model. The
differences in the results between species richness of dry grassland
specialists and richness of all grassland species is also in line with
hypothesis 8 in Tscharntke et al. (2012), which states that different management strategies are required to enhance functional
diversity and ecosystem services in agricultural landscapes on the
one hand and to protect endangered species on the other. However, our results suggest that these two aims can be reconciled with
each other; in other words, diversication of the agricultural landscape by setting aside habitat and by providing short-grass area will
benet all grassland species, whereas additional quality-enhancing
measures may be necessary at the patch scale (see Section 4.2) to
protect endangered dry grassland specialists (Table 3).
Species richness of all species in the patch, however, appeared
to be signicantly inuenced by spillover effects from arable land,
not only reected by the high proportion of non-grassland species
in the patches (Table 2), but also the signicant effect of the variable AGRIC representing the proportion of agricultural edge. As
Blitzer et al. (2012) conclude in their review, such spillover effects
have been underestimated for a long time, but as demonstrated in
our study, they are highly inuential on the species assemblage of
grassland patches. Evidently, species richness patterns in agricultural landscapes with grassland patches appear to be a complex
superposition of several responses types, requiring further concepts in addition to the island-ocean paradigm for a comprehensive
description (Fahrig et al., 2011).
4.2. Local factors
Patch area is one of the central quantities of the habitat fragmentation paradigm (Debinski and Holt, 2000). Even if species richness
is measured with a standardised sampling effort across patches,
local species density of patch-dependent species should be higher
in larger patches, since extinction is less likely and immigration is
more likely than in smaller patches, all else being equal (Connor
and McCoy, 1979). Our results for dry grassland specialists were in
line with these predictions only after two important adjustments:
(1) We found that historical patch area was a better predictor than
current patch area. (2) Area measures weighted by habitat quality
performed much better than patch area as calculated from aerial
photography.
The result that historical patch area (AR50) was a better predictor than current area agrees with recent ndings for plant species
richness in several regions (e.g. Eriksson et al., 2002; Helm et al.,
2006; Kuussaari et al., 2009; Lindborg and Eriksson, 2004; Piqueray
et al., 2011). Species with long generation times and populations
near their extinction thresholds are particularly prone to incur an
extinction debt (Hanski and Ovaskainen, 2002; Kuussaari et al.,

33

2009). However, in our analyses, historical area had a higher independent variance explanation for animals (in particular for ground
beetles and true bugs) than for vascular plants.
Regarding plants, current habitat quality measures (PHYT4,
GRAV) and quality-weighted area measures (SH00, SH50) were
more powerful predictors of species richness than area per se (AR00,
AR50). The independent effect of phytomass on the species richness of specialists was about four times higher than the effect of
historical patch area (Fig. 2). Since extinction events are typically
triggered by a reduction of the patch carrying capacity plus environmental uctuations, habitat degradation can have quick and
detrimental consequences on population sizes and extinction risks
(Mortelliti et al., 2010). In some earlier studies, habitat quality measures have emerged as variables that add unnecessary complexity
to the description of patch occupancy (Moilanen and Hanski, 1998).
Later, evidence has accumulated that habitat parameters can play
a major role in determining survival in patchy landscapes (Collinge
et al., 2003; Fleishman et al., 2002). Thomas et al. (2001) found that
patch habitat quality variation explained patch occupancy much
better than patch area for three species of British grassland butteries. Patches with optimal habitat may have buttery equilibrium
densities more than 200 times higher than sub-optimal yet still
occupied patches (Thomas, 1984). Independent effects of quality
predictors for vascular plant specialist species richness eclipsed
those of all other variables (Fig. 2). With respect to animals, independent effects of historical area and patch quality variables were of
similar magnitude. As shown in the analyses of individual animal
taxa, not all animal groups responded positively to short swards
and gravel in the soil.
Among the factors affecting habitat quality in the patches, edge
effects are considered important (Saunders et al., 1991). The high
proportion of generalist species found in the patches (Table 2) supports this prediction. However, we found no evidence that patch
shape (PFD, SHAPE) and the habitat types bordering the patch
(AGRIC, SHADE) signicantly affected the species richness of dry
grassland specialist species in the centre of the patches (cf. Yamaura
et al., 2008). By contrast, AGRIC was a signicant predictor of total
species richness. A sampling scheme more specically targeted at
this purpose might be necessary to further clarify possible edge
effects.
4.3. Landscape factors
In our study, landscape variables were inuential for all dry
grassland specialist taxa combined and also for many taxa analysed individually. For vascular plant grassland specialist richness,
the landscape quality measure SHGRA had the highest independent effect. For mobile animal taxa, other landscape measures, e.g.
MAINL, were also important and their independent effects were
often at least as high as those of local factors (Fig. 2). This contrasts
with earlier studies that landscape factors are often insignicant
species richness predictors for invertebrates (Mazerolle and Villard,
1999) and is in line with more recent ndings that highlight the
importance of the surrounding landscape for species survival in
patches (e.g. Batry et al., 2007; Janisov et al., 2013; Reitalu et al.,
2012; Steffan-Dewenter and Tscharntke, 1999).
Habitat quality was important not only at the patch scale, but
also at the landscape scale. The independent effect of SHGRA, a metric quantifying the area of short-grass grassland around the patch,
was often more than twice as high as the independent effect of
EXTEN, a metric quantifying a variety of landscape elements with
low or no agricultural use (Fig. 2). Consequently, whether landscape factors are found to be signicant predictors in fragmentation
studies might not only depend on the selection of isolation measures, but also, and apparently to high degree, on the delineation
of what counts as potentially connecting landscape element within

