Fragmentation and dispersal in freshwater wetlands

Fragmentatie en zaadverspreiding in zoetwaterwetlands ISBN: 978-94-6108-290-9

Cover: Hester Soomers

Grafische vormgeving: Gildeprint Enschede

Figuren: Geomedia, Faculteit Geowetenschappen, Universiteit Utrecht Foto’s: Hester Soomers

Printed by: Gildeprint, Enschede Printed on FSC certified paper

© 2012 Alle rechten voorbehouden. Niets uit deze uitgave mag worden verveelvoudigd, opgeslagen in een geautomatiseerd gegevensbestand, of openbaar gemaakt, in enige vorm of op enig wijze, hetzij elektronisch, mechanisch, door fotokopieën, opnamen, of op enig andere manier, zonder voorafgaande schriftelijke toestemming van de rechthebbende. Fragmentation and in freshwater wetlands

Fragmentatie en zaadverspreiding in zoetwaterwetlands

(met een samenvatting in het Nederlands)

Proefschrift

ter verkrijging van de graad van doctor aan de Universiteit Utrecht op gezag van de rector magnificus, prof.dr. G.J. van der Zwaan, ingevolge het besluit van het college voor promoties in het openbaar te verdedigen op woensdag 30 mei 2012 des middags te 4.15 uur

door

Hester Soomers

geboren op 5 augustus 1977 te Heerlen Promotoren: Prof.dr. M.J. Wassen Prof.dr. J.T.A. Verhoeven

Co-promotor: Dr. P.A. Verweij Contents

Chapter 1 Introduction 7

Chapter 2 The effect of and abiotic factors on fen plant occurrence 29

Chapter 3 Factors influencing the seed source and sink functions of a floodplain nature reserve in the Netherlands 51

Chapter 4 The dispersal and deposition of hydrochorous plant in drainage ditches 75

Chapter 5 Linking habitat suitability and seed dispersal models in order to analyse the effectiveness of hydrological fen restoration strategies 99

Chapter 6 Wind and water dispersal of wetland plants across fragmented landscapes 127

Chapter 7 Synthesis 163

Abstract 181 Samenvatting 187 Dankwoord 195 Curriculum Vitae 201 Chapter 1 Chapter 1

Introduction

8 | Chapter 1

R1 R2 R3 R4 R5 R6 R7 R8 R9 R10 R11 R12 R13 R14 R15 R16 R17 R18 R19 R20 R21 R22 R23 R24 R25 R26 R27 R28 R29 R30 R31 R32 R33 R34 R35 R36 R37 R38 R39 Introduction | 9

1.1 sCoPe R1 R2 The focus of this thesis is on plant seed dispersal via surface water in fragmented freshwater R3 wetlands. R4 In the next sections, I argue why habitat fragmentation is a threat to plant R5 and why it is relevant to study seed dispersal in the context of habitat fragmentation. Also, R6 I explain why freshwater wetlands were chosen as a study . This chapter ends with R7 the aims and outline of this thesis, where a preview of the executed research is given and the R8 research approach is explained. R9 R10 1.2 Human InduCed envIronmental tHreats to Plant bIodIversIty R11 R12 Human impact on nature has been severe in the past centuries. The rate of biodiversity loss R13 massively increased since the beginning of the industrial period, and is now more than hundred R14 times higher than in the pre-industrial period, for well studied taxonomic groups (Vitousek R15 et al., 1997; Rockström et al., 2009). The Royal Botanic Gardens in Kew estimated more than R16 20% of all plant species to be threatened with extinction (Royal Botanic Gardens 2010) and R17 May (2011) states that approximately 70% of the evaluated angiosperm plants are estimated R18 to be threatened. There is evidence that a rich biodiversity contributes to the stability of R19 (e.g. Folke et al. 2004, Tilman et al. 2006)) and that reduced plant species diversity R20 may negatively affect ecosystem functioning (Naeem et al. 1994, Isbell et al. 2011). R21 Land use change is seen as the main driver of the loss of biodiversity worldwide (Vitousek R22 et al. 1997, Sala et al. 2000, Millennium Ecosystem Assessment 2005, Rockström et al. 2009). A schematisation of the effects of land use change on plant habitat is given in Figure 1.1. R23 Conversion of natural ecosystems into agricultural land or urban areas has led to habitat loss R24 for many plant species; croplands and pastures now occupy ca 40% of the land surface (Foley R25 et al. 2005). R26 Additionally, habitat quality of the remaining natural ecosystems is often negatively affected R27 by land use change. Fertilizers used in modern agriculture are a major source of excess nitrogen R28 and phosphorus in surface water and also cause leaching of nutrients to groundwater and R29 increased atmospheric nitrogen deposition (Foley et al. 2005). Consequently, the resulting R30 of terrestrial and aquatic natural habitats has led to higher plant productivity R31 and thereby often to reduced plant diversity (Grime 1979, Moore and Keddy 1988, Moore R32 et al. 1989). Apart from eutrophication, human land use may lead to desiccation of natural R33 habitats. Drainage of wetlands to enable agricultural activities and groundwater extraction for R34 drinking water negatively affect groundwater dependent ecosystems (Grootjans et al. 1988, R35 Wassen et al. 1990). In addition, river regulation for energy supply and flood protection has led to loss or degradation of wetlands (Kareiva et al. 2007). Also, pollution and acidification R36 due to human activity leads to habitat degradation (Rockström et al. 2009). R37 Furthermore, land use change typically results in a fragmented distribution of the remaining R38 habitat for many plant and animal species (Stockwell et al., 2003) (Figure 1.1). Habitat R39 10 | Chapter 1

