Effects of Rotational Shepherding on Plant
Dispersal and Gene Flow in Fragmented
Calcareous Grasslands
by
Yessica Rico Mancebo del Castillo
A thesis submitted in conformity with the requirements
for the degree of Doctor of Philosophy
Department of Ecology and Evolutionary Biology
University of Toronto
© Copyright by Yessica Rico Mancebo del Castillo, 2012
Effects of Rotational Shepherding on Plant Dispersal and Gene Flow in
Fragmented Calcareous Grasslands
Yessica Rico Mancebo del Castillo
Doctor of Philosophy
Department of Ecology and Evolutionary Biology University of Toronto
2012
Abstract
Understanding dispersal and gene flow in human-modified landscapes is crucial for effective
conservation. Seed dispersal governs colonization, recruitment, and distribution of plant
species, whereas both pollen and seed dispersal determine gene flow among populations. This
PhD thesis tests the effect of rotational shepherding on seed dispersal and gene flow in
fragmented calcareous grasslands. Calcareous grasslands (Gentiano-Koelerietum pyramidatae
vegetation) in Central Europe are semi-natural communities traditionally used for rotational
grazing that experienced a decline of plant species during the 20th century due to abandonment
of shepherding. This PhD profits from a management project started in 1989 in Bavaria,
Germany to reconnect previously abandoned calcareous grasslands in three non-overlapping
shepherding systems. Two vegetation surveys in 1989 and 2009 revealed colonizations in
previously abandoned grasslands reconnected by shepherding. First, I propose a comprehensive
approach to identify determinants of community-level patch colonization rates based on 48
habitat specialist plants by testing competing models of pre-dispersal and dispersal effects and
accounting for post-dispersal effects. Mean source patch species occupancy in 1989, and
structural elements in focal patches related to establishment explained community-level patch
ii
colonization rates. Secondly, by adapting the community analysis to all 31 individual species of
the same community with sufficient data, I corroborate the role of shepherding to support
dispersal for a range of species, even if they lack seed morphological traits related to zoochory.
Thirdly, for the habitat specialist Dianthus carthusianorum, I genotyped 1,613 individuals from
64 populations at eleven microsatellites to test the effect of dispersal by sheep on spatial genetic
structure at the landscape scale. Genetic distances between grazed patches of the same herding
system were related to distance along herding routes, whereas ungrazed patches showed
isolation by geographic distance. Lastly, within individual grassland patches, shepherding
significantly decreases the degree of relatedness among neighboring individuals (kinship
structure) and increases genetic diversity. My thesis contributes towards understanding the
effects of zoochory on spatial dynamics in plant populations across scales.
iii
Acknowledgments
My PhD thesis would not be possible without the support from people involved in all aspects of
my life such as my professional career, my family, and friends. First of all, I am deeply thankful
to my advisor and mentor Dr. Helene Wagner because working with her represents a turning
point in my professional development. Helene has guided me through my entire PhD program,
supporting me in difficult moments of my PhD when the field and lab work, the results from
data analyses, or my writing didn’t flow well and in the planned direction. More importantly, I
have learned from her passion how to do good science and that there is always an exciting story
to tell about the way we understand the complexity of nature. I am also thankful to my
committee members Dr. Marie-Josée Fortin, Dr. Sasa Stefanovic, and Dr. Rolf Holderegger for
their thoughtful comments, criticisms, and advice on my PhD research. In particular, thanks for
the suggestions and encouragement from Marie-Josée to finalize my PhD thesis, and to Rolf for
his helpful comments about the analyses and interpretation of my genetic data.
My PhD research benefited greatly from the collaboration with Rolf Holderegger during
two stays at the WSL Swiss Federal Research Institute. At the WSL, I had valuable learning
experiences about lab techniques and analysis of genetic data, besides it being a great place
where I profited from the interaction with enthusiastic scientists related to my research interests.
Special thanks to Dr. Juergen Boehmer, who is a fundamental collaborator of my PhD research
since his early work as an ecologist in my study area provided the research basis of this PhD
project; as well as for hosting me twice at the Interdisciplinary Latin America Center ILZ at the
University of Bonn in Germany. I greatly appreciate that Juergen shared with me his passion
and knowledge of the flora and social-natural history of the beautiful landscape of Franconia in
Bavaria, by which I was captivated since the first day of my field work.
Also thanks to the people involved in the conservation project on calcareous grasslands
in Franconia, Karlheinz Dadrich and Doris Baumgartner (Untere Naturschutzbehörde,
Landkreis Weissenburg-Gunzenhausen), Bernd Raab (Landesbund für Vogelschutz in Bayern),
Stefanie Haacke (Landschaftspflegeverband Mittelfranken), Jens Sachteleben (PAN, Munich),
and the shepherds Erich Beil, Erich Neulinger, and Alfred Grimm for providing valuable
information of the management practices in the area. Special thanks to Henry Lehnert for field
support and René Graf for support during my lab work experience at the WSL. Thanks to the
iv
Biology Department and to the Micro-Electronics group at the University of Toronto,
Mississauga for their assistance.
Financial support to do my PhD studies at the University of Toronto was provided by
the National Council on Science and Technology of Mexico, CONACYT and by the Secretary
of Public Education of Mexico, SEP. I also received funding for my research stay in Germany
by the German Academic Exchange Service, DAAD. Additional funding was provided by the
Natural Sciences and Engineering Research Council of Canada, NSERC, through a Discovery
grant to Helene H. Wagner, and by the Department of Biology and Ecology and Evolutionary
Biology at the University of Toronto.
I feel thankful and lucky to have as colleagues and friends Ilona and Shekhar for sharing
with me their minds and hearts and for walking together the ups and downs along these four
significant years of my life. Thanks to many good friends that I have made at different places,
especially to Carmen, Liz, Chang, To, Gil, Maria Luisa, Carolina, Adrián, Pera, and friends
from San Cristóbal, Chiapas, for listening my experiences and thoughts, and cheering me up
when I most needed it.
Lastly but not less importantly I want to thank to all the members of my family who
have always supported my decisions for pursuing my dreams. Especially to my parents, Martha
and Luis, my brothers Luis and Abraham, my grandparents Purecita and Manuel, because no
matter that I am not there with them at home, they always encouraged and reminded me that I
can accomplish whatever goal I propose to do.
v
Table of Contents
Contents
Acknowledgments .......................................................................................................................... iv
Table of Contents ........................................................................................................................... iv
List of Tables ................................................................................................................................. ix
List of Figures ................................................................................................................................. x
List of Appendices ......................................................................................................................... xi
Chapter 1 ....................................................................................................................................... 1
1.1 Effect of Habitat Fragmentation in Plants .......................................................................... 1
1.2 Investigating Functional Connectivity in Plants .................................................................. 3
1.3 Importance of Directed Dispersal for Plant Genetic Connectivity ...................................... 5
1.4 Study System: Calcareous Grasslands ................................................................................. 9
1.5 Study organism: Dianthus carthusianorum L. ................................................................... 11
1.6 Objectives and Outline of the Thesis ................................................................................. 12
Chapter 2 ..................................................................................................................................... 17
2.1 Abstract .............................................................................................................................. 17
2.2 Introduction ........................................................................................................................ 18
2.3 Methods .............................................................................................................................. 22
2.3.1 Study site .................................................................................................................. 22
2.3.2 Actual functional connectivity data ......................................................................... 23
2.3.3 Management records ............................................................................................... 24
2.3.4 Models of potential functional connectivity ............................................................. 25
2.3.5 Pre-dispersal effects ................................................................................................. 26
2.3.6 Post-dispersal effects ............................................................................................... 27
2.3.7 Data analysis ........................................................................................................... 27
2.4 Results ................................................................................................................................ 27
2.5 Discussion .......................................................................................................................... 29
2.6 Conclusions ........................................................................................................................ 32
vi
Chapter 3 ..................................................................................................................................... 41
3.1 Abstract .............................................................................................................................. 41
3.2 Introduction ........................................................................................................................ 42
3.3 Methods .............................................................................................................................. 45
3.3.1 Study area and species occupancy data ................................................................... 45
3.3.2 Functional connectivity model ................................................................................. 47
3.3.3 Species traits ............................................................................................................ 49
3.3.4. Species occupancy analysis .................................................................................... 49
3.3.5 Genetic sample collection and microsatellite analysis ............................................ 50
3.3.6 Genetic data analysis ............................................................................................... 52
3.4 Results ................................................................................................................................ 53
3.4.1 Functional connectivity models at the species level ................................................ 53
3.4.2 Connectivity effects on patch occupancy and gene flow in Dianthus
carthusianorum ...................................................................................................... 54
3.5. Discussion ......................................................................................................................... 55
3.5.1 Connectivity by shepherding supports dispersal of calcareous grassland plants ... 55
3.5.2 Contribution of source and focal patch properties to landscape species
occupancy ............................................................................................................. 58
3.5.3 Consistency of ecological and genetic data in functional connectivity
assessments of Dianthus carthusianorum .............................................................. 59
3.6 Conclusions ........................................................................................................................ 60
Chapter 4 ..................................................................................................................................... 69
4.1 Abstract .............................................................................................................................. 69
4.2 Introduction ........................................................................................................................ 70
4.3 Methods .............................................................................................................................. 73
4.3.1 Study site and species ............................................................................................... 73
4.3.2 Sampling and microsatellite analysis ...................................................................... 74
4.3.3 Analysis of landscape-scale patterns of genetic structure ....................................... 75
4.3.4 Isolation by geographic distance and connectivity by shepherding ........................ 76
4.4 Results ................................................................................................................................ 77
4.4.1 Landscape-scale patterns of genetic structure ........................................................ 77
vii
4.4.2 Isolation by geographic distance (IBD) and directed dispersal by sheep ............... 78
4.5 Discussion .......................................................................................................................... 79
4.5.1 Effect of directed dispersal by shepherding on genetic structure at the
landscape .............................................................................................................. 79
4.5.2 Effect of IBD vs. directed dispersal by shepherding on genetic connectivity .......... 81
Chapter 5 ..................................................................................................................................... 88
5.1 Abstract .............................................................................................................................. 88
5.2 Introduction ........................................................................................................................ 89
5.3 Methods .............................................................................................................................. 93
5.3.1 Study site .................................................................................................................. 93
5.3.2 Study species and sampling ..................................................................................... 94
5.3.3 Microsatellite analysis ............................................................................................. 95
5.3.4 Quantification of SGS, genetic diversity, and inbreeding ........................................ 96
5.3.5 Statistical test of SGS, genetic diversity, and inbreeding coefficients among
groups ................................................................................................................... 98
5.4 Results ................................................................................................................................ 99
5.4.1 Strength of fine-scale spatial genetic structure (SGS) ............................................. 99
5.4.2 Estimates of neighborhood size and gene dispersal .............................................. 100
5.4.3 Estimates of genetic diversity and inbreeding ....................................................... 100
5.5 Discussion ........................................................................................................................ 102
5.6 Conclusions ...................................................................................................................... 107
Chapter 6 ................................................................................................................................... 115
6.1 Directed Dispersal by Shepherding Promotes Functional Connectivity in Calcareous
Grasslands ....................................................................................................................... 115
6.2 Shepherding Effects on Seed-Mediated Gene Flow across Spatial Scales in Dianthus
carthusianorum ............................................................................................................... 118
6.3 Conservation Implications and Future Directions............................................................ 121
References .................................................................................................................................. 125
viii
List of Tables
Table 1. Distance models description……………………………………………………….….34
Table 2. Variation partitioning and sensitivity analysis of the final model……………………36
Table 3. Description of dispersal traits………………………………………………………....61
Table 4. Best performing Si connectivity indices of each species……………………………...62
Table 5. Analysis of molecular variance among shepherding systems and ungrazed patches in
D. carthusianorum……………………………………………………………………………...87
Table 6. Estimates of SGS across 49 populations in D. carthusianorum…………………….108
Table 7. Estimates of gene dispersal and neighborhood size for six treatment groups in D.
carthusianorum …………………………………………………………………………….....111
Table 8. Estimates of genetic diversity for six treatment groups in D. carthusianorum……...112
ix
List of Figures
Fig. 1 Model species, Dianthus carthusianorum……………………………………………....16
Fig. 2 Conceptual model of plant functional connectivity…………………………………......38
Fig. 3 Relative importance of each parameters in the Si connectivity models…………………39
Fig. 4 Side-by-side boxplots of the community-level patch colonization rates……………......40
Fig. 5 Conceptual diagram of genetic and demographic connectivity in plants…………….....66
Fig. 6 Percentage of best ranked dispersal models by dispersal mode………………………...67
Fig. 7 Relative importance of Si connectivity parameters in D. carthusianorum……………...68
Fig. 8 Ordination plot of PCA of population allele frequencies in D.carthusianorum...............84
Fig. 9 Pie charts of population genetic membership scores for two genetic clusters in D.
carthusianorum……………………………………………………………………..........85
Fig. 10 Isolation by geographic distance and distance along shepherding routes on population
genetic distances in D. carthusianorum…………..…………...........................................86
Fig. 11. Sketch of spatial locations of populations analyzed for SGS in D.
carthusianorum.........................................................................................................113
Fig. 12 Interaction plot of the Sp statistics as a function of sheep grazing and population size
factors in D. carthusianorum ………………………………………………………....114
x
List of Appendices
A1. Distribution of calcareous grassland patches in the study area…………………………...152
A2. Optimized α-values for each dispersal model and for each analyzed species…………….153
A3. Genetic diversity indices per loci in D. carthusianorum…………………………….........155
A4. Multiple comparisons with Tukey’s HSD of interclass PCA axes…………………….....156
A5. Plot of DIC average values and Posterior estimates of cluster memberships………….....157
A6. Genetic diversity estimates, inbreeding coefficients, and location coordinates of D.
carthusianorum populations ………….….................................................................................158
xi
1
Chapter 1
General Introduction
1.1. Effect of Habitat Fragmentation in Plants
Habitat loss and fragmentation, due to human activities, change a formerly continuous
distribution of populations into disconnected habitat patches varying in size and spatial
isolation, causing plant biodiversity decline at the scales of individual patches and the entire
landscape (Saunders et al. 1991). The degree to which these processes will affect plant
population persistence will depend on the composition and spatial arrangement of the
landscape elements, the quality of the remaining habitat, the effective dispersal of seeds
through the landscape matrix to reach suitable habitat patches and on the exchange of genes
among remaining populations (Taylor 1993; Tischendorf and Fahring 2000; Moilanen and
Hanski 2001).
In plants, increased habitat isolation is likely to decrease effective seed dispersal and
gene flow among populations (where a population is defined ecologically as all individuals of
a species in a habitat patch) (Sork et al. 1999). On one hand, lack of gene flow through pollen
and seeds will “erode” genetic diversity within populations by drift and inbreeding, which will
be stronger in small and isolated populations (Templeton et al. 1990; Young et al. 1996;
Young and Clark 2000; Frakham 2005; Aguilar et al. 2008). On the other hand, limited seed
2
dispersal may prevent small isolated populations from being “rescued” from extinction due to
lack of seed immigration (Nathan and Muller-Landau 2000). Importantly, patch re-
colonization after local extinction can only occur through seed dispersal, which is fundamental
for the long-term species persistence in fragmented landscapes (Cain et al. 2000; Levin et al.
2003; Nathan 2006; Levey et al. 2008).
Restoring habitat connectivity is thus expected to enhance dispersal and gene flow
among fragmented plant populations having a positive effect on the maintenance of landscape
biodiversity (Opdam et al. 2006). As pointed out by Fischer and Lindenmayer (2007) in a
review of habitat fragmentation effects on biodiversity, species conservation research will
benefit from addressing the following six research priorities: (1) focusing on species dispersal
by a combination of ecological and genetic approaches, (2) assessing changes in biodiversity
patterns not only at the patch level but through entire modified landscapes, (3) implementing
and evaluating natural experiments to study larger spatial and temporal scales than typically
investigated, (4) focusing on plants and invertebrates to overcome the bias in connectivity
studies in birds, mammals, and amphibians, (5) investigating cascading effects of landscape
change; and (6) evaluating trade-offs between biodiversity conservation goals and land-use
benefits in collaboration with conservation practitioners and policy makers. My PhD, which
investigates determinants of seed dispersal and gene flow in calcareous grasslands, addresses
the first fourth points based on “natural experiment” of a landscape management project that
has been ongoing for more than 20 years.
3
1.2. Investigating Functional Connectivity in Plants
Conservation planning in fragmented landscapes often focuses on restoring landscape
connectivity, assuming that dispersal and gene flow will be enhanced by linking habitat
patches (Baguette and Van Dyck 2007). To effectively restore connectivity, it is essential to
distinguish two components: (1) structural connectivity that refers to the spatial pattern, such
as habitat area, the distance between habitat patches and the composition of the intervening
matrix, and (2) functional connectivity, which refers to the species interaction with pollen and
seed vectors and the matrix resulting in effective dispersal and gene flow (Taylor et al. 1993;
Uezu et al. 2005). Two aspects of functional connectivity can be further distinguished as: (1)
potential functional connectivity, which refers to models of species’ (or their pollen or seed
vectors) behavioral response to the matrix, and (2) actual functional connectivity, which
corresponds to the quantification of effective dispersal, such as observed dispersal rates,
colonization events, or estimates of gene flow (Calabrese and Fagan 2004; Fagan and
Calabrese 2006).
To effectively inform conservation efforts, it is necessary to apply more comprehensive
approaches not only focusing on the process of species dispersal per se (potential functional
connectivity), but also on the importance of habitat site characteristics influencing pre- and
post-dispersal processes and thus actual functional connectivity in terms of realized seed
dispersal and gene flow (Nathan and Muller-Landau 2000; Mortelliti et al. 2010). Source
patches varying in population size may unequally contribute with emigrant seeds and amount
of pollen flow, while environmental characteristics of focal patches, i.e. relating to habitat
quality, may affect probability of establishment and growth (e.g., Honnay et al. 1999;
Fleishman et al. 2002; Johansson and Ehrlen 2003).
4
While functional connectivity is assumed to be a species-specific property (Taylor et
al. 2006) it is logistically and economically unfeasible to investigate and manage connectivity
for each species across the landscape. Alternatively, large-scale empirical studies at the
community-level of analysis may reveal general patterns of species responses to habitat
fragmentation (Lindenmayer 2009; Minor et al. 2009; Schleicher et al. 2011), which may
provide useful information to conservation managers aiming to protect multiple species
simultaneously.
Another aspect of consideration when modeling connectivity is the trade-off between
the nature of the data and information gain (Fagan and Calabrese 2006). Most empirical
studies at the plant-community level have modeled connectivity simply as a function of the
physical distances between habitat patches without considering the influence of the matrix and
the role of vectors (Nathan et al. 2000; Erik and Priya 2003; Jordano et al. 2007). Such a
simplistic approach may lead to the erroneous conclusion that connectivity is not a key factor
if dispersal and gene flow rates do not depend on distance alone but on matrix resistance, or on
the interaction of the vectors with the matrix.
Neutral genetic markers, such as nuclear microsatellites, are a common approach to
analyze the effect of landscape structure on rates of gene flow (Ouborg et al. 1999). For
instance, to infer genetic connectivity across populations scattered in the landscape, either pair-
wise estimates of genetic differentiation (FST) are regressed against inter-patch distances, or
patch-level estimates of population genetic diversity are regressed against patch connectivity
indices, assuming that populations within structurally well connected patches are likely to
experience high rates of gene flow (e.g., Keyghobadi 2005; Honnay et al. 2007; Bizoux et al.
2008; Neel 2008). Such, inferences of gene flow assessed with nuclear markers are affected by
5
both seed dispersal and pollen flow, which may differ considerably in their rates and spatial
scales (Ellstrand 1992; Ennos 1994; McCauley 1997; Petit et al. 2005). However, genetic
connectivity at the landscape scale may largely depend either on pollen or seeds, depending
upon their interaction with dispersal vectors and the matrix. The ratio of pollen- vs. seed-
mediated gene flow can be quantified from combining markers with different mode of
inheritance such as nuclear and chloroplast DNA microsatellites. For instance, in most
angiosperm species chloroplast DNA (cpDNA) is maternally inherited only via seeds. The
differences in genetic structure between markers thus allow disentangling the contributions of
seeds vs. pollen to rates of gene flow (McCauley 1997). However, the above approach is often
limited since cpDNA polymorphism is often insufficient to reveal enough genetic structure
within and among populations at the landscape scale (Holderegger and Wagner 2008).
Alternatively, indirect information may be gained by contrasting demographic data of species
dispersal (i.e. colonization rates) and estimates of gene flow based on nuclear markers against
alternative and competing models of dispersal that are based on alternative assumptions of
vector interactions and matrix effects. A combination of approaches is needed to fully
understand plant functional connectivity in human modified landscapes.
1.3. Importance of Directed Seed Dispersal for Plant Genetic
Connectivity
Gene flow in plants is a scale-dependent process (McCauley 1997) that is largely determined
by the dispersal ability of seeds and pollen (Loveless and Hamrick 1984; Ennos 1994).
According with the prevalent view, pollen-mediated gene flow at the landscape scale is
expected to be more substantial than seed-mediated gene flow (Cruse-Sanders and Hamrick
6
2004; Petit et al. 2005), whereas at local scales, i.e. within habitat patches, the restricted
dispersal of seeds is expected to create fine-scale spatial genetic structure (e.g., Hamilton
1999; Heuertz et al. 2003; Trapnell and Hamrick 2004; 2005; Cruse-Sanders and Hamrick
2004; Escudero et al. 2006; Bittencourt and Sebbenn 2007; Wang et al. 2011). However, a
growing number of empirical studies revealed that the contribution of seed- relative to pollen-
mediated gene flow across scales depend to a large extent on seed dispersal mechanisms and
landscape heterogeneity (e.g., Cruse-Sanders and Hamrick 2004; Jordano et al. 2007; Zhou et
al. 2007; Freeland et al. 2012).
Plants exhibit a large variety of seed dispersal vectors, influencing the distance,
direction, and destination at which seeds are deposited away from the source to reach suitable
sites for establishment and growth (Shupp 1993; Nathan and Muller Landau 2000; Shupp
2011). For instance, plants without morphological seed dispersal adaptations are likely to be
dispersed over very short distances (often < 1 m), resulting in clustering of related genotypes
within the vicinity of the source (e.g., Hamrick and Nason 1996; Epperson and Alvarez-Buylla
1997; Cruse-Sanders and Hamrick 2004). On the other hand, seeds adapted to dispersal by
wind will travel longer distances (Tackenberg 2003; Soons et al. 2004), but the random
direction of dispersal and thus the chance to arrive in suitable sites for establishment is highly
stochastic (Nathan 2000; 2006). Directed seed dispersal by animal vectors can also occur over
long distances, but in contrast to wind dispersal, seed deposition in suitable sites with high
probability of establishment can be substantial (Howe and Smallwood 1982; Wenny 2001).
Directed seed dispersal with animal vectors (zoochory) has been shown to influence the
structure of populations and plant communities (Wenny 2001; Aukema and del Rio 2002;
Purves et al. 2005; Briggs et al. 2009) and to determine, spatial patterns of genetic structure
7
within and among populations (e.g., García et al. 2007; Jordano et al. 2007; Karubian et al.
2010; Kloss et al 2011). Available evidence from tree and shrub species dispersed by
frugivores found that directed dispersal may create strong spatial clustering of related
genotypes even with extensive pollen flow (Grivet et al. 2005; García et al. 2009; Torimaru et
al. 2007). The resulting pattern of fine-scale spatial genetic structure was explained by the
clumped seed deposition in preferable microsites used by the vector, regardless of the long-
distance dispersal away for the source (e.g., Torimaru et al. 2007). In contrast, directed
dispersal may also result in a homogenization of the spatial distribution of genetic variation by
the recurrent mixing of seeds from several population sources (e.g., Karubian et al. 2010).
These conflicting results likely are related to the behavior of the animal vector and its
interaction with the landscape structure.
In the case of frugivory, plants invest in the production of fruits that are attractive to
their animal vectors, and animals deliberately consume on these fruits (Traveset and Perez
2008; Lorts et al. 2008). In contrast, animals may inadvertently transport seeds, e.g. attached to
their fur or hooves (epizoochory) or if seeds are ingested during foraging (endozoochory)
(Janzen 1984; Fisher et al. 1996; Iravini et al. 2011). Some plants have developed
morphological seed adaptations to promote epizoochory, e.g. with awns or bristles (Hughes et
al. 1994; Lorts et al. 2008). In agricultural landscapes, grazing by herds of domestic ungulates
such as sheep, horses, or cattle may thus support dispersal of a range of herbaceous species
(Fischer et al. 1996; Couvreur et al. 2004; Cosyns et al. 2005; Bruun and Poschlod 2006;
Auffret et al. 2012). In grazed calcareous grasslands communities, rotational shepherding is
presumed to be one of the key factors sustaining local and regional species richness (Fischer et
al. 1996; Schrautzer et al. 2009; Kuiters and Huiskes 2010; Piqueray et al. 2011; Wagner et al.
8
2012). In contrast to seed dispersal by wind (anemochory), directed seed dispersal by
rotational shepherding increases the chances of successful long-distance dispersal, as seeds
may be retained for extended periods of time and travel several kilometers to suitable sites as
herds move between pastures (Manzano and Malo 2006). However, whether rates of seed
dispersal among spatially isolated plant populations connected along shepherding routes can
be substantial enough to influence landscape-scale patterns of genetic structure has not yet
been shown.
At a local scale, seed dispersal determines the spatial distribution of individuals within
a habitat patch and thus is the major determinant of fine-scale spatial genetic structure
(hereafter SGS) (Heywood 1991; Epperson 1993). SGS is characterized by the degree and
spatial extent of relatedness between adjacent plants (kinship structure), which arises from
restricted seed dispersal near to the source. If in addition, pollen flow is restricted, biparental
inbreeding would increase. Thus SGS influences the outcomes of subsequent ecological
processes such as mating patterns, effective population size, selection, and progeny fitness
(Loveless and Hamrick 1984; Heywood 1991; Ouborg and Van Treuren 1994). Life history
traits such as reproduction type (selfing vs. outcrossing), and the type of seed dispersal vector,
and demographic factors such as population size and history, are likely to affect the strength of
SGS (Loveless and Hamrick 1984; Hamrick et al. 1993; Hamrick and Nason 1996; Epperson
2003; Vekemans and Hardy 2004; Jones et al. 2006; Troupin et al. 2006; Hamrick and
Trapnell 2011).
If grazing by domestic ungulates promotes seed dispersal into populations by long-
distance dispersal (Fisher et al. 1996; Manzano and Malo 2006; Rico et al. 2012) while also
moving seeds within grassland patches, these effects are likely to modify spatial patterns of
9
relatedness within plant populations (Kleij and Steingner et al. 2002; Kloss et al. 2011). In
general, zoochory tends to increase the variance of seed dispersal and deposition patterns
(Jordano et al. 2007). Increased variance in dispersal distances tend to increase the overlap of
offspring (seed shadows) from different mother plants, causing a gradual decay of relatedness
over distance (Degen et al. 2001; Chung et al. 2002; Karubian et al. 2010).This may increase
effective population size, as nearby individuals become more genetically different on average
(Vakemans and Hardy 2004; Hamrick and Trapnell 2011). While long-distance dispersal by
shepherding is recognized as a major determinant of spatial dynamics for grazed calcareous
grasslands (Adler et al. 2001; Kuiters and Huiskers et al. 2010; Reitalu et al. 2010; Wagner et
al. 2012), its potential to contribute to seed-mediated gene flow and to influence spatial genetic
structure across scales has been little investigated.
1.4. Study System: Calcareous Grasslands
Calcareous grasslands (Gentiano-Koelerietum pyramidatae vegetation; Oberdorfer 1978) in
Central Europe are semi-natural communities of medieval origin traditionally used for sheep
grazing and hay production (Ellenberg 1996). They have high floristic and faunistic species
richness of important conservation value (WallisDeVries et al. 2002). Abandonment of
transhumance shepherding during the early 20th century, followed by shrub encroachment and
natural reforestation (Hakes 1987), led to a dramatic decline in the number and extent of
calcareous grasslands with subsequent loss of biodiversity (Poschlod and WallisDeVries 2002;
Butaye et al. 2005). Secondary succession after abandonment of traditional land use practices
typically leads to relatively species-poor beech forests with abundant species that are of
10
limited interest for conservation, whereas calcareous grasslands have high numbers of habitat
specialist plants and arthropod. Within this anthropogenic landscape, which was partly
deforested and subject to traditional land use since Roman times, calcareous grasslands have
become important refugia for numerous endangered species and hotspots for biodiversity at the
landscape scale. As a consequence, this anthropogenic habitat type has received protection
status at regional, national, and European levels.
Habitat specialist plants of calcareous grasslands are highly vulnerable to current trends
of habitat fragmentation (Poschlod and WallisDeVries 2002) as most species do not form a
persistent seed bank and do not disperse far (Coulson et al. 2001; Fenner and Thompson 2005;
Joshi et al. 2006; Zeiter et al. 2006). Thus, it is doubtful whether impoverished patches where
many species have gone locally extinct after abandonment may restore their species richness
without artificial sowing (Poschlod et al. 1996; Kahmen et al. 2002).
Available evidence from connectivity studies in calcareous grasslands showed no
consistent trends. For instance, some studies have found that species richness is mostly
explained by the physical distance among patches (e.g., Brunn et al. 2000; Geertsema 2005;
Adriaens et al. 2006; Joshi et al. 2006; Bruckman et al. 2010), whereas other studies have
found that patch area is the main predictor (e.g., Krauss et al. 2004; Bisteau and Mahy 2005),
while only a few have incorporated the effect of connectivity by shepherding on species
richness (e.g., Reitalu et al. 2010; Wagner et al. 2012).