34

K.P. Zulka et al. / Agriculture, Ecosystems and Environment 182 (2014) 2536

the surrounding landscape. This might explain the widely differing

results for the importance of landscape factors in fragmentation
studies.
Interestingly, measures quantifying linear elements (LINEA,
SHLIN) showed no signicant relationship to dry grassland richness in single regression analyses and frequently a negative effect
in multiple regression analyses. What might be considered a corridor in a structural sense need not act as a corridor in a functional
sense, or worse, it may even act as a habitat sink (Krewenka
et al., 2011). In accord with similar ndings (Liira et al., 2008),
our results cast doubt on the possibilities of a cheap mitigation
of fragmentation risks by eld and road margins in agricultural
landscapes.
4.4. Comparison of taxa
The response of dry grassland specialist plant species was characterised by a dominance of patch and landscape element quality
variables (Fig. 2). This response also inuenced the pattern of all
taxa combined. In contrast, the response patterns of animals were
more balanced between quantitative and qualitative variables, both
at the patch and the landscape scale. Apparently, the wider variety
of habitat requirements, life history and dispersal strategies in the
many animal taxa involved in this study led to a more even response
pattern with independent effects of the main predictor variables
explaining a similar level of variance (Fig. 2). Overall, animal taxa
conformed better to the island-ocean paradigm, whereas plant
species richness was better described by habitat quality parameters. Vascular plants may resist extinction even in small isolated
populations for some time as long as habitat conditions are preserved (Eriksson, 1996).
The hypothesis that sedentary taxa should be more inuenced
by local factors than mobile taxa (ckinger et al., 2009) is broadly
supported by our ndings. Local factors were most important for
vascular plants, moderately important also for animal taxa with
mixed dispersal capacities, such as ground beetles, and the least
important for butteries, burnets, leafhoppers and planthoppers.
By contrast, the long-distance landscape variable MAINL was among
the best predictors for mobile animal taxa. Given the often widely
diverging species richness patterns across taxa (cf. Turtureanu
et al., 2013), it might have been reasonable to expect some conicting differences in the responses at least for some of the taxa.
However, for most variables, the signs were identical across taxa,
only their relevance, measured as independent effects, varied
considerably.
4.5. Conclusions and application
The results of this study indicate that a network of high quality
patches, not too far apart from each other, may be the best conservation strategy in fragmented landscapes. Linear elements in
the surroundings were either unimportant or even counterproductive. By contrast, size mattered, if measured as area of short-grass
quality habitat. This overall outcome is in remarkable agreement
with the results of a microcosm study on microarthropod metacommunities (Chisholm et al., 2011). Unfettered by the restrictions
of mensurative experiments at the landscape scale, Chisholm
et al. (2011) highlighted the overall importance of habitat quality and the limitations of linear elements, acting mostly as a sink
habitat.
The goal of a high quality grassland network reconciles traditional management concepts with modern theoretical approaches:
The rst and foremost conservation goal should be the protection of short-grass dry grassland where it still exists, since these
patches not only harbour the most patch-dependent specialist
plant species, but also a number of very demanding animal species.