R1 fragmentation is defined as a process during which “a large expanse of habitat is transformed R2 into a number of smaller patches of smaller area, isolated from each other by a matrix of R3 habitats unlike the original” (Wilcove et al., 1986). In practice, habitat loss always accompanies R4 habitat fragmentation, because the ‘removal of habitat’ causes a fragmented distribution of R5 the remaining habitat patches situated around the converted area (see: Figure 1.1). R6 R7 R8 R9 R10 R11 R12 R13 R14 R15 R16 R17 R18 R19 R20 R21 R22 R23 R24 R25 R26 R27 R28 R29 R30 R31 R32 Figure 1.1. Simplified visualisation of the process of land use change (A) and the resulting habitat R33 fragmentation (B). Land use change might lead to habitat loss, habitat degradation and habitat R34 fragmentation. Habitat fragmentation is reflected by decreased total habitat area, decreased average patch size, increased habitat edge/area ratio and increased patch isolation. R35

R36 In the ecological literature, habitat fragmentation is sometimes referred to as the change in R37 landscape configuration (increase in number of patches, decrease in patch sizes, and increase R38 in isolation of patches) regardless of habitat loss. In this thesis, we will refer to this concept R39 Introduction | 11

as ‘habitat fragmentation per se’ (Fahrig, 2003). It is valuable to investigate the additional R1 effects of habitat fragmentation per se on biodiversity or species viability. In the next section, R2 the different processes and factors associated with habitat fragmentation, and its relation R3 with seed dispersal are explained. R4 R5 R6 1.3 HabItat FragmentatIon R7

R8 1.3.1 Population size R9 Habitat fragmentation leads to a decreased total habitat area and decreased average habitat patch size (Figure 1.1). A decreasing patch size is very likely to lead to a decrease in population R10 size of the plant species occupying the patch. With decreasing population size, random effects R11 become increasingly important for local population dynamics (Hanski and Gaggiotti 2004). R12 Population dynamics of plant populations are influenced by demographic and environmental R13 stochasticity. Demographic stochasticity implies the random variation in ‘birth’ and death of R14 individuals. In very small populations, demographic stochasticity leads to higher extinction R15 risks than in larger populations (Hanski and Gaggiotti 2004). R16 Furthermore, every sub-population is subject to environmental stochasticity, meaning the R17 yearly random variation in the environment due to, for instance, variation in temperature R18 and precipitation. Environmental stochasticity is a major cause of local extinction and the local R19 extinction risk due to environmental stochasticity increases with decreasing population size R20 (Hanski and Gaggiotti 2004). R21 Additionally, diminished population size caused by habitat fragmentation results in reduced R22 genetic variation at (sub-)population level, because the remaining individuals represent only a fraction of the original gene pool (Young et al., 1996). Subsequently, in the longer term, R23 reduced population size and isolation can lead to genetic drift (i.e. a change in allele frequency R24 from one generation to the next, caused by random sampling of alleles) and increased R25 inbreeding, which again cause a reduced genetic variation within sub-populations or within R26 individuals (i.e. reduced heterozygosity) respectively (Saunders et al. 1991, Ellstrand and Elam R27 1993, Matthies et al. 2004, Leimu et al., 2010). Reduced heterozygosity within individuals of R28 a sub-population can reduce individual fitness (e.g. Oostermeijer et al. 1995, Gaggiotti and R29 Hanski 2004) by an increased expression of deleterious recessive alleles, and thereby reduce R30 sub-population viability. In the long term, reduced genetic variation within sub-populations R31 may limit species’ ability to respond to changing abiotic conditions. Therefore, reduced R32 genetic variation (both within sub-populations and within individuals) can lead to increased R33 probability of local extinction. Increased local extinction, again, leads to a reduced genetic R34 variation within the population as a whole (Young et al., 1996). All these processes may R35 increase the extinction risk of small sub-populations of plant species (see: Joshi et al. 2006, Collins et al. 2009). R36 R37 R38 R39 12 | Chapter 1

R1 1.3.2 Edge effects R2 The breaking apart of a large population into several smaller ones not only leads to smaller R3 habitat patches but also to a larger habitat edge/-area ratio, compared to that in the original R4 population. Therefore, plants occurring in fragmented habitats are often more subjected to R5 the so called ‘edge effect’. Moreover, linear habitat remnants have proportionally more edge R6 than round ones (Saunders et al., 1991). Edges of habitat patches are often ecologically and physically different from the patch interior. R7 In addition to a difference in soil moisture and nutrient availability due to differing abiotic R8 conditions in the surrounding ‘matrix’ (i.e. unsuitable area for the focal species, for instance R9 agricultural fields or urban area), edges can also differ from the patch interior regarding, for R10 instance, wind speed, temperature and radiation. This can have positive, negative or neutral R11 effects on patch biodiversity (Ries et al., 2004). An increased amount of edge can aggravate R12 the negative effect of detrimental abiotic conditions on species inside the patch (Debinski and R13 Holt, 2000), leading to a decreasing biodiversity closer to the edge. On the other hand, when R14 the habitat patch borders a non-habitat area with resources complementary to those in the R15 habitat patch, a higher abundance of certain species depending on resources in both areas R16 can be expected near edges (Ries et al., 2004). R17 Plant biodiversity in herbaceous vegetation tends to be highest under intermediate productivity R18 and usually decreases above a certain level of nutrient availability and productivity (Grime R19 1979, Moore et al. 1989, Wassen et al. 2005). When natural herbaceous ecosystems become surrounded by agricultural fields, fertilization of these fields will cause increased nutrient R20 availability in (at least) the edges of the nature areas. Consequently, fast growing, tall species R21 will gain competitive advantage and may become dominant, thus decreasing biodiversity R22 (Grime 1979). Therefore, it can be expected that for ecosystems located in or near intensive R23 agricultural areas, the edge-effect is generally rather negative than positive. R24 R25 1.3.3 Isolation R26 Another consequence of habitat fragmentation is that remnant habitat patches that remain R27 in the fragmented landscape become more isolated from each other by distance or by inter- R28 patch barriers (Saunders et al. 1991). Increased spatial isolation between sub-populations R29 reduces gene flow (i.e. the transfer of alleles from one population to another) between sub- R30 populations (Young et al., 1996), due to diminished seed and exchange, leading to R31 erosion of genetic variance. The negative effects of reduced genetic variance on population R32 viability were already discussed in section 1.2.1. Additionally, several empirical studies report reduced individual seed production in smaller populations (e.g. Morgan 1999, Kéry et al. 2000, R33 Collins et al. 2009). If this is the case, habitat remnants become even more isolated after R34 fragmentation than they would be due to the distance effect only. R35 A group of several spatially separated populations in a region might act as a metapopulation. R36 Metapopulations often occur in fragmented landscapes (Opdam 1991, Hanski and Ovaskainen R37 2003). Hanksi and Gilpin (1991) define metapopulations as ensembles of interacting R38 populations with a finite lifetime. Thus, in a metapopulation, there is an ongoing process of R39 Introduction | 13