In the study area in the Southern Franconian Alb (Bavaria, Germany), calcareous
grasslands typically are found along the steep slopes and on un-tillable wasteland of the karstic
plateau. The natural vegetation of the area would be characterized by orchid-rich beech forest
(Cephalanthero-Fagion community) or sedge-rich beech (Carici-Fagetum community;
11
Adalbert 1978). Calcareous grasslands (Gentiano-Koelerietum pyramidatae vegetation;
Oberdorfer 1978) within the study area were grazed since medieval times in approximately ten
communal shepherding systems following defined routes around municipal land (Hornberger
1959; Jacobeit 1961). By the 19th century, regional transhumance shepherding systems
involved the grasslands along the steep slopes of the Franconian Alb, while calcareous
grasslands located on the plateau were still grazed as communal land. During the first half of
the 20th century, due to socio-economic land use changes, calcareous grassland were
progressively abandoned (“previously abandoned patches”), with an estimated decline in the
study area from 970 ha to 302 ha by the early 1990 (Dolek and Geyer 2002). Some of the
larger and more easily accessible grassland patches remained grazed in local herding systems
(“core areas”), and these cores areas typically have high species richness of habitat specialist
plants.
1.5. Study species: Dianthus carthusianorum L.
Dianthus carthusianorum (Caryophyllaceae) is a mainly outcrossing, diploid, and perennial
herb, and a habitat specialist of Central European calcareous grasslands (A1 see Appendix). D.
carthusianorum forms a rosette of unramified shoots of 30 to 35 cm in height, and has tubular-
shaped flowers. Flowering occurs from June to October. Pollination is carried out by
specialized Lepidoptera species such as Satyrus ferula, Papilio machaon, Macroglossum
stellarathum, Thymelicus sylvestri (Bloch et al. 2005). Dianthus carthusianorum do not form a
persistent seed bank, and it reproduces by seed while also vegetatively (Klotz et al. 2002).
Seeds are likely to be dispersed by gravity (authochory) or by wind guts (boleochory) since
12
they lack morphological adaptations to anemochory or zoochory (Klotz et al. 2002). I selected
D. carthusianorum as a model species because it is a characteristic forb species of calcareous
grasslands in the study region (Boehmer et al. 1990), it occurs in many patches in the study
area but never in high densities, and showed numerous colonizations of previously abandoned
patches, increasing its occurrence in such patches from 40% in 1989 to 90% in 2009. In
addition, nuclear microsatellite markers were available for the species or for closely related
species.
1.6. Objectives and Outline of the Thesis
My PhD research profits from an ongoing long-term landscape management project as a
natural experiment. In 1989, the County of Weißenburg-Gunzenhausen initiated a landscape
management project aimed to restore calcareous grasslands by reconnecting core areas with
previously abandoned calcareous grasslands in three non-overlapping herding systems (A1,
Appendix). Large-flocks of approximately 400 to 800 ewes are herded along predefined routes
up to 55 km and covering around 140 ha of calcareous grasslands (Wagner et al. 2012).
Previous empirical evidence for the study area based on a baseline survey from 1989/1990 and
an evaluation survey in 2008/2009 revealed numerous species colonizations in previously
abandoned patches that were reconnected by shepherding as compared to ungrazed patches
(Lehnert 2008). These documented colonizations provided an excellent research opportunity to
study the effects of seed dispersal by rotational shepherding.
The main objective of this research was to investigate the role of directed dispersal in
plant functional connectivity in fragmented calcareous grasslands. By assessing functional
13
connectivity in herbaceous plant communities and individual species, this research helps
overcome the bias towards mammals, amphibians, and birds in empirical studies of functional
connectivity (Fischer and Lindenmayer 2007), and as well as the bias towards tree and shrub
species in landscape genetics studies of plants. The remainder of the thesis comprises four
chapters that each address a main research question as outlined below, followed by a final
synthesis chapter.
Chapter 2 investigates which are the main determinants of functional connectivity
for the community of calcareous grassland plants. Based on mean patch colonization rates
of 48 habitat specialist plants I quantify the relative contribution of (i) pre-dispersal effects
(source patch properties) by including patch area and mean species occupancy; (ii) dispersal
effects measured by five competing models of potential functional connectivity: geographic
proximity between patches, matrix resistance, and three alternative models of dispersal by
shepherding; and (iii) post-dispersal effects (focal patch properties) measured by the number of
dynamic structural elements present in focal patches. This chapter highlights that actual
functional connectivity in plants, as quantified by patch-level colonization rates, cannot be
approximated by structural connectivity based on physical distance alone, but depends on the
number of patches traversed by sheep along shepherding routes, and that predictions of actual
functional connectivity need to consider pre- and post-dispersal processes.
Chapter 3 builds on chapter 2 to contrast the predictions from the community
level assessment of functional connectivity with analysis for individual species of the
same plant community and with molecular data. The specific goals are (i) to assess whether
shepherding effectively promotes dispersal at the landscape scale for a majority of habitat
14
specialist plants irrespective of seed morphological adaptation to zoochory, and (ii) to evaluate
whether pre- and post-dispersal processes importantly contributed to explain patch occupancy
at the species level. For one characteristic species, Dianthus carthusianorum, I obtained
population genetic data from nuclear microsatellite markers (iii) to assess whether assessments
with patch occupancy and genetic data consistently reveal the potential of seed dispersal by
sheep using ecological patch occupancy and population genetic diversity data. By assessing
the consistency between common approaches to quantify connectivity in plants, this chapter
illustrates that when a dispersal vector is shared between species, community-level assessment
may be sufficient to reveal general patterns for a range of species. In addition, this chapter
highlights that contrasting ecological and genetic data in connectivity models can provide a
better understanding of the mechanisms of dispersal and gene flow in plant populations across
the landscape.
Chapter 4 determines the effect of seed-mediated gene flow by large-flock
shepherding on landscape-scale patterns of population genetic structure in Dianthus
carthusianorum. Specifically, I (i) test if there is significant genetic differentiation between
the three non-overlapping shepherding systems, and (ii) contrast the effects of isolation by
geographic distance (IBD) and (iii) determine distance along shepherding routes on population
genetic structure. This chapter provides evidence for a substantial contribution of seed-
mediated gene flow by directed dispersal from shepherding on landscape-scale patterns of
population genetic structure in a calcareous grassland plant.
Chapter 5 focuses on assessing the effect of directed seed dispersal by rotational
shepherding on fine-scale spatial genetic structure (SGS) in Dianthus carthusianorum.
Specifically, I assess if (i) populations of grazed grassland patches show higher genetic
15
diversity and weaker SGS compared to populations of ungrazed grasslands. Moreover, I test if
(ii ) large populations will show higher genetic diversity and weaker SGS than small
populations Furthermore, for populations of grazed grasslands I compare genetic diversity and
SGS between populations of grassland patches colonized since 1989 and pre-existing
populations to infer (iii) if recently colonized populations were founded by colonists from a
variety of populations sources. By contrasting multiple populations (replicates) within a
landscape, this chapter provides evidence on the role of directed dispersal by shepherding on
spatial patterns of genetic structure within populations of D. carthusianorum.
16
Fig. 1 Dianthus carthusianorum L.
17
Chapter 2
Determinants of Actual Functional Connectivity
for Calcareous Grassland Communities Linked
by Rotational Shepherding
Contents of this chapter have been published in the Journal of Landscape Ecology. Permission to use this published material in this dissertation has been obtained from
the publisher (license copyright number 2951391258260): Rico, Y., H. J. Boehmer, H. H. Wagner. 2012. Determinants of actual functional
connectivity for calcareous grassland communities linked by rotational sheep grazing. Landscape Ecology 27: 199-209
A link to the published paper can be found at:
http://www.springerlink.com/content/h4l30247813222t5/
2.1. Abstract
In fragmented landscapes, plant species persistence depends on functional connectivity in
terms of pollen flow to maintain genetic diversity within populations, and seed dispersal to re-
colonize habitat patches following local extinction. Connectivity in plants is commonly
modeled as a function of the physical distance between patches, without testing alternative
dispersal vectors. In addition, pre- and post-dispersal processes such as seed production and
establishment are likely to affect patch colonization rates. Here, we test alternative models of
18
potential functional connectivity with different assumptions on source patch effects (patch area
and species occupancy) and dispersal (relating to distance among patches, matrix composition,
and sheep grazing routes) against empirical patch colonization rates at the community level
(actual functional connectivity), accounting for post-dispersal effects in terms of structural
elements providing regeneration niches for establishment. Our analyses are based on two
surveys in 1989 and in 2009 of 48 habitat specialists in 62 previously abandoned calcareous
grassland patches in the Southern Franconian Alb in Bavaria, Germany. The best connectivity
model Si, as identified by multi-model inference, combined distance along sheep grazing routes
including consistently and intermittently grazed patches with mean species occupancy in 1989
as a proxy for pre-dispersal effects. Community-level patch colonization rates depended to
equal degrees on connectivity and post-dispersal processes. Our study highlights that actual
functional connectivity of calcareous grassland communities cannot be approximated by
structural connectivity based on physical distance alone, and modeling of functional
connectivity needs to consider pre- and post-dispersal processes.
2.2. Introduction
Habitat fragmentation threatens the persistence of plant populations by reducing habitat
connectivity and thus affecting dispersal of pollen and seeds between habitat fragments (Sork
and Smouse 2006). While pollen flow may be sufficient to maintain genetic diversity and
avoid inbreeding effects (Young and Clark 2000; Keller and Weller 2002), habitat re-
colonization after local extinction can only occur by propagule dispersal (Josht et al. 2002).
Given that in most grassland plants propagules are dispersed within 1m in the vicinity of the
19
source (Coulson et al. 2001; Fenner and Thompson 2005), lack of connectivity by seed
dispersal will limit species’ ability to reach empty patches and establish new populations
(Soons et al. 2004). Moreover, the colonization process may increase population genetic
differentiation if propagules come from a few sources only (Withlock and McCauley 1990;
Panell and Dorken 2006), which is likely to be the case with increased fragmentation.
Restoring connectivity is thus expected to avoid deleterious demographic and genetic effects
mostly in small isolated populations (Frakham 2005; Aguilar et al. 2008).
Connectivity comprises two components: structural connectivity, which is defined by
the spatial habitat configuration without reference to organism movement behavior, and
functional connectivity, which refers to the individual behavioral response to the landscape
pattern, including the scaling of inter-patch distances by maximum dispersal distance or the
transversability of different land-covers in the intervening matrix, and the resulting dispersal
and gene flow (Taylor et al. 1993; 2006). Assessing functional connectivity in plants remains a
methodological challenge, because even for a single species it is not feasible to observe
dispersal events over long distances, for more than a single source population, or over multiple
seasons and even less so for entire species assemblages (Fischer and Lindenmayer 2007).
Commonly, empirical studies of plant communities model connectivity as a function of
patch area and geographic distance between patches alone and test model predictions against
species occupancy data, as empirical data on colonization rates are often lacking (but see
Soons et al. 2004; Herrera et al. 2011). These approaches are unlikely to capture the
mechanism behind functional connectivity. As recommended by Murphy and Lovett-Dousst
(2004), it is necessary to incorporate a functional approach for modeling landscape
connectivity including variables associated with effective dispersal trough the matrix (i.e.,
20
dispersal vectors, predation) and with local establishment (i.e., resource availability,
competition), which together determine long-term species persistence. Empirical studies thus
need to combine an assessment of landscape structure with biological assumptions on
organism dispersal into realistic models of potential functional connectivity, and test these
models against empirical estimates of effective dispersal (e.g., colonization rates or migration
events inferred by assignment test with genetic data) that correspond to actual functional
connectivity (Fig. 2; Calabrese and Fagan 2004; Fagan and Calabrese 2006).
Beyond dispersal per se, pre- and post-dispersal processes are likely to influence patch
colonization success (Nathan and Muller-Landau 2000). For instance, population size in
source patches may determine the quality and the potential number of emigrant propagules.
The patch connectivity index, Si, of the incidence function model (IFM, Hanski 1994) includes
the area of source patches as a proxy for population size assuming that the carrying capacity of
the focal patch is proportional to its area (Ovaskainen and Hanski 2004). While most studies
include source patch variables at least in terms of patch area, focal patch properties affecting
establishment have rarely been considered in landscape connectivity models (Clobert et al.
2004). For instance, availability of resources in focal patches would influence seedling
establishment and thus colonization success. Hence, most measures of actual functional
connectivity are likely to confound dispersal effects with post-dispersal processes when testing
predictions of potential functional connectivity. To effectively inform management efforts, it is
crucial to disentangle the contribution of dispersal per se from pre- and post-dispersal
processes. We propose a comprehensive approach to assess and disentangle determinants of
actual functional connectivity for plant species at the community level (Fig. 2). Our approach
considers that actual functional connectivity (colonization rates) depends on the emigrant pool
21
in source patches, a dispersal function reflecting the main dispersal vector, and the
establishment probability of propagules in focal patches.
We apply this framework to study plant community connectivity of calcareous
grasslands in Germany. Calcareous grasslands are semi-natural communities traditionally used
for sheep grazing or hay production (Ellenberg 1996). They typically are nutrient poor,
unfertilized and free of herbicide or pesticide application, resulting in a high floristic and
faunistic species richness of important conservation value (WallisDeVries et al. 2002). In
Central Europe, since the late 19th century abandonment of transhumance sheep grazing
followed by encroachment of natural forest succession led to a dramatic decline in calcareous
grasslands with a consequent loss of biodiversity (Poschlod and WallisDeVries 2002; Bender
et al. 2005). Paradoxically, preservation of natural succession is not a goal of nature
conservation here because it typically leads to relatively species poor beech forests with
predominantly ubiquitous species that are of no particular interest for conservation, whereas
calcareous grasslands are very diverse with high numbers of habitat specialist arthropod and
plant species. Within this old cultural landscape, which was deforested and subject to
traditional land use since Roman times, calcareous grasslands have become important refugia
for numerous endangered species and hotspots for biodiversity at the landscape scale.
In 1989 in the Franconian Alb in Bavaria, Germany, a conservation project was
established aimed to restore abandoned calcareous grasslands and re-connect them with
existing core areas by rotational sheep grazing. Due to extensive baseline data from 1989 for
all previously abandoned patches (Boehmer et al. 1990), this project represents a unique
research opportunity to study connectivity of grassland communities. Based on an evaluation
survey in 2009, empirical colonization rates were observed for all 48 habitat specialist plants.
22
Wagner et al. (2012) showed that rotational sheep grazing significantly increased species
richness of previously abandoned patches, although structural connectivity based on the
physical distance between patches had no effect on species richness. Here, we use patch
colonization rates at the community level as a measure of actual functional connectivity to test
multiple competing models of potential functional connectivity, including pre- and post-
dispersal effects (Fig. 3). By pooling colonization events across the 48 habitat specialist
species, we are able to overcome data limitations that would prevent statistical analysis at the
species level.
2.3. Methods
2.3.1. Study site
The study area of approximately 10 km x 15 km in the Southern Franconian Alb near
Weissenburg, Bavaria, Germany, comprises valleys and limestone plateaus with agricultural
fields, forests, grasslands, orchards, and settlements. Between 1900 and 1960, pasture
abandonment led to a dramatic regional decrease of calcareous grassland cover from 15% to
1% today (Bender et al. 2005). In the study area, calcareous grasslands declined from 970 ha
in 1830 to 302 ha by the early 1990s (Dolek and Geyer 2002). In 1989, the County of
Weissenburg-Gunzenhausen initiated a pilot project aimed to preserve and reconnect
calcareous grasslands by implementing three independent rotational grazing systems, which
connected larger, consistently grazed patches (“core areas”) with previously abandoned
patches experiencing secondary succession (“abandoned patches”; A1 Appendix).
23
2.3.2. Actual functional connectivity data
All previously abandoned calcareous grasslands of at least 25 m2 that had remnants of
Gentiano-Koelerietum pyramidatae vegetation were surveyed as a basis for implementing the
conservation project (Boehmer et al. 1990). During summer and fall of 1989 and spring of
1990, complete species lists of vascular plants were recorded with Braun-Blanquet abundance
information on all 62 previously abandoned calcareous grasslands in the study area (baseline
survey). All 48 habitat specialist species were surveyed again during summer 2008 and spring
and fall 2009 (evaluation survey) on all previously surveyed patches and in all core areas. Core
areas were fully surveyed in the evaluation survey but only aggregate data are available for the
baseline survey, consisting of the frequency of occurrence of each species among 11 sampled
core areas Only previously abandoned patches were included in the statistical analysis, but
core areas were included in the calculation of Si connectivity models.
Consistency between baseline and evaluation surveys was high as both were led by the
same scientist. Comparisons between independent surveys by two different observers
confirmed that species were reliably detected even if reproductive structures were absent,
possibly with the exception of Allium oleraceum. As summer visits for the baseline survey
were primarily done during late summer 1989, this may have affected the detectability of four
plant species that flower during early summer: Leontodon hispidus, Ranunculus bulbosus,
Ajuga genevensis, and Linum catharticum. To assess sensitivity to detectability, we repeated
all analyses without the above-mentioned five species.
We calculated mean patch colonization rate CRi = Ci / (48 – Richness.1989i) at the
community level including all 48 habitat specialist plants. i.e., for each previously abandoned
24
patch i, the number Ci of species present in the evaluation survey, but absent in the baseline
survey was divided by the number of species absent in patch i in the baseline survey. We thus
scaled the observed number of colonization events in each patch by the maximum possible net
number of colonizations among the 48 specialist species.
The interpretation of CRi as community-level patch colonization rate relies on the
assumptions that species were reliably detected in both surveys and that colonization did not
occur from the seed bank. The majority of habitat specialist species in this system is known to
have a transient or short-term persistent seed bank. However, long-term persistent seed bank
has been reported for eight species: Ranunculus bulbosus, Medicago lupulina, Lotus
corniculatus, Sanguisorba minor, Thymus pulegioides, Gentiana cruciata, Euphorbia
cyparissias, and Linum catharticum (Poschlod et al. 2003). To assess sensitivity to seed bank
persistence, we repeated all analyses without these eight species.
2.3.3. Management records
The grazing regime 1989 - 2009 of each of the 62 previously abandoned patches was classified
based on archived management records and current maps of grazing routes combined with
shepherd interviews. Thus, 26 of the previously abandoned patches were consistently grazed
since the beginning of the conservation project, which means that 400 – 800 ewes were herded
through each patch 3-5 times per year; 13 patches were intermittently grazed, i.e., not grazed
all years from 1990 to 2009, or only later in the season, or they were only grazed initially for a
few years after the start of the project. The remaining 23 patches were not included in the three
grazing systems and thus remained ungrazed from 1989 to 2009 (A1, Appendix).
25
2.3.4 Models of potential functional connectivity
We parameterized a patch connectivity index, Si, to test competing dispersal models and
source patch effects. The Si index quantifies distances (dij) between focal patch i and each
source patch j using a negative exponential dispersal kernel with a constant scaling parameter
α (Hanski 1994; Ovaskainen and Hanski 2004). Parameter Aj refers to the area (ha) of source
patch j, and in a single species model, parameter pj indicates source patch occupancy (present
= 1, absent = 0). For our community-level model, we averaged baseline species occupancy pjk
over all 48 species k, resulting in pj = Σk pjk/48, so that 0 ≤ pj ≤ 1. All core areas received the
value of pj = 0.75 derived from the aggregate baseline data. Substituting patch-level occupancy
data from the evaluation survey did not improve model fit.
We modeled the effect of potential dispersal vectors with five alternative dispersal
models (Table 1) by modifying the distance parameter dij in the Si index. The simplest model,
geographic distance (Table 1a), is a null model that assumes seeds are dispersed by wind as a
simple diffusion process without an effect of the landscape matrix. The second model, matrix
resistance (Table 1b), assumes that seeds are dispersed by simple diffusion, but seeds are
intercepted by forest in the intervening matrix. The remaining three models assume sheep to be
the main dispersal vector: the consistently grazed model (Table 1c) assumes that grazing needs
to be consistent, i.e., every year and throughout the season, to effectively transport seeds
between patches along the grazing route; the model implies that distance in terms of the
number of patches traversed between two patches i and j matters (i.e., distance effect). The
consistently or intermittently grazed model (Table 1d) is similar to the previous model, but
26
some degree of grazing is assumed sufficient to effectively transport seeds. The grazed within
the same system model (Table 1e) is a null model for dispersal by sheep as it includes no
distance effect, i.e., it assumes that seeds transported by sheep are equally likely to reach all
grazed patches within the same grazing system.
Dispersal capacity accounted for by parameter α is unknown for our set of 48 specialist
species. In a sensitivity analysis varying α from 0.1 to 50 with increments of 0.1, we found that
α = 0.2 resulted in the best or second best fit between patch colonization rates, CRi, and each
of the five dispersal models (i.e., highest positive Pearson correlation), hence we used this
value for the final models.
2.3.5. Pre-dispersal effects
For each distance measure, dij, we calculated four alternative Si connectivity indices with
different assumptions on source patch effects. Model 1 does not include source patch effects Aj
or pj, and thus assumes that all patches are equal sources of propagules, so that patch
connectivity Si depends only on distance dij between patches. Model 2 includes patch area (Aj),
assuming that seed production is proportional to habitat area. Model 3 includes only mean
patch species occupancy pj. Model 4 includes both source patch parameters, assuming that the
emigrant propagule pool depends on habitat area (Aj) and species occupancy (pj).
27
2.3.6. Post-dispersal effects
We recorded in each previously abandoned grassland how many types of dynamic structural
elements were present that are likely to create regeneration niches for the establishment of
habitat specialist species (post-dispersal effect): rock debris, erosion, ant hills, and small
mammal burrows.
2.3.7. Data analysis
We used multi-model inference with the function dredge in the R library MuMIn to rank each
of the 20 Si connectivity models for explaining patch colonization rates, CRi. For each
parameter in the Si connectivity index, we summed Aikake model weights wm over all Si
models that contained the parameter of interest to assess its relative importance. For the best
performing Si connectivity index, we performed significance tests and residual analysis and
assessed model fit with adjusted R2. Subsequently, for the full regression model including the
best Si connectivity index and the number of structural elements in 2009, we applied variation
partitioning (Legendre and Legendre 1998) to assess the unique and shared contributions of
each factor.
2.4. Results
Based on multi-model inference, consistently or intermittently grazed was by far the best
dispersal model explaining patch colonization rates (CRi: relative importance = 0.96, Fig. 3A).
For the pre-dispersal effects, the best model was Model 3 with mean species occupancy in
28
1989 (pj: relative importance = 0.63, Fig. 3B), followed by Model 1 (null: relative importance
= 0.35, Fig. 3B) that only incorporated distance effects. Interestingly, when source patch area
(Aj) was included (Model 2 and 4) the model performed worse than without pre-dispersal
effects (Fig. 3B). Thus, the best Si connectivity index included dij as consistently or
intermittently grazed and mean species occupancy pj as pre-dispersal effect. This model (AIC
= -62.6, w = 0.46) had well behaved residuals and explained 24% (R2adj, df = 1 and 59, F =
19.5, p = 0.001) of the variation in community-level patch colonization rates CRi. Neither the
geographic distance model (p = 0.28) nor the matrix resistance model (p = 0.15), both
including mean species occupancy pj, were statistically significant for explaining patch
colonization rates CRi.
Post-dispersal effects in terms of the number of structural elements present alone
explained 26% of the variation in CRi (R2adj, df = 1 and 59, F = 22.3, p = 0.0001; Fig. 4).
Combining Si and the number of structural elements in the full regression model significantly
increased the variance explained to 37% (R2adj, df = 2 and 58, F = 18.3, p = 0.0001). This
model had well-behaved residuals without influential outliers. Variation partitioning showed
that unique contribution of the number of structural elements was 13% of the total variation
and the unique contribution of the Si connectivity index was 11%, with a shared variance of
13%. Repeating the analysis without five species with potential issues of detectability between
surveys and eight species likely presenting a long-term persistent seed bank did not change the
nature or statistical significance of the results (Table 2).
The consistently and intermittently grazed model potentially confounds the effects of
connectivity by grazing per se and a distance effect. Omitting the ungrazed patches from the
final model estimation reduced the variance explained overall and both by the connectivity
29
model Si and the structural elements, but all effects remained statistically significant (Table 2).
In contrast, further omitting the distance effect by substituting the grazed within the same
system distance model resulted in an overall non-significant regression model. Similarly,
omitting the pre-dispersal effect pj resulted in non-significant contributions of both the
connectivity model Si and the structural elements (Table 2).
2.5. Discussion
Our results clearly demonstrate the importance of connectivity in terms of sheep as dispersal
vector for patch colonization at the community level for grassland plant species. Wagner et al.
(2012) found no significant effect of connectivity as calculated from the physical distance
between patches, which corresponds to our geographic distance model without pre-dispersal
effects. We tested alternative dispersal models and found strong evidence that dispersal does
depend on distance but in terms of the number of patches that sheep need to traverse between
two sites along grazing routes. This effect remained significant when omitting the variation
due to grazed vs. ungrazed patches. Although we cannot rule out that a species may have been
overlooked in a survey or colonized a patch from the seed bank, our results appear to be robust
as omitting species with potential issues of detectability or ability to form a long-term
persistent seed bank did not change the nature or statistical significance of the results.
Available evidence from empirical studies in calcareous grasslands based on patch
occupancy data showed no consistent trends regarding the effects of habitat loss and isolation.
For instance, some studies found that species richness is mostly explained by geographic
proximity between patches (e.g., Geertsema 2005; Adriaens et al. 2006; Joshi et al. 2006;
30
Bruckman et al. 2010), whereas other studies found that patch area is the main predictor of
patch species richness (e.g., Krauss et al. 2004; Bisteau and Mahy 2005). The lack of robust
and consistent trends may be due to an oversimplified assessment of habitat connectivity and
the use of indirect measures of actual functional connectivity, such as patch occupancy
patterns instead of colonization rates. For these studies, functional connectivity was assumed
to depend exclusively on the source patch area and the physical distances between patches,
which ignored the biological processes behind actual functional connectivity (Taylor et al.
1993; 2006).
Probability of patch re-colonization after extinction will decrease as habitat isolation
increases (Geertsema 2005; Joshi et al. 2006). However, for plants, dispersal vectors that are
likely to transport propagules over longer distances enable plants to partly overcome such
limitation (Bruun and Fritzbøger 2002; Nathan et al. 2008). For habitat specialists of
calcareous grassland pastures (Gentiano-Koelerietum pyramidatae vegetation), sheep are
assumed to act as the main dispersal vector (Fischer et al. 1996). Empirical evidence, however,
is limited to small field experiments measuring seed adhesive potential and seed distance
traveled on few tamed sheep or with experimental coats (Moussie et al. 2005; Manzano and
Malo 2006), by germination experiments of dung samples (Kuiters and Huiskes 2010), or
indirectly inferred by contrasting species occupancy data between grazed patches with varying
grazing history (Reitalu et al. 2010). Our results thus confirm with empirical data at the
landscape scale the role of sheep as dispersal vector for the maintenance of calcareous
grassland biodiversity.
From the strong support for connectivity in terms of consistent or intermittent grazing,
we conclude that some amount of rotational sheep grazing may be sufficient for many species.
31
However, further research is needed to assess whether early flowering species were equally
likely to be dispersed to patches that are grazed only later in the season after crop harvesting in
surrounding fields. The clear support for a distance effect in terms of number of patches
traversed by sheep between two patches suggests that most seeds dispersed by sheep do not
stay on the sheep for a long time. This is consistent with previous experimental results, where
most seeds fell off the wool within the first days, although both morphologically adapted and
non-adapted species were found to persist in the wool for over a month (Fischer et al. 1996).
Repeating our analysis for ungrazed patches only showed no significant effect of
connectivity, neither for simple diffusion models (geographic distance model, p = 0.8) nor for
the models that assume seeds to be intercepted by intervening forest (matrix resistance, p =
0.9). Since some ungrazed patches experienced colonization events, these events may depend
on turbulent wind conditions, where the distance and direction of dispersal may be
unpredictable (Soons et al. 2004; Bolli 2009), or other dispersal vectors such as machinery,
wild or domestic animals, or humans. Further research using molecular methods may be able
to identify likely sources of known colonizations and thus provide further insight into
connectivity.
Source patch area is widely used as a proxy of population size and thus of seed
production (Moilanen and Hanski 2006). However, including patch area decreased rather than
increased explanatory power of the Si connectivity index, suggesting that habitat area may not
be a good proxy for population size for most calcareous grassland plants. Here we did not
include population size data as this would require modeling at the species level with binary
response data, for which a considerably larger data set would be needed.
32
In contrast to habitat area, mean species occurrence in 1989 (pj) significantly improved
the model fit of the Si connectivity index. It is reasonable to expect that if source patches
provide a more diverse propagule pool, community-level patch colonization rates of nearby
connected patches will be higher. This positive association supports existing empirical
evidence showing that after local species extinction, restoration success of calcareous
grasslands depends on diversity of the species pool in nearby patches (Kahmen et al. 2002).
Once a viable seed arrives in a patch of suitable habitat, seedling establishment will be
influenced by species interactions (e.g., predation, competition, population density; Orrok et
al. 2006) and local environmental conditions (regeneration niches, disturbance; Boehmer 1994;
Rusch and Fernandez-Palacios 1995; Willems and Bik 1998). Accounting for post-dispersal
processes in terms of the number of dynamic structural elements providing regeneration niches
improved model fit R2adj markedly from 0.24 to 0.37. While connectivity and post-dispersal
effects were correlated, their unique contributions were of the same magnitude (0.11 and 0.13,
respectively). Given the highly stochastic nature of the dispersal process and the additional
variation that is likely introduced by further post-dispersal factors affecting establishment,
growth and mortality (Clobert et al. 2004), the R2adj of 0.37 can be considered rather high for a
community-level analysis. Further analysis is needed to assess to what degree these results
depend on species traits.
2.6. Conclusions
Our results show that patch colonization rates at the community level for habitat specialist
plant species of calcareous grasslands depend on: (1) the availability of propagules in source
33
patches, (2) the presence of sheep as dispersal vector, at least intermittently, and distance
related to the time consumed by sheep herds between source and focal patches, and (3) the
number of structural elements providing a variety of regeneration niches for propagule
establishment. Modeling connectivity for plant communities based on the physical distance
between patches alone (structural connectivity) without considering dispersal vectors and how
they respond to landscape structure (potential functional connectivity) may lead to erroneous
conclusions about the determinants and importance of functional connectivity in plants. It is
equally important to recognize that measures of actual functional connectivity like
colonization rates are the result of dispersal and post-dispersal processes. Thus post-dispersal
effects may introduce noise when the focus of interest is on dispersal per se. Based on our
comprehensive approach (Fig. 2), it was possible to assess the unique contribution of each
process to actual functional connectivity, which is an important concern for species
conservation.