Second, the high importance of phytomass as a signicant predictor of species richness at the patch scale and the similar importance
of short-grass area at the landscape scale make the reduction
of grass biomass a primary candidate for management intervention. This could be done by grazing or mowing, and is probably
the fastest and most straightforward way to improve the habitat
conditions for many dry grassland specialists. Third, landscapes
with dense networks of short-grass patches should be protected
or restored. Fourth, indications of an extinction debt call for the
restoration of the patches to their historical size. In many cases, this
might be implemented by the clearing of shrubs and subsequent
grazing.
By contrast, general enhancement measures in the agricultural
landscape might increase heterogeneity at the landscape scale and
thus favour some generalist grassland species. This might be a
complementary strategy to increase overall species richness in
agricultural landscapes and to foster ecosystem services such as
pollination and pest control. Extensication of grassland use alone
would be insufcient to protect dry grassland specialists and generalist grassland species alike.
None of our conclusions on management strategies could have
been derived from the analysis of single animal taxa alone, and
biased conclusions would have been obtained by focusing only
on vascular plants. We thus underline the recommendation by
Sderstrom et al. (2001), who strongly advised against basing
conservation strategies and the design of management actions
in grassland patches on single-taxon studies. However, it might
have been expected that conservation strategies derived from
the responses of all taxa analysed together might have led to
unfavourable consequences for single taxa within this pool; for
example, vegetation-dwelling taxa such as true bugs or owervisiting taxa such as butteries might not benet from reduction
of sward height as suggested by the responses of vascular plant
specialists and total specialist richness. Yet, we detected no such
conservation strategy conicts. This nding is in accord with
Watling and Donnelly (2006), who concluded from their review
of fragmentation studies that differences in response patterns
between taxa were limited.
Considering further research on habitat fragmentation, the
results of the present investigation indicate that habitat quality
variables need particular emphasis, both at the patch and the landscape scale. Analysing species with similar traits (e.g. ckinger et al.,
2009, 2010) appears to be more informative than analysing particular taxa with a mixture of habitat requirements and dispersal
strategies. However, an even better way to disentangle the complex responses of communities to habitat fragmentation would be
to analyse the responses of individual species with nely tuned
habitat and matrix variables and then to aggregate the results to
obtain a general picture.

Acknowledgements
The study was nanced by the Austrian Federal Ministry
of Education, Science and Culture within the framework
Austrian Landscape Research (Kulturlandschaftsforschung,
http://www.bmwf.gv.at/startseite/forschung/national/programme
-schwerpunkte/kulturlandschaftsforschung). We are grateful to
Christian Smoliner and Karolina Begusch-Pfefferkorn for their
continuous support, Hannes Paulus and Georg Grabherr for provision of work facilities at their departments, Hans Ambach for
help with ant identication, Christine Jakomini for botanical eld
work, Alexander Pernstich () and Thomas Messner for help with
the pitfall sampling, and John Plant for checking and polishing the
English. Two anonymous reviewers and the editor of the Special

K.P. Zulka et al. / Agriculture, Ecosystems and Environment 182 (2014) 2536

Issue gave valuable feedback on an earlier draft that helped to

improve the analyses and the manuscript substantially.

Appendix A. Supplementary data

Supplementary data associated with this article can be found, in
the online version, at http://dx.doi.org/10.1016/j.agee.2013.11.016.