populations becoming extinct and patches becoming colonised by seed dispersal followed by R1 and establishment. Metapopulations can persist because of this balance between R2 extinction and colonisation. However, if patches become increasingly isolated and colonisation R3 rates decrease severely, regional survival of the metapopulation might be at risk (e.g. Hanski R4 1999). Although metapopulation research mainly focuses on animals, metapopulation R5 dynamics have been reported for plant species (for a review, see: Eriksson 1996). Source-sink R6 populations can be seen as a special case of metapopulations, and have also been found R7 for plant species (Eriksson 1996). Sink populations (i.e. populations in low-quality habitat R8 dependent on propagule supply from a source population) (Eriksson 1996), which depend R9 on nearby source populations for their persistence, might go extinct if isolation increases. Thus, effective seed dispersal is of vital importance for the persistence of fragmented plant R10 populations. Several studies reported dispersal limitation as a mechanism restricting plant R11 distribution (e.g. Primack and Miao 1992, McKenna and Houle 2000, Leng et al. 2009). In the R12 11-year seed sowing experiment of Ehrlén et al. (2006), suitable but empty patches, at which R13 populations could be sustained if seed was provided artificially, existed. These results suggest R14 that for forest herbs, seed dispersal was limiting species distribution. R15 The distance between patches, the degree of hostility or roughness of the matrix and the R16 dispersal capacity of the considered species determine the degree of isolation of local R17 populations. Knowledge on dispersal capacity of plant species in a range of fragmented R18 habitats differing in size and spatial arrangement of patches and quality of the matrix is R19 thus important for understanding metapopulation dynamics and for spatial optimisation of R20 restoration measures. Only few studies have addressed plant species dispersal in relation to R21 landscape patterns (e.g. Pearson and Dawson 2005, Lehouck et al. 2009, Hu et al. 2010). R22 R23 1.4 seed dIsPersal R24 R25 As explained above, gene-flow and colonization are of vital importance for metapopulation R26 survival. R27 Gene-flow can be realised by two mechanisms: pollen dispersal and propagule dispersal. R28 Propagules are defined as any part of the plant, for instance a seed, that is used for plant R29 propagation (i.e generative or vegetative reproduction of a plant). In a population-genetic R30 study, Bacles et al. (2006) showed that for the tree species Fraxinus excelsior, seed dispersal R31 instead of pollen dispersal was the main vector of gene flow among fragmented habitat R32 remnants. Other studies found that seed dispersal was not the main cause of gene flow but R33 still substantially contributed to it, and that pollen flow/seed flow ratios differed per species R34 (Ennos 1994, Oddou-Muratorio and Klein 2008, Kamm et al. 2009). R35 Colonisation, in contrast to gene-flow, can only be realised by propagule dispersal, not by pollen dispersal. For colonisation events to be effective, germination and establishment R36 after dispersal is necessary. Nevertheless, propagule dispersal is a prerequisite for successful R37 colonisation. From here onwards, generative dispersal will be referred to as seed dispersal. R38 R39 14 | Chapter 1

R1 Thus, to understand metapopulation dynamics and spread and persistence of species in R2 fragmented landscapes, it is necessary to understand seed dispersal mechanisms in these R3 landscapes. R4 Dispersal ability differs among different species, dispersal vectors and landscapes, but skewed R5 leptokurtic dispersal kernels, with a majority of the seeds being dispersed over short distances R6 only, are often reported (Fleming and Heithaus 1981, Alvarez-Buylla and Martinez-Ramos 1990, Carey and Watkinson 1993, Laman 1996, Nathan et al. 2001). R7 Plant seeds can be dispersed by different vectors: wind (anemochory), surface water R8 (hydrochory), animals (zoochory) or humans (anthropochory), or by the plant itself (autochory). R9 Some plant seeds are specifically adapted to one type of dispersal, whereas others may be R10 dispersed by several vectors. R11 R12 Species typically adapted to wind dispersal have seeds with low terminal velocity (i.e. falling R13 speed of seed in still air), caused by low seed weight or morphological adaptations such as R14 plumes or wing-like structures that decrease fall speed (Bouman et al. 2000). Wind dispersal R15 has been studied intensively over the last decades, both via experimental (e.g. Hensen and R16 Müller 1997, Stewart et al. 1998, Skarpaas et al. 2004, Jongejans et al. 2007) and modelling R17 approaches (e.g. Andersen 1991, Nathan et al. 2002, Tackenberg 2003, Soons et al. 2004, Katul R18 et al. 2005a). These studies for instance revealed that seed uplifting by upward vertical wind R19 velocity under unstable atmospheric conditions (turbulence) enables long distance dispersal (Soons et al. 2004, Kuparinen 2006, Nathan et al. 2011). Furthermore, horizontal wind speed, R20 seed release height, vegetation height and seed terminal velocity are recognised as important R21 determinants of seed dispersal distance (Soons et al. 2004). Bohrer et al. (2005) investigated R22 wind dispersal in the context of metapopulations in fragmented habitats, and conclude R23 that long-distance dispersal events increase metapopulation survival for populations with R24 intermediate local-extinction probabilities. R25 R26 Many wetland plants have floating seeds, and thus have the ability for long-distance dispersal R27 by surface water (van den Broek et al. 2005). Hydrochory is reported to be an important dispersal R28 mechanism for wetland plants growing along rivers (e.g. Boedeltje et al. 2003, Boedeltje et al. R29 2004, Vogt et al. 2004, Merritt et al. 2010). Although experiments investigating the effect of R30 hydrochorous dispersal via rivers on the riparian vegetation have been commonly performed R31 over the last decade (e.g. Andersson et al. 2000b, Nilsson et al. 2002, Goodson et al. 2003, R32 Vogt et al. 2004, Jansson et al. 2005) experiments investigating dispersal distances by water are rare, especially for slow flowing to stagnant water bodies, and modelling studies on this R33 topic are even rarer; to my knowledge, no process based spatially explicit hydrochory models R34 are developed until now. Such models could provide insight into consequences of hydrochory R35 for metapopulation dynamics in riparian species in different fragmented landscapes. R36 R37 Zoochory can be subdivided into endozoochory (dispersal of seeds via ingestion by animals) R38 and epizoochory (seed transport on the outside of animals). are the most frequently R39 Introduction | 15