This study fills an important research gap regarding the determinants of actual
functional connectivity in plant communities (Erik and Priya 2003; Fischer and Lindenmayer
2007) by testing competing connectivity models of seed dispersal with empirical colonization
data at the community level. This direct approach is a considerable improvement over indirect
methods based on patch occupancy data (Fagan and Calabrese 2006). Comprehensive
landscape connectivity assessments that use direct estimates of actual functional connectivity,
such as colonization rates and consider pre- and post-dispersal effects are much needed to
effectively inform conservation efforts aimed to mitigate, revert or prevent biodiversity loss in
fragmented landscapes.
34
Table 1. Description of the distance models included for the estimation of each incidence
function model (IFM) of the patch connectivity Si.
Table 1. Continue on the following page
Distance measure
(dij)
Description Units
a) Geographic distance
Straight line distance from the center of the
focal patch i to the center of each other source
patch j
Km
b) Matrix resistance
Straight line distance cutting through forest
from the center of the focal patch i to the
center of each other source patch j
Km
c) Consistently grazed
Number of patch-to-patch steps from focal
“consistently” grazed patch i to each other
“consistently” grazed patch j within each
grazing system.
A value of 100 was assigned to ungrazed
patches, to “intermittently” grazed patches,
and to grazed patches from different grazing
systems.
Integer
35
Distance measure (dij) Description Units
d) Consistently or
intermittently grazed
Number of patch-to-patch steps from
“consistently” or “intermittently” grazed
focal patch i to each other “consistently” or
“intermittently” grazed patch j within the
same grazing system.
A value of 100 was assigned to ungrazed
patches and to grazed patches from
different grazing systems.
Integer
e) Grazed within the
same system
Same value of 1assigned to all grazed
patches within the same grazing system,
whereas a value of 100 was assigned to
ungrazed patches or grazed patches from
different grazing systems.
1 or 100
36
Table 2. Variation partitioning and sensitivity analysis of the final model. Each line shows the
total variance explained (R2adj), the unique variance explained by post-dispersal effects
(Structural elements) and by the connectivity model (Si only) as well as their shared variance
explained (Shared). Models differ by the species or patches included (Data) and by the
inclusion of pre-dispersal effect pj in the connectivity model Si. Asterisks indicate statistical
significance of the regression model (R2adj) and of partial regression coefficients (based on
Type II sums of squares) for post-dispersal effects (Structural elements) and connectivity (Si
only) (‘***’: p < 0.001; ‘**’: p < 0.01; ‘*’: p < 0.05; ‘.’: p< 0.1, ‘n.s.’: p ≥ 0.1).
Data Distance dij pj R2adj Structural
elements Shared
Si only
All consistently or intermittently grazed
Yes 0.37*** 0.13*** 0.13 0.11***
Without 5 species with detectability issues
consistently or intermittently grazed
Yes
0.31***
0.12**
0.11
0.07**
Without 8 species with long-term persistent seed bank
consistently or intermittently grazed
Yes
0.36***
0.16***
0.13
0.07**
Table 2. Continue on the following page
37
Data Distance dij pj R2adj Structural
elements Shared Si only
Only grazed patches
Grazed within the same system
Yes 0.08 (n.s.) 0.08 0 0
Only grazed patches
consistently or intermittently grazed
No
0.12*
0.07
0.02
0.03
38
Fig. 2 Conceptual model of functional connectivity including effects relating to pre-dispersal (e.g., seed production) and post-dispersal
processes (e.g., availability of regeneration niches) that interact with potential functional connectivity to determine actual functional
connectivity.
Emigrant pool
Dispersal Pre-dispersal
Potential functional connectivity model
Establishment
probability
Post-dispersal
Actual functional connectivity
Landscape structure
Species dispersal behavioral model
39
null pj Aj Aj pj
A B
Fig. 3 Relative importance of parameters in the Si connectivity model. Each bar shows for one
version of the distance parameter dij. (A) or pre-dispersal effects (B): null = no pre-dispersal
effect, pj = mean patch occupancy in baseline survey, Aj = patch area, and Apj = Aj * pj) the sum
of Akaike model weights wm of all candidate models m of connectivity Si containing that
parameter. Abbreviations of the distance models dij: geographic distance (Eu), matrix resistance
(Matrix), consistently grazed (Shecte), consistently and intermittently grazed (Sheint), and
grazed within the same grazed system (Shenu).
Eu Matrix Shecte Sheint Shenu
Aka
ike
mo
del
wei
ght (w
m)
40
(n= 7) (n= 13) (n= 21) (n= 17) (n= 3)
Number of structural elements
0 2 3 1
4
Fig. 4 Side-by-side boxplots of community-level patch colonization rates CRi for different
numbers of structural elements providing regeneration niches (rock debris, small mammal
burrows, erosion, ant hills) present in focal habitat patches in 2009.
Pat
ch c
olo
niza
tion
s ra
tes
41
Chapter 3
Plant Connectivity Assessment at the Community,
Species, and Genetic Level Consistently Reveals
Dispersal by Shepherding
3.1. Abstract
Functional connectivity is species-specific and reflects species response to habitat fragmentation.
Therefore any community-level assessment of connectivity may be insufficient to protect
multiple species simultaneously. Yet when a dispersal vector is shared within a community, the
community-level assessment can provide valuable information about community response to
fragmentation. This chapter evaluates the extent to which the results of species-level assessments
are consistent with the predictions of functional connectivity at the community level based on
aggregate species data for 48 calcareous grassland plants. For 31 species of the same plant
community, I tested competing models of dispersal, as well as source and focal patch effects to
explain patch occupancy using presence-absence data of two surveys from 1989 and 2009. I
evaluated if effects of connectivity by shepherding were limited to species with dispersal
adaptations. Species-level assessments showed patterns consistent with community-level results,
identifying a distance-dependent effect of shepherding connectivity for almost all species.
Population size and patch area of source patches were, in most cases, not important predictors of
42
patch occupancy. While all zoochorous species strongly responded to connectivity by
shepherding, dispersal by sheep was also supported for most species without zoochory
adaptations. Molecular genetic analysis for a selected species, Dianthus carthusianorum, using
eleven microsatellite markers showed consistency between connectivity models based on patch
occupancy and genetic data, identifying sheep as main seed dispersal vector. This study
illustrates that if an effective dispersal vector is shared between species, the community-level
assessment may be sufficient to reveal general patterns and determinants of functional
connectivity for a range of species, even if they vary in dispersal-related traits.
3.2. Introduction
For plants in fragmented landscapes, increased habitat isolation of remnant patches commonly
decreases effective seed dispersal and pollen flow between populations, with negative effects on
population dynamics (Young et al. 1996). Seed dispersal limitation may prevent small, isolated
plant populations from being “rescued” from extinction by seed immigration, and may
jeopardize species persistence by the lack of habitat re-colonization after local extinction (Fig. 5;
Cain et al. 2000; Piessens et al. 2005). Hence, seed dispersal is a crucial process determining
population demographic connectivity and species persistence at the landscape scale (Fig. 5). On
the other hand, population genetic connectivity is provided by the exchange of genes from the
dispersal of both seeds and pollen. Theoretical predictions and empirical evidence suggests that
genetic connectivity via seeds is expected to be lower than the contribution from pollen flow
(Ellstrand 1992; Petit et al. 2005), and pollen is therefore assumed to support genetic
connectivity at the landscape scale (McCauley 1997; Sork et al. 1999; Fig. 5). However, this may
43
not be the case for all plants, for instance, in the case of species whose seeds are effectively
dispersed through zoochory (e.g., Herrera and Jordano 1981; Jordano et al. 2007; Zhou et al.
2007). Lack of gene flow and thus genetic connectivity may lead to increased population genetic
differentiation and to impoverishment of genetic diversity within populations due to increased
rates of inbreeding and drift (Young et al. 1996; Lowe et al. 2005; Aguilar et al. 2008; Vranckx
et al. 2011).
Plant species’ ability to effectively exchange pollen and seeds between populations, i.e.,
plant functional connectivity, largely depends on the interaction of dispersal vectors with the
landscape matrix (Taylor et al. 1993). Functional connectivity is expected to be a species-
specific property, because dispersal vectors and response to the matrix may vary between species
(Taylor et al. 2006). This implies that predictions from functional connectivity assessments may
not be generalizable from one species to another. However, assessing functional connectivity for
each species in an ecological community is impractical, especially for very rare or abundant
species for which statistical analysis is often not feasible. Assessment of connectivity at the
community level, which would aggregate responses of all species in a community, may
overcome the challenge of collecting large amounts of data for individual species. Instead of
fitting a large number of single species models, connectivity may be analyzed by contrasting
species richness among habitat patches varying in spatial isolation (e.g., Lindborg and Eriksson
2004; Geertsema et al. 2005; Brudvig et al. 2009), or by classifying species according to life
history traits associated with dispersal (e.g., Dupre and Ehrlen 2002; Kolb and Diekmann 2005;
Schleicher et al. 2011). Community-level assessment may reveal general patterns of plant species
response to habitat modification and fragmentation (Minor et al. 2009; Schleicher et al. 2011)
and thus provide valuable information for conservation managers aiming to protect multiple
44
species simultaneously. However, whether the predictions at the community level effectively
identify determinants of functional connectivity for a range of species requires empirical testing.
Understanding the mechanisms and patterns of dispersal and gene flow in human
modified landscapes is crucial for managing and conserving plant populations. Commonly,
inference of genetic connectivity is evaluated by the amplification of nuclear markers to compare
the degree of genetic differentiation or genetic diversity between populations varying in spatial
isolation (e.g., Keyghobadi 2005; Honnay et al. 2007; Bizoux et al. 2008; Neel 2008). In plants
estimates of gene flow using nuclear markers reflect gene flow events of both pollen and seeds.
Consequently, the assessment of population demographic connectivity may be an insufficient
predictor of genetic connectivity and vice versa, in the situation where pollen-mediated gene
flow is more extensive than that of seeds. Moreover, assessment of habitat connectivity using
estimates of species dispersal to infer and protect genetic variation at the landscape-scale may
misinform conservation decisions, for instance, of reserve design (Grenwald 2010).
Calcareous grasslands are semi-natural ecosystems traditionally used for sheep grazing,
and are of special conservation concern since they present the highest floristic and faunistic
species richness of Central Europe, including Germany (Poschlod and WallisDeVries 2002). For
most calcareous grassland plants of the Gentiano-Koelerietum pyramidatae vegetation type
(Oberdorfer 1978), sheep have been suggested to promote seed dispersal at the landscape scale
(Fischer et al. 1996; Poschlod et al. 1998; Kuiters et al. 2010). In particular, Rico et al. (2012)
(chapter 2) tested alternative seed dispersal models at the community level and found that
functional connectivity in terms of mean patch colonization rates was explained by a distance-
dependent effect of dispersal along shepherding routes. Furthermore, habitat area of source
patches had no effect, whereas the presence of structural elements in focal patches, which is
45
expected to affect probability of establishment, also contributed to explain patch colonization
rates.
Here, I applied the community-level assessment of functional connectivity of calcareous
grassland specialist plants from chapter 2 (Rico et al. 2012) to all individual species of the same
plant community with sufficient data for statistical analysis. Specifically, I aimed to assess the
extent to which the results from the species-level assessment are consistent with the results of
Rico et al (2012) at the community level. Since calcareous grassland habitat specialist plants
differ in a range of morphological traits related to seed dispersal, I assessed whether all species,
including those without adaptations to zoochory were dispersed by sheep. Subsequently, I
evaluated if patch area and population size in source patches, and the presence of structural
elements in focal patches, importantly contributed to explain patch occupancy for each species.
Finally, for one characteristic calcareous grassland species, Dianthus carthusianorum, I obtained
population genetic data from nuclear microsatellite markers to assess whether connectivity
assessments from patch occupancy and genetic data reflect different patterns of connectivity.
3.3. Methods
3.3.1. Study area and species occupancy data
The study area located in the Southern Franconia Alb near Weissenburg, Bavaria, Germany has
an extent of approximately 10 x 15 km and includes 96 patches of calcareous grassland
embedded in a heterogeneous matrix of settlements, agricultural fields, meadows, orchards, and
forest. In the area, abandonment of traditional shepherding caused a rapid decline of calcareous
46
grasslands from 970 ha in 1830 to 302 ha by 1989 (Dolek and Geyer 2002). Local conservation
agencies initiated a landscape-scale management project in 1989 aiming at restoring and
reconnecting previously abandoned calcareous grasslands (abandoned at least since 1960) with
remaining “core areas” (i.e., calcareous grasslands that were not abandoned) in three non-
overlapping rotational shepherding systems (A1, Appendix). Each herd has approximately 400 to
800 ewes, which are herded along predefined routes of up to 55 km and covering up to 140 ha of
calcareous grasslands (Wagner et al. 2012). Of 62 previously abandoned calcareous grasslands,
26 were grazed three to five times per year since 1989 (“consistently grazed”), 13 were grazed
only later during the season or just for a few years (“intermittently grazed”), and 23 remained
ungrazed; whereas core areas (n = 34) were never abandoned.
All 62 previously abandoned grasslands with remnants of the Gentiano-Koelerietum
pyramidatae vegetation were surveyed in summer and fall 1989 and spring 1990, establishing
complete plant species lists (1989 survey; Boehmer et al. 1990; for further details refer to
Wagner et al. 2012). Vegetation surveys were repeated (by the same scientist) for 48 habitat
specialist plants in all 62 previously abandoned grasslands and all 34 core areas in summer 2008
and spring and fall 2009 (2009 survey). Species presence-absence data of the two surveys were
used in logistic regressions as follows: for previously abandoned patches, each species was
assigned a value of 0 (absent 1989 and 2009), a value of 1 (present 1989 or 2009), or a value of 2
(present in both 1989 and 2009). Core areas were individually surveyed only in 2009, and thus
each species in core areas was assigned either a value of 0 (absent 2009) or 2 (present 2009). To
check the effect of replacing the missing data of 1989 with the data of 2009 for core areas,
statistical analyses were repeated for all 96 patches using the 2009 data only. I did not observe
differences in overall trends for each species (i.e., change in connectivity model selected) and
47
thus I present the results of the 1989-2009 survey data. From the 48 habitat specialist plants,
analyses were performed for only those 31 species for which sufficient data for statistical
analysis was available. These 31 species had an occurrence between 15% and 85% in the 2009
survey.
Population size and patch area in source patches were measured to assess the effect of
source patch properties on patch occupancy. Population size in the 2009 survey was quantified
using four ordinal size classes: 1 = 1-3 individuals, 2 = 4- 39 individuals, 3 = 40- 99 individuals
and 4 = ≥100 individuals. Area of patches was digitized from orthophotos in ArcGis 9.3 to
estimate patch area (ha). Focal patch properties were quantified by counting the number of four
dynamic structural elements present in each patch: rock debris, anthills, small mammal burrows,
and erosion. The presence of structural elements has been suggested to provide microsites for
establishment and thus increasing patch colonization rates of habitat specialist species (Rico et
al. 2012; Wagner et al. 2012).
3.3.2. Functional connectivity model
To determine functional connectivity of populations, I fitted the same Si connectivity models for
each species as proposed in chapter 2 (Rico et al. 2012): �� = ∑ ����−∝ ��� �� · ����� , where
Si is the connectivity of the focal patch i, Aj refers to area of patch j, pj indicates source patch
occupancy (absent = 0, present = 1), dij is the distance between the focal patch i and patch j, and
α is a constant scaling parameter (Hanski 1994). Alternatively, I substituted Aj for the parameter
Nj that refers to population size of patch j.
48
Species dispersal (αdij) was modeled by five alternative dispersal models. (1) The
geographic distance model assumes that seeds are dispersed by a simple diffusion process
without matrix effects (Euclidean distance). (2) The matrix resistance model assumes that seeds
are dispersed by simple diffusion, but seeds are intercepted by intervening forest. (3) The
consistently grazed model assumes that seeds are dispersed along shepherding routes and
includes a distance effect in terms of the number of patches that sheep traverse from patch i to
patch j. The model assumes that grazing needs to occur yearly and three to five times throughout
the season to effectively transport seeds. (4) The intermittently or consistently grazed model
includes all grazed patches not distinguishing between grazing treatments. The model also
includes a distance effect, but intermittent grazing during a few years only or for a few times at
the end of the season is sufficient to effectively transport seeds. The last model (5) grazed within
the same grazing system, differs from the previous two models as it does not include a distance
effect, thus seeds are equally likely to be transported anywhere within the same herding system.
For each dispersal model αdij, I analyzed three models of source patch effects. Model 1
included parameter pj, assuming that all occupied patches are equal sources of seeds. Model 2
included patch area Aj · pj, (in ha) assuming a positive relationship between area and population
size. In model 3, instead of including Aj, I used population size Nj (ordinal categories: 0, 1, 2, 3,
4) from the 2009 survey. I tested additional transformations with interval midpoints for
population size (values of 2, 20, 70, and 150) to assess whether the classification (1 to 4) had an
influence on the results of logistic regressions. I did not observe differences in overall trends,
therefore I present only the results of population size with the four ordinal categories.
The constant α of the Si index is a species-specific dispersal constant that was optimized
for each species by varying α from 0.1 to 10 with increments of 0.1. Because α is a constant
49
scaling the distance parameter dij, the value of α was optimized separately for each dispersal
model using the Si index with the occupancy parameter pj. The α-value with the lowest deviance
of logistic regressions with the two-time survey occupancy data (1989 and 2009) was selected
for parameterizing each dispersal model dij (A2, Appendix). The combination of the five distance
models with the three source patch properties resulted in 15 Si connectivity models for each
species.
3.3.3. Species traits
Species were group by traits related to dispersal: release height, seed shape, seed length,
vegetative form, zoochory, and anemochory dispersal syndrome (Table 3). Seed shape was
estimated as the ratio between seed width and seed length. There was no significant correlation
between traits (pairwise Pearson correlation between quantitative variables, point bi-serial
correlation between quantitative and binary variables), thus I included all selected traits in
statistical analysis. Trait data were obtained from two plant databases: BIOLFLOR (Klotz et al.
2002) and LEDA (Kleyer et al. 2008; Table 3).
3.3.4. Species occupancy analysis
For each species, I fitted logistic regressions using patch occupancy data from the two surveys in
1989 and 2009 as defined above. I used multi-model inference with the function dredge in the R-
library MuMIn (Burnham and Anderson 2002) to rank and summarize each of the 15 Si
connectivity models for predicting occupancy for each species. Akaike model weights wm were
50
summed for each parameter over all candidate Si connectivity models (m) containing the
repetitive parameter to determine its relative importance. I used one-sided significance tests of
the regression coefficient to evaluate only positive associations.
I applied multiple logistic regressions of the best performing Si connectivity index and the
number of dynamic structural elements present on focal patches to test the effect of post-
dispersal processes on patch occupancy. Subsequently, I applied variation partitioning (Legendre
and Legendre 1998) to determine the unique contributions of both predictors to patch occupancy.
I performed a chi-square test to compare the proportion of the best ranked dispersal
models for zoochorous, anemochorous, and species with no specific dispersal adaptations to test
whether the effect of shepherding connectivity was associated with dispersal adaptions. In
addition, to assess if there were dispersal traits that largely contributed to connectivity by
shepherding, I performed multiple linear regressions using the pseudo-R2 of the Si connectivity
index of each species as a measure of the strength of the response to connectivity (response
variable). I applied stepwise model selection with AIC to rank and select the best full linear
regression model. For the best performing model, I performed significance tests and residual
analysis and assessed model fit with adjusted R2. All statistical tests were conducted using R (R
Development Core Team 2010).
3.3.5. Genetic sample collection and microsatellite analysis
I selected Dianthus carthusianorum L. for genetic analysis as this species showed 18 patch
colonizations between 1989 and 2009, and nuclear microsatellite markers were available. D.
51
carthusianorum has no persistent seed bank and no specialized adaptations to wind or animal
seed dispersal (Klotz et al. 2002). In summer 2009, I collected leaf samples from 1,613
individuals from 64 patches (considered here as populations) including core areas, previously
abandoned patches, and six grass verges along roads or forest edges where D. carthusianorum
was also present. In the southern half of the study area, all patches were sampled, whereas in the
northern half, all but 11 patches were sampled. Leaf tissue samples were immediately dried in
silica.
Genomic DNA was extracted following the DNeasy 96 Plant kit protocol (QIAGEN). I
amplified 15 microsatellite loci developed for related Dianthus species (MS-DINMADSBOX,
MS-DCDIA30, MS-DCAMCRBSY, MS-DINCARACC, DCA 221, DCD 224, DCB140,
Smulders et al. 2000; DCB109, Smulders et al. 2003; CB018a, CB057a; CB004a, CB027a,
CF003a, CB011a, CB020a, Kimura et al. 2009). Microsatellite amplifications were done using
the QIAGEN Multiplex kit as follows: 0.2 to 0.4 µl of each primer (5µM), 4.7 µl of Master-mix,
and 5-10 ng of genomic DNA in a total reaction volume of 10 µl. PCR conditions followed those
described in Smulders et al. (2003). Fluorescent labeled PCR products were run on an ABI
3730X Automated Sequencer (Applied Biosystems) with 500 LIZ size standard.
Electrophenograms were analyzed using GENMARKER 1.91(Softgenetics). I detected only a
high proportion of non-amplifications likely due to null alleles at three microsatellite loci
(DCA221, DCD224, DCD140), which were discarded from analysis.
52
3.3.6. Genetic data analysis
Departure from Hardy-Weinberg (HW) equilibrium and linkage disequilibrium (LD) for each
population were tested using probability tests with 10000 Markov chain iterations in GENEPOP
3.4 (Raymond and Rousset 1995). Locus DC020a significantly deviated from HW equilibrium in
50 populations and was discarded from further statistical analyses. Some locus combinations
showed statistically significant LD in some populations, but there were no recurrent patterns
across locus pairs. The final genetic dataset included 1,613 individuals genotyped for eleven
polymorphic microsatellite loci (A3 Appendix). Genetic diversity estimates such as allelic
richness (Ar), observed (Ho) and expected heterozygosity (He) were estimated using FSTAT
(Goudet 2002) and GENALEX (Peakall and Smouse 2006). I corrected for the effect of sample
size in allelic richness measures using rarefaction in HP-RARE (Kalinowoski 2005).
I used the rarified corrected measure of allelic richness (Ar) as response in linear
regression analyses since it was the genetic diversity estimate producing the best model fit (i.e.,
homoscedasticity and residual distribution), and the highest R2 coefficient. Although statistical
inferences were similar (i.e. positive linear relationship) among the three genetic diversity
estimates, logarithmic or square root transformations on Ho and He did not improve model fit in
comparison with models using Ar estimates. Populations with less than four individuals were
excluded from statistical analysis. I applied multi-model inference of multiple linear regressions
to rank the relative importance of each Si parameter on genetic diversity. I performed
significance tests and residual analysis for the best Si connectivity index. Lastly, I added the
number of dynamic structural elements in focal patches with the best performing Si connectivity
index in a full linear regression model. Small populations (n ≤ 4 individuals) were discarded
from statistical analysis.
53
3.4. Results
3.4.1. Functional connectivity models at the species level
Almost all species had significant positive regression coefficients for the best ranked Si
connectivity models with pseudo-R2 values ranging from 0.06 (Ajuga genevensis) to 0.3
(Scabiosa columbaria, Table 4), except for Hippocrepis comosa which had a negative coefficient
for the matrix resistance model. Distance models (dij) of dispersal by sheep were the best ranked
models for 28 of the 31 species tested (Table 4). Specifically, consistently grazed was the most
supported model for 15 species, while consistently or intermittently grazed was the best ranked
model for 13 species. Only for two species, Ononis repens and Euphorbia verrucosa, the
geographic model was the best-ranked dispersal model (Table 4). For 15 out of the 31 species,
patch occupancy pj was the best supported source patch parameter, population size (Nj) was the
best parameter for six species, and patch area (Aj) only for one species (on the basis of relative
importance values higher than 0.6; Table 4). Thus, the best two performing Si connectivity
indices included dij as consistently grazed, followed by consistently or intermittently grazed, both
with pj.
Post-dispersal effects quantified by the number of structural elements in focal patches
significantly increased the model variance explained in most species (except Phleum phleoides),
resulting in pseudo-R2 values ranging from 0.15 (Pulsatilla vulgaris and Koeleria pyramidata) to
0.38 (Cirsium acaule, Table 4). Variation partitioning showed that Si connectivity alone had a
larger contribution than the number of structural elements for 22 of the 31 species. The opposite
was shown for three species (Prunella grandiflora, Carlina acaulis, and Phleum pheloides;
Table 4).
54
There were no significant differences in the proportion of the best ranked models selected
among species with adaptations to zoochory, anemochory, or without dispersal adaptations (X 2 =
7.4, df = 6, p = 0.28; Fig. 6). I further aimed to predict the strength of association with
connectivity Si from dispersal-related traits. The best lineal regression model based on stepwise
model selection of the presudo-R2 of the Si connectivity index of each species (response variable)
included zoochory and seed length as predictors. This model was significant and showed the
lowest AIC (-160.3) with R2adj = 0.17 (df = 2 and 27, F = 4.07, p = 0.03), and it had well-behaved
residuals without influential outliers. However, only zoochoory was statistically significant.
3.4.2. Connectivity effects on patch occupancy and gene flow in Dianthus
carthusianorum
Multi-model inference showed that dispersal models of shepherding were the best ranked models
for both patch-occupancy and genetic diversity data of D. carthusianorum (Fig. 7). For both data
types, the consistently or intermittently grazed model was the best ranked dispersal model
(relative importance value wm of 0.92 and 0.75; Fig. 7a and 7c respectively). For the patch-
occupancy data, model selection was not conclusive for neither of the source patch properties
(Fig. 7b), whereas for the genetic diversity data, presence of the species pj had higher support (wm
= 0.6) than population size (wm = 0.37) and with no support for patch area (wm = 0.02; Fig. 7d).
Results from linear regression analysis of the best performing Si connectivity index with
genetic diversity data (consistently and intermittently grazed with population size) showed a
significant positive relationship of patch connectivity Si and allelic richness (Ar) (Fig. 7). This
model had well-behaved residuals and explained 17% of the total variation (R2adj = 0.17, F =
55
12.47, 1 and 55 df, p = 0.0008). Including the presence of the number of dynamic structural
elements in focal patches in a full regression model had no effect (R2adj = 0.05 df = 2 and 45, F =
2.2, p = 0.12).
3.5. Discussion
3.5.1. Connectivity by shepherding supports dispersal of calcareous grassland
plants
For most of the 31 species for which species-level models could be fitted, I found substantial
consistency with the functional connectivity assessment at the community level (Rico et al.
2012) identifying: (i) shepherding as the main dispersal vector, with a distance-dependent effect,
(ii) a lack of association between focal patch occupancy with source patch area and population
size, which suggests that species merely have to be present in the landscape to act as source for
habitat colonization along sheep grazing routes, and (iii) the occurrence of structural elements as
a significant post-dispersal effect for establishment in focal patches.
While consistently or intermittently grazed was the best ranked model at the community
level (Rico et al. 2012), at the species level the consistently grazed model was selected for 48%
of the species, whereas consistently or intermittently grazed was the best ranked model in 42% of
the species. In contrast, for most of the 31 species, neither geographic isolation nor the context of
the intervening matrix (forest) explained patch occupancy. The two selected shepherding models
contained a distance-dependent effect in terms of the number of patches that sheep traverse along
the route. The clear support of a distance-dependent effect suggests that most seeds do not stay
56
on the sheep for a long time. This interpretation is consistent with previous experimental results
where a high proportion (> 50%) of seeds fell off the wool within the first days (Fischer et al.
1996), although some proportion of seeds were found to persist adhered in the wool for more
than a month (Fischer et al. 1996; Manzano and Malo 2006) and to travel long distances over
400 km (Manzano and Malo 2006). For species in which the consistently grazed model was the
best predictor of patch occupancy, intermittent grazing may not be effective for propagule influx.
However, this interpretation needs to be taken with caution since I cannot rule out other
confounding effects, for instance, related to site characteristics.
Species with adaptation to zoochory and with longer seeds tended to show a stronger
response to dispersal models of shepherding than other species, though only zoochory had a
statistically significant effect. Observational and experimental studies have shown a higher
probability of seed attachment to the fur by specialized seed appendages related to anemochory
or zoochory (e.g., hooks, hairs, sticky coats) and by longer seeds that are morphologically more
likely to be caught in wool and thus remain attached to the fur (Fisher et al. 1996; Couvrer et al.
2005). Although there is a strong association for zoochorous species, this result does not imply
that species without adaptations to zoochory are not transported by sheep. The above is evident
from the lack of significant differences between the proportions of dispersal models selected for
zoochorous, anemochorous, and species without specific adaptations to dispersal. In fact, more
than half of the latter species responded strongly to dispersal by shepherding (Fig. 6). These
findings of connectivity provided by shepherding regardless of dispersal mode is in agreement
with experimental studies that have found seeds without dispersal adaptations attached to the fur
in substantial numbers (Fisher et al. 1996; Couvrer et al. 2004; Rommerman et al. 2005).
Remarkably, I did not find an effect of release height. Adhesion to the wool is less likely for
57
plants that release seeds at height lower than 40 cm (Mouissie et al. 2005). However given that
there was no such effect, seeds may also be effectively dispersed by hooves or dung of sheep
(Fisher et al. 1996).
For 35% of the species, no species-level models could be fitted due to insufficient
numbers of presences or absences in habitat patches, which highlights data limitations for
individual species for statistical analysis. For instance, habitat specialist species such as
Brachyopodium pinnatum or Festuca ovina were very abundant and occurred in more than 90%
of the patches, whereas e.g., Campanula glomerata and Gentiana germanica were rare species
that occurred in less than 5% of the patches. Among the 17 species for which dispersal models
could not be fitted, nine (all zoochorous) were frequent species (occurring in ≥ 85% of the
patches), and four (two anemochorous and two without dispersal adaptations) were rare (< 15%
of the patches).