References
Abensperg-Traun, M., Smith, G.T., Arnold, G.W., Steven, D.E., 1996. The effects
of habitat fragmentation and livestock-grazing on animal communities in
remnants of gimlet Eucalyptus salubris woodland in the Western Australian
wheatbelt. I. Arthropods. J. Appl. Ecol. 33, 12811301.
S., 2007. Effects of local
Batry, P., Orci, K.M., Bldi, A., Kleijn, D., Kisbenedek, T., Erdos,
and landscape scale and cattle grazing intensity on Orthoptera assemblages of
the Hungarian Great Plain. Basic Appl. Ecol. 8, 280290.
Bender, D.J., Fahrig, L., 2005. Matrix structure obscures the relationship between
interpatch movement and patch size and isolation. Ecology 86, 10231033.
Bieringer, G., Grinschgl, F., 2001. Von der Steppe zum Ballungsraum. Stapa 77,
93100.
Bieringer, G., Berg, H.-M., Sauberer, N. (Eds.), 2001. Die vergessene Landschaft.
Beitrge zur Naturkunde des Steinfeldes. Stapa 77, 1313.
Bignal, E.M., McCracken, D.I., 1996. Low-intensity farming systems in the conservation of the countryside. J. Appl. Ecol. 33, 413424.
Blitzer, E.J., Dormann, C.F., Holzschuh, A., Klein, A.-M., Rand, T.A., Tscharntke, T., 2012.
Spillover of functionally important organisms between managed and natural
habitats. Agric. Ecosyst. Environ. 146, 3443.
Bransby, D.I., Tainton, N.M., 1977. The disc pasture meter: possible applications in
grazing management. Proc. Grassl. Soc. S. Afr. 12, 115118.
Brckmann, S.V., Krauss, J., Steffan-Dewenter, I., 2010. Buttery and plant specialists
suffer from reduced connectivity in fragmented landscapes. J. Appl. Ecol. 47,
799809.
Chevan, A., Sutherland, M., 1991. Hierarchical partitioning. Am. Stat. 45, 9096.
Chisholm, C., Lindo, Z., Gonzalez, A., 2011. Metacommunity diversity depends
on connectivity and patch arrangement in heterogeneous habitat networks.
Ecography 34, 415424.
Collinge, S.K., Prudic, K.L., Oliver, J.C., 2003. Effects of local habitat characteristics and
landscape context on grassland buttery diversity. Conserv. Biol. 17, 178187.
Connor, F., McCoy, E.D., 1979. The statistics and biology of the speciesarea relationship. Am. Nat. 113, 791833.
Dabney, A., Storey, J.D., Warnes, G.R., 2013. qvalue: Q-value Estimation for
False Discovery Rate Control. R Package Version 1.36.0, Available from:
http://www.bioconductor.org/packages/release/bioc/html/qvalue.html
(accessed 01.11.13).
Dauber, J., Hirsch, M., Simmering, D., Waldhardt, R., 2003. Landscape structure as
an indicator of biodiversity: matrix effects on species richness. Agric. Ecosyst.
Environ. 98, 321329.
Debinski, D.M., Holt, R.D., 2000. A survey and overview of habitat fragmentation
experiments. Conserv. Biol. 14, 342355.
Drgeloh, W.G., 2002. Calibrating a disc pasture meter to estimate above-ground
standing biomass in Mixed Bushveld, South Africa. Afr. J. Ecol. 40, 100102.
Elkie, P., Rempel, R., Carr, A., 1999. Patch Analyst Users Manual. Northwest Science
& Technology, Thunder Bay, Ontario.
Eriksson, O., 1996. Regional dynamics of plants: a review of evidence for remnant,
source-sink and metapopulations. Oikos 77, 248258.
Eriksson, O., Cousins, S.A.O., Bruun, H.H., 2002. Land-use history and fragmentation
of traditionally managed grasslands in Scandinavia. J. Veg. Sci. 13, 743748.
Fahrig, L., 2003. Effects of habitat fragmentation on biodiversity. Annu. Rev. Ecol.
Syst. 34, 487515.
Fahrig, L., Baudry, J., Brotons, L., Burel, F.G., Crist, T.O., Fuller, R.J., Sirami, C., Siriwardena, G.M., Martin, J.-L., 2011. Functional landscape heterogeneity and animal
biodiversity in agricultural landscapes. Ecol. Lett. 14, 101112.
Fleishman, E., Ray, C., Sjgren-Gulve, P., Boggs, C.L., Murphy, D.D., 2002. Assessing the roles of patch quality, area, and isolation in predicting metapopulation
dynamics. Conserv. Biol. 16, 706716.
Forman, R.T.T., 1995. Land Mosaics. The Ecology of Landscapes and Regions. Cambridge University Press, Cambridge.
Haila, Y., 2002. A conceptual genealogy of fragmentation research: from island biogeography to landscape ecology. Ecol. Appl. 12, 321334.
Hanski, I., 1998. Metapopulation dynamics. Nature 369, 4149.
Hanski, I., 1999. Metapopulation Ecology. Oxford University Press, Oxford.
Hanski, I., Ovaskainen, O., 2002. Extinction debt at extinction threshold. Conserv.
Biol. 16, 666673.
Harrison, S., Bruna, E., 1999. Habitat fragmentation and large-scale conservation:
what do we know for sure? Ecography 22, 225232.
Helm, A., Hanski, I., Prtel, M., 2006. Slow response of plant species richness to
habitat loss and fragmentation. Ecol. Lett. 9, 7277.
Herrera, L.P., Laterra, P., 2011. Relative inuence of size, connectivity and disturbance
history on plant species richness and assemblages in fragmented grasslands.
Appl. Veg. Sci. 14, 181188.