reported endozoochorous dispersal vector (e.g. Clausen et al. 2002, Soons et al. 2008, Brochet R1 et al. 2010). For some plant species, gut passage of seeds even enhances germination (Soons R2 et al. 2008). Morphological seed adaptations known to promote endozoochory are fleshy R3 fruits or elaiosomes (i.e. fleshy structures attached to the seed) (Willson 1993). Epizoochory R4 mostly takes place by seeds with, for instance, hook-, spine- of barb structures sticking to the R5 fur of such as large (e.g. Constible et al. 2005, Couvreur et al. 2008). R6 Additionally, humans are reported to disperse seeds (unintentionally), for instance when R7 moving through the landscape (e.g. with shoes (Wichmann et al. 2009) or motor vehicles R8 (Hodkinson and Thompson 1997) or with mowing machinery (Strykstra et al. 1997). Because R9 animal movement behaviour is difficult to predict, zoochorous dispersal events are highly stochastic and therefore the effects of these dispersal mechanisms on metapopulation R10 dynamics are variable and difficult to predict. R11 R12 Some plants are able to shoot their own seeds over a certain distance (=autochory) (Benvenuti R13 2007). An example of such species are Geranium species, which fruit capsules are joined to R14 beak-like structures that, when the fruit is ripe, act as a catapult shooting away the seeds. R15 Dispersal distances reached by autochory will typically not exceed several meters (Bouman et R16 al. 2000). R17 R18 When comparing the amount of published literature addressing the various dispersal vectors R19 it becomes evident that much attention has been paid to anemochory leading to an extensive R20 body of knowledge already available on anemochory. Furthermore, zoochory is a largely R21 stochastic process, making it extremely difficult to quantify, and autochory operates on such R22 a small scale that this process probably is irrelevant for metapopulation dynamics. Together with the apparent lack of knowledge considering hydrochory, this made me decide that it R23 is meaningful to study hydrochory in the context of habitat fragmentation. Although many R24 studies have shown the importance of hydrochory for plant population and community R25 patterns (for an overview, see the review of Nilsson et al. 2010), and thus for biodiversity, the R26 precise mechanisms determining hydrochorous dispersal capacity of plant seeds in different R27 landscapes are underexposed in the literature. R28 Therefore, in this thesis, I focus mainly on hydrochorous dispersal of (semi-)terrestrial species R29 in different types of fragmented freshwater wetlands. Hereafter, I will explain hydrochorous R30 dispersal in more detail, and indicate the knowledge gaps in literature. R31 R32 R33 1.5 HydroCHory R34 R35 Hydrochorous dispersal can both take place by vegetative and generative propagules. Dispersal by vegetative units can take place by, for instance, shoot fragments or rhizomes R36 (i.e. horizontal part of the stem that is below ground, often sending out roots and R37 shoots from its nodes) that detached from the mother plant and are transported by R38 R39 16 | Chapter 1

R1 the surface water. Generative hydrochourous dispersal takes place by seeds or . R2 Boedeltje et al. (2003) trapped diaspores in a lowland stream in the Netherlands, and R3 found that 97.1% of the trapped diaspores of riparian species were of generative origin. R4 This is an important reason why I focus in this thesis primarily on generative dispersal. R5 R6 Many seeds of riparian species can float at least for some time, but some seeds have special adaptations to hydrochory (Nilsson et al. 2010). Such adaptations can be: cork-like, air-filled R7 tissue, resulting in a low relative density or a hydrophobic seed coat (ref: zie vorige versie). R8 Species with air-filled tissue are for instanceCalla palustris, Menyanthes trifoliata, Iris R9 pseudacorus, Rumex hydrolapathum, Mentha aquatica and Sparganium spp (Barrat-Segretain R10 1996, Bouman et al. 2000, Thorsen et al. 2009). Nymphoides peltata (Barrat-Segretain 1996) R11 and orchid species (Weston et al. 2005) are examples of species with a hydrophobic seed coat. R12 R13 For riparian plant species, the hydrochorous dispersal path is as follows (see Figure 1.2): First, R14 seeds are released from the plant and may end up in the water. If plants grow at the river/ R15 ditch- bank and hang over to the water (f.i. like Carex pseudocyperus or Iris pseudacorus) or R16 if they grow in shallow water along the bank, seeds that fall straight down from the plant R17 might immediately end up in the water by gravity. If plants grow in wetlands somewhat R18 further from the water, seeds may primarily be dispersed by gravity, wind or animals and R19 either be deposited on land or end up in water. If seeds have entered the water, they are transported on the waters’ surface, assuming that they float. During the transport process, R20 they may be deposited in the riparian zone (Moggridge and Gurnell 2010), for instance in R21 curves, or in aquatic vegetation or other obstacles in which transporting seeds can strand (e.g. R22 Schneider and Sharitz 1988, Johansson and Nilsson 1993). They may either be deposited there R23 permanently and potentially germinate at that location, or they may be further dispersed R24 by the surface water later (for instance when the water level rises) and potentially become R25 permanently deposited at another location (Figure 1.2). R26 R27 In rivers, the transport process is driven by water currents. In slow flowing or stagnant surface R28 water like ditches or lakes, hydrochorous seed transport might be partially or completely R29 driven by wind-shear stress (i.e. the shear stress exerted by wind on the water’s surface). R30 However, clear evidence for the mechanisms playing a role in transport in the latter ecosystems R31 is lacking until now (but see: Sarneel 2010). Because river currents can be fast, and seed R32 floating times of several months are no exception for wetland plants (van den Broek et al. 2005), one might expect distances covered by hydrochorous seeds via rivers to exceed several R33 kilometres. However, the hydrochorous dispersal kernel (i.e. the function that describes the R34 probability of dispersal to different distances) is little studied in rivers (Nilsson et al. 2010), R35 and even less or not at all in lentic water bodies. Many studies investigating the importance of R36 hydrochory for riparian vegetation placed nets in the river water column or Astroturf mats on R37 the banks to catch propagules present in the water (e.g. Boedeltje et al. 2003, Boedeltje et al. R38 2004, Merritt and Wohl 2006, Gurnell et al. 2008). These studies provide valuable information R39 Introduction | 17