It is possible that grazing per se may affect ecological conditions of individual patches,
such as gap creation by trampling, removal of biomass, or others, thus indirectly influencing
establishment, a previous analysis in this same study system teased apart both confounding
effects by showing that sheep grazing connectivity (as quantified by the Si index from Rico et al.
2012) rather than grazing treatments had a higher power to explain the increase in species
richness in previously abandoned patches based on the surveys of 1989 and 2009 (Wagner et al.,
2012). The results from the individual species assessments thus confirm the important effect of
sheep grazing to support species dispersal of calcareous grassland specialist plants.
58
3.5.2. Contribution of source and focal patch properties to landscape species
occupancy
Similar to the community-level assessment (Rico et al. 2012), including patch area as a proxy of
seed production reduced model fit compared to patch occupancy alone. Interestingly, the results
showed that population size was also not a good source patch predictor. In the field data, I did
not find a strong relationship between species-specific population size and patch area (results not
shown). Taken together, these results indicate that independently of patch area or population
size, presence of a species in source patches was sufficient to sustain colonization of focal
patches (Hanski et al. 2004; Tremlova and Munzbergova 2007). This suggests that landscape-
scale occupancy patterns for most calcareous grassland plants are largely related to the
distribution of patches acting as potential sources, rather than local abundances in source
patches; a fact that simplifies parameterization of landscape connectivity models.
The presence of the number of dynamic structural elements in focal patches, related to
post-dispersal effects, increased the total variance in patch occupancy explained for most species.
This result was in agreement with the community-level analysis (Rico et al. 2012). However,
variation partitioning showed that Si connectivity (including pre-dispersal and dispersal
processes) had a larger contribution to explain patch occupancy for 88% of the species, except
for Ajuga genevensis, Carlina acaulis, and Prunella grandiflora in which structural elements in
focal patches had a higher contribution. This result underscores that patch connectivity supported
by shepherding was the main predictor of patch occupancy (Piessens et al. 2005; Bruckmann et
al. 2010).
59
3.5.3. Consistency of ecological and genetic data in functional connectivity
assessments of Dianthus carthusianorum
Both connectivity models for D. carthusianorum, with patch-occupancy and with genetic
diversity based on the corrected rarified measure of mean allelic richness (Ar), identified the
‘consistently or intermittently grazed’ model as the best-ranked dispersal model. Although
estimates of population genetic diversity are an indirect measure of gene flow between
populations, the clear support of a seed dispersal model by sheep suggests the potential influence
of seed dispersal on genetic connectivity for D. carthusianorum. This finding was remarkable
given that genetic diversity estimates from nuclear markers incorporate gene flow from both
seeds and pollen, so that we might expect a lack of association with seed dispersal models in the
case of a large contribution of pollen flow to genetic connectivity. In conjunction with recent
evidence, the results emphasize that gene flow mediated by seed dispersal at the landscape-scale
may be more important than commonly expected (e.g., Bacles et al. 2006; Iwaizumi et al. 2010;
Freeland et al. 2012). However, these interpretations are limited since I cannot directly estimate
the relative contribution of pollen vs. seeds to gene flow.
The higher regression coefficient of the connectivity model that included as a response Ar
instead of Ho and He is likely related to the higher sensitivity of the number of alleles to
contemporary changes of habitat fragmentation, while measures that take into account the
relative frequency of alleles such as Ho and He, are expected be affected at slower rates than Ar.
This difference has previously been reported in similar studies of habitat fragmentation effects
on population genetic diversity in plants (Honnay and Jacquemyn 2006; Aguilar et al 2008). On
the other hand, presence of structural elements in focal patches had a significant effect on patch
occupancy, but no influence on population genetic diversity. While the presence of dynamic
60
structural elements may be related to establishment of seeds (Wagner et al. 2012), this not
necessarily implies that larger numbers of established seeds in local patches would have a direct
effect on allelic richness within populations.
3.6. Conclusions
To effectively protect species and maintain species diversity at the landscape scale, we need to
understand what determines functional connectivity. Performance and predictive ability of
connectivity models are constrained by a trade-off between data amount (i.e., number of
populations sampled), scale of analysis (i.e., local, landscape scale), and choice and number of
target species. This study illustrates that if a dispersal vector is shared among plant species,
community-level assessment may be sufficient to identify determinants of functional
connectivity for a broad range of species. However, I recognize that connectivity is a species-
specific property (Taylor et al. 2006) and variation in species-specific responses should not be
neglected, especially for species of conservation concern.
Contrasting ecological and genetic data can provide insights in the mechanisms of plant
dispersal and gene flow at the landscape scale. The results on the basis of genetic diversity and
patch-occupancy data for D. carthusianorum revealed that dispersal by sheep not only explained
patch occupancy, but may also have contributed substantially to population genetic diversity.
Further research is needed to assess the relative contributions of gene flow by seeds vs. pollen to
population genetic structure.
61
Table 3. Description of dispersal traits included in Pearson correlation analyses
Dispersal trait Values Sample size Source
Seed length Continuous (mm) 30 Klotz et al. 2002
Seed shape Continuous (mm) 30 Klotz et al. 2002
Anemochory dispersal 1: present
0: absent
8
22
Klotz et al. 2002
Zoochory dispersal 1: present
0: absent
13
18
Klotz et al. 2002
Vegetative propagation 1: below ground
0: above ground
14
4
Klotz et al. 2002
Release height class 1: max < 30cm
2: min < 30 and max ≥ 30cm
3: max & min ≥ 30cm
8
6
17
Kleyer et al. 2008
62
Table 4. Best performing Si connectivity index parameters as ranked by the sum of Akaike model weights, and relative
variance contribution of each factors to the total variance of the final model.
Best Si index c Final model
d
Species Dispersal mode
a
N b Distance dij Source
Si
z-coefficiente
Si Elem Pseudo-R
2 e
Sanguisorba minor Animal 79 Sheint (0.97) pj (0.77) 1.07*** 0.04 0.07 0.24*
Arabis hirsute Animal 40 Shecte (0.9) pj (0.93) 1.047*** 0.22 0.02 0.30**
Centaurea jacea Animal 76 Sheint (0.88) Nj (0.47) 0.62** 0.03 0.0 0.07*
Koeleria pyramidata Animal 66 Sheint (0.99) pj (0.85) 0.716*** 0.06 0.03 0.15**
Linum catharticum Animal 48 Shecte (0.92) pj (0.68) 1.175*** 0.18 0.05 0.33**
Medicago lupulina Animal 62 Shecte (0.93) pj (0.84) 0.772*** 0.09 0.03 0.19**
Plantago media Animal 77 Shecte (1.0) Nj (0.7) 1.45*** 0.11 0.05 0.24**
Polygala comosa Animal 48 Shecte (0.76) pj (0.88) 1.152*** 0.15 0.08 0.35*
Table 4. Continue on the following page
63
Best Si index c Final model d
Species Dispersal mode a N
b Distance dij Source Si
z-coefficiente
Si Elem Pseudo-R
2 e
Prunella grandiflora Animal 73 Sheint (0.9) pj (0.62) 0.886*** 0.07 0.13 0.29**
Ranunculus bulbosus Animal 71 Shecte (0.92) Nj (0.59) 0.976*** 0.16 0.04 0.30*
Salvia pratensis Animal 75 Sheint (0.63) Aj (0.61) 0.932*** 0.07 0.05 0.16**
Scabiosa columbaria Animal 67 Shecte (1.0) Nj (0.99) 1.614*** 0.24 0.02 0.30**
Hieracium pilosella Wind 57 Shecte (0.98) pj (0.86) 1.332*** 0.21 0.04 0.37**
Leontodon hispidus Wind 54 Sheint (0.97) Nj (0.8) 1.352*** 0.20 0.05 0.35**
Anthyllis vulneraria Wind 28 Shecte (0.98) pj (0.91) 0.894*** 0.18 0.01 0.21 **
Campanula rotundifolia Wind 77 Shecte (0.75) pj (0.66) 0.78*** 0.07 0.08 0.22 **
Carlina acaulis Wind 62 Shecte (0.76) Nj (0.4) 0.66*** 0.04 0.07 0.17 *
Table 4. Continue on the following page
64
Best Si index c Final model d
Species Dispersal a mode N
b Distance dij Source Si
z-coefficiente
Si Elem Pseudo-R
2
Cirsium acaule Wind 52 Sheint (0.99) Nj (0.99) 1.401*** 0.26 0.04 0.38**
Pulsatilla vulgaris Wind 44 Shecte (0.63) Nj (0.59) 0.796*** 0.12 0.01 0.15**
Ajuga genevensis None 52 Shecte (0.84) pj (0.43) 0.586*** 0.02 0.07 0.20*
Asperula cynanchica None 49 Sheint (0.98) pj (0.98) 1.038*** 0.20 0.06 0.20**
Dianthus carthusianorum None 74 Sheint (0.97) pj (0.79) 0.919*** 0.10 0.06 0.20**
Euphorbia cyparissias None 77 Sheint (0.8) Nj (0.78) 1.01*** 0.08 0.02 0.14**
Euphorbia verrucosa None 26 Eu (0.9) pj (0.61) 1.014*** 0.20 0.0 0.22**
Hippocrepis comosa None 66 Matrix (0.99) pj (0.77) n.s NA NA N.A
Ononis repens None 59 Eu (0.44) Nj (0.52) 0.698*** 0.10 0.06 0.19**
Ononis spinosa None 17 Shecte (0.77) Nj (0.92) 1.2*** 0.30 0.02 0.33***
Table 4. Continue on the following page
65
Best Si index c Final model d
Species Dispersal mode a N
b Distance dij Source Si
z-coefficiente
Si Elem Pseudo-R
2 e
Onobrychis viciifolia None 39 Sheint (0.99) pj (0.62) 0.123*** 0.12 0.020 0.17**
Phleum phleoides None 34 Shenu (0.49) pj (0.52) 0.447*** 0.01 0.08 0.12n.s.
Stachys recta None 19 Sheint (0.89) pj (0.45) 0.92*** 0.17 0.03 0.20**
Trifolium montanum None 41 Shecte (0.95) pj (0.45) 1.143*** 0.19 0.03 0.31**
a Number of occupied patches in the 2009 vegetation survey for each of 31 species. b Dispersal mode of each species (Klotz et al. 2002). c Relative importance values shown in parentheses and z-coefficient of logistic regressions of best performing Si index and patch
occupancy. Definitions of distance dij models: geographic distance (Eu), consistently grazed (Shecte), consistently or intermittently
grazed (Sheint), grazed within the same grazing system (Shenu), and matrix resistance (matrix). Source patch parameters: patch
occupancy pj, population size Nj, and patch area Aj. d Variation partitioning analysis of best connectivity index (Si), number of structural elements present (Elem); Pseudo-R2 value of
the final model. e Statistical significance of one-sided tests ***p < 0.0001; **p < 0.001; *p < 0.01; n.s. p ≥ 0.1.
66
Fig. 5 Conceptual diagram showing the main roles of pollen flow (genetic connectivity) and seed dispersal (demographic
connectivity) to determine landscape connectivity. Grey boxes indicate determinants and resulting effects, while white boxes
represent the biological processes of dispersal and gene flow with resulting effects on species persistence. Direct arrows indicate
effects and the line width main contribution (i.e., role at the landscape-scale).
Pollen flow/ // dispersal
Range expansion
Extinction rescue
Gene flow
Gene flow
Colonization
Recruitment
Genetic connectivity
Demographic
connectivity
Vectors and matrix interactions
Seed dispersal
Genetic diversity maintenance/ reduced genetic differentiation
Genetic diversity maintenance/ reduced genetic differentiation
67
Fig. 6 Percentage of the best ranked dispersal models for zoochorous (n = 12) and anemochorous
(n = 7) species, and for species with no specific adaptations to dispersal (n = 11). Dispersal
model abbreviations: geographic distance (Eu), consistently grazed (Shecte), consistently or
intermittently grazed (Sheint), and grazed within the same grazing system (Shenu). For none of
the 31 species was matrix resistance selected as the best ranked dispersal model with significant
positive regression coefficients.
0%
10%
20%
30%
40%
50%
60%
70%
80%
90%
100%
Zoochory Anemochory None
Shenu Eu Sheint ShecteP
erce
nta
ge
of s
pec
ies
68
Fig. 7 Relative importance of each parameter in the Si connectivity index for Dianthus
carthusianorum based on the sum of Akaike model weights (wm) over all Si candidate models
(m) containing the same parameter. Each bar shows one version of the distance (dij) or source
patch parameters (pj, Aj, and Nj) for models of patch-occupancy data (A and B, n = 96) and for
models on genetic diversity (C and D, n = 58). Distance model abbreviations: geographic
distance (Eu), consistently grazed (Shecte), consistently or intermittently grazed (Sheint), grazed
within the same grazing system (Shenu), and matrix resistance (Matrix).
Eu Matrix Shecte Sheint Shenu
Eu Matrix Shecte Sheint Shenu
Aka
ike
mo
del
wei
ght (w
m)
Aka
ike
mo
del
wei
ght (w
m)
pj Aj pj Nj
pj Aj pj Nj
A) B)
C) D)
69
Chapter 4
Directed dispersal by grazing affects seed-
mediated gene flow
4.1. Abstract
Directed seed dispersal by animal vectors can have a large effect on the structure and dynamics
of plant populations and has been found to influence genetic structure in plants dispersed by
frugivores. Yet, empirical data are lacking on the potential of directed seed dispersal by grazing
of domestic animals to mediate gene flow across the landscape. Here, I investigated the potential
effect of large-flock shepherding on landscape-scale genetic structure in the calcareous grassland
plant Dianthus carthusianorum. Based on eleven nuclear microsatellite loci, I found a significant
pattern of genetic structure differentiating calcareous grassland patches of three non-overlapping
shepherding systems and populations of ungrazed patches. Among ungrazed patches, I found a
significant and strong effect of isolation by distance (IBD; Mantel correlation = 0.55, p = 0.001).
In contrast, genetic distance between grazed patches within the same herding system was
unrelated to geographic distance but significantly related to distance along shepherding routes,
i.e., the number of intervening patches traversed by sheep (Mantel correlation =0.45, p = 0.001).
The distance-dependent effect of shepherding connectivity suggests that gene flow occurs mostly
between adjacent populations similar to the stepping stone model of gene flow. While this study
70
used nuclear markers that integrate gene flow from pollen and seeds, the significant differences
in genetic structure between ungrazed patches and patches connected by large-flock shepherding
indicate a substantial potential of directed dispersal by grazing to mediate patterns of gene flow
across the landscape-scale.
4.2. Introduction
Seed dispersal and pollen flow are central processes influencing ecological and evolutionary
dynamics of plant populations across the landscape. Seed dispersal supports colonization and
recruitment, and influences the spatial structure of populations (Nathan and Muller-Landau 2000;
Levin et al. 2003; Clobert et al. 2004), whereas both seed dispersal and pollen flow contribute to
gene flow between populations (Levin 1981; Hamrick and Trapnell 2011). Because pollen is
likely to travel across the landscape over larger distances and in larger numbers, gene flow
mediated by pollen is expected to be much more substantial than seed-mediated gene flow
(Ellstrand 1992; Ennos 1994; McCauley 1997; Petit et al. 2005). However, empirical evidence
from a growing number of species suggests that rates of seed-mediated gene flow depend to a
large extent on seed dispersal mechanisms (e.g., Cruse-Sanders and Hamrick 2004; Jordano et al.
2007; Zhou et al. 2007; Freeland et al. 2012).
Understanding the link between spatial patterns of genetic structure and dispersal
mechanisms is fundamental for the long-term conservation of plant populations in human-
modified landscapes. Plants exhibit a large variety of seed dispersal vectors, which affect the
distance, direction, and destination at which seeds are deposited away from the source. For
instance, seeds of wind-dispersed species can travel longer distances (Tackenberg 2003; Soons et
71
al. 2004), but the chance to arrive at suitable sites is often stochastic (Nathan 2000; 2006). In
contrast, seed-dispersal by animals often is non-random (Kollman 2000; Spiegel and Nathan
2007), and some vectors may disperse seeds in substantial numbers into sites with high
probability of establishment (Howe and Smallwood 1982; Wenny 2001). This process known as
directed dispersal has been shown to influence spatial dynamics of plant populations (Aukema
and del Rio 2002; Purves et al. 2005; Briggs et al. 2009), but less evidence has been gathered on
its effects to influence spatial patterns of gene flow across the landscape. Most empirical studies
have focused on species dispersed by endozoochory, especially trees and shrubs adapted to
frugivory (Godoy and Jordano 2001; Jordano et al. 2007). Several studies showed that directed
dispersal tends to generate strong spatial genetic structure resulting from the aggregated
deposition of related individuals in particular microsites for establishment, even in cases where
there is extensive pollen flow (Grivet et al. 2005; García et al. 2007; 2009; Torimaru et al. 2007).
Yet, the opposite effect has also been found, where recurrent directed long-distance dispersal by
frugivores over time resulted in spatial homogenization of genetic variation (Karubian et al.
2010). Such contrasting patterns highlight the complexity of interactions between plants and their
dispersal vector.
Empirical data on the effect of directed dispersal on gene flow are biased towards
frugivore-dispersed tree species of temperate and tropical ecosystems. However, zoochory by
large flocks of domestic ungulates may play an important role for maintaining grassland
connectivity, as in the highly fragmented semi-natural calcareous grasslands in Central Europe
(Fisher et al. 1996; Couvrer et al. 2005). In these low-intensity agricultural systems, large flocks
of sheep, horses, or goats have the potential of transporting a large number of seeds over long
distances for a range of habitat specialists, which include mostly forbs and grasses (Janzen 1984;
72
Moussie et al. 2005; Manzano and Malo 2006). Since herds move along predefined grazing
routes, seeds dispersed either by seed attachment to the fur or hooves or by seed consumption
have a higher chance to be deposited in distant grasslands across the landscape (Fischer et al.
1996; Cosyns et al. 2005; Bruun and Poschlod 2006; Auffret et al. 2012). Despite the importance
of directed dispersal to largely influence the structure and composition of grassland plant
communities (Bruun and Fritzbøger 2002; Schrautzer et al. 2009; Kuiters and Huiskes 2010;
Piqueray et al. 2011; Wagner et al. 2012), to date there is no empirical information on the effects
of directed dispersal by grazing on patterns of gene flow across a landscape.
In this study chapter, I assessed whether directed dispersal by large-flock shepherding has
affected landscape-scale patterns of genetic structure. As study species, I selected the calcareous
grassland habitat specialist Dianthus carthusianorum. In the study area, most calcareous
grasslands are grazed by sheep in three non-overlapping herding systems (between 400-800
ewes) following defined routes up to 55 km and covering approximate 140 ha of calcareous
grasslands (Wagner et al. 2012). If directed dispersal by shepherding is substantial, I
hypothesized that (i) spatial patterns of genetic structure are in association with shepherding
systems. Subsequently, by comparing genetic distances of populations of grazed vs. ungrazed
patches, I hypothesized that (ii) genetic structure in ungrazed patches can be explained by
isolation by geographic distance (IBD) resulting from a lack of connectivity from seed dispersal
by shepherding, and that (iii) populations connected within the same herding system show
population genetic structure associated with directed dispersal along shepherding routes.
73
4.3. Methods
4.3.1. Study site and species
The study area was located in the Southern Franconian Alb near Weissenburg, Bavaria Germany,
and covered approximately 10 x 15 km. In this region, calcareous grasslands of the Gentiano-
Koelerietum pyramidatae vegetation association (Oberdorfer 1978) declined from 970 ha in 1830
to 302 ha by 1989 due to the abandonment of traditional shepherding practices (Dolek and Geyer
2002). In 1989, a conservation project was started to reconnect previously abandoned calcareous
grassland patches (abandoned at least since 1960) with still grazed high-quality grasslands (“core
areas”) in three non-overlapping shepherding systems. Today, sheep flocks of approximately 400
to 800 ewes are herded in both directions following predefined routes (A1 Appendix1). Grazing
season lasts from March until early November. Of the 62 previously abandoned calcareous
grasslands, 26 were grazed three to five times per year since 1989, 13 were only grazed later in
the season or only for a few years, and the remaining 23 grasslands were not grazed. Core areas
(n = 34) had never been abandoned.
Dianthus carthusianorum L. (Caryophyllaceae) is a perennial herbaceous habitat specialist of
calcareous grasslands in the region (Oberdorfer 1978). The species is diploid and has an
outcrossing mating system (Bloch et al. 2005). There is no persistent seed bank (Klotz et al.
2002). Flowering time is from June to October. Pollination is by few specialized Lepidoptera
species (Bloch et al. 2005), and the species has no specialized adaptations for seed dispersal by
wind or animals (Klotz et al. 2002). However, since release height is 30-35 cm and seed
production falls within the main period of grazing in the area, seeds dispersal by attachment to
74
the fur is possible, as well as attachment to hooves or transportation via ingestion and dung
deposition (endozoochory).
4.3.2. Sampling and microsatellite analysis
A total of 58 calcareous grassland patches (here considered as populations) including core areas
and previously abandoned patches were sampled, representing about 86% of the identified
populations of D. carthusianorum in the study area. In the southern half of the study area, all
existing populations were sampled (n = 38). In addition to the above 58 populations in calcareous
grassland patches, D. carthusianorum also occurred and was sampled along six grass verges
along the roads or along forest edges. I collected leaf material from 30 to 40 individuals sampled
across the patch for grasslands with more than 40 individuals, whereas all individuals were
sampled from patches with less than 40 individuals. Leaf tissue samples were immediately dried
in silica. Genomic DNA was extracted following the DNeasy 96 Plant kit protocol (QIAGEN). I
amplified 15 microsatellite loci developed for related Dianthus species (MS-DINMADSBOX,
MS-DCDIA30, MS-DCAMCRBSY, MS-DINCARACC, DCA 221, DCD 224, DCB140;
Smulders et al. 2000; DCB109, Smulders et al. 2003; CB018a, CB057a; CB004a, CB027a,
CF003a, CB011a, CB020a; Kimura et al. 2009). Microsatellite amplifications were done with the
QIAGEN Multiplex kit as follows: 0.2 to 0.4 µl (5µM) of each primer, 4.7 µl of Master-mix, and
5-10 ng of genomic DNA in a total reaction volume of 10 µl. PCR conditions followed those
described in Smulders et al. (2003). Fluorescent labeled PCR products were run on an ABI
3730X Automated Sequencer (Applied Biosystems) with 500 LIZ size standard.
Electrophenograms were analyzed using GENMARKER 1.91(Softgenetics). I detected a high
75
proportion of non-amplifications at three microsatellite loci (DCA221, DCD224, DCD140),
which were discarded from analysis.
Departure from Hardy-Weinberg (HW) equilibrium and linkage disequilibrium (LD) for
each population were tested using probability tests with 10000 Markov chain iterations in
GENEPOP 3.4 (Raymond and Rousset 1995). Locus DC020a significantly deviated from HW
equilibrium in 50 populations and was discarded from further statistical analyses. Some locus
combinations showed statistically significant LD in some populations, but there were no
recurrent patterns across locus pairs. The final genetic dataset included 1,613 individuals
genotyped at eleven polymorphic microsatellite loci (A3 Appendix).
4.3.3. Analysis of landscape-scale patterns of genetic structure
I performed principal component analysis (PCA) of population allele frequencies to assess
patterns of population genetic structure at the landscape scale as constrained by shepherding
systems using the function dudi.pca of the adegenet package (Jombart 2008) in the R software
(R development core team 2010). The first three PCA axes containing the largest variance were
retained for an interclass PCA among four groups: populations within each of the three
shepherding systems and populations in ungrazed patches. The interclass PCA was applied using
the function bca of the ade4 package (Dray and Dufour 2007) in the R software. Because the
first interclass PCA axes maximize the variance explained constrained by the defined groups, I
used the PCA scores to test whether populations from the three shepherding systems and
ungrazed patches significantly differ. Pairwise comparisons with Tukey’s HSD were performed
using as the response the three interclass PCA scores. I applied Bonferroni corrections to adjust
76
for the multiple tests performed as alpha = 1 - 0.05/k, where k denotes number of tests.
Additionally, I performed analysis of molecular variance (AMOVA) as implemented in
ARLEQUIN (Excoffier et al. 2005) by partition the genetic variance among the three
shepherding systems and ungrazed patches to test for significance genetic differences among
groups using 1000 permutations.
In addition to PCA analysis I performed Bayesian clustering methods using TESS 2.3
(Chen et al. 2007), which incorporates the spatial locations of individuals by constructing a
neighbourhood network to assess spatial patterns of genetic structure. I performed 40 runs with
the admixture model, using the default spatial parameter ψ at = 0.6, burn-in lengths of 100,000
and with 5000 sweeps for of each kmax ranging from two to ten. The lowest values of the
deviance information criterion (DIC) were used to select the optimum number of clusters k as
suggested by Chen et al. (2007).
4.3.4. Isolation by geographic distance and connectivity by shepherding
To assess if spatial patterns of genetic structure were explained by directional dispersal along
shepherding routes or by isolation of geographic distance (IBD), I performed Mantel and partial
Mantel correlation tests using two predictor matrices: inter-patch geographic distances and
distance along shepherding routes. The distance along shepherding routes was calculated as the
number of patches that sheep traverse along the route from patch i to path j, including all grazed
patches without distinguishing between consistently and intermittently grazed patches (see
chapter 2). As the response variable, a matrix of genetic distances among populations was
calculated using the Cavalli-Sforza and Edwards (1967) chord distance (Dc), which has been
77
shown to be more appropriate for microsatellite data than other types of genetic distances
(Takezaki and Nei 1996). I performed a Mantel test between population genetic distances, Dc,
distances and Euclidean distances to test for isolation by geographic distance (IBD) for pairs of
populations in ungrazed patches (n = 12). For populations of patches connected by shepherding
(n = 47), I tested for significant differences in the regression slopes and intercepts among the
three shepherding grazing systems by including grazing system as a random (blocking) factor in
the multiple linear regression. To partial out the effect of geographic distance from the distance
along shepherding routes, I performed a partial Mantel test controlling for Euclidean distance.
Significance of Mantel correlation coefficients was tested by permuting observations within each
shepherding system 1000 times. To account for non-linearity Spearman rank correlation
coefficients (rho) were calculated. Although the use of Mantel test has been criticized in
landscape genetics due to lower statistical power compared to traditional linear models
(Legendre and Fortin 2010), so far it is the most appropriate method when hypotheses testing are
clearly defined in terms of distances, such as inferences of gene flow through neutral markers
and landscape structure (i.e. inter-patch distances or distance as tested here in terms of
shepherding connectivity; Cushman and Landguth 2010).
4.4. Results
4.4.1. Landscape-scale patterns of genetic structure
I found clear patterns of population genetic structure among the three non-overlapping
shepherding systems, as revealed by an interclass principal component analysis (PCA) that
factored in the three non-overlapping herding systems and a fourth group containing all ungrazed
78
patches (Fig. 8). Although the genetic variation explained by the three retained interclass PCA
axes was not high (axis 1: 6.25%; axis 2: 3.2%; axis 3: 2.2%), the first interclass PCA axis
significantly distinguished herd 2 populations from those grazed by herds 1 and 3 (Tukey-HSD:
p = 0.0001; A4 Appendix), the second interclass PCA axis significantly distinguished herd 3
populations from herd 1 (Tukey-HSD: p = 0.0001; A4 Appendix), and the third axis significantly
distinguished ungrazed populations from those in the three herding systems (Fig. 8; Tukey-HSD:
p = 0.0001; A4 Appendix). According to the results of AMOVA the proportion of the genetic
variation explained between the three herding systems and ungrazed patches is relative low
(0.6%), while most of the genetic variation was found within populations (97%). However, there
was significant genetic differentiation among herding systems and ungrazed patches (Table 5).
Bayesian clustering analysis, which did not take into account herding systems, identified two
genetic groups separating populations in the east (herd 2) and the west (herd 1 and 3) of the study
area (Fig. 9; A5 Appendix). This east-west genetic differentiation was consistent with the results
shown from the first interclass PCA axis.
4.4.2. Isolation by geographic distance (IBD) and directed dispersal by sheep
There was clear evidence of IBD effects for populations in ungrazed patches, as there was a
strong and significant positive association between inter-patch geographic distances with
population genetic distances (Mantel correlation = 0.55, p= 0.001; Fig. 10A). There were no
significant differences in the regression slopes among herding systems for Euclidean distances
(ANOVA test of overall effect of interaction between Euclidean distance and shepherding
system: t= 1.17, p = 0.18; herd 2: t = -0.52, p = 0.8; herd 3: t = -0.06, p = 0.9) and shepherding
distances (overall: t= 0.86, p = 0.28; herd 2: t = 1.15, p = 0.57; herd 3: t = 0.51, p = 0.78).
79
Similarly there were no significant differences among the regression intercepts among
shepherding systems for the relationship of genetic distances with Euclidean distances (overall:
t= 2.12, p = 0.19; herd 2: t = 0.75, p = 0.19; herd 3: t = -2.17, p = 0.8), but I found significant
differences for the regression intercept of herd 3 from herd 1 for the shepherding distances
(overall: t= 8.2, p = 0.001, herd 2: t = -0.53, p = 0.19; herd 3: t = -2.8, p = 0.001). Hence, the
final model contained a common slope and different intercepts for shepherding systems.
I found no significant effect of IBD on population genetic distances (Mantel correlation = 0.18, p
= 0.14; Fig. 10B). Instead, the distance along shepherding routes (measured as the number of
intervening patches traversed by the sheep) had a strong and significant effect on population
genetic distances (Mantel correlation = 0.43, p = 0.001; Fig. 10C). In fact, this relatively strong
association even increased slightly after partialling out the effect of IBD to account for potential
confounding of geographic distance and distance along shepherding routes (partial Mantel
correlation = 0.45, p = 0.001).
4.5. Discussion
4.5.1. Effect of directed dispersal by shepherding on genetic structure at the
landscape
I found evidence to support the main hypothesis that large-flock shepherding influences
landscape-scale patterns of genetic structure through directed seed dispersal in the calcareous
grassland forb Dianthus carthusianorum. In calcareous grasslands, traditional land use by
rotational sheep grazing is recognized to influence the structure and composition of this plant
80
community, as many habitat specialist plants are effectively dispersed among patchy grasslands
by sheep (Fisher et al. 1996; Kahmen et al. 2002; Kuiters and Huiskers et al. 2010; Reitalu et al.