35

Hickling, R., Roy, D.B., Hill, J.K., Fox, R., Thomas, C.D., 2006. The distributions of a
wide range of taxonomic groups are expanding polewards. Global Change Biol.
12, 450455.
Hodgson, J.G., Grime, J.P., Wilson, P.J., Thompson, K., Band, S.R., 2005. The impacts
of agricultural change (19632003) on the grassland ora of Central England:
processes and prospects. Basic Appl. Ecol. 6, 107118.
Holzner, W., Horvatic, E., Kllner, E., Kppl, W., Pokorny, M., Scharfetter, E., Schramayr, G., Strudl, M., 1986. sterreichischer Trockenrasenkatalog. Steppen,
Heiden, Trockenwiesen, Magerwiesen: Bestand, Gefhrdung, Mglichkeiten
ihrer Erhaltung. Bundesministerium fr Gesundheit und Umweltschutz, Wien.
Janisov, M., Michalcov, D., Bacaro, G., Ghisla, A., 2014. Landscape effects on diversity of semi-natural grasslands. Agric. Ecosyst. Environ. 182, 4758.
Kohler, B., Rauer, G., Wendelin, B., 1994. Landschaftswandel. In: Dick, G., Dvorak, M.,
Grll, A., Kohler, B., Rauer, G. (Eds.), Vogelparadies mit Zukunft? Ramsar-Bericht
3 Neusiedler See Seewinkel. Umweltbundesamt, Wien, pp. 2134.
Krauss, J., Steffan-Dewenter, I., Tscharntke, T., 2003. How does landscape context
contribute to effects of habitat fragmentation on diversity and population density of butteries? J. Biogeogr. 30, 889900.
Krewenka, K.M., Holzschuh, A., Tscharntke, T., Dormann, C.F., 2011. Landscape elements as potential barriers and corridors for bees, wasps and parasitoids. Biol.
Conserv. 144, 18161825.
Kuussaari, M., Bommarco, R., Heikkinen, R.K., Helm, A., Krauss, J., Lindborg, R.,
ckinger, E., Partel, M., Pino, J., Roda, F., Stefanescu, C., Teder, T., Zobel, M.,
Steffan-Dewenter, I., 2009. Extinction debt: a challenge for biodiversity conservation. Trends Ecol. Evol. 24, 564571.
Liira, J., Schmidt, T., Aavik, T., Arens, P., Augenstein, I., 2008. Plant functional group
composition and large-scale species richness in European agricultural landscapes. J. Veg. Sci. 19, 314.
Lindborg, R., Eriksson, O., 2004. Historical landscape connectivity affects present
plant species diversity. Ecology 85, 18401845.
Lindenmayer, D.B., Fischer, J., 2007. Tackling the habitat fragmentation panchreston.
Trends Ecol. Evol. 22, 127132.
Mac Nally, R., 2000. Regression and model building in conservation biology, biogeography and ecology: the distinction between and reconciliation of predictive
and explanatory models. Biodivers. Conserv. 9, 655671.
MacArthur, R.H., Wilson, E.O., 1967. The Theory of Island Biogeography. Princeton
University Press, Princeton.
Marini, L., Fontana, P., Scotton, M., Klimek, S., 2008. Vascular plant and Orthoptera
diversity in relation to grassland management and landscape composition in the
European Alps. J. Appl. Ecol. 45, 361370.
Mazerolle, M.J., Villard, M.A., 1999. Patch characteristics and landscape context
as predictors of species presence and abundance: a review. Ecoscience 6,
117124.
McCullagh, P., Nelder, J.A., 1983. Generalized Linear Models. Chapman & Hall, London.
McGarigal, K., Marks, B.J., 1995. Spatial Pattern Analysis Program for Quantifying
Landscape Structure. USDA Forest Service, Portland, OR.
McIntyre, S., Barrett, G.W., 1992. Habitat variegation, an alternative to fragmentation. Conserv. Biol. 6, 146147.
McIntyre, S., Hobbs, R., 1999. A framework for conceptualizing human effects on
landscapes and its relevance to management and research models. Conserv. Biol.
13, 12821292.
Moilanen, A., Hanski, I., 1998. Metapopulation dynamics: effects of habitat quality
and landscape structure. Ecology 79, 25032515.
Mortelliti, A., Amori, G., Boitani, L., 2010. The role of habitat quality in fragmented
landscapes: a conceptual overview and prospectus for future research. Oecologia
163, 535547.
Mucina, L., Kolbek, J., 1993. Festuco-Brometea. In: Mucina, L., Grabherr, G., Ellmauer,
T. (Eds.), Die Panzengesellschaften sterreichs. Teil 1: Anthropogene Vegetation. Gustav Fischer, Jena, pp. 420492.
ckinger, E., Franzn, M., Rundlf, M., Smith, H.G., 2009. Mobility-dependent
effects on species richness in fragmented landscapes. Basic Appl. Ecol. 10,
573578.
ckinger, E., Schweiger, O., Crist, T.O., Debinski, D.M., Krauss, J., Kuussaari, M.,
Petersen, J.D., Poyry, J., Settele, J., Summerville, K.S., Bommarco, R., 2010.
Life-history traits predict species responses to habitat area and isolation: a
cross-continental synthesis. Ecol. Lett. 13, 969979.
Paar, M., Schramayr, G., Tiefenbach, M., Winkler, I., 1993. Naturschutzgebiete sterreichs. Band 1: Burgenland, Niedersterreich, Wien. Umweltbundesamt, Wien.
Paar, M., Tiefenbach, M., Winkler, I., 1994. Trockenrasen in sterreich. Umweltbundesamt, Wien.
Pallmann, H., Eichenberger, E., Hasler, A., 1940. Eine neue Methode der Temperaturmessung bei kologischen und bodenkundlichen Untersuchungen. Ber.
Schweiz. Bot. Ges. 50, 337362.
Piessens, K., Honnay, O., Nackaerts, K., Hermy, M., 2004. Plant species richness
and composition of heathland relics in north-western Belgium: evidence for
a rescue-effect? J. Biogeogr. 31, 16831692.
Piqueray, J., Bisteau, E., Cristofoli, S., Palm, R., Poschlod, P., Mahy, G., 2011.
Plant species extinction debt in a temperate biodiversity hotspot: community, species and functional traits approaches. Biol. Conserv. 144, 1619
1629.
Pollard, E., 1977. A method for assessing changes in the abundance of butteries.
Biol. Conserv. 12, 115134.
Poschlod, P., WallisDeVries, M.F., 2002. The historical and socioeconomic perspective of calcareous grasslands: lessons from the distant and recent past. Biol.
Conserv. 104, 361376.