on the number and type of species and the abundance of seeds for which surface water is R1 an important transport vector, but they do not give information on the potential dispersal R2 distances and thus the metapopulation structure and colonisation potential of riparian R3 species along rivers. Some studies released seeds or seed mimics in rivers or tidal wetlands R4 and revealed that indeed distances of more than a kilometre can be covered by hydrochorous R5 seeds in these ecosystems. Johansson et al. (1993), for instance, reported that most released R6 Ranunculus lingua seeds in a Swedish river were deposited within 1500 m. Griffith and Forseth R7 (2002) found distances up to 2600 m for hydrochorously transported Aeschynomene virginica R8 seeds in a tidal wetland. Helianthus annuus seeds released in a Swedish river by Andersson et R9 al. (2000b) dispersed up to 23 km. To my knowledge, only one study attempted to investigate dispersal distances in slow flowing ditches (Beltman et al. 2005). They found a maximum of R10 500 m for hydrochorously dispersing Carex elata seeds. R11 Despite the long dispersal distances found in free-flowing river reaches, many studies revealed R12 that river fragmentation by dams or sluices, which is a widespread phenomenon nowadays, R13 hampers hydrochorous dispersal, and that these effects are reflected in the riparian vegetation R14 composition (e.g. Andersson et al. 2000a, Jansson et al. 2000, Merritt and Wohl 2006, Merritt R15 et al. 2010). To my knowledge, the effect of obstructions in the water on hydrochory in slow R16 flowing or stagnant waters has not been investigated previously. R17 R18 Surface water is not the only dispersal vector in freshwater wetlands. As explained above, R19 primary dispersal of seeds ending up in a waterbody might take place by, for instance, wind. R20 Some seeds in wetlands will never enter the waterbody and wind or animals will remain R21 the only dispersal vector (see: Soons 2006). Anemochorous and zoochorous dispersal have an R22 additional value to riverine plant seeds, besides hydrochory: whereas hydrochorous dispersal in rivers is unidirectional and takes place in downstream direction only, anemochory or zoochory R23 may enable inter-catchment dispersal and upstream dispersal (see: Pollux et al. 2009). R24 Some experimental studies have investigated the importance of hydrochory versus the R25 importance of anemochory for the dispersal of wetland seeds in river ecosystems or tidal R26 marshes by placing traps that could only catch water dispersed seeds and traps that could only R27 catch wind dispersed seeds (Schneider and Sharitz 1988, Neff and Baldwin 2005, Moggridge R28 and Gurnell 2010). However, from such studies, no conclusions can be drawn on the dispersal R29 distances covered via both dispersal modes and thus on the effect of both dispersal modes R30 on metapopulation dynamics. Population genetic analyses can provide indirect evidence for R31 the importance of unidirectional hydrochorous dispersal versus other means of dispersal in R32 river ecosystems, but application of these techniques cannot provide answers to questions R33 regarding the importance of different dispersal strategies in systems in which the water R34 surface can flow in multiple directions, such as in ditches or lakes. Spatially explicit models can R35 provide answers to such questions. Several models simulating seed dispersal by wind have been developed (Soons 2003, Soons et al. 2004, Katul et al. 2005b, Nathan et al. 2011). However, to R36 my knowledge, no spatially explicit process based models simulating hydrochorous dispersal R37 have been developed until now. R38 R39 18 | Chapter 1

R1 Naturally, a prerequisite for hydrochorous dispersal is the presence of surface water. Wetlands R2 are ecosystems in which surface water is abundant. Because I focus on hydrochorous dispersal R3 in this thesis, the study systems considered are wetland ecosystems. I restrict myself to R4 freshwater wetlands, which excludes oceans and (brackish) tidal marshes. R5 R6 R7 R8 R9 R10 R11 R12 R13 R14 R15 R16 R17 Figure 1.2. Schematisation of dispersal processes in uplands and in and along riparian zones. Seeds that fall from the plant might be dispersed over short (1a & 1c) or longer (1b) distance by wind. If R18 seeds of riparian species fall into the water (1a) they are transported through the water (2), might R19 be captured in the riparian zone temporarily (3) and be transported further later (5), or might become permanently captured in the riparian zone (4), and potentially germinate and establish. R20 Figure adapted after Figure 2 in Nilsson et al. 2010. R21 R22 R23 1.6 HabItat FragmentatIon In FresHwater wetlands R24 R25 Freshwater wetlands are areas fed by freshwater where the water table is “at, near or above R26 the lands surface long enough to promote hydric soils, hydrophytic vegetation and biological R27 activities adapted to wet environments” (cf. (Price and Waddington 2000). R28 Freshwater wetlands provide important ecosystem services, such as water filtration, carbon R29 storage (Price and Waddington 2000, Kayranli et al. 2010) and flood protection (Houlahan et R30 al. 2006, Keddy et al. 2009). Wetlands are rich in biodiversity, with many species confined to R31 wetlands only (Wheeler 1980a, b, c, Hails 1996). However, over 50% of wetland ecosystems R32 have been lost nowadays in Europe (Joosten and Clarke 2002, Keddy et al. 2009). Conversion of freshwater wetlands to agriculture is a major cause of wetland loss (Brinson and Malvarez R33 2002, Verhoeven et al. 2008). Additionally, drainage and subsequent lowering of the R34 groundwater table for the benefit of agriculture has eliminated large areas of wetlands in R35 the USA and Europe while eutrophication caused by agricultural activities reduces biodiversity R36 in remaining wetlands (Brinson and Malvarez 2002, Houlahan et al. 2006). Furthermore, R37 canalisation of rivers for flood control resulted in loss of floodplain habitat (Verhoeven et al. R38 2008). R39 Introduction | 19