2010). This is the first study to show that directed dispersal by grazing can result in sufficient
rates of seed-mediated gene flow to affect spatial patterns of genetic structure. This result is
especially important as D. carthusianorum lacks any seed morphological adaptations for
zoochory (Klotz et al. 2002).
Only two empirical studies have investigated patterns of population genetic structure in
association with shepherding practices in calcareous grassland plants. Willerdirg and Poschlod
(2002) studied the grass Bromus erectus, but did not find a clear association of patterns of
population genetic structure with shepherding routes, although seeds of B. erectus were found to
be transported by sheep in large numbers (Fischer et al. 1996). This lack of association might be
explained by the small number of populations studied (n = 12), and the generally large
population size of this species in calcareous grasslands. Honnay et al. (2006) investigated the
association of population genetic diversity and structure with temporal patterns of landscape
connectivity (as assessed from maps from 1850 to 1984) in the calcareous grassland plant
Anthyllis vulneraria, but did not find an effect. These authors speculated that directed dispersal
by shepherding may have homogenized spatial genetic variation in this species. In contrast to
these studies, I found significant population genetic structure related shepherding systems within
a relatively small study area (10 x 15 km).
81
4.5.2. Effect of IBD vs. directed dispersal by shepherding on genetic connectivity
When comparing populations within shepherding systems and controlling for the effect of
geographic distance between grazed populations, I found a significant and strong correlation of
population genetic distances with distance along shepherding routes. Specifically, there is no
data from observational or experimental studies reporting the presence of D. carthusianorum
seeds to be attached in the fur of animals, but other similar calcareous grasslands specialist
species lacking dispersal adaptations to zoochory (e.g. Asperula cynanchica; Fischer et al 1996)
have been found to become attached to the fur of sheep and travelling long distances up to 400
km (Fischer et al. 1996; Couvreur et al. 2004; Manzano and Malo 2006). Frequent long-distance
dispersal would be expected to homogenize spatial genetic structure across the landscape.
However, the finding of a distance-dependent effect similar to IBD but instead associated with
the distance along shepherding routes, suggests that gene flow is spatially restricted and mostly
occurring between neighboring populations along shepherding routes, thus supporting the
stepping-stone model of gene flow (Kimura and Weiss 1964). As required by local conservation
agencies, sheep are kept in designated paddocks for rumination, where consumed seeds are most
likely to be deposited. Sheep flocks move back and forth along grazing routes and animals do not
stay within a grassland patch for more than three days on smaller patches. Therefore, directed
dispersal into calcareous grassland patches may depend largely on epizoochory In a field
experiment, a large proportion of seeds of a variety of calcareous grassland specialists fell off
from the sheep wool within a few hours after grazing (Fischer et al. 1996; Couvrer et al. 2005),
which may explain the pattern of genetic structure found in populations connected along
shepherding routes.
82
Remarkably, the substantial effect of IBD of populations in ungrazed patches suggests
that in the absence of directed dispersal by sheep, seeds of D. carthusianorum may not travel far.
Experimental studies have documented that seeds of most calcareous grassland plants are
dispersed over short distances, i.e., often within 1m of the source plant (Coulson et al. 2001).In
the context of species conservation, a lack of connectivity by rotational sheep grazing may
endanger the long-term persistence of calcareous grassland specialist plants such as D.
carthusianorum by reducing the species’ ability to colonize grasslands after local extinction
(Soons et al. 2004; Rico et al. 2012) and by inhibiting gene flow and thus increasing the risk of
deleterious effects of inbreeding and genetic drift in local populations (Young et al. 1999).
Although the data do not allow quantification of the relative contributions of pollen- and
seed-mediated gene flow, the contrast between spatial genetic structure among ungrazed patches
and among patches within the same herding system, all within the same study area, suggests that
shepherding increased rates of seed-mediated gene flow markedly. Yet, the relative contribution
of seed-mediated gene flow to genetic diversity and spatial genetic structure depends also on the
rate of pollen flow. On one hand, pollen-mediated gene flow may be high in D. carthusianorum
as the species is mainly outcrossing and pollinated by specialized butterflies (Bloch et al. 2005),
which may travel long distances. On the other hand, habitat fragmentation has been shown to
negatively affect the effectiveness of pollinators for transferring pollen among plant populations
(Allen-Wardell 1998). The study system that is characterized by a heterogeneous matrix of
forest, intensive agricultural fields, orchards, and settlements is likely to have an effect on the
behavior of pollinators. In such a situation, a shortage of pollinators of D. carthusianorum due to
habitat fragmentation might lower rates of pollen flow, both for grazed and ungrazed patches,
thus increasing the importance of seed-mediated gene flow.
83
4.6. Conclusions
Overall, this study involving directed dispersal by grazing of domestic ungulates confirms
previous evidence found from plants dispersed by frugivores (e.g., Bacles et al. 2006; Freeland et
al. 2012). The significant differences found in genetic structure between ungrazed patches and
patches connected by large-flock shepherding indicate a substantial potential of directed seed
dispersal by grazing to mediate spatial patterns of gene flow across the landscape. Further
research is needed to assess whether the landscape-scale genetic structure found in D.
carthusianorum is typical for other characteristic species of calcareous grasslands, and to
elucidate the effect of traditional land-use such as sheep herding on spatial genetic structure in
plants.
84
Axi
s 2
Axis 1
Axi
s 3
Axis 1
Fig. 8 Ordination of the first two interclass PCA axes based on population allele frequencies
constrained by three non-overlapping shepherding systems and ungrazed patches (n = 59). Inset
shows ordination of the first and third interclass PCA axes differentiating ungrazed patches. Grey
squares populations in ungrazed patches, black diamond’s populations of herd 1, white triangles
populations of herd 2, and white circles populations of herd 3.
Herd 1
Herd 2 Herd 3 Ungrazed
85
Fig. 9 Pie charts of population membership scores for two genetic clusters based on TESS.
Background map shows in light grey the forested areas of the Upper Fraconia Jura plateau and
the colored lines correspond to the three non-overlapping herding routes connecting calcareous
grasslands (circles). The inset map shows the distribution of calcareous grasslands (grey areas) in
Germany and the location of the study area (orange box, map modified from Beinlich and
Plachter 1995).
86
Fig. 10 Effects of isolation by geographic distance (IBD) and distance along shepherding routes
on population genetic distances. A) IBD for pairs of population in ungrazed patches (n = 12); B)
IBD for pairs of population in grazed patches within the same herding system (n = 47), and C)
distances along shepherding routes for pairs of population in grazed patches within the same
herding system (n = 47). Symbols on plots B and C differentiate shepherding systems: herd 1=
open circles, herd 2= blue crosses, and herd 3= red triangles.
C)
Geographic distance (km)
0 2 4 6 8 10 12 14
0.1
0.2
0.3
0.4
0.5 A)
Gen
etic
dis
tanc
e
Dc
Gen
etic
dis
tanc
e
Dc
Sheep grazing distance
0 5 10 15
0.15
0.25
0.35
0.45
Gen
etic
dis
tanc
e
Dc
Sheep grazing distance
B)
0 5 10 15
0.15
0.25
0.35
0.45
r = 0.55, p= 0.001 r = 0.18, p= 0.14
r = 0.45, p= 0.001
87
Table 5. Analysis of molecular variance (AMOVA) and F-statistics by factor. Groups are
defined by three shepherding systems and ungrazed patches with Dianthus carthusianorum.
Source of variation
Sum of Squares
Variance component
Percentage of variation
F-statistc
Among herding systems and ungrazed
65.8 0.019 0.56 0.006*
Among populations within groups
393.4 0.07 2.12 0.021*
Within populations 10263.5 3.26 97.32 0.027*
Total 10722.7 3.35
*p < 0.05
88
Chapter 5
Genetic Consequences of Directed Seed
Dispersal by Shepherding in Fragmented
Calcareous Grasslands
5.1. Abstract
In fragmented landscapes, increased habitat isolation is likely to decrease rates of seed dispersal
and pollen flow among plant populations. Lack of gene flow will erode genetic diversity due to
genetic drift and inbreeding, which are stronger in small and isolated populations. Within
fragments, restricted seed dispersal at shorter distances will create spatial clustering of
genetically related individuals (fine-scale spatial genetic structure, SGS). However, directed seed
dispersal by zoochory promoting effective dispersal across the landscape will increase the
mixing of seeds from a variety of mother plants, thus affecting levels of genetic diversity and
SGS at the local scale, within a population. Here, I investigated the effects of directed seed
dispersal by rotational shepherding on the strength of SGS and genetic diversity, using eleven
nuclear microsatellites for 1,613 individuals from 49 populations of the calcareous grassland
specialist Dianthus carthusianorum. Populations connected by shepherding showed significantly
weaker SGS and significantly higher genetic diversity than populations in ungrazed grasslands,
89
suggesting that seed dispersal among spatially isolated grasslands is a major determinant
reducing plant relatedness within grassland patches. Independent of grazing treatment, small
populations (< 40 individuals) showed significantly stronger SGS and lower genetic diversity
than larger populations, likely due to genetic drift and inbreeding, which may further reduce
effective population size in small populations. A lack of significant differences in the strength of
SGS and genetic diversity between populations that were recently colonized and pre-existing
populations suggested that populations colonized after the reintroduction of shepherding were
likely founded by colonists from diverse source populations, reducing relatedness during the
initial colonization process. I conclude that directed long-distance dispersal by rotational
shepherding has the potential to increase genetic diversity and reduce SGS within D.
carthusianorum populations. This study thus highlights the importance of considering the
mechanisms of seed dispersal across spatial scales to better understand the nature of spatial
genetic structure in plant populations.
5.2. Introduction
Human modification of the landscape often changes a formerly continuous distribution of a
population into a set of smaller and spatially disconnected populations (Taylor et al. 2006). In
plants, increased habitat fragmentation is likely to decrease rates of seed dispersal and pollen
flow among remaining populations such that gene flow may be insufficient to counteract the loss
of genetic diversity caused by increased rates of drift and inbreeding in small and isolated plant
populations (Young et al. 1996; Lowe et al. 2005; Aguilar et al. 2008; Vranckx et al. 2011). In
isolated fragments, plant spatial dynamics are expected to be stronger, as restricted pollen flow
90
will increase rates of inbreeding, while restricted seed dispersal near to the mother plant will lead
to higher spatial overlap of offspring, increasing kinship structure and the probability of mating
among relatives (Heywood 1991; Wells and Young 2002). These effects will be more
pronounced in small populations, which may experience rapid loss of genetic diversity (Leimu et
al. 2006), thus potentially decreasing effective population size that will in turn affect offspring
fitness and population viability (Heywood 1991; Ellstrand and Ellam 1993; Sork et al. 1999;
Young and Clarke 2000; Keller and Weller 2002). Investigating spatial patterns of genetic
structure is fundamental for understanding ecological and evolutionary dynamics of plant
populations, such as demography, mating patterns, selection, and long-term persistence (Ouborg
and Van Treuren 1994; Hamrick and Nason 1996).
Restricted dispersal of seeds creating fine-scale spatial genetic structure, i.e., non-random
distribution of genetically related individuals (hereafter SGS; Heywood 1991; Epperson 1995;
2003), has been reported from many plant populations (e.g., Loiselle et al. 1995; Epperson 2000;
Vekemans and Hardy 2004; Trapnell and Hamrick 2004; 2005; Escudero et al. 2006; Sato et a.
2006; De-Lucas et al. 2009; Barluenga et al. 2010; Volis et al. 2010; Hamrick and Trapnell 2011;
Ndiade-Bourobou et al. 2011; Sebbenn et al. 2011; Wang et al. 2011). In general, evidence from
analyses across species suggests that the strength of the association of relatedness over physical
distance is largely determined by seed dispersal mode. Species with seed dispersal by gravity are
prone to show higher SGS than species with extensive seed dispersal provided by wind or animal
vectors (Hamrick and Nason 1996; Hamrick and Trapnell 2011). In the case of directed dispersal
by animals, the strength of SGS can decrease or increase as a function of the vector’s behavior.
Frugivores (endozoochory) that are specialized to forage on few plant sources and cache high
proportions of genetically related seeds in microsites suitable for establishment, will increase
91
SGS despite long-distance dispersal (Grivet et al. 2005; García et al. 2009; Torimaru et al. 2007).
In contrast, animals that forage on varied plant sources (endozoochory) or that inadvertently
transport seeds by attachment to the fur or hooves (epizoochory) will lead to higher mixing of
seeds from different mother sources, thus decreasing SGS within a population (Sezen et al. 2005;
Karubian et al. 2010; Kloss et al. 2011).
Directed long-distance dispersal by zoochory may also influence occurrence of SGS
resulting from colonization events. If populations are founded by numerous and unrelated
colonists from varied population sources, SGS will be rather low at first than if only a few seed
sources provided the colonists of the founded population (Withlock and McCauley 1990; Panell
and Charlesworth 2000). Colonization sets up the initial template of SGS on which seed and
pollen dispersal will further act, influencing subsequent mating patterns (Kalisz et al. 1999;
2001; Chung et al. 2003). Thus, to better understand the processes and consequences of SGS in
plant populations, it is important to consider the mechanisms determining seed dispersal at both
the landscape and local scales.
In agricultural landscapes, directed seed dispersal by domestic ungulates, such as sheep,
is recognized to provide directed long-distance dispersal for a range of calcareous grassland
plants (Fischer et al. 1996; Couvrer et al. 2005; Manzano and Malo 2006). As shepherding
supports effective seed dispersal among calcareous grasslands (Fischer et al. 1996; Auffret et al.
2012; Rico et al. 2012; Wagner et al. 2012), the likely mixing of seeds from different sources
will tend to decrease genetic relatedness among individuals, thus decreasing SGS within plant
populations. There are few studies on the effects of grazing on spatial patterns of genetic
structure, which showed no consistent trends (e.g., Kleijn and Steinger 2002; Rudman et al.
2007; Reisch and Poschlod 2009; Smith et al. 2009; Kloss et al. 2011). For instance, increased
92
aggregation of identical genotypes has been found to be enhanced by grazing (Kleijn and
Steingner 2002), while the opposite effect of lower SGS has also been reported (Kloss et al.
2011). However, since detailed data on management practices are often lacking (i.e., grazing
history, grazing routes), so far there is no empirical evidence showing the link between the
spatial patterns of genetic structure observed within grassland patches and directed long-distance
seed dispersal by shepherding.
Here I investigate the effects of directed seed dispersal by shepherding on SGS in the
perennial herb and specialist of calcareous grasslands Dianthus carthusianorum (Oberdorfer
1978). Although, D. carthusianorum lacks seed-dispersal traits related to zoochory (Klotz et al.
2002), previous empirical evidence has shown that the species responds to dispersal by
shepherding by showing increased habitat colonization after the reintroduction of sheepherding
in the study system in 1989 (see chapter 3) and by showing spatial patterns of population genetic
structure at the landscape scale associated with shepherding routes (see chapter 4). Specifically, I
hypothesize (i) that populations in grazed grassland patches will show weaker SGS and higher
genetic diversity than populations of ungrazed grasslands due to effective long-distance seed
dispersal increasing seed mixing within populations. Also, I expect that (ii ) large populations will
show higher genetic diversity and weaker SGS than small populations, where genetic drift and
inbreeding are expected to be stronger to reduce effective population size. For grazed grasslands,
I further compare SGS and genetic diversity between populations colonized since 1989 and pre-
existing populations. I expect to find (iii) weak SGS and moderately high genetic diversity due to
effective long-distance seed dispersal from varied seed population sources.
93
5.3. Methods
5.3.1. Study site
The study was conducted in the Southern Franconian Alb near Weissenburg, Bavaria Germany
covering an area of approximately 10 x 15 km. Calcareous grasslands of the Gentiano-
Koelerietum pyramidatae vegetation association (Oberdorfer 1978) in the study area are mainly
located on the steep slopes between the Upper Franconia Jura plateau and the valleys. In the 20th
century, progressive abandonment of rotational sheep grazing of calcareous grasslands led to a
substantial decrease in species richness of habitat specialist plants caused by local extinction due
to shrub encroachment and reforestation (Hakes 1987; Butaye et al. 2005). A conservation
management project was initiated in 1989 to reconnect remaining consistently grazed grassland
patches (“core areas”) with patches abandoned at least since 1960 by sheep grazing in three non-
overlapping herding systems. Flocks of about 400 to 800 ewes are herded along defined routes
up to 55 km and covering up to 140 ha of calcareous grasslands (Wagner et al. 2012). From 62
previously abandoned grassland patches, 26 are now grazed three to five times annually
throughout the season, 13 are grazed only towards the end of the season or were grazed only
during a few years at the beginning of the project, and 23 grassland patches remained ungrazed.
In the study area, sheep are kept in designated paddocks for rumination as prescribed by
calcareous grassland conservation management. In this situation, consumed seeds are unlikely to
be deposited in grassland patches and thus dispersal may depend mostly on epizoochory.
94
5.3.2. Study species and sampling
Dianthus carthusianorum L. (Caryophyllaceae) is a perennial herb of 30-45 cm in height. It has a
predominantly outcrossing mating system (Bloch et al. 2005), does not form a persistent seed
bank, and reproduces both sexually and vegetatively, although clonal shoots do not detach from
the rosette (Klotz et al. 2002). Flowering and seed shed occurs from June to October. Hence seed
dispersal by sheep is possible as the grazing season in the study area lasts from March until early
November. Pollination is carried out by specialized Lepidoptera species (Bloch et al. 2005) and
seeds lack morphological adaptations to dispersal by wind or animals (Klotz et al. 2002). In a
1989 baseline survey, D. carthusianorum occurred in 40 % of the 62 previously abandoned
patches and in 90 % of the core areas (Boehmer et al. 1990). An evaluation survey in 2009 (Rico
et al. 2012; Wagner et al. 2012) showed that successful colonization increased the species’
occurrence to 90 % in previously abandoned patches, whereas occurrence in unconnected
(ungrazed) patches remained at 40 %.
Leaf material was sampled in spring of 2009. I sampled all 38 grassland patches with
occurrence of D. carthusianorum (here denoted as populations) in the southern half of the study
area, where most occurrences of the species are concentrated, and additional 20 grassland
patches from the northern part of the study area. I also sampled D. carthusianorum from six
grass verges along forest edges, resulting in a total of 64 sampled patches, i.e. populations,
including 50 from grazed calcareous grasslands and eight from ungrazed grasslands. I sampled
leaf material from all individuals in patches with less than 40 individuals (here defined as small
populations), whereas I sampled across a patch approximately 30 to 40 leaf samples from
individuals in patches with more than 40 individuals (large populations). The latter implied that
sampling was spread out within patches of large populations such that the nearest sampled
95
individuals are not likely to be the actual nearest neighbors and thus the estimation of pairwise
kinship coefficients at the shortest distances are likely to be underestimated. Leaf tissue was
preserved in silica gel. The geographic coordinates of each sampled individual were recorded
using a Trimble® GeoXT™ 2008 GPS receiver with sub-meter resolution based on differential
GPS post-processing.
5.3.3. Microsatellite analysis
Genomic DNA was extracted following the DNeasy 96 plant kit protocol (QIAGEN). I amplified
15 microsatellite loci developed for related Dianthus species (MS-DINMADSBOX, MS-
DCDIA30, MS-DCAMCRBSY, MS-DINCARACC, DCA 221, DCD 224, DCB140, Smulders et
al. 2000) and specifically for Dianthus caryophyllus (DCB109; Smulders et al. 2003; CB018a,
CB057a, CB004a, CB027a, CF003a, CB011a, CB020a; Kimura et al. 2009). Microsatellite
amplifications were carried out with the QIAGEN Multiplex kit as follows: 0.2 to 0.4 µl (5µM)
of each primer 4.7 µl of Master-mix, and 1 µl of genomic DNA (approximately 5-10 ng/µl) in a
total reaction volume of 10 µl. PCR conditions were as described in Smulders et al. (2003).
Fluorescently labeled PCR products were run on an ABI 3730X Automated Sequencer (Applied
Biosystems) with 500 LIZ as size standard. Electrophenograms were analyzed using
GENMARKER v1.3. I only detected a high proportion of null amplifications, likely due to null
alleles in three microsatellite loci (DCA221, DCD224, DCD140), which were consequently
discarded from analysis.
Departure from Hardy-Weinberg equilibrium (HWE) and linkage disequilibrium (LD) for
each population was assessed using probability tests with 10000 Markov chain iterations in
96
GENEPOP v. 3.4 (Raymond and Rousset 1995). Locus DC020a significantly deviated from
HWE in 50 populations and was discarded from further analyses. Few locus combinations were
statistically significant for LD in some populations, but there were no congruent patterns.
Identical genotypes (< 2%), which likely corresponded to clones were excluded from analysis.
The final genetic data set included 1,613 individuals from 64 populations genotyped at eleven
polymorphic microsatellite loci (A3 Appendix). The same data set was previously used for
analysis of chapter 3 and 4.
5.3.4. Quantification of SGS, genetic diversity, and inbreeding
To evaluate the patterns of SGS, I used a multilocus measure of spatial genetic structure at the
individual level, the pairwise kinship coefficient of Ritland (1996) Fij as implemented in
SPAGEDI (Hardy and Vakemans 2002). The Fij coefficient measures the probability that two
alleles are identical by descent. Negative values of Fij may occur if allele frequencies between
two compared individuals differ more than expected at random from the entire data set. To assess
the strength of SGS, I estimated the Sp statistic of Vekemans and Hardy (2004) for each patch.
The Sp statistic has been widely applied in empirical studies of SGS since it provides an estimate
of SGS under isolation by distance (higher values of Sp indicate higher SGS; Vekemans and
Hardy 2004). If Fij tends to decrease linearly with the logarithm of distance, the strength of SGS
is quantified by the Sp statistic as bF / (1- F1), where bF represents the regression slope of the
pairwise Fij relatedness coefficients between individuals against the logarithmic of the distance,
while F1 is the mean Fij between pairs of individuals in the first distance class (Hardy and
Vekemans 2004). Since there were differences in sampling of individuals between small and
large populations, such that there is high variation among patches in the range of distances
97
between pairs of individuals within a patch, equal distance classes to compare all patches as
traditionally implemented in most studies could not be used. Instead F1 was defined as the
regression intercept, i.e., the value where the logarithm of distance is 0, which corresponds to the
value at a distance of 1 m. Higher values of the Sp statistic indicate a strong relationship of
pairwise Fij relatedness over distance. Significance of the regression slope in each population
was tested by permuting the pairwise Fij coefficients within populations 1000 times. Analyses
were implemented using R (R Development core team, 2010).
When SGS is the result of isolation by distance and assuming drift-dispersal equilibrium
in two-dimensional space, neighborhood size Nb and gene dispersal σ2 can be estimated from the
parameters of the Sp statistics (Hardy and Vakemans 1999). Under the above assumption I
estimated Nb, which represents the unit of genetic structure (Wright 1943), for each treatment
group as follows: Nb = (1- F1)/bF. Similarly, gene dispersal σ2 representing half the average of
the square axial parent-offspring distance was derived from the Sp statistic as: σ = 1/(4πSpδ),
where δ represents the estimated effective mating density of a population (although typically it
is only estimated by counting the number of individuals per square meter, which likely
overestimate effective mating densities). Since I did not have estimates of population density for
all 64 sampled patches, I estimated an average population density using the abundance data from
patches with complete census of individuals (i.e. patches having small populations). Plant density
thus was estimated as 0.028 ± 0.04 individuals per m2. For calculations of σ2 and Nb I only
included populations with significant values of the regression slope bF, as non-significant
populations produced extreme values for both parameters.
Genetic diversity indices including allelic richness (Ar) as well as observed (Ho) and
unbiased expected heterozygosity (He) were estimated for each population using GENALEX
98
(Peakall and Smouse 2006). I corrected the measures of allelic richness for the effect of sample
size using the rarefaction procedure in HP-RARE (Kalinowoski 2005). Inbreeding coefficients
FIS for each population were calculated in FSTAT v2.9.3.2 (Goudet 1995). Indices of genetic
differentiation among populations FST (Weir and Cockerham 1984) and their standard deviation
were estimated by jackknifing loci over populations using SPAGEDI (Hardy and Vakemans
2002).
5.3.5. Statistical test of SGS, genetic diversity, and inbreeding coefficients among
groups
To test for main effects and interaction of the two factors grazing and population size on genetic
diversity indices (He, Ho, and Ar), FIS inbreeding coefficients, and the Sp statistics, I performed a
two-way ANOVA for each response variable. To account for an unbalanced design, I used Type
II sums of squares. To normalize the distribution of residuals, we applied square-root
transformation to the Sp statistic, while FIS and genetic diversity indices were not transformed.
Negative values of the Sp statistic were set to zero. Model assumptions of normal distribution of
residuals and homogeneity of variances were checked using Anderson-Darling and Bartlett tests.
In addition, I tested for statistical differences in the degree of spatial isolation among grazed and
ungrazed patches using as a predictor the patch Si connectivity index based on inter-patch
geographic distances previously estimated in chapter 2 (Table 7). Since population history (i.e.,
presence of D. carthusianorum in 1989) is not confirmed for all core areas, statistical analysis of
population history was only performed for populations with known history. Due to this reduced
sample size, history was not included as an additional factor in the two-way ANOVA with the
99
factors grazing and population size. Additionally, I excluded populations with less than ten
individuals from analysis, one population with a minimum distance between neighboring
individuals > 10m, and four populations from patches that were grazed only during a few years
but remained ungrazed for the last 15 years. Thus, statistical analyses were based on 49
populations (Table 6, A6 Appendix; Fig. 11). All statistical analyses were performed in R (R
Development Core Team 2010).
5.4. Results
5.4.1. Strength of fine-scale spatial genetic structure (SGS)
Estimates of SGS for each population are shown in Table 6. The strength of SGS as quantified
by the Sp statistic was significantly related to both grazing and population size and without a
significant interaction (two-way ANOVA: sheep grazing: F1, 45= 4.74, p= 0.035; population size:
F1, 45 = 4.28, p = 0.044; interaction F1, 45 = 0.003, p = 0.95). The model showed normally
distributed and homoscedastic residuals (Anderson-Darling’s A = 0.73, p = 0.05; Bartlett's K2 =
3.75, p = 0.28). Specifically, grazed populations showed significantly weaker SGS than ungrazed
populations (i.e., lower Sp values), and independent of grazing, small populations showed
significantly higher Sp values than large populations (Fig. 12). Moreover, the regression slopes
bF of all ungrazed patches, independent of population size, were significant, while, for grazed
grasslands small populations showed a higher number of significant cases of SGS than large
populations (small: 6 out of 11, large: 13 out of 32 populations; Table 6). On the other hand,
there were no significant differences in the degree of spatial isolation between populations of
ungrazed and grazed grassland patches (ANOVA, F1, 47=1.05, p = 0.3).
100
Within grazed grasslands, I found no significant differences between the strength of SGS
in colonized and pre-existing populations (ANOVA, F1, 14 = 0.7, p = 0.41). This result did not
change when accounting for population size (two-way ANOVA population size: F1, 12 = 2.7, p =
0.12; interaction: F1, 12 = 1.26, p = 0.28). The model showed normally distributed and
homoscedastic residuals (A = 0.35, p = 0.4; K2 = 0.57, p = 0.49).
5.4.2. Estimates of neighborhood size and gene dispersal
Assuming that plant density is constant within patches, grazed grasslands showed higher
estimates of gene dispersal (small σ = 14.35, large σ = 19.67) than ungrazed grasslands (small σ
= 12.8, large σ = 15.6; Table 7). A similar pattern was observed for estimates of neighborhood
size, with the highest value for large populations in grazed grasslands (Nb = 150.94). Comparing
the values for small populations in which I had complete sampling, as expected small
populations of grazed patches showed a larger estimate of neighborhood size (Nb = 78.5; Table
7) than small populations in ungrazed patches (Nb = 61.9; Table 7). For colonized and pre-
existing populations, both gene dispersal and neighborhood size were very similar (Table 7).
5.4.3. Estimates of genetic diversity and inbreeding
Average estimates of genetic diversity, Ho, He, and Ar tended to be higher in populations of
grazed patches than in populations of ungrazed patches (except for Ho in ungrazed-small vs.
grazed-small; Table 8), but only differences in mean allelic richness were statistically significant,
where both sheep grazing and population size had a significant effect for allelic richness (two-
101
way ANOVA: sheep grazing: F1,45 = 7.09, p = 0.01; population size: F1,45 = 6.3, p = 0.01). The
model showed homoscedastic residuals (K2 = 1.26, p = 0.74), whereas the normality test was
significant (A = 0.91, p = 0.02). However, visual inspection showed no specific patterns or
influential outliers; hence residual distribution could not be improved by transformation.
No significant differences were observed for Ho (two-way ANOVA: sheep grazing: F1, 45
= 4.0, p = 0.05; population size: F1, 45 = 0.85, p = 0.36) and He (two-way ANOVA: sheep grazing:
F1, 45 = 0.48, p = 0.49; population size: F1, 45 = 3.17, p = 0.08). Colonized populations had higher
mean values of He (0.6) and mean allelic richness (Ar = 4.32) than pre-existing populations (He=
0.58, Ar = 4.14; Table 8), but no significant differences were found (ANOVA: Ar: df = 2, F =
0.39, p = 0.54; Ho: df = 2, F = 1.12, p = 0.31; He: df = 2, F = 0.26, p = 0.62).
The values of FST were higher for populations of ungrazed patches than for populations in
grazed patches independent of population size (small-ungrazed FST = 0.057 vs. small-grazed FST
= 0.041; Table 7). Average inbreeding FIS in D. carthusianorum was rather low (FIS = 0.095).
However and contrary to expectation, small ungrazed populations showed lowest inbreeding (FIS
= 0.06; Table 8). The two-way ANOVA showed a significant interaction between population
size and grazing (F1, 45 = 8.2, p = 0.006). Pairwise comparisons with Tukey’s HSD revealed that
small populations of ungrazed grasslands had significantly lower FIS values than all other groups.