36

K.P. Zulka et al. / Agriculture, Ecosystems and Environment 182 (2014) 2536

Pyry, J., Paukkunen, J., Heliola, J., Kuussaari, M., 2009. Relative contributions of local
and regional factors to species richness and total density of butteries and moths
in semi-natural grasslands. Oecologia 160, 577587.
R Core Team, 2013. R: A Language and Environment for Statistical Computing. R Foundation for Statistical Computing, Vienna, Austria, Available from:
http://www.R-project.org.
Rabotnov, T.A., 1984. On the development of scales in vegetation science, with
emphasis on Drudes scale. In: Knapp, R. (Ed.), Sampling methods and taxon
analysis in vegetation science. Handbook of Vegetation Science No 4. Dr. W.
Junk, The Hague, pp. 5559.
Reitalu, T., Purschke, O., Johansson, L.J., Hall, K., Sykes, M.T., Prentice, H.C., 2012.
Responses of grassland species richness to local and landscape factors depend
on spatial scale and habitat specialization. J. Veg. Sci. 23, 4151.
Ricketts, T.H., 2001. The matrix matters: effective isolation in fragmented landscapes. Am. Nat. 158, 8799.
Roback, P.J., Askins, R.A., 2005. Judicious use of multiple tests. Conserv. Biol. 17,
261267.
Sauberer, N., Zulka, K.P., Abensperg-Traun, M., Berg, H.-M., Bieringer, G., Milasowszky, N., Moser, D., Plutzar, C., Pollheimer, M., Storch, C., Trstl, R.,
Zechmeister, H., Grabherr, G., 2004. Surrogate taxa for biodiversity in agricultural landscapes of eastern Austria. Biol. Conserv. 117, 181190.
Saunders, D., Hobbs, R.J., Margules, C.R., 1991. Biological consequences of ecosystem
fragmentationa review. Conserv. Biol. 5, 1832.
Sderstrom, B., Svensson, B., Vessby, K., Glimskar, A., 2001. Plants, insects and birds
in semi-natural pastures in relation to local habitat and landscape factors. Biodivers. Conserv. 10, 18391863.
Steffan-Dewenter, I., Tscharntke, T., 1999. Effects of habitat isolation on pollinator
communities and seed set. Oecologia 121, 432440.
Steffan-Dewenter, I., Tscharntke, T., 2000. Buttery community structure in fragmented habitats. Ecol. Lett. 3, 449456.
Storey, J., Taylor, J.E., Siegmund, D., 2004. Strong control, conservative point estimation and simultaneous conservative consistency of false discovery rates: a
unied approach. J. R. Stat. Soc. B 66, 187205.
Thomas, J.A., 1984. The conservation of butteries in temperate countries: past
efforts and lessons for the future. In: Vane-Wright, R.I., Ackery, P. (Eds.), Biology of Butteries. Symposium of the Royal Entomological Society of London, 11.
Academic Press, London, pp. 333353.
Thomas, J.A., Bourn, N.A.D., Clarke, R.T., Stewart, K.E., Simcox, D.J., Pearman, G.S.,
Curtis, R., Goodger, B., 2001. The quality and isolation of habitat patches both