Remaining wetland patches show a spatially fragmented distribution (Verhoeven et al. 2008). R1 Around and in between agricultural fields, patches of wetland nature reserves remained, R2 surrounded by the matrix of agricultural fields, and often dissected or surrounded by drainage R3 ditches. Examples of these kinds of areas are the Mississippi River Basin in the USA (Moore R4 and Kröger 2011), the Norfolk Broads in England (George, 1992), the Camargue (France) R5 (Chauvelon, 1998), the Northern upper Rhine floodplain (Germany) (Hölzel and Otte, 2003), R6 the Danube Delta (Romania) (Pringle et al., 1993), drained wetlands in Denmark (Hoffmann R7 and Baattrup-Pedersen, 2007) and Dutch fen meadows (Lamers et al., 2002). Reclamation R8 of wetlands for agriculture created a different type of surface water system in comparison R9 to natural wetlands (Mitsch and Gosselink 2007). Surface water is present in these human- dominated areas in the form of an extensive network of ditches and canals that are used to R10 control the ground water level and in which water flow is slow to stagnant. At places were R11 eutrophication is limited, for instance along extensively managed agricultural fields, banks of R12 these drainage ditches may still serve as refuges for characteristic wetland species (Blomqvist R13 et al. 2006). Along rivers, small, isolated floodplains in between the canalised parts of the river R14 have remained at some places (Verhoeven et al. 2008, Krause et al. 2011). R15 R16 Many studies investigated the effects of different measures that aim to restore abiotic R17 conditions in degraded wetlands. These measures for instance include rewetting (e.g. R18 Pfadenhauer and Grootjans 1999, Van Duren 2000, Mälson et al. 2008), top soil removal (e.g. R19 Grootjans et al. 2002, Hausman et al. 2007, Klimkowska et al. 2007), and liming (e.g. Van Duren R20 et al. 1998, Beltman et al. 2001). Such studies on wetland restoration projects often show that R21 biodiversity is not restored, in spite of the restoration of abiotic conditions. Seed dispersal R22 is thought to be limiting the success of wetland restoration projects in these fragmented ecosystems (Middleton 1999, Pfadenhauer and Grootjans 1999, Bischoff 2002, Boedeltje et al. R23 2004, Verhoeven et al. 2008). R24 R25 R26 1.7 Problem deFInItIon, aIms and outlIne oF tHe tHesIs R27 R28 Habitat fragmentation and habitat quality are factors well studied, but mostly not in R29 combination. From the above, it became clear that habitat fragmentation and thus also R30 dispersal limitation may threaten biodiversity in freshwater wetlands. However, the relative R31 effect of habitat fragmentation compared to habitat degradation on freshwater wetland R32 biodiversity remains unclear until now. Although many studies recognise habitat fragmentation R33 as a serious threat to biodiversity (e.g. Hanski 2005, Piessens et al. 2005, Helm, Hanski et al. R34 2006), other studies did not find significant effects of connectivity on the occurence of species R35 in fragmented landscapes (see review of Fahrig 2003). In the context of research on habitat fragmentation and metapopulation ecology, it is R36 important to gain insight in dispersal capacity of plant species in different landscapes. R37 Although dispersal via surface water is expected to be important for vegetation structure R38 R39 20 | Chapter 1

R1 and biodiversity in freshwater wetlands, hitherto, the majority of research on seed dispersal, R2 especially modelling research, has focused on dispersal by wind. The relative importance of R3 water dispersal versus wind dispersal for metapopulation dynamics in freshwater wetlands R4 is unknown until now. Moreover, hydrochorous dispersal kernels have not been well studied R5 at all. Furthermore, the existing field- or experimental studies on hydrochory focus almost R6 exclusively on river ecosystems, in which water flow is relatively fast. As a result, there is a lack of knowledge on seed dispersal in landscapes dominated by agriculture with slow flowing R7 ditches, which is nowadays a widespread ecosystem that can provide refuges for characteristic R8 wetland species. The ditches in these ecosystems might act as efficient hydrochorous dispersal R9 corridors for wetland species. To gain more knowledge on metapopulation dynamics and the R10 restoration prospects of fragmented freshwater wetlands, it is of great importance that the R11 mechanisms of seed dispersal by water and wind in such human made landscapes for different R12 landscape configurations are investigated. R13 R14 the aim of this thesis is to shed light on the contribution of hydrochory to seed dispersal in R15 freshwater wetlands impacted by human activity and on the effect of landscape configuration R16 and -characteristics of the water body on realised dispersal distances. The wetland types R17 considered herein are rich fens, floodplains and ditch bank vegetation. R18 For this purpose I use an innovative combination of different research approaches including R19 fieldwork, analyses of large datasets, experimental research and model development and simulations. The outline of the thesis and the approach used in each chapter is summarised R20 below. R21 R22 Chapter 2: R23 To check whether habitat fragmentation may indeed negatively influences species R24 distribution in freshwater wetlands, the relative effect of isolation, habitat size and R25 habitat edge compared to the effect of habitat quality on plant occurrence was R26 investigated for 6 characteristic fen plant species. For this purpose, large datasets on R27 habitat quality and species distribution in a Dutch semi-natural fen area were used. Best R28 subsets logistic regression analyses were used to assess the importance of each factor. R29 main research question: R30 1. what is the relative effect of habitat fragmentation, compared to abiotic R31 factors, for the occurrence of six characteristic fen plant species in a dutch R32 semi-natural fen area?