Residuals showed normal distribution and homogeneity of variance (A = 0.23, p = 0.79; K2 =
3.37, p = 0.33). There were no significant differences in FIS between colonized and pre-existing
populations (ANOVA: df = 2, F = 0.36, p = 0.7).
102
5.5. Discussion
Analysis of SGS in D. carthusianorum, a typical calcareous grassland plant species, showed
significant patterns of SGS in 26 out of 49 populations. This result indicates that restricted gene
flow within populations’ affects genetic structure at local scales following isolation by distance
(Wright 1943), resulting from restricted seed dispersal close to the mother plant (Crawford 1984;
Epperson 1993; Epperson and Alvarez-Buylla 1997). However, as expected I found that the
strength of SGS across populations within a landscape was affected by the seed vector, as
populations of grazed grassland patches showed significantly weaker SGS than ungrazed
patches. This result was based on comparing only small populations, which unlikely large
populations (with 40≥ individuals) did not suffer from potential underestimation of the shortest
distances between neighboring plants since all individuals were sampled. These results are
consistent with previous empirical evidence showing that species dispersed by animals (but
mostly frugivores) have weak SGS, which is explained by effective seed dispersal over larger
distances (Degen et al. 2001; Fuchs and Hamrick 2010; Barluenga et al. 2011; Hamrick and
Trapnell 2011; Wang et al. 2011). Here, I showed that this also holds true for seed dispersed in
the fur of animals, such as directed seed dispersal by shepherding.
The above result was in agreement with related empirical studies in grassland plants that
have found that grazed plots present weaker patterns of SGS (Smith et al. 2009; Kloss et al.
2011). In contrast to these previous studies, by comparing levels of genetic diversity and SGS
within populations known to be connected within rotational shepherding systems, it is possible to
provide inferences on the ways in which grazing can decrease SGS in D. carthusianorum. First,
seed dispersal by sheep within patches will increase the variance of seed dispersal distances,
leading to higher overlap of seed shadows from different mothers and hence higher seed mixing
103
within sites. Secondly, directed dispersal by shepherding promotes effective long-distance
dispersal among grasslands connected along herding routes (see chapter 4). This process will
likely increase and maintain within-patch genetic diversity by seed-mediated gene flow, while
also reducing genetic relatedness of established seeds by interspersing seeds from external
sources within the patch . According to my results showing that populations of grazed grasslands
had higher genetic diversity than populations of ungrazed grasslands, long-distance seed
dispersal is likely an important determinant.
Populations of grazed patches clearly showed higher estimates of gene dispersal (σ)
within populations than ungrazed patches. Larger dispersal distances, also are expected to lead to
larger neighborhood sizes Nb in populations of grazed patches as sheep will increase the variance
of seed dispersal distances within sites. However, inferences made to estimate gene dispersal and
neighborhood size assume drift-dispersal equilibrium among populations (Wright 1946).
Contemporary land-use changes in the study system, such as increased calcareous grassland
isolation due abandonment of shepherding and recent colonization events, may prevent
populations from reaching drift-dispersal equilibrium (Vakemans and Hardy 2004). However,
since information on plant density was lacking for most large populations and I thus used the
assumption that plant density was equal in all patches, my estimates of gene dispersal and
neighborhood size need to be interpreted with caution and are only indicative of the likely
variation among populations determined by factors of seed dispersers, plant demography, and
population history.
In addition I found that population size is another factor modifying SGS and genetic
diversity. Independent of grazing treatment, small populations had significantly stronger SGS
and significantly lower genetic diversity than large populations. Small populations are more
104
prone to suffer from drift, which reduce effective population size, and thus erode genetic
diversity. Evidence based on meta-analyses across studies in plants showed that population size
is an important factor associated with levels of population genetic diversity (Honnay and
Vackemyn 2007; Aguilar et al. 2008; Vranckx et al. 2011). Furthermore, evidence showed that
small populations with reduced genetic diversity experience reduced fitness and population
viability (Leimu et al. 2006). Although I found that small populations have stronger SGS than
large populations, which may be due to reduced effective population size, this finding is partially
limited since individuals in large populations were not completely sampled but spread out across
the patch, and thus the nearest sampled individuals may not be the actual nearest neighbors. This
may have an effect on the estimation of the Sp statistics in large populations of both grazed and
ungrazed patches, thus limiting my interpretations on the effect of population size on SGS.
Further research will be needed to include estimates of plant density, instead of population size,
to analyze likely changes on SGS by population density within a patch.
In accordance with my expectations, I did not find differences in the strength of SGS and
genetic diversity between pre-existing and recently colonized grazed populations. This result
provides some information on the colonization process after the reintroduction of shepherding by
conservation management in 1989. Theoretical genetic models of colonization predict that when
populations are founded by colonists from a single seed source, population genetic diversity will
be low and genetic differentiation among recently colonized populations will be higher as
compared to older populations, whereas higher genetic diversity and lower genetic differentiation
will be found if colonists are from diverse seed sources (Withlock and McCauley 1990; Panell
and Charlesworth 2000; Pannell and Dorken 2006). The lack of significant differences in genetic
diversity, SGS, and FST values between recently colonized and pre-existing populations in grazed
105
patches, suggests that historical gene flow, at least partly mediated through directed seed
dispersal by shepherding was high enough to maintain similar levels of genetic diversity within
populations and low levels of genetic differentiation among populations connected by
shepherding.
It is important to contrast estimates of SGS, genetic diversity, and inbreeding to infer
patterns of gene flow within populations, and to understand gene flow effects in small and
isolated populations. Contrary to expectation, small ungrazed populations showed significantly
lower levels of inbreeding than all other groups. Inbreeding in plants, here measured by FIS and
theoretically defined by non-random mating reducing He as expected under HW, could be the
result of bi-parental inbreeding or selfing (Mimura and Aitken 2004; Gonzales et al. 2006). In the
absence of further information (e.g., parentage analysis) to distinguish whether inbreeding is
mainly due to selfing or biparental inbreeding, one can approximate the question by comparing
values of F1 and FIS (Vakemans and Hardy 2004). If F1 is higher than FIS, biparental inbreeding
may explain the pattern of inbreeding, whereas if F1 is lower than FIS, selfing is suggested
(Vakemans and Hardy 2004). Although the level of inbreeding was not particularly high in D.
carthusianorum (average FIS = 0.095), FIS was higher than F1 in all treatments groups (Table 6),
suggesting that selfing might be important.
However, evidence from studies based on meta-analysis in outcrossing plants (Leimu et
al. 2006), have found that inbreeding as measured by the FIS coefficient is often not associated
with population size as expected as large populations showed higher levels of FIS than small
populations. The above is explained by a primary effect of genetic drift rather than inbreeding
resulting in a more rapid loss of alleles than the decrease of He (Leimu et al. 2006). Specifically
the FIS values observed within all six compared groups in D. carthusianorum (Table 8) were
106
below 0.15 (overall FIS= 0.1). Values above 0.15 correspond to a kinship structure of half-sibs,
while values above 0.25 indicate relatedness of full-sibs (Wright 1921). It is important to
consider that the FIS coefficient as calculated based on estimates of mean He within
subpopulations represents a measure of inbreeding over several generations, and may not
necessarily reflect the current situation of the population.
D. carthusianorum is partially outcrossing and insect-pollinated by specialized
butterflies (Bloch et al. 2005). In the study system, calcareous grasslands are embedded in a
heterogeneous matrix composed of forest, intensive agriculture, meadows, orchards, and
settlements. Habitat fragmentation may negatively affect the effectiveness of pollinators to
transfer pollen among patchy plant populations (Allen-Wardell 1998; Kearns et al. 1998).
Butterfly pollinators that are specialists of calcareous grasslands have been shown to be affected
by increased habitat isolation (Bruckmann et al. 2010). Reduced rates of pollen transfer between
grasslands by pollinators may increase biparental inbreeding, thus reducing the contribution of
pollen to gene flow, while seed-mediated gene flow by shepherding may have a major
importance in D. carthusianorum. The lack of significant differences in the geographic isolation
among grasslands does not necessarily imply that movement of pollinators across the landscape
may not be affected by the composition of the matrix. However, with nuclear markers, I cannot
estimate the relative contribution of pollen- vs. seed-mediated gene flow to determine whether
gene flow by pollen was less important in D. carthusianorum than gene flow by seed (Ellstrand
and Ellam 1993; Ennos 1994). Estimating the ratio of pollen to seed flow will add important
information relevant for species conservation (Leimu et al. 2006; Aguilar et al. 2008). For plants,
increased rates of selfing and biparental inbreeding will reduce genetic diversity within
107
populations with negative consequences for offspring fitness and the long-term persistence of
plant populations (Young et al. 1999; Keller and Weller 2002; Frakham 2005).
5.6. Conclusions
The majority of empirical studies of fine-scale spatial genetic structure have investigated one or
few populations, assuming that patterns of SGS are homogeneous across populations within a
landscape. By contrasting multiple populations of D. carthusianorum in calcareous grassland
patches, I showed substantial variation in SGS among populations. More importantly, my results
revealed that directed seed dispersal by sheep is an important factor reducing SGS within
populations. Another factor affecting genetic diversity within populations and SGS is population
size (Leimu et al. 2006). The lack of differences in the strength of SGS between grazed
populations varying in population age suggests that colonized populations are likely to be the
result of effective seed dispersal among physically disconnected calcareous grasslands after the
reintroduction of sheep grazing in 1989. This result adds evidence of the successful restoration of
abandoned grassland patches by reestablishing rotational shepherding in the study area. Further
research is needed to investigate if differential rates of inbreeding in populations of grazed
grasslands as compared to ungrazed grasslands may be the result of different rates of selfing and
biparental inbreeding.
108
Table 6. Estimates of spatial genetic structure SGS across 49 populations of D. carthusianorum: N, number of individuals genotyped;
F1, kinship coefficient at ln(distance = 1m); slope (bF) of the regression kinship coefficient on ln(distance); Sp statistic reflecting the
intensity of SGS; larger values of Si indicates higher connectivity. Significant p-values (α = 0.05) of the regression slopes in bold.
Population Minimum distance
N Connectivity index Si
Grazing Population size
Population history
F1
Slope
bF Sp
P-value
A03 0.3 29 0.943 grazed small colonized 0.064 -0.0126 0.0135 0.001
A05 0.3 26 1.497 grazed small colonized 0.054 -0.0127 0.0134 0.001
A08 0.5 23 1.599 grazed small colonized 0.011 0.0003 -0.0003 0.564
A26 0.3 26 2.346 grazed small colonized 0.032 -0.0062 0.0064 0.026
A31 0.3 23 2.773 grazed small colonized 0.050 -0.0099 0.0104 0.001
A33 0.4 35 1.805 grazed small colonized 0.097 -0.0180 0.0200 0.002
A45 0.7 15 1.262 grazed small colonized 0.021 -0.0020 0.0021 0.272
C08 0.3 30 0.819 grazed small unknown 0.025 -0.0015 0.0015 0.203
A12 0.3 12 1.325 grazed small pre-existing 0.055 -0.0150 0.0159 0.148
A25 0.5 15 1.572 grazed small pre-existing 0.103 -0.0444 0.0495 0.001
E07 2.2 15 0.092 grazed Small unknown -0.068 0.0328 -0.0307 0.91
A06 0.5 30 1.673 grazed Large colonized 0.038 -0.0100 0.0104 0.017
Table 6. Continue on the following page
109
Table 6. Continue on the following page
Population Minimum distance
N Connectivity index Si
Grazing Population size
Population history
F1
Slope
bF Sp
P-value
A07a 2.0 30 1.207 grazed large colonized 0.008 0.0019 -0.0019 0.678
N05 0.5 28 0.119 grazed large colonized 0.021 -0.0016 0.0016 0.334
A14 0.4 16 1.714 grazed large pre-existing 0.128 -0.0346 0.0397 0.001
A18 1.2 29 1.560 grazed large pre-existing 0.081 -0.0125 0.0136 0.005
A28 0.4 33 2.574 grazed large pre-existing 0.014 0.0004 -0.0004 0.558
A29 0.8 28 2.644 grazed large pre-existing 0.012 -0.0034 0.0034 0.012
A38 2.4 30 1.041 grazed large pre-existing 0.010 -0.0006 0.0006 0.456
E01 0.3 36 0.456 grazed large unknown 0.040 -0.0047 0.0049 0.004
G01 0.6 29 0.322 grazed large unknown 0.051 -0.0051 0.0053 0.11
G13 0.3 41 0.092 grazed large unknown 0.014 -0.0010 0.0011 0.363
G16 1.4 29 0.904 grazed large unknown 0.018 -0.0038 0.0038 0.01
G20 1.8 30 0.556 grazed large unknown 0.006 -0.0001 0.0001 0.477
G21 0.3 30 0.895 grazed large unknown 0.034 -0.0064 0.0066 0.001
G23 2.8 28 0.905 grazed large unknown 0.065 -0.0104 0.0112 0.001
G26 0.6 38 1.586 grazed large unknown 0.026 -0.0039 0.0040 0.015
G28 0.3 30 0.790 grazed large unknown 0.002 0.0000 0.0000 0.535
G29 3.3 30 1.447 grazed large unknown -0.007 0.0039 -0.0039 0.808
110
Population Minimum
distance
N Connectivity
index Si
Grazing Population
size
Population
history
F1
Slope
bF
Sp
P-value
G30 0.3 30 1.098 grazed large Unknown 0.036 -0.0079 0.0082 0.005
G31 0.7 30 1.411 grazed large Unknown -0.002 0.0015 -0.0015 0.786
G37 0.3 27 1.322 grazed large Unknown -0.011 0.0028 -0.0028 0.831
G40 0.3 31 0.716 grazed large Unknown 0.003 0.0004 -0.0004 0.593
G45 5.2 31 0.646 grazed large Unknown 0.009 -0.0001 -0.0001 0.523
G46 0.3 30 1.193 grazed large Unknown 0.002 -0.0004 0.0004 0.432
G47 2.0 28 1.426 grazed large Unknown -0.001 0.0007 -0.0007 0.589
G48 0.3 29 1.340 grazed large Unknown 0.028 -0.0048 0.0050 0.056
G49 0.5 30 1.269 grazed large Unknown 0.019 -0.0014 0.0014 0.343
G50 0.3 28 1.030 grazed large Unknown 0.085 -0.0235 0.0257 0.001
G05a 0.8 27 1.036 grazed large Unknown 0.047 -0.0060 0.0063 0.163
Gzim 0.3 30 0.211 grazed large Unknown 0.017 -0.0026 0.0027 0.041
N03 0.3 29 0.302 grazed large Unknown 0.044 -0.0104 0.0109 0.001
E03 0.3 15 0.941 ungrazed small Unknown 0.091 -0.0166 0.0182 0.001
E04 0.3 30 0.421 ungrazed small Unknown 0.052 -0.0179 0.0189 0.005
E09 0.3 32 1.338 ungrazed small Unknown 0.047 -0.0083 0.0087 0.001
Nroth 0.4 14 0.573 ungrazed small Unknown 0.106 -0.0362 0.0404 0.001
A37 0.3 31 1.205 ungrazed Large Colonized 0.067 -0.0121 0.0130 0.001
N10 0.3 25 0.733 ungrazed Large Unknown 0.030 -0.0103 0.0106 0.001
111
Table 7. Estimates of gene dispersal σ, and neighborhood size Nb, F1 kinship coefficient at
ln(distance = 1m), for the six treatment groups in Dianthus carthusianorum. SD indicates
standard deviation of the mean.
Group σ (SD) Nb (SD) F1 (SD)
Ungrazed- small 12.8 ± 3.9 61.9 ± 38.0 0.074 ± 0.03
Ungrazed- large 15.6 ± 1.1 85.79 ± 12.1 0.048 ± 0.03
Grazed- small 14.35 ± 4.5 78.48 ± 45.7 0.040 ± 0.04
Grazed-large 19.67 ± 6.7 150.94 ± 92.7 0.028 ± 0.03
Colonized-grazed
Pre-established-grazed
15.84 ± 3.0
14.79 ± 9.8
91.07 ± 35.8
102.2 ± 127.6
0.042 ± 0.03
0.039 ± 0.03
112
Table 8. Summary of genetic diversity measures and fixation indices of the six treatment groups
in Dianthus carthusianorum. N, sample size; Ho observed heterozygosity; He expected
heterozygosity; rarefied Ar allelic richness; FIS, inbreeding coefficient, FST fixation index of
genetic differentiation. SD standard deviation of the mean.
Group N Ho (SD) He (SD) Ar (SD) FIS (SD) FST (SD)
Ungrazed- small 4 0.53 ± 0.05 0.56 ± 0.04 4.11 ± 0.28 0.04 ± 0.05 0.057 ± 0.02
Ungrazed- large 2 0.50 ± 0.003 0.56 ± 0.02 4.02 ± 0.08 0.12 ± 0.02 0.032 ± 0.01
Grazed- small 11 0.52 ± 0.04 0.58 ± 0.02 4.21 ± 0.22 0.12 ± 0.05 0.041 ± 0.00
Grazed-large 32 0.54 ± 0.02 0.59 ± 0.03 4.44 ± 0.23 0.09 ± 0.03 0.018 ± 0.00
Colonized-grazed
Pre-established-
grazed
9
7
0.53 ± 0.03
0.54 ± 0.04
0.60 ± 0.02
0.58 0.02
4.32 ± 0.20
4.14 ± 0.31
0.097 ± 0.07
0.098 ± 0.04
0.03 ± 0.004
0.04 ± 0.009
Total 49 0.54 ± 0.03 0.59 ± 0.03 4.34 ± 0.27 0.095 ± 0.02 0.03 ± 0.002
113
Fig 11. Sketch of spatial locations of patches (populations) analyzed for fine-scale spatial genetic
structure (SGS) in Dianthus carthusianorum. Black circles denote larger populations and white
circles correspond to small populations (n < 40 individuals). Lines indicate shepherding routes
connecting calcareous grassland patches in three non-overlapping herding systems. Ungrazed
populations thus are showed as not connected by lines.
Map produced by Adrián Sarabia Rangel
114
Fig. 12 Interaction plot of the mean square-root transformed Sp statistic as a function of the
factors grazing and population size in populations of Dianthus carthusianorum. The dashed line
indicates small population size (n < 40 individuals) and the full line denotes large population size
(n ≥ 40 individuals).
115
Chapter 6
Synthesis and Future Directions
6.1. Directed Dispersal by Shepherding Promotes Functional
Connectivity in Calcareous Grasslands
To effectively protect and maintain plant biodiversity in human-modified landscapes, we need to
understand what regulates functional connectivity. For plants, investigating determinants of
effective seed dispersal (Shupp 1993) is critical for species conservation, as seed dispersal
supports colonization, recruitment, range expansion, and ultimately species persistence (Nathan
and Muller-Landau 2000; Levin et al. 2003; Clobert et al. 2004). Animal vectors are important
determinants of connectivity because they can effectively disperse seeds through the matrix to
deposit them in suitable sites for establishment (directed dispersal) (Shupp 1993; 2011).
Despite the importance of animal vectors for effective long-distance dispersal of seeds,
most empirical studies of fragmented plant communities have modeled connectivity only as a
function of the physical distances between habitat patches, ignoring the role of seed dispersal
vectors, and thus failing to approximate functional connectivity. This has been the typical case of
most empirical studies of connectivity in calcareous grassland communities (e.g., Krauss et al.
2004; Geertsema 2005; Adriaens et al. 2006; Joshi et al. 2006; Bruckman et al. 2010). Contrary
to previous studies, my PhD research (chapter 2) shows that functional connectivity of
116
calcareous grassland communities cannot be comprehensively understood by quantifying only
structural connectivity as a function of inter-patch distances alone, e.g., with simple dispersal
kernels or distance thresholds. In chapter 2, by testing competing models of potential functional
connectivity with different assumptions of dispersal relating the physical distances between
patches, vegetation compositions of the matrix, and dispersal by shepherding, I showed that
functional connectivity in terms of mean patch colonization rates is associated with directed seed
dispersal by large-flock shepherding. Furthermore, the analysis implemented in chapter 2 using
estimates of actual functional connectivity such as patch-level colonization rates represents an
important improvement over studies using indirect estimates of functional connectivity such as
data of patch-level species richness (Fagan and Calabrese 2006).
Modeling functional connectivity at the community level offers general insights in
species’ responses to habitat fragmentation (Minor et al. 2011; Schleicher et al. 2011). However,
by pooling colonization events across species, species-specific responses may be neglected
(Taylor et al. 2006), such as in the case of species varying in dispersal-related traits. In
calcareous grasslands, traditional land-use by rotational shepherding has been suggested to
support dispersal of habitat specialist plants (Fischer et al. 1996; Poschlod et al. 1998; Schrautzer
et al. 2009; Kuiters and Huiskes 2010; Reitalu et al. 2010; Piqueray et al. 2011). However,
habitat specialists of calcareous grasslands vary in seed dispersal syndromes, such that not all
species may respond to connectivity by shepherding. In chapter 3, by contrasting the results of
the community-level assessment (chapter 2) with analyses of 31 individual species of the same
plant community, I showed that rotational shepherding effectively supports dispersal for most of
these species regardless of the presence of seed morphological adaptations to zoochory (e.g.,
bristles, hooks, awns). Based on the results from chapters 2 and 3, I found evidence
117
corroborating the pivotal role of large-flock shepherding to maintain species richness in
calcareous grassland communities by supporting effective seed dispersal among physically
isolated calcareous grasslands.
The development of comprehensive approaches for assessing functional connectivity in
plants requires investigating not only the processes governing seed dispersal per se, but in
addition focusing on the factors that determine the migrant pool (pre-dispersal) and that facilitate
establishment and growth (post-dispersal) (Murphy and Lovett-Doust 2004). By testing
alternative source patch effects, in terms of patch area, population size (both proxies of seed
production), and species presence in source patches, I found that the diversity of seeds of the
source patch at the community level (chapter 2) or species occurrence at the individual level of
analysis (chapter 3) were sufficient predictors of connectivity. On the other hand, ecological
factors of focal patches affecting establishment and thus colonization rates have not commonly
been considered (Clobert et al. 2004). By including the number of dynamic structural elements
present in focal patches (i.e., small mammal burrows, rock debris, erosion, and ant hills) I found
a significant increase of model performance for both, community- and species-level assessments.
Importantly, since colonization rates as a measure of actual functional connectivity reflect the
outcome of dispersal and post-dispersal processes, assessing the relative importance of these
processes is essential for effective landscape management (Nathan and Muller Landau 2000).
Based on the comprehensive approach proposed in chapter 2, dispersal and post-dispersal
processes were found to be equally important predictors of patch colonization rates at the
community level. Therefore, maintenance of calcareous grasslands not only depends on the
presence of shepherding to support dispersal, but also on the diversity of microsites facilitating
establishment within patches (Wagner et al. 2012).
118
Understanding maintenance of genetic diversity in fragmented populations is another
important element of species conservation (Vellend and Geber 2005; Leimu et al. 2006; Hughes
et al. 2008). Genetic diversity within populations is the result of several processes including gene
flow, mutation, drift, and demographic processes (Slatkin 1987). Gene flow tends to homogenize
the spatial genetic variation among populations and maintain similar levels of genetic diversity
within populations (Levin et al. 2003). Connectivity models with ecological (presence-absence)
and genetic data (nuclear microsatellites) in D. carthusianorum showed that dispersal by
shepherding not only influenced patch occupancy across the landscape, but also is likely to
influence genetic diversity (as observed by estimates of mean allelic richness) within populations
through seed-mediated gene flow.
6.2. Shepherding Effects on Seed-Mediated Gene Flow across
Spatial Scales in Dianthus carthusianorum
As suggested by the results in chapter 3, rotational shepherding is likely an important
determinant of seed-mediated gene flow in D. carthusianorum. In chapter 4, I specifically tested
specific hypotheses relating to the effect of directed seed dispersal by shepherding on spatial
patterns of genetic structure at the landscape scale. In the study area, most calcareous grasslands
have been grazed in three non-overlapping herding systems since 1989 (after the progressive
abandonment of shepherding during the first half of the 20th century). If directed dispersal by
sheep substantially contributes to seed immigration into patches connected by shepherding, I
expected to find genetic differentiation among non-overlapping herding systems and ungrazed
patches. Remarkably and as supported consistently with different analyses (e.g., PCA and
Bayesian clustering methods), there was clear evidence of landscape-scale genetic differentiation
119
among herding systems and ungrazed patches given the short time elapsed since introduction of
the herding systems in 1989. Although sheep grazing has been recognized to determine species
composition and community structure of calcareous grassland communities (Kahmen et al. 2002;
Reitalu et al. 2010; Piqueray et al. 2011; Wagner et al. 2012), the analysis presented in chapter 4
is the first to show that directed long-distance dispersal by sheep can result in sufficient rates of
seed-mediated gene flow to determine spatial patterns of genetic structure at the landscape scale
in a calcareous grassland plant.
Furthermore, by specifically disentangling the confounding effects of IBD and
shepherding connectivity on population genetic distances within each herding system, the
analysis carried out in chapter 4 goes beyond to other related studies (e.g., Willendirg and
Poschold 2002). Interestingly, the association of genetic distances and shepherding connectivity,
in terms of the number of patches that sheep need to traverse between two sites along grazing
routes, suggests that gene flow occurs mostly between nearby populations, whereas geographic
distance per se showed no significant association with population genetic distances.
Most empirical studies of seed-mediated gene flow related to directed dispersal have
focused on endozoochorous species, mostly trees and shrubs dispersed by frugivores (e.g., Grivet
et al. 2005; Jordano et al. 2007; Garcia et al. 2009; Karubian et al. 2010). Dispersal by domestic
ungulates like sheep and goats likely occurs mostly by epizoochory. In particular, in this study
system sheep do not stay long within a patch since animals are herded along grasslands, while
they are kept to ruminate in specific paddocks, where most dung is deposited. Hence,
endozoochory likely plays a considerably smaller role than epizoochory is this system. Spatial
patterns of genetic structure resulting from directed dispersal by epizoochory may depend to
some extent on the adhesive interaction between seeds and furs (Courvrer et al. 2004). Species
120
with adhesive seed appendages or specialized structures to promote epizoochory are prone to be
transported in higher numbers to travel over larger distances (Fisher et al. 1996). In contrast, D.
carthusianorum lacks seed morphological adaptations to zoochory (Klotz et al. 2002), and most
seeds may not stay long in the sheep wool, leading to a distance-dependent effect of shepherding
connectivity along grazing routes at the landscape scale.
At the local scale, i.e., within a patch, seed dispersal is a major determinant of genetic
structure as it influences the spatial distribution of the offspring, while also dispersing both
maternal and paternal alleles (Heywood 1991; Epperson 1993; Ouborg 1999). Given that
directed seed dispersal by large flocks of sheep has the potential to support effective seed
immigration from connected grasslands into a population (chapter 4) and to move seeds within a
patch, I expected to find an effect on fine-scale genetic structure (SGS) in D. carthusianorum
populations. Chapter 5 specifically addressed this hypothesis. Analysis of SGS across multiple
populations differing in grazing treatment suggested that sheep grazing is a major determinant of
SGS, as it significantly decreased the degree of spatial relatedness among conspecifics.
Interestingly, populations of grazed patches had significantly higher levels of genetic diversity
(allelic richness) than ungrazed patches, which suggests that effective seed dispersal among
grasslands is likely to play a major role in seed mixing and reducing relatedness of neighboring
individuals (kinship structure).
Other intrinsic factors of a population can also modify patterns of SGS. Plant
demography, determining effective population size, is known to influence SGS (Hardy and
Vekemans 2004). I found that independent of grazing treatment, large populations had
significantly weaker SGS and higher genetic diversity than small populations. This result is
likely to be related to the combined effects of genetic drift and inbreeding (biparental inbreeding
121
or selfing) acting more strongly in small populations, which are likely to reduce effective
population size and increase the strength of SGS (Ellstrand and Ellam 1993; Young and Clarke
2000; Keller and Weller 2002; Charlesworth 2003; Hardy and Vakemans 2004).
6.3. Conservation Implications and Future Directions
Calcareous grasslands in the Franconian Alb are listed in Bavaria and in Germany as an
endangered plant community (Walentowski et al. 1991). The conservation management project
for calcareous grassland protection initiated in the study area in 1989 opted to reintroduce
shepherding as the main restoration practice. Results of connectivity assessments at the
community- and species-levels emphasize the important role of shepherding to support effective
dispersal and thus restore species composition of previously abandoned patches by enabling
habitat recolonization after local extinction. These results were consistent with a scientific
evaluation of this 20 year-old management project, which showed the improvement in species
richness of previously abandoned calcareous grasslands that were reconnected by shepherding
since 1989 compared to grasslands that remained ungrazed (Wagner et al. 2012).
Habitat fragmentation not only will affect plant demography and probability of habitat
colonization by limiting seed dispersal, but in addition will have genetic consequences for plant
populations (Aguilar et al. 2008; Vranckx et al. 2011). Thus, ecological data alone (presence-
absence data or colonization rates) cannot tell whether gene flow is sufficient to maintain
genetically diverse populations. In this study system, the effectiveness of landscape management
by shepherding to restore dispersal and gene flow among previously abandoned grasslands can
be inferred from the genetic data in D. carthusianorum by contrasting patterns of SGS between
122
grasslands colonized since 1989 with pre-existing populations. Importantly, the lack of
significant differences in SGS and genetic diversity between colonized and pre-existing
populations in grazed grasslands indicates that recently colonized populations were founded by
seeds from a variety of source populations, whereas ungrazed patches generally lacked
colonization.
Another important finding for species conservation management refers to the result that
population genetic structure of D. carthusianorum in ungrazed patches was significantly
associated with isolation by distance. In addition, ungrazed patches showed significantly stronger
patterns of kinship structure, lower genetic diversity, and the highest FST index of population
genetic differentiation. These results suggest that in the absence of sheep promoting effective
seed dispersal among physically isolated grasslands, seeds of D. carthusianorum do not disperse
far. These data are particularly important since D. carthusianorum lacks dispersal traits related to
anemochory or zoochory. Thus, dispersal by shepherding has important genetic implications for
the conservation of this species by preventing loss of genetic diversity and development of
genetic differentiation among populations by supporting effective seed dispersal among spatially
isolated grasslands.