determine where butteries persist in fragmented landscapes. Proc. R. Soc. Lond.

B 268, 17911796.
Tscharntke, T., Tylianakis, J.M., Rand, T.A., Didham, R.K., Fahrig, L., Batry, P., Bengtsson, J., Clough, Y., Crist, T.O., Dormann, C.F., Ewers, R.M., Frnd, J., Holt, R.D.,
Holzschuh, A., Klein, A.M., Kleijn, D., Kremen, C., Landis, D.A., Laurance, W., Lindenmayer, D., Scherber, C., Sodhi, N., Steffan-Dewenter, I., Thies, C., van der
Putten, W.H., Westphal, C., 2012. Landscape moderation of biodiversity patterns
and processeseight hypotheses. Biol. Rev. 87, 661685.
Turtureanu, P.D., Palpurina, S., Becker, T., Dolnik, C., Ruprecht, E., Sutcliffe, L.M.E.,
Szab, A., Dengler, J., 2014. Scale- and taxon-dependent biodiversity patterns
of dry grassland vegetation in Transylvania (Romania). Agric. Ecosyst. Environ.
182, 1524.
WallisDeVries, M.F., Poschlod, P., Willems, J.H., 2002. Challenges for the conservation
of calcareous grasslands in northwestern Europe: integrating the requirements
of ora and fauna. Biol. Conserv. 104, 265273.
Walsh, C., Mac Nally, R., 2013. Package hier.part. Hierarchical Partitioning.
Version 1.0-4. R Foundation for Statistical Computing, Vienna, Austria, Available at http://cran.r-project.org/web/packages/hier.part/hier.part.pdf (accessed
30.10.13).
Walter, H., Lieth, H., 1967. Klimadiagramm-Weltatlas. Gustav Fischer, Jena.
Watling, J.I., Donnelly, M.A., 2006. Fragments as islands: a synthesis of faunal
responses to habitat patchiness. Conserv. Biol. 20, 10161025.
Wettstein, W., Schmid, B., 1999. Conservation of arthropod diversity in montane
wetlands: effect of altitude, habitat quality and habitat fragmentation on butteries and grasshoppers. J. Appl. Ecol. 36, 363373.
Willner, W., Jakomini, C., Sauberer, N., Zechmeister, H.G., 2004. Zur Kenntnis kleiner Trockenraseninseln im Osten sterreichs. Tuexenia 24,
215226.
Willner, W., Sauberer, N., Staudinger, M., Schratt-Ehrendorfer, L., 2013. Syntaxonomic revision of the Pannonian grasslands of Austriapart I: introduction and
general overview. Tuexenia 33, 399420.
Yamaura, Y., Kawahara, T., Iida, S., Ozaki, K., 2008. Relative importance of the
area and shape of patches to the diversity of multiple taxa. Conserv. Biol. 22,
15131522.
Zechmeister, H.G., 1998. Annual growth of four pleurocarpous moss species and
their applicability for biomonitoring heavy metals. Environ. Monit. Assess. 52,
441451.
Zechmeister, H.G., Moser, D., 2001. The inuence of agricultural land-use intensity
on bryophyte species richness. Biodivers. Conserv. 10, 16091625.