R33 Chapter 3: R34 The aim of this chapter was to determine how species traits and abiotic factors influence R35 the extent of hydrochorous dispersal into and out of floodplains. Hydrochorous seeds were R36 captured before and after a small and isolated floodplain of a Dutch river throughout a year. R37 Composition and numbers of trapped seeds before and after the floodplain were related to R38 species traits, species growing location and river water level. R39 Introduction | 21

main research questions: R1 2. does the considered floodplain function as a net seed source or sink? R2 3. which plant traits and growing location characteristics are related to seed R3 inflow into the floodplain and to seed outflow? R4 Chapter 4: R5 To gain insight into the mechanisms by which hydrochorous seeds are transported in drainage R6 ditches in landscapes dominated by agriculture, the effects of the velocity of wind and water R7 on the rate of transport of three wetland species (Carex pseudocyperus, Iris pseudacorus and R8 Sparganium erectum) were investigated using an experimental set-up. Furthermore, in release R9 and retrace experiments with painted C. pseudocyperus seeds, a number of factors potentially determining the probability of seed deposition were investigated. R10 main research questions: R11 4. what are the relative contributions of wind and water flow to the rate of R12 transport of floating seeds in drainage ditches? R13 5. what characteristics of ditches and ditch banks determine seed deposition? R14 R15 Chapter 5: R16 A simple linked habitat suitability-seed dispersal model was developed to investigate the R17 importance of adding seed dispersal models to habitat suitability models for optimising spatial R18 planning of restoration strategies. This model predicts potential species distribution as a R19 function of current species distribution, species-specific dispersal traits, dispersal infrastructure R20 and habitat configuration. When dispersal modules of such models would be improved, such R21 models may help to investigate the effect of different hydrological measures on characteristic R22 fen species viability. research question: R23 6. are species distribution models considerably improved by the explicit R24 modelling of dispersal? R25 R26 Chapter 6: R27 The aim of this chapter was to investigate 1) the relative contribution of wind dispersal R28 and dispersal via surface water to total seed dispersal distances in agricultural landscapes R29 with slow flowing ditches and, 2) to what extent drainage ditches are effective corridors for R30 dispersal of riparian species and how the dispersal success is related to properties of landscape R31 and the ditches and to species traits. For these purposes, an innovative, coupled anemochory- R32 hydrochory model was developed and validated, and a sensitivity analyses was performed. R33 Different model scenarios, representing different landscape configurations and ditch states R34 were simulated for a riparian species typically adapted to wind dispersal (Phragmithes R35 australis) and a riparian species adapted to hydrochorous dispersal (Carex pseudocyperus). main research questions: R36 7. what is the relative contribution of wind- and water dispersal to seed dispersal R37 in landscapes with different configurations of drainage ditches? R38 R39 22 | Chapter 1

R1 8. to what extent do system characteristics of the landscape (f.i. ditch direction, R2 -density, -roughness, obstructions) determine the dispersal distances of seeds? R3 R4 Chapter 7: R5 In the last chapter, results of the previous chapters are synthesised and discussed. Furthermore, R6 directions for future research and implications of the results for wetland management are outlined. R7

R8 R9 R10 R11 R12 R13 R14 R15 R16 R17 R18 R19 R20 R21 R22 R23 R24 R25 R26 R27 R28 R29 R30 R31 R32 R33 R34 R35 R36 R37 R38 R39 Introduction | 23

reFerenCes R1 R2 Alvarez-Buylla, E. R., and M. Martinez-Ramos. 1990. Seed bank versus seed rain in the regeneration R3 of a tropical pioneer tree. Oecologia 84:314-325. Andersen, M. 1991. Mechanistic models for the seed shadows of wind-dispersed plants. American R4 Naturalist 137:476-497. R5 Andersson, E., C. Nilsson, and M. E. Johansson. 2000a. Effects of river fragmentation on plant dispersal and riparian flora. River Research and Applications 16:83-89. R6 Andersson, E., C. Nilsson, and M. E. Johansson. 2000b. Plant dispersal in boreal rivers and its relation R7 to the diversity of riparian flora. Journal of Biogeography 27:1095-1106. R8 Bacles, C. F. E., A. J. Lowe, and R. A. Ennos. 2006. Effective seed dispersal across a fragmented landscape. Science 311:628. R9 Barrat-Segretain, M. H. 1996. Strategies of reproduction, dispersion, and competition in river plants: R10 A review. Vegetatio 123:13-37. Beltman, B., T. Van Den Broek, A. Barendregt, M. C. Bootsma, and A. P. Grootjans. 2001. Rehabilitation R11 of acidified and eutrophied fens in The Netherlands: Effects of hydrologic manipulation R12 and liming. Ecological Engineering 17:21-31. R13 Beltman, B., T. van den Broek, and P. Vergeer. 2005. Het beperkte succes van Laagveenrestauratie. Landschap 20:173-179. R14 Benvenuti, S. 2007. Weed seed movement and dispersal strategies in the agricultural environment. R15 Weed Biology and Management 7:141-157. Bischoff, A. 2002. Dispersal and establishment of floodplain grassland species as limiting factors in R16 restoration. Biological Conservation 104:25-33 R17 Blomqvist, M. M., W. L. M. Tamis, J. P. Bakker, and E. van der Meijden. 2006. Seed and (micro) R18 site limitation in ditch banks: Germination, establishment and survival under different management regimes. Journal for Nature Conservation 14:16-33. R19 Boedeltje, G., J. P. Bakker, R. M. Bekker, J. M. Van Groenendael, and M. Soesbergen. 2003. Plant R20 dispersal in a lowland stream in relation to occurrence and three specific life-history traits of the species in the species pool. Journal of Ecology 91:855-866. R21 Boedeltje, G., J. P. Bakker, A. Ten Brinke, J. M. Van Groenendael, and M. Soesbergen. 2004. Dispersal R22 phenology of hydrochorous plants in relation to discharge, seed release time and buoyancy R23 of seeds: the flood pulse concept supported. 92:786-796. Bohrer, G., R. Nathan, and S. Volis. 2005. Effects of long-distance dispersal for metapopulation R24 survival and genetic structure at ecological time and spatial scales. Journal of Ecology R25 93:1029-1040. Bouman, F., D. Boesewinkel, R. Bregman, N. Devente, and G. Oostermeijer. 2000. Verspreiding van R26 zaden. KNNV Uitgeverij, Utrecht. R27 Brinson, M. M., and A. I. Malvarez. 2002. Temperate freshwater wetlands: Types, status, and threats. R28 Environmental Conservation 29:115-133. Brochet, A. L., M. Guillemain, H. Fritz, M. Gauthier-Clerc, and A. J. Green. 2010. Plant dispersal R29 by teal (Anas crecca) in the Camargue: Duck guts are more important than their feet. R30 Freshwater Biology 55:1262-1273. Carey, P. D., and A. R. Watkinson. 1993. The dispersal and fates of seeds of the winter annual grass R31 Vulpia ciliata. Journal of Ecology 81:759-767. R32 Clausen, P., B. A. Nolet, A. D. Fox, and M. Klaassen. 2002. Long-distance endozoochorous dispersal R33 of submerged macrophyte seeds by migratory waterbirds in northern Europe - A critical review of possibilities and limitations. Acta Oecologica 23:191-203. R34 Collins, C. D., R. D. Holt, and B. L. Foster. 2009. Patch size effects on plant species decline in an R35 experimentally fragmented landscape. Ecology 90:2577-2588. Constible, J. M., R. A. Sweitzer, D. H. Van Vuren, P. T. Schuyler, and D. A. Knapp. 2005. Dispersal of R36 non-native plants by introduced bison in an island ecosystem. Biological Invasions 7:699- R37 709. R38 R39 24 | Chapter 1