Using nuclear microsatellites, it is not possible to quantify the relative contribution of
pollen- and seed-mediated gene flow to spatial patterns of genetic structure (Ennos 1994). D.
carthusianorum has a mixed mating system and is pollinated by specialized butterflies (Bloch et
al. 2005). The agricultural landscape of the Southern Franconian Alb is characterized by a
heterogeneous matrix composed of intensive agricultural fields, beech and pine forest, meadows,
orchards, and settlements. Habitat fragmentation may disrupt plant-insect interactions such as
pollination (Allen-Wardell 1998; Kearns et al. 1998). Pollinator species of calcareous grasslands
123
such as bees and butterfly species have been shown to be affected by habitat fragmentation
(Goverde et al. 2002; Dewenter and Tscharntke 2002). Specifically, butterfly specialists of the
ecological conditions of calcareous grasslands are vulnerable to habitat fragmentation, which
may lead to a population decline or a change in their foraging behavior (Bruckmann et al. 2010).
A shortage of pollinators of D. carthusianorum will likely reduce rates of pollen transfer between
grasslands, thus leading to relatively lower rates of pollen-mediated gene flow than seed-
mediated gene flow, in the case, where grasslands are connected by shepherding. Although
inbreeding coefficients were not high across populations in D. carthusianorum, surprisingly
average inbreeding was lowest within small ungrazed patches. Thus, investigating the
contribution of pollen flow within and among populations to overall gene flow would be an
important research step to complement our knowledge regarding determinants of functional
connectivity in calcareous grasslands.
For instance, using paternity analysis make feasible to test if selfing or biparental
inbreeding is occurring within populations, and more importantly, identifying which landscape
factors affect rates of pollen flow across the landscape. Equally important, would be to test the
potential of shepherding dispersal for seed-mediated gene flow in other habitat specialist plants.
Corroborating the findings of spatial patterns of genetic structure at the landscape scale (i.e. west
and east genetic differentiation) in other calcareous grassland species would show whether such
patterns are exclusive of the biology of D. carthusianorum or to the management history of this
landscape. Furthermore, it will be interesting to contrast if the result of a distance-dependent
effect similar to IBD but in terms of the number of patches traversed by sheep along shepherding
routes depends on dispersal syndrome. For instance, paternity analysis would allow testing
whether selfing or biparental inbreeding occur within populations, and more importantly,
124
identifying which landscape factors affect rates of pollen flow across the landscape. Equally
important would be to test the potential of seed dispersal by sheep for seed-mediated gene flow
in other habitat specialist plants. Corroborating the findings of spatial patterns of genetic
structure at the landscape scale (i.e. west and east genetic differentiation) in other calcareous
grassland species would show whether such patterns are exclusive of the biology or population
history of D. carthusianorum or related to the management history of this landscape.
Furthermore, it will be interesting to contrast if the finding of a distance-dependent effect along
shepherding routes on spatial genetic structure depends on dispersal syndrome. For instance,
species with seeds adapted to zoochory may likely not show such a pattern, as higher rates of
seed dispersal may tend to homogenize the spatial genetic variation across the landscape. On the
other hand, studying flagship species of this calcareous grassland community such as the
endangered Pulsatilla vulgaris would provide important information on the likely effects of
habitat fragmentation in species on conservation concern.
In conclusion, this PhD study contributes towards understanding what directs functional
connectivity in plants using calcareous grasslands as a model system. This study addresses most
of the research needs identified by Fischer and Lindenmayer (2007; see chapter 1) as by a
combination of genetic and ecological data, this research provides comprehensive empirical
evidence on patterns and mechanisms of dispersal and gene flow in plants at the landscape scale.
Specifically, I found evidence both at the community and species levels that highlights the
important role of directed seed dispersal by large-flock shepherding for determining spatial
patterns of species occupancy in a range of habitat specialist species, as well as patterns of seed-
mediated gene flow at the spatial scales of the landscape and within individual patches as shown
for the habitat specialist forb Dianthus carthusianorum.
125
References Adler, P. B., D. A. Raff, and W. K. Launroth. 2001. The effect of grazing on the spatial
heterogeneity of vegetation. Oecologia. 128: 465-479.
Adriaens, D., O. Honnay, and M. Hermy. 2006. No evidence of a plant extinction debt in highly
fragmented calcareous grasslands in Belgium. Biological Conservation 133: 212-224.
Aguilar, R., M. Quesada, L. Ashworth, Y. Herrerias-Diego, and J. Lobo. 2008. Genetic
consequences of habitat fragmentation in plant populations: susceptible signals in plant
traits and methodological approaches. Molecular Ecology 17: 5177-5188.
Allen-Wardell, G. et al. 1998. The potential consequences of pollinator declines on the
conservation of biodiversity and stability of food crop yields. Conservation Biology 12:
8-17.
Auffret, A. G., R. Schmucki, J. Reimark, and S. A. O. Cousins. 2012. Grazing networks provide
useful functional connectivity for plants in fragmented systems. Journal of Vegetation
Science DOI: 10.1111/j.1654-1103.2012.01413
Aukema, J. E., and C. M. del Rio. 2002. Where does a fruit-eating bird deposit mistletoe seeds?
Seed deposition patterns and an experiment. Ecology 83: 3489-3496.
Bacles, F. E. C., A. J. Lowe, and R. A. Ennos. 2006. Effective seed dispersal across a fragmented
landscape. Science 311: 628.
Baguette, M. and H. Van Dyck. 2007. Landscape connectivity and animal behavior: functional
grain as a key determinant for dispersal. Landscape Ecology 22: 1117–1129.
126
Barluenga, M., F. Austerlitz, J. A. Elzinga, S. Teixeira, J. Goudet, and G. Bernasconi. 2011.
Fine-scale spatial genetic structure and gene dispersal in Silene latifolia. Heredity 106: 13-
24.
Beinlich, B. and H. Plachter, 1995. Ein Naturschutzkonzept für die Kalkmagerrasen der
Mittleren Schwäbischen Alb (Baden-Württemberg): Schutz, Nutzung und Entwicklung.
Beihefte zu den Veröffentlichungen fürNaturschutz und Landschaftspflege in Baden
Württemberg 83, 520 pp.
Bender, O., H. J. Boehmer, D. Jens, and K. P.Schumacher. 2005. Analysis of land-use change in
a sector of Upper Franconia (Bavaria, Germany) since 1850 using land register records.
Landscape Ecology 20:149-163.
Bisteau, E. and G. Mahy. 2005. Vegetation and seed bank in a calcareous grassland restored
from a Pinus forest. Applied Vegetation Science 8:167-174.
Bizoux, J. P., K. Dainou, O. Raspe, S. Lutts, and G. Mahy. 2008. Fitness and genetic variation of
Viola calaminaria, an endemic metallphyte: implications of population structure and
history. Plant Biology 10: 684-693.
Bloch, D., N. Werdenberg, and A. Erhardt. 2005. Pollination crisis in the butterfly-pollinated
wild carnation Dianthus carthusianorum? New Phytologist 169: 699-706.
Boehmer, H. J, L. Janeck, S. Steidler, and B. Raab. 1990. Verbundsystem Halbtrockenrasen.
Trittstein- und Refugialbiotope im östlichen Landkreis Weißenburg-Gunzenhausen,
Munich: Bayerisches Landesamt für Umweltschutz.
127
Boehmer, H. J. 1994. Die Halbtrockenrasen der Fränkischen Alb - Strukturen, Prozesse,
Erhaltung. Mitteilungen der Fraenkischen Geographischen Gesellschaft 41:323-343.
Bolli, J. C. 2009. A multiscale assessment of plant dispersal: How functional traits and landscape
characteristics interact. Dissertation Nr. 18380, Swiss Federal Institute of Technology
(ETH Zurich), URL: http://e-collection.library.ethz.ch/eserv/eth:41916/eth-41916-01.pdf
Briggs, J. S., B. V. W. Stephen, and H. J. Stephen. 2009. Forest rodents provide directed
dispersal of Jeffrey pine seeds. Ecology 90: 675–687
Bruckmann, S. V., J. Krauss, and I. Steffan-Dewenter. 2010. Butterfly and plant specialist suffer
from reduced connectivity in fragmented landscapes. Journal of Applied Ecology 47:
799-809.
Brudvig, L. A., E. I. Damschen, J. K. Tewksbury, N. M. Haddad, and D. J. Levey. 2009.
Landscape connectivity promotes plant biodiversity spillover into non-target habitats.
Proceedings of the National Academy of the United States of America 23: 9328-9332.
Bruun, H. H, and B. Fritzbøger. 2002. The past impact of livestock husbandry on dispersal of
plant seeds in the landscape of Denmark. Ambio 31: 425-431.
Bruun, H. H. and P. Poschlod. 2006. Why are small seeds dispersed through animal guts: large
numbers or seed size per se? Oikos 113: 402-411.
Burnham, K. P. and D. R., Anderson. 2002. Model selection and multimodel inference: a
practical information-theoretic approach. Springer, Heidelberg.
128
Butaye, J., D. Adriaens, and O. Honnay. 2005. Conservation and restoration of calcareous
grasslands: a concise review of the effects of fragmentation and management on plant
species. Biotechnology, Agronomy, Society and Environment 9: 111-118.
Cain, M. L., B. K. Milligan, and A. E. Strand. 2000. Long-distance dispersal in plant
populations. American Journal of Botany 87: 1217-1227.
Calabrese, J. M., and W. F. Fagan. 2004. A comparison shoppers' guide to connectivity metrics:
Trading off between data requirements and information content. Frontiers in Ecology and
Environment 2: 529-536.
Cavalli-Sforza L. L., and A. W. F. Edwards. 1967. Phylogenetic analysis: models and estimation
procedures. Evolution 32: 550–570. .
Charlesworth, D. 2003. Effects of inbreeding on the genetic diversity of populations.
Philosophical Transactions of the Royal Society of London 358: 1051-1070.
Chen, C., E. D. Forbes, and F. O. Francois. 2007. Bayesian clustering algorithms ascertaining
spatial population structure: a new computer program and a comparison study. Molecular
Ecology Notes 7: 747–756.
Chung, M. G., M. Y. Chung, G. S. Oh, and B. K. Epperson. 2000. Spatial genetic structure in
Neolitsea sericea population (Lauraceae). Heredity 85: 490-497.
Clobert, J., R. A. Ims, and F. Rousset. 2004. Causes, mechanism and consequences of dispersal.
In: Hanski I, Gaggiotti OE (eds) Ecology, genetics and evolution of metapopulations,
Elsevier Academic Press, Amsterdam, The Netherlands, pp 307-335.
129
Cosyns, E., S. Claerbout, I. Lammot, and M. Hoffmann. 2005. Endozoochorous seed dispersal by
cattle and horse in spatially heterogeneous landscape. Plant Ecology 178: 149-162.
Coulson, S. J, J. M. Bullock, M. J. Stevenson, and R. F. Pywell. 2001. Colonization of grassland
by sown species: dispersal versus microsite limitation in response to management. Journal
of Applied Ecology 38: 204-216.
Couvreur, M., B. Vandenberghe, K. Verheyen, and M. Hermy. 2004. An experimental
assessment of seed adhesivity on animal furs. Seed Science Research 14: 147-159.
Couvreur, M., K. Verheyen, and M. Hermy. 2005. Experimental assessment of plant seed
retention times in fur of cattle and horse Flora 200: 136-147.
Crawford, T. J. 1984. The estimation of neighborhood parameters for plant populations. Heredity
53: 273–283.
Cruse-Sanders, J. M., and J. L. Hamrick. 2004. Spatial and genetic structure within populations
of wild Americana Ginseng (Panax quinquefolius). Journal of Heredity 4: 309-321
Cushman, S.A., and E. L. Landguth. 2010. Spurious correlations and inference in landscape
genetics. Molecular Ecology 19: 3592-3602.
Degen , B., H. Caron, E. Bandou, L. Maggia, M. H. Chevallier, A. Leveau, and A. Kremer .
2001. Fine-scale spatial genetic structure of eight tropical tree species as analyzed by
RAPDs. Heredity 87: 497 – 507.
Dewenter, I. S., and T. Tscharntke. 2002. Insect communities and biotic interactions on
fragmented calcareous grasslands- a mini review. Biological Conservation 104: 275-284.
130
Dolek, M. and A, Geyer. 2002. Conserving biodiversity on calcareous grasslands in the
Franconian Jura by grazing: a comprehensive approach. Biological Conservation 104:
351-360.
Dray, S. and A. B. Dufour. 2007. The ade4 package: implementing the duality diagram for
ecologists. Journal of Statistical Software 22: 1-20.
Dupre, C, and J. Ehrlen. 2002. Habitat configuration, species traits and plant distributions.
Journal of Ecology 86: 922-933.
Ellenberg, H .1996. Vegetation Mitteleuropas mit den Alpen, 5th Edition. Stuttgart.
Ellstrand, N. C. 1992. Gene flow by pollen: implication for plant conservation genetics. Oikos
64: 77-86.
Ellstrand, N.C. and D. R. Elam. 1993. Population genetic consequences of small population size.
Implications for plant conservation. Annual Review of Ecology and Systematics 24: 217-
243.
Ennos, R. A. 1994. Estimating the relative rates of pollen and seed migration among plant
populations. Heredity 72: 250–259.
Epperson, B. I. 1993 Recent advances in correlation studies of spatial patterns of genetic
variation. Evolutionary Biology 27: 95-155.
Epperson, B. K. 1995. Spatial distribution of genotypes under isolation by distance. Genetics
140: 1431-1440.
131
Epperson, B. K. 2000. Spatial genetic structure and non-equilibrium demographics within plant
populations. Plant Species Biology 15: 269-279.
Epperson, B. K., and E. R. Alvarez-Buylla. 1997. Limited seed dispersal and genetic structure in
life stages of Cecropia obtusifolia. Evolution 51: 275–282.
Erik, S. J., and S. Priya. 2003. A broader ecological context to habitat fragmentation: Why
matrix habitat is more important than we thought. Journal of Vegetation Science 14: 459-
464.
Fagan, W. F., and J. M. Calabrese. 2006. Quantifying connectivity: balancing metric
performance with data requirements. In: Crooks KR, and M. A. Sanjayan (eds)
Connectivity conservation, Cambridge University Press, Cambridge, pp 297-317.
Fenner, M., and K. Thompson. 2005. The ecology of seeds. Cambridge University Press,
Cambridge.
Fischer, J., and D. B. Lindenmayer. 2007. Landscape modification and habitat fragmentation: a
synthesis. Global Ecology and Biogeography 16: 265-280.
Fischer, S. F., P. Poschlod, and B. Beinlich. 1996. Experimental studies on the dispersal of plants
and animals on sheep in calcareous grasslands. Journal of Applied Ecology 33: 1206–1222.
Fleishman, E., C. Ray, Sjogren-Gulve, P., Boggs, C. L. Murphy, D.D. 2002. Assessing the roles
of patch quality, area, and isolation in predicting metapopulation dynamics. Conservation
Biology 16: 706-712.
Frakham, R. 2005. Genetics and extinction. Biological Conservation 126: 131–140.
132
Freeland, J. R., P. Biss, and J. Silvertown. 2012. Contrasting patterns of pollen and seed flow
influence the spatial genetic structure of sweet vernal grass (Anthoxanthum odoratum)
populations. Journal of Heredity 103: 28-35.
Fuchs, E. J., and J. L. Hamrick. 2010. Spatial genetic structure within size classes of the
endangered tropical tree Guaiacum sanctum (Zygophyllaceae). American Journal of
Botany 97: 1200-1207.
Garcia C., P. Jordano, and J. A. Godoy. 2007. Contemporary pollen and seed dispersal in Prunus
mahaleb population: patterns in distance and direction. Molecular Ecology 16: 1947-1955.
Garcia, C., P. Jordano, J. M. Arroyo, and J. A. Godoy. 2009. Maternal genetic correlations in the
seed rain: effects of frugivore activity in heterogeneous landscapes. Journal of Ecology 97:
1424-1435.
Geertsema, W. 2005. Spatial dynamics of plant species in an agricultural landscape in the
Netherlands. Plant Ecology 178: 237-347.
Godoy, J. A. and P. Jordano. 2001. Seed dispersal by animals: exact identification of source trees
with endocarp DNA microsatellites. Molecular Ecology 10: 2275–2283
Gonzalez, E., J. L. Hamrick, P. E. Smouse, and R. J. Dyer. 2006. Pollen mediated gene dispersal
within continuous and fragmented populations of a forest understory species, Trillium
cuneatum. Molecular Ecology 15: 2047-2058.
Goudet, J. 1995. FSTAT (Version 1.2): A Computer Program to Calculate F-Statistics. Journal of
Heredity 86: 485-486.
133
Goverde, M., K. Scheweizer, B. Baur, and A. Erhardt. 2002. Small-scale habitat fragmentation
effects on pollinator behavior: experimental evidence from the bumblebee Bombus
veteranus on calcareous grasslands. Biological Conservation 104: 293-299.
Greenwald, K.R. 2010. Genetic data in population viability analysis: case studies with
ambystomatid salamanders. Animal Conservation 13:115–122.
Grivet, D., J. J. Robledo-Arnuncio, P. E. Smouse, and V. L. Sork. 2005. Relative contribution of
contemporary pollen and seed dispersal to the effective parental size of seedling population
of California valley oak (Quercus lobata, Ne´e). Molecular Ecology 18: 3967–3979
Hakes, W. 1987. Einfluß von Wiederbewaldungsvorgängen in Kalkmagerrasen auf die
floristische Artenvielfalt und Möglichkeiten der Steuerung durch Pflegemaßnahmen.
Dissertationes Botanicae 109, 151pp.
Hamilton, M. B.1999 Tropical tree gene flow and seed dispersal. Nature 401: 129–130.
Hamrick, J. L, and D. W. Trapnell. 2011. Using population genetic analyses to understand seed
dispersal patterns. Acta Oecologica 37: 641-649.
Hamrick, J. L, and J. D. Nason.1996. Consequences of dispersal in plants. In: Population
Dynamics in Ecological Space and Time (eds Rhodes OE Jr, Chesser RK, Smith MH), pp.
203–236. University of Chicago Press, Chicago, Illinois.
Hamrick, J.L., D. A. Murawski, and J. D. Nason.1993. The influence of seed dispersal
mechanisms on the genetic structure of tropical tree populations. Vegetatio 108: 281–297.
134
Hanski, I. 1994. A practical model of metapopulation dynamics. Journal of Animal Ecology 63:
151-162.
Hanski, I., C. Eralahti, M. Kankare, O. Ovaskainen, and H. Siren. 2004. Variation in migration
propensity among individuals maintained by landscape structure. Ecology Letters 7: 958-
966.
Hardy, O. J., and X. Vekemans. 2002. SPAGeDi: a versatile computer program to analyse spatial
genetic structure at the individual or population levels. Molecular Ecology Notes 2: 618-
620.
Herrera, C. M., and P. Jordano. 1981. Prunus mahaleb and birds: the high efficiency seed
dispersal system of a temperate fruiting tree. Ecological Monographs 51: 203–221.
Heywood, J. S. 1991. Spatial analysis of genetic variation in plant populations. Annual Review
of Ecology and Systematics 22: 335-355
Holderegger, R. and Wagner H. Helene. 2008. Landscape genetics. Bioscience 58: 199-207.
Honnay, O., and H. Jacquemyn. 2006. Susceptibility of common and rare plants species to the
genetic consequences of habitat fragmentation. Conservation Biology 21: 823-832
Honnay, O., D. Adriaens, E. Coart, H. Jacquemyn, and I. Roldan-Ruiz. 2007. Genetic diversity
within and between remnants populations of the endangered calcareous grassland plant
Globularia bisnagarica L. Biological Conservation 8: 293-303.
135
Honnay, O., E. Coart, J. Butaye, D. Adriaens, S. V. Glabeke, and I. Roldán-Ruiz. 2006. Low
impact of present and historical landscape configuration on the genetics of fragmented
Anthyllis vulnearia populations. Biological Conservation 127: 411-419.
Hornberger, T., 1959. Die kulturgeographische Bedeutung der Wanderschaferei in
Suddeutschland. Suddeutsche Tranhumanz. Forschungen zur Deutschen Landeskunde 10:
1–173.
Howe, H. F. and J. Smallwood. 1982. Ecology of seed dispersal. Annual Review of Ecology,
Evolution, and Systematics 13: 201–28.
Hughes, A. R., B. D. Inouye, M. T. J. Johnson, N. Underwood, and M. Vellend. 2008. Ecological
consequences of genetic diversity. Ecology Letters 11: 609-623.
Iwaizumi, M., G. M. Takahashi, A. Watanabe, and M. Ubukata. 2010. Simultaneous evaluation
of paternal and maternal immigrant gene flow and the implications for the overall genetic
composition of Pinus densiflora dispersed seeds. Journal of Heredity 101: 144-153.
Jacobeit, W., 1961. Schafhaltung und Schafer in Zentraleuropa bis zum Beginn des 20.
Jahrhunderts. Akademie-Verlag, Berlin.
Janzen, D. H. 1984. Dispersal of small seeds by big herbivores: foliage is the fruit. The American
Naturalist 123: 338-353.
Johansson, P. and J. Ehrlen J. 2003. Influence of habitat quantity, quality and isolation on the
distribution and abundance of two epiphytic lichens. Journal of Ecology 91: 213-221.
136
Johst, K., R. Brandl, and S. Eber. 2002. Metapopulation persistence in dynamic landscapes: the
role of dispersal distance. Oikos 98: 263–270
Jombart, T. 2008. Adegenet: a R package for the multivariate analysis of genetic markers.
Bioinformatics 24: 1403-1405.
Jones, F. A., J. L. Hamrick, C. J. Peterson, and E. R. Squiers. 2006. Inferring colonization history
from analyses of spatial genetic structure within populations of Pinus strobus and
Quercus rubra. Molecular Ecology 15: 851-861.
Jordano, P., C. García, J. A. Godoy, and J. L. García-Castaño. 2007. Differential contribution of
frugivores to complex seed dispersal patterns. Proceedings National Academy of Sciences
USA 104: 3278-3282
Jordano, P., C. García, J.A. Godoy, and J. L. García-Castaño. 2007. Differential contribution of
frugivores to complex seed dispersal patterns. Proceedings National Academy of
Sciences USA 104: 3278-3282
Joshi, J., P. Stoll, H. P. Rusterholz, B. Schmid, C. Dolt, and B. Baur. 2006. Small-scale
experimental habitat fragmentation reduces colonization rates in species rich-grasslands.
Oecologia 148: 144-152.
Kahmen, S., P. Poschlod, and K. Friedrich-Schreiber. 2002. Conservation management of
calcareous grasslands. Changes in plant species composition and response of functional
traits during 25 years. Biological Conservation 104: 319-328.
Kalinowoski, S.T. 2005. HP-Rare: a computer program for performing rarefaction on measures
of allelic diversity. Molecular Ecology Notes 5: 187-189.
137
Kalisz, S., F. H. Hazawa, S. J. Tonsor, D. A. Thiede, and S. Voigt. 1999. Ant-mediated seed
dispersal alters patterns of relatedness in a population of Trillium gradiflorum. Ecology 80:
2620-2634.
Kalisz, S., J. D. Nason, F. M. Hanzawa, and S. J. Tonsor. 2001. Spatial population genetic
structure in Trillium gradiflorum: The roles of dispersal, mating, history, and selection.
Evolution 55: 1560-1568.
Karubian J., V. L. Sork, T. Roorda, R. Duraes, and T. B. Smith. 2010. Destination-based seed
dispersal homogenizes genetic structure of a tropical palm. Molecular Ecology 19: 1745-
1753.
Kearns, C.A., D. W., Inouye, and N. M., Waser 1998. Endangered mutualisms: the conservation
of plant–pollinator interactions. Annual Review of Ecology and Systematics 29: 83–112.
Keller, L. F., and D. W. Weller 2002. Inbreeding effects in wild populations. Trends in Ecology
and Evolution 117: 230-241.
Keller, L.F, and D. M. Waller. 2002. Inbreeding effects in wild populations. Trends in Ecology
and Evolution 17: 230–241.
Keyghobadi, N., J. Roland, S. F. Matter, and C. Strobeck. 2005. Among- and within-patch
components of genetic diversity respond to different rates to habitat fragmentation: an
empirical demonstration. Proceedings of the Royal Society of Biological Sciences 1562:
553-560.
Kimura, M., and G. H. Weiss. 1964. The stepping stone model of population structure and the
decrease of genetic correlation with distance. Genetics 49: 561-76.
138
Kimura, T., M. Yagi, C, Nishitani, T. Onozaki, Y. Ban, and T. Yamamoto. 2009. Development
of SSR markers in Carnation (Dianthus caryophyllus). Journal of the Japanese Society of
Horticultural Sciences 78: 115-123.
Kleyer, M., Bekker, R. M., Knevel, I. C., Bakker, J. P., Thompson, K., Sonnenschein, M.,
Poschlod, P., van Groenendael, J. M., Klimeš, L., Klimešová, J., Klotz, S., Rusch, G. M.,
Hermy, M., Adriaens, D., Boedeltje, G., Bossuyt, B., Dannemann, A., Endels, P.,
Götzenberger, L., Hodgson, J. G., Jackel, A., Kühn, I., Kunzmann, D., Ozinga, W. A.,
Römermann, C., Stadler, M., Schlegelmilch, J., Steendam, H. J., Tackenberg, O.,
Wilmann, B., Cornelissen, J. H. C., Eriksson, O., Garnier, E. Peco, B. 2008. The LEDA
traitbase: a database of life-history traits of the Northwest European flora. Journal of
Ecology 96: 1266-1274.
Klotz, S., I. Kuhn, and W. Durka. 2002. BIOLFLOR : eine Datenbank mit biologisch-
ökologischen Merkmalen zur Flora von Deutschland. Bundesamt fur Naturschutz, Bonn.
Kolb, A., and M. Diekeman. 2005. Effect of life-history traits on responses of plant species to
forest fragmentation. Conservation Biology 19: 929-938.
Kollmann, J. 2000. Dispersal of fleshy-fruited species: a matter of spatial scale? Perspectives in
Plant Ecology, Evolution, and Systematics 3: 29–51.
Krauss, J., A. M. Klein, I. Steffan-Dewenter, and T. Tscharntke. 2004. Effects of habitat area,
isolation, and landscape diversity on plant species richness of calcareous grasslands.
Biodiversity Conservation 13: 1427-1439.
139
Kuiters, A. T., and H. P. J. Huiskes. 2010. Potential of endozoochorous seed dispersal by sheep
in calcareous grasslands: correlations with seed traits. Applied Vegetation Science
13: 163–172.
Legendre, P., and L. Legendre. 1998. Numerical ecology. 2nd English edition. Elsevier Science
BV, Amsterdam.
Legendre, P., and M.J., Fortin. 2010. Comparison of the Mantel test and alternative approaches
for detecting complex multivariate relationships in the spatial analysis of genetic data.
Molecular Ecology Resources 10: 831-544.
Lehnert, H. 2008. Long-term evaluation of management strategies for calcareous grasslands in
the Altmühltal region (Bavaria, Germany). Master Thesis. University of Bonn. Germany.
Leimu, R., P. Mutikainen, J. Koricheva, and M. Fischer. 2006. How general and positive
relationships between plant population size, fitness and genetic variation? Journal of
Ecology 94: 942-952
Levey, D. J., J. J. Tewksbury, and B. M. Bolker. 2008. Modelling long-distance seed dispersal in
heterogeneous landscapes. Journal of Ecology 96: 599–608.
Levin, D. A. 1981. Dispersal versus gene flow in plants. Annals of the Missouri Botanical
Garden 68: 233-253.
Levin, S. A., H. C. Muller-Landau, R. Nathan, and J. Chave. 2003. The ecology evolution of
seed dispersal: a theoretical perspective. Annual Review of Ecology, Evolution, and
Systematics 34: 575-604.
140
Lindborg, R., and O. Eriksson. 2004. Historical landscape connectivity affects present plant
species diversity. Ecology 85: 1840-1845.
Loiselle, B. A., V. L. Sork, J. Nason, and C. Graham. 1995. Spatial genetic structure of a tropical
understory shrub, Phychotria officinalis (Rubiaceae). American Journal of Botany 82:
1420-1425.
Lorts, C. M., T. Briggeman, and T. Sang. 2008. Evolution of fruit types and seed dispersal: A
phylogenetic and ecological snapshot. Journal of Systematics and Evolution 46: 396-404.
Loveless, M. D. and J. L. Hamrick. 1984. Ecological determinants of genetic structure in plant
populations. Annual Review of Ecological and Systematics 15: 65-95.
Lowe, A. J., D., Boshier, M., Ward, C. F. E. Bacles, and C. Navarro. 2005. Genetic resource
impacts of habitat loss and degradation; reconciling empirical evidence and predicted
theory for neotropical trees. Heredity 95: 255–273.
Manzano, P., and J. Malo. 2006. Extreme long-distance seed dispersal via sheep. Frontiers in
Ecology and the Environment 4: 244-248.
McCauley, D. E. 1997. The relative contributions of seed and pollen movement to the local
genetic structure of Silene alba. Journal of Heredity 88: 257-263.
Mimura, M., and S. N. Aitken. 2007. Increase selfing and decrease effective pollen donor
number in peripheral relative to central populations of Picea sitchensis (Pinaceae).
American Journal of Botany 94: 991–998.
141
Minor, E. S., S. M. Tessel, K. A. M. Engelhardt, and T. R. Lookingbill. 2009. The role of
landscape connectivity in assembling exotic plant communities: a network analysis.
Ecology 90: 1802-1809.
Moilanen, A, and I. Hanski. 2006. Connectivity and metapopulation dynamics in highly
fragmented landscapes. In: Crooks KR, Sanjayan MA (eds) Connectivity conservation,
Cambridge University Press, Cambridge, pp 44-71.
Mortelliti, A., G., Amori, and L., Boitani. 2010. The role of habitat quality in fragmented
landscapes: a conceptual overview and prospectus for future research. Oecologia 163:
535-547.
Moussie, A. M., W. Lengkeek, and R. Van Diggelen. 2005. Estimating adhesive seed dispersal
distances: field experiments and correlated random walks. Functional Ecology 19: 478-
486.
Murphy, H.T, and J. Lovett-Doust. 2004. Context and connectivity in plant metapopulations and
landscape mosaics: does the matrix matter? Oikos 105: 3-14.
Nathan R., J. M. Bullock, R. Ophelie, and F. M. Schurr. 2009. Seed dispersal. ELS.
DOI: 10.1002/9780470015902.a0021225.
Nathan, R. 2006. Long-distance dispersal of plants. Science 313: 786-788.
Nathan, R., and H. C. Muller-Landau. 2000. Spatial patterns of seed dispersal, their
determinants and consequences for recruitment. Trends Ecology and Evolution 15: 278-
285.
142
Nathan, R., F. M. Schurr, O. Spiegel, O. Steinitz, A. Trakhtenbrot, and A. Tsoar. 2008.
Mechanisms of long-distance seed dispersal. Trends in Ecology and Evolution 23: 638-
646.
Ndiade-Bourobou, D., O. J. Hardy, B. Favreau, H. Moussavou, E. Nzengue, A. Mignots, and J.
M. Bouvet. 2010. Long-distance seed and pollen dispersal inferred from spatial genetic
structure in the very low-density rainforest tree, Baillonella toxisperma Pierre, in Central
Africa. Molecular Ecology 19: 4949-4962.
Neel, M. C. 2008. Patch connectivity and genetic diversity conservation in the federally
endangered and narrowly endemic plant species Astragulus albens (Fabaceae). Biological
Conservation 141: 938-955.
O’Connell, L. M., A. Mosseler, and O. P. Rajora. 2006. Impacts of forestfragmentation on the
mating system and genetic diversity of white spruce (Picea glauca) at the landscape
level. Heredity 97: 418–426.
Oberdorfer, E. 1978. Süddeutsche Pflanzengesellschaften, vol. 2. Gustav Fischer Verlag.
Orrock, J.L., D. J. Levey, B. J. Danielson, and E. I. Damschen. 2006. Seed predation, not seed
dispersal, explains the landscape-level abundance of an early-successional plant. Journal
of Ecology 94: 838-845.
Ouborg, N. J. and R. Van Treuren. 1994. The significance of genetic erosion in the process of
extinction. IV. Inbreeding load and heterosis in relation to population size in the
mint Salvia pratensis. Evolution 48: 996–1008.
143
Ouborg, N.J., Y. Piquot, and J. M. Van Groenendael. 1999. Population genetics, molecular
markers and the study of dispersal in plants. Journal of Ecology 87: 551-568.
Ovaskainen, O., and I. Hanski. 2004. Metapopulation dynamics in highly fragmented landscapes.
In: Hanski I, Gaggiotti OE (eds) Ecology, genetics and evolution of metapopulations,
Elsevier Academic Press, Amsterdam, The Netherlands, pp 73-104.
Panell, J. R. and M. E. Dorken. 2006. Colonisation as a common denominator in plant
metapopulations and range expansions: effects on genetic diversity and sexual systems.
Landscape Ecology 21: 837-848.
Pannell, J. R. and B. Charlesworth. 2000. Effects of metapopulation processes on measures of
genetic diversity. Philosophical Transactions of the Royal Society 355: 1851-1864.
Peakall R., and P. E. Smouse 2006. GENALEX 6: genetic analysis in excel. Population genetic
software for teaching and research. Molecular Ecology Notes 6: 288–295.
Petit R, J. Duminil, S. Fineschi, A. Hampe, D. Salvini , and G. Vendramin. 2005. Comparative
organization of chloroplast, mitochondrial and nuclear diversity in plant populations.
Molecular Ecology 14: 689–701.
Piessens K, O. Honnay, and M. Hermy. 2005. The role of fragment area and isolation in the
conservation of heathland species. Biological Conservation 122:61-69.
Piqueray, J., G. Bottin, L. Delescaille, E. Bisteau, G. Colinet and G. Mahy. 2011. Rapid
restoration of a species-rich ecosystem assessed from soil and vegetation indicators: The
case of calcareous grasslands restored from forest stands. Ecological Indicators 11: 724-
733.
144
Poschlod, P. and M. F. WallisDeVries. 2002 The historical and socioeconomic perspective of
calcareous grasslands: lessons from the distant and recent past. Biological Conservation
104: 361-376.
Poschlod, P., M. Kleyer, A. K. Jackel, A. Dannemann, and O. Tackenberg. 2003. BIOPOP: a
database of plant traits and internet application for nature conservation. Folia
Geobotanica 38: 263-271.
Poschlod, P., S. Kiefer, U. Tränkle, S. Fischer, and S. Bonn. 1998. Plant species richness in
calcareous grasslands as affected by dispersability in space and time. Applied Vegetation
Science 1: 75-90.
Purves, D. W., and J. Dushoff. 2005. Directed seed dispersal and metapopulation response to
habitat loss and disturbance: Application to Eichhornia paniculata. Journal of Ecology 93:
658-669.
R Development Core Team 2010. R: A language and environment for statistical computing. R
Foundation for Statistical Computin, Vienna, Austria. ISBN 3-900051-07-0. http://www.R-
project.org.
Raymond, M., and F. Rousset. 1995. GENEPOP (version 1.2): population genetics software for
exact tests and eumeniscm. Journal of Heredity 86: 248-249.
Reitalu, T., L. T. Johansson, M. T. Sykes, K. Hall, and H. H. Prentice. 2010. History matters:
village distances, grazing and grassland species diversity. Journal of Applied Ecology 47:
1216-1224.
145
Rico, Y., H. J. Boehmer, H. H. Wagner. 2012. Determinants of actual functional connectivity for
calcareous grassland communities linked by rotational sheep grazing. Landscape Ecology.
27: 199-209.
Ritland, K.1996. Estimators for pairwise relatedness and individual inbreeding coefficients.
Genetical Research Cambridge 67: 175–185.
Robledo-Arnuncio, J. J., R. Alıa, and L. Gil. 2004. Increased selfing and correlated paternity in a
small population of a predominantly outcrossing conifer, Pinus sylvestris. Molecular
Ecology 13: 2567–2577.
Romermann, C., O. Tackenberg, and P. Poschlod. 2005. How to predict attachment potential of
seeds to sheep and cattle coat from simple morphological seed traits. Oikos 110: 219-230.
Rusch, G., and J. M. Fernandez-Palacios. 1995. The influence of spatial heterogeneity on
regeneration by seed in a limestone grassland. Journal of Applied Ecology 6: 417-426.
Sato T., Y. Isagi, H. Sakio, K. Osumi , S. Goto. 2006. Effect of gene flow on spatial genetic
structure in the riparian canopy tree Cercidiphyllum japonicum revealed by microsatellite
analysis. Journal of Heredity 96: 79-84.
Saunders, D. A., R. J. Hobbs, and C. R. Margules. 1991. Biological consequences of ecosystem
fragmentation: a review. Conservation Biology 5: 18-32.
Schleicher, A., R. Biedermann, and M. Kleyer. 2011. Dispersal traits determine plant response to
habitat connectivity in an urban landscape. Landscape Ecology 26: 529-540.
146
Schrautzer, J., D. Jansen, M. Breuer, and O. Nelle. 2009. Succession and management of
calcareous dry grasslands in the Northern Franconian Jura, Germany. Tuexenia 29: 339-
557 351.
Schupp, E. W. 1993. Quantity, quality and the effectiveness of seed dispersal by animals.
Vegetatio 108: 15–29.
Sebbenn, A. M., A. C. M. Carvalho, M. L. M. Freitas, S. M. B. Moraes, A. P. S. C. Gaino, J. M.
da Silva, C. Jolivet, and M. L. T. Moraes. 2011. Low levels of realized seed and pollen
gene flow and strong spatial genetic structure in a small, isolated and fragmented
population of the tropical tree Copaifera langsdorffii Desf. Heredity 106: 134-145.
Sezen, U. U., R. L. Chazdon, and K. E. Holsinger. 2005. Genetic consequences of tropical
second-growth forest regeneration. Science 307: 891
Shupp, E. W., P. Jordano, and J. M. Gomez. 2011. Seed dispersal effectiveness revisited: a
conceptual review. New Phytologist 188: 333-353.
Slatkin, M . 1987. Gene flow and the geographic structure of natural populations. Science 236:
787-92.
Smulders, M. J. M., Y. Nordijk, W. Rus-Kortekaas, G. M. M. Bredemeijer, and B. Vosman.
2003. Microsatellite genotyping of carnation varieties. Theoretical and Applied Genetics
106: 1191-1195.
Smulders, M. J. M., W. Rus-Kortekaas, and V. Vosman. 2000. Microsatellite markers useful
throughout the genus Dianthus. Genome 43: 208-210.
147
Soons, M. B, R. Nathan, and G. G. Katul. 2004. Human effects on long-distance wind
dispersal and colonization by grassland plants. Ecology 85: 3069-3079.
Sork, V. L., J. Nason, D. R. Campbell, and J. F. Fernandez. 1999. Landscape approach to
historical and contemporary gene flow in plants. Trends in Ecology and Evolution 14: 218-
224.
Sork, V., and P. E. Smouse. 2006. Genetic analysis of landscape connectivity in tree populations.
Landscape Ecology 21: 821–836.
Spiegel, O. and R. Nathan. 2007. Incorporating dispersal distance into the disperser effectiveness
framework: frugivorous birds provide complementary dispersal to plants in a patchy
environment. Ecology Letters 10: 718-728.
Tackenberg, O. 2003. Modeling long distance dispersal of plant diaspores by wind. Ecological
Monographs 73:173–89.
Takezaki, N. and M. Nei. 1996. Genetic distances and reconstruction of phylogenetic trees from
microsatellite data. Genetics 144: 389-399.
Taylor, P. D, L. Fahrig, and K. A.With.2006. Landscape connectivity: a return to the basics. In:
Crooks KR, Sanjayan MA (eds). Connectivity conservation, Cambridge University Press,
Cambridge, pp 31-43.
Taylor, P. D., L. Fahrig, K. Henein, and G. Merriam. 1993. Connectivity is a vital element of
landscape structure. Oikos 69: 571–573.
148
Templeton, A.R., K. Shaw, E. Routman, S. K. Davis. 1990. The genetic consequences of habitat
fragmentation. Annals of the Missouri Botanical Garden 77: 13-027.
Tischendorf, L. and L, Fahrig. 2000. On the usage and measurement of landscape connectivity.
OIKOS 90: 7-19.
Torimaru T, N. Tani, Y. Tsumura, N. Nishimura, and N. Tomaru. 2007. Effects of kin-structured
seed dispersal on the genetic structure of the clonal dioecious shrub Ilex leucoclada.
Evolution 61: 1289–1300.
Trapnell, D. W. and J. L. Hamrick. 2004. Three-dimensional fine scale genetic structure of the
neotropical epiphytic orchid, Laeila rubescens. Molecular Ecology 13: 1111-1118.
Trapnell, D.W. and J. L. Hamrick. 2005. Mating patterns and gene flow in the neotropical
epiphytic orchid, Laelia rubescens. Molecular Ecology 14: 75-84.
Traveset, A. and J. Rodríguez-Pérez. 2008. Seed dispersal. Instituto Mediterraneo de Estudios
Avanzados (CSIC-UIB), Sporles, Spain.
Tremlova, K., and Z. Munzbergova. 2008. Importance of species traits for species distribution in
fragmented landscapes. Ecology 88: 965-977.
Troupin, D., R. Nathan, and G. G. Vendramin. 2006. Analysis of spatial genetic structure in an
expanding Pinus halepensis population reveals development of fine-scale genetic
clustering over time. Molecular Ecology 15: 3617-3630.
149
Uezu, A., J.P. Metzger, and J. M. E. Vielliard. 2005. Effects of structural and functional
connectivity and patch size on the abundance of seven Atlantic forest birds species. Biol
Conservation 123: 507-519.
Vanden-Broeck A., R. Gruwez, K. Cox, S. Adriaenssens, I. M. Michalczyk, and K. Verheyen.
2011. Genetic structure and seed-mediated dispersal rates of an endangered shrub in a
fragmented landscape: a case study of Juniperus communis in northwestern Europe.
Genetics 12: 73-89.
Vekemans, X. and O. J. Hardy. 2004. New insights from fine-scale spatial genetic structure
analyses in plant populations. Molecular Ecology 13: 921–935.
Vellend, M., and M. A. Geber. 2005. Connections between species diversity and genetic
diversity. Ecology Letters 8: 767-781.
Volis, S., M. Zaretsky, and I. Shulgina. 2010. Fine-scale spatial genetic structure in a
predominantly selfing plant: role of seed and pollen dispersal. Heredity 105: 384-393.
Vranckx, G., H. Jacquemyn, B. Muys, and O. Honnay. 2011. Meta-analysis of susceptibility of
woody plants to loss genetic diversity through habitat fragmentation. Conservation Biology
26: 228-237
Wagner, H. H., Y. Rico, H. Lehnert and H. J. Boehmer. 2012. Process-based long-term
evaluation of an ecological network of calcareous grasslands connected by sheep
herding. Ecography 35: 001-009.
Walentowski, H., B. Raab and W. A. Zahlheimer. 1991: Vorläufige Rote Liste der in Bayern
nachgewiesenen oder zu erwartenden Pflanzengesellschaften. - II. Wirtschaftswiesen und
150
Unkrautgesellschaften. - Berichte der Bayerischen Botanischen Gesellschaft zur
Erforschung der heimischen Flora, Beiheft, Vol.62, 1, pp.1-85, München
WallisDeVries, M. F., P. Poschlod, and J. H. Willems. 2002. Challenges for the conservation of
calcareous grasslands in northwestern Europe: integrating the requirements of flora and
fauna. Biological Conservation 104: 265–273
Wang, R, S. G. Compton, and X. Y. Chen. 2011. Fragmentation can increase spatial genetic
structure without decreasing pollen-mediated gene flow in a wind-pollinated tree.
Molecular Ecology 20: 4421-4432.
Weir, B. S., and C. C. Cockerman. 1984. Estimating F-statistics for the analysis of population
structure. Evolution 38: 1358-1370.
Wenny, D. G. 2001. Advantages of seed dispersal: a re-evaluation of directed dispersal.
Evolutionary Ecology Research 3: 51–74.
Whitlock, M. C. and D. E. McCauley. 1990 Some population genetic consequences of colony
formation and extinction: genetic correlations within founding groups. Evolution 44: 1717-
1724.
Willems, J. H, and L. P. M. Bik.1998. Restoration of high species density in calcareous
grassland: the role of seed rain and soil seed bank. Applied Vegetation Science 1: 91-100.
Willendirg, C., and P. Poschlod. 2002. Does dispersal by sheep affect population genetic
structure of the calcareous grassland species Bromus erectus? Biological Conservation 104:
329-337.
151
Wright, 1946. Isolation by distance under diverse systems of mating. Genetics 31: 39-59
Wright, S. 1921. Systems of mating. II. The effects of inbreeding on the genetic composition of a
population. Genetics 6: 124-143.
Young, A. C, and G. M., Clarke. 2000. Genetics, Demography and Viability of Fragmented
Populations. Cambridge University Press, Cambridge.
Young, A. G., T. Boyle, and T. Brown. 1996. The population genetic consequences of habitat
fragmentation for plants. Trends in Ecology and Evolution 11: 413–418.
Young, A.C, and G. M. Clarke. 2000. Genetics, demography and viability of fragmented
populations. Cambridge University Press, Cambridge.
Zhao, R., X. Hanbing, and L. Bao-Rong. 2009. Fine-scale genetic structure enhances biparental
inbreeding by promoting mating events between more related individuals in wild soybean
(Glycine soja; Fabaceae) populations. American Journal of Botany 96: 1138-1147.
Zhou H., J. Chen, and F. Chen. 2007. Ant-mediated seed dispersal contributes to the local spatial
pattern and genetic structure of Globba lancangensis (Zingiberaceae). Journal of Heredity
98: 317-324.
152
Appendices A1. Distribution of calcareous grasslands in the study area located in the Southern Fraconian Alb
(map from Wagner et al. 2012). Square symbols correspond to previously abandoned grassland
patches that were reconnected with core areas (red polygons) in three non-overlapping
shepherding systems (red, blue, and green bold lines). Sheep grazing treatments (consistently and
intermittently) since 1989 and patches that had remained ungrazed are indicated by the shading
pattern of the square symbols. Inset map shows the distribution of calcareous grasslands (grey
areas) in Germany, while the orange box indicates the location of the study area. This map has
previously been published by the Journal of Ecography: Wagner et al. 2012. Permission to
use this published material in this dissertation has been obtained from the publisher
153
A2. Optimized α-values for each dispersal model and for each analyzed calcareous grassland plants. Bold numbers indicate the α-value for
the best performing dispersal model.
Table A2. Continue on the following page
Species Geographic distance (Eu)
Matrix resistance (Matrix)
Consistently grazed (Shecte)
Consistently and intermittently grazed (Sheint)
Grazed within the same grazed system (Shenu)
Hieracium pilosella 0.85 1.56 0.72 0.81 2.42
Leontodon hispidus 2.50 0.97 1.16 1.52 0.67
Sanguisorba minor 2.50 2.50 2.50 2.50 2.26
Arabis hirsute 1.19 2.50 1.57 0.85 1.05
Centaurea jacea 2.50 1.68 1.36 2.50 1.53
Koeleria pyramidata 0.66 0.47 2.50 1.96 1.78
Linum catharticum 0.72 1.34 0.67 0.56 1.78
Medicago lupulina 0.91 0.10 0.55 1.00 2.10
Plantago media 2.50 0.53 0.83 0.32 1.80
Polygala comosa 0.78 2.50 0.92 1.06 1.99
Prunella grandiflora 0.10 1.35 0.86 0.87 1.36
Ranunculus bulbosus 2.50 2.06 0.65 0.54 1.02
Salvia pratensis 2.30 2.50 0.11 0.10 2.15
Scabiosa columbaria 2.50 2.50 2.50 1.25 1.09
154
Species Geographic distance (Eu)
Matrix resistance (Matrix)
Consistently grazed (Shecte)
Consistently and intermittently grazed (Sheint)
Grazed within the same grazed system (Shenu)
Anthyllis vulneraria 1.07 0.66 2.50 2.50 2.22
Campanula rotundifolia 0.89 1.50 0.67 0.80 2.26
Carlina acaulis 1.84 1.53 0.63 0.72 1.51
Cirsium acaule 0.83 0.10 1.81 1.56 1.04
Pulsatilla vulgaris 0.87 2.50 2.50 2.50 1.92
Ajuga genevensis 2.50 0.40 0.37 2.50 1.04
Asperula cynanchica 0.60 2.29 2.50 1.13 1.02
Dianthus carthusianorum 1.86 2.20 0.65 1.45 2.43
Euphorbia cyparissias 2.50 2.50 2.50 2.40 1.58
Euphorbia verrucosa 0.68 0.45 2.50 0.37 2.30
Hippocrepis comosa 0.51 2.50 0.52 0.72 1.14
Ononis repens 0.28 0.53 0.91 0.46 2.09
Ononis spinosa 0.41 0.67 2.50 2.50 1.63
Onobrychis viciifolia 0.47 0.43 2.50 0.87 1.93
Phleum phleoides 0.41 0.43 0.13 2.50 1.58
Stachys recta 1.06 0.53 2.46 1.48 0.74
Trifolium montanum 2.50 1.40 0.71 0.64 1.57
155
A3. Genetic diversity indices of eleven loci amplified for 1,613 individuals from 64 populations
in Dianthus carthusianorum. Number of alleles per loci, Ho, observed heterozygosity; He,
expected heterozygosity, Ar mean allelic richness, FST, fixation index. SD, standard deviation of
the mean.
Locus No. alleles Ho (SD) He (SD) Ar (SD) FST
DC109 12 0.68 ± 0.02 0.69 ± 0.01 4.40 ± 1.0 0.035
MSACC 13 0.20 ± 0.02 0.2 ± 0.01 1.83 ± 0.39 0.039
MSBOX 6 0.48 ± 0.02 0.53 ± 0.01 2.40 ± 0.44 0.042
MSA30 7 0.35 ± 0.02 0.37 ± 0.02 2.56 ± 0.63 0.035
MSBSY 21 0.65 ± 0.02 0.80 ± 0.01 5.18 ± 1.1 0.037
CB004 11 0.67 ± 0.02 0.68 ± 0.02 4.08 ± 0.93 0.046
CB057 18 0.81 ± 0.01 0.87 ± 0.01 6.27 ± 1.4 0.044
CF003 19 0.77 ± 0.01 0.88 ± 001 6.60 ± 1.4 0.044
CB011 2 0.04 ± 0.01 0.04 ± 0.01 1.20 ± 0.25 0.045
CB027 11 0.65 ± 0.02 0.71 ± 0.07 4.10 ± 0.77 0.038
CB018 13 0.69 ± 0.02 0.76 ± 0.01 4.50 ± 1.0 0.056
Total 133 0.55 ± 0.25 0.59 ± 0.28 4.04 ± 1.7 0.042 ± 0.002
156
A4. Multiple comparisons with Tukey’s HSD (family-wise significance level of alpha = 0.05) of
mean scores for the three interclass PCA axes accounting for three shepherding systems and
populations in ungrazed patches. Significant values are in bold.
Pairwise groups Difference Lower limit Upper limit p-value
adjusted
PCA axis 1:
herd 2 vs. herd 3 -5.485 -7.750 -3.219 0.0001
herd 1 vs. herd 3 -0.460 -2.889 1.969 0.936
ungrazed vs.herd 3 -2.599 -5.140 -0.059 0.014
herd 1 vs. herd 2 5.024 3.112 6.937 0.0001
ungrazed vs. herd 2 2.885 0.833 4.937 0.0003
ungrazed vs. herd 1 -2.139 -4.371 0.092 0.023
PCA axis 2:
herd 2 vs. herd 1 -3.570 -4.922 -2.218 0.0001
herd 3 vs. herd 1 -5.703 -7.153 -4.254 0.0001
ungrazed vs. herd 3 -2.902 -4.418 -1.386 0.0001
herd 3 vs. herd 2 -2.133 -3.274 -0.992 0.0001
ungrazed vs. herd 2 0.668 -0.556 1.892 0.344
ungrazed vs. herd 3 2.801 1.469 4.132 0.0001
PCA axis 3:
Herd 2 vs. herd 3 0.426 -0.984 1.837 0.788
Herd 1vs. herd 3 0.729 -0.783 2.242 0.454
Ungrazed vs. herd 1 4.121 2.539 5.704 0.0001
Herd 1 vs. herd 2 0.302 -0.888 1.493 0.862
Ungrazed vs. herd 2 3.695 2.416 4.973 0.0001
Ungrazed vs. herd 3 3.392 2.002 4.782 0.0001
157
A5. A) Plot of the lowest values of the deviance information criterion (DIC) averaged over 10
run using TESS with admixture models varying K from 2 to 10, ψ = 0.6, burn-in lengths of
100,000 and with 5000 sweeps. Although the plateau starts at kmax = 4, we observed that
additional clusters of runs kmax = 3 to 10 were ambiguously identified as they had very low
membership probabilities. B) Posterior estimates of cluster memberships for kmax 2 to 5. The
two main clusters at the east and the west of the study area remained relatively consistent
among runs, with little variation of membership scores of the added clusters for each k.
A)
B)
A)
B)
kmax = 2
kmax = 3
kmax = 4
kmax = 5
158
A6. Estimates of genetic diversity calculated for eleven polymorphic loci amplified in 49 Dianthus carthusianorum populations
defined by population size, population history, and shepherding system. Populations with less than ten individuals are not included. N,
sampled size; Ho, observed heterozygosity; He, expected heterozygosity, Ar allelic richness corrected by rarefaction; FIS, inbreeding
coefficient.
Population Latitude Longitude N
Population
size
Population
history
Shepherding
system Ho He Ar FIS
E01 48º99’85.6”N 10º 96’01.9”E 34 big unknow herd 1 0.553 0.613 4.54 0.099
G01 49º 01’37.0”N 10º96’85.0”E 29 big unknow herd 1 0.566 0.607 4.27 0.03
G100 48º 99’44.5”N 10º96’56.9”E 58 big unknow herd 1 0.547 0.603 4.45 0.123
G16 48º 97’39.7”N 10º97’01.3”E 20 big unknow herd 1 0.524 0.598 4.49 0.093
G20 48º 97’09.0”N 10º95’56.3”E 29 big unknow herd 1 0.566 0.602 4.5 0.107
N03 48º 98’75.5”N 10º95’67.5”E 27 big unknow herd 1 0.545 0.637 4.7 0.108
N05 49º 02’68.4”N 11º00’51.6”E 28 big unknow herd 1 0.513 0.573 4.33 0.066
A45 48º 97’44.4”N 10º94’83.7”E 16 small colonized herd 1 0.592 0.603 4.32 0.02
A06 48º 99’08.2”N 11º05’60.3”E 29 big colonized herd 2 0.542 0.638 4.58 0.097
A07a 48º 98’37.7”N 11º06’08.5”E 27 big colonized herd 2 0.548 0.592 4.35 0.085
A14 48º 95’93.0”N 11º02’21.7”E 16 big pre-existing herd 2 0.528 0.542 3.56 0.11
A18 48º97’18.8”N 11º00’32.4”E 27 big pre-existing herd 2 0.567 0.591 4.35 0.045
Table A6. Continue on the following page
159
Population Latitude Longitude N
Population
size
Population
history
Shepherding
system Ho He Ar FIS
E07 49º00’36.0”N 11º06’41.2”E 15 big unknow herd 2 0.596 0.477 4.45 0.102
G05a 49º04’65.6”N 11º03’19.9”E 27 big unknow herd 2 0.538 0.598 4.45 0.145
G13 49º00’79.1”N 11º06’30.4”E 41 big unknow herd 2 0.536 0.599 4.67 0.069
G37 48º98’29.6”N 11º01’59.2”E 26 big unknow herd 2 0.515 0.606 4.6 0.10
G40 48º97’79.6”N 11º03’93.8”E 30 big unknow herd 2 0.563 0.570 4.26 0.153
G45 48º97’42.5”N 11º05’57.4”E 29 big unknow herd 2 0.487 0.566 4.17 0.111
G46 48º97’67.4”N 11º02’53.1”E 28 big unknow herd 2 0.521 0.583 4.44 0.05
G47 48º97’74.3”N 11º01’50.2”E 26 big unknow herd 2 0.549 0.591 4.42 0.121
G48 48º98’93.5”N 11º06’06.7”E 26 big unknow herd 2 0.558 0.613 4.56 0.108
G49 48º98’61.2”N 11º05’50.1”E 30 big unknow herd 2 0.511 0.576 4.26 0.065
G50 48º98’06.8”N 11º05’52.2”E 29 big unknow herd 2 0.544 0.582 4.45 0.126
A03 48º99’78.1”N 11º05’99.4”E 26 small colonized herd 2 0.517 0.571 3.89 0.096
A05 48º99’27.3”N 11º05’62.1”E 25 small colonized herd 2 0.509 0.577 4.21 0.09
A08 48º97’01.8”N 11º02’60.1”E 20 small colonized herd 2 0.486 0.582 4.27 0.146
A12 48º96’61.4”N 11º02’60.8”E 12 small pre-existing herd 2 0.462 0.587 4.28 0.22
C08 48º95’56.7”N 11º02’89.5”E 30 small unknow herd 2 0.494 0.525 3.84 0.14
A28 48º95’56.8”N 10º94’13.8”E 33 big pre-existing herd 3 0.565 0.598 4.21 0.056
A29 48º95’47.4”N 10º94’26.2”E 27 big pre-existing herd 3 0.571 0.600 4.38 0.048
Table A6. Continue on the following page
160
Population Latitude Longitude N
Population
size
Population
history
Shepherding
system Ho He Ar FIS
A38 48º93’66.0”N 10º98’19.3”E 29 big pre-existing herd 3 0.552 0.598 4.45 0.078
G21 48º96’60.8”N 10º97’35.2”E 29 big unknow herd 3 0.554 0.612 4.68 0.127
G23 48º96’17.8”N 10º96’53.9”E 27 big unknow herd 3 0.558 0.632 4.49 0.096
G26 48º95’57.3”N 10º95’09.4”E 35 big unknow herd 3 0.555 0.597 4.63 0.037
G28 48º94’86.8”N 10º95’66.9”E 26 big unknow herd 3 0.539 0.589 4.55 0.044
G29 48º95’04.2”N 10º94’93.4”E 29 big unknow herd 3 0.509 0.575 4.41 0.072
G30 48º94’22.0”N 10º97’77.0”E 28 big unknow herd 3 0.512 0.609 4.44 0.13
G31 48º94’23.4”N 10º98’35.6”E 30 big unknow herd 3 0.544 0.604 4.55 0.117
Gzim 48º92’48.2”N 10º98’96.8”E 28 big unknow herd 3 0.564 0.637 4.92 0.099
A26 48º95’91.9”N 10º94’27.4”E 25 small colonized herd 3 0.538 0.603 4.52 0.11
A31 48º95’37.8”N 10º94’51.0”E 23 small colonized herd 3 0.514 0.594 4.38 0.138
A33 48º94’85.9”N 10º93’68.9”E 35 small colonized herd 3 0.542 0.602 4.37 0.101
A25 48º96’04.1”N 10º94’21.8”E 11 small pre-existing herd 3 0.521 0.582 4.05 0.108
A37 48º94’75.9”N 10º98’46.7”E 30 big colonized ungrazed 0.499 0.576 4.1 0.136
N10 48º95’99.9”N 10º97’91.1”E 23 big unknow ungrazed 0.505 0.540 3.93 0.102
E03 48º94’52.7”N 10º93’75.8”E 15 small unknow ungrazed 0.577 0.577 4.1 0.0
E04 48º95’01.3”N 11º01’52.8”E 30 small unknow ungrazed 0.578 0.579 4.01 0.034 E09 48º97’02.8”N 10º95’03.9”E 32 small unknow ungrazed 0.531 0.605 4.37 0.034
Nroth 49º04’22.4”N 11º00’16.6”E 14 small unknow ungrazed 0.429 0.475 3.64 0.107
Top Related