Couvreur, M., K. Verheyen, M. Vellend, I. Lamoot, E. Cosyns, M. Hoffmann, and M. Hermy. 2008. R1 Epizoochory by large herbivores: merging data with models. Basic and Applied Ecology R2 9:204-212. R3 Ehrlen, J., Z. Munzbergova, M. Diekmann, and O. Eriksson. 2006. Long-term assessment of seed limitation in plants: results from an 11-year experiment. Journal of Ecology 94:1224-1232. R4 Ennos, R. A. 1994. Estimating the relative rates of pollen and seed migration among plant R5 populations. Heredity 72:250-259. Eriksson, O. 1996. Regional dynamics of plants: A review of evidence for remnant, source-sink and R6 metapopulations. Oikos 77:248-258. R7 Fleming, T. H., and E. R. Heithaus. 1981. Frugivorous , seed shadows, and the structure of tropical R8 forests. Biotropica 13. Foley, J. A., R. DeFries, G. P. Asner, C. Barford, G. Bonan, S. R. Carpenter, F. S. Chapin, M. T. Coe, G. C. R9 Daily, H. K. Gibbs, J. H. Helkowski, T. Holloway, E. A. Howard, C. J. Kucharik, C. Monfreda, R10 J. A. Patz, I. C. Prentice, N. Ramankutty, and P. K. Snyder. 2005. Global consequences of land use. Science 309:570-574. R11 Folke, C., S. Carpenter, B. Walker, M. Scheffer, T. Elmqvist, L. Gunderson, and C. S. Holling. 2004. R12 Regime shifts, resilience, and biodiversity in ecosystem management. Pages 557-581 R13 Annual Review of Ecology, Evolution, and Systematics. Goodson, J. M., A. M. Gurnell, P. G. Angold, and I. P. Morrissey. 2003. Evidence for hydrochory and R14 the deposition of viable seeds within winter flow-deposited sediments: The River Dove, R15 Derbyshire, UK. River Research and Applications 19:317-334. Griffith, A. B., and I. N. Forseth. 2002. Primary and secondary seed dispersal of a rare, tidal wetland R16 annual, Aeschynomene virginica. 22:696-704. R17 Grime, J. P. 1979. Plant strategies and vegetation processes. John Wiley and Sons, New York. R18 Grootjans, A. P., J. P. Bakker, A. J. M. Jansen, and R. H. Kemmers. 2002. Restoration of brook valley meadows in the Netherlands. Hydrobiologia 478:149-170. R19 Grootjans, A. P., R. van Diggelen, M. J. Wassen, and W. A. Wiersinga. 1988. The effects of drainage on R20 groundwater quality and plant species distribution in stream valley meadows. Vegetatio 75:37-48. R21 Gurnell, A., K. Thompson, J. Goodson, and H. Moggridge. 2008. Propagule deposition along river R22 margins: Linking hydrology and ecology. Journal of Ecology 96:553-565. R23 Hails, A. J., editor. 1996. Wetlands, Biodiversity and the Ramsar Convention: the role of the convention on wetlands in the conservation and wise use of biodiversity. Ramsar Convention Bureau, R24 Gland, Switzerland. R25 Hanski, I. 1999. Metapopulation Ecology. Oxford University Press, Oxford. Hanski, I., and O. E. Gaggiotti. 2004. Ecology, Genetics and Evolution of Metapopulations Elsevier R26 Academic Press, London, UK. R27 Hanski, I., and M. Gilpin. 1991. Metapopulation dynamics: brief history and conceptual domain. R28 Biological Journal of the Linnean Society 42:3-16. Hanski, I., and O. Ovaskainen. 2003. Metapopulation theory for fragmented landscapes. Theoretical R29 Population Biology 64:119-127. R30 Hausman, C. E., L. H. Fraser, M. W. Kershner, and F. A. De Szalay. 2007. Plant community establishment in a restored wetland: Effects of soil removal. Applied Vegetation Science 10:383-390. R31 Hensen, I., and C. Müller. 1997. Experimental and structural investigations of anemochorous R32 dispersal. Plant Ecology 133:169-180. R33 Hodkinson, D. J., and K. Thompson. 1997. Plant dispersal: The role of man. Journal of Applied Ecology 34:1484-1496. R34 Houlahan, J. E., P. A. Keddy, K. Makkay, and C. S. Findlay. 2006. The effects of adjacent land use on R35 wetland species richness and community composition. Wetlands 26:79-96. Hu, L. J., K. Uchiyama, H. L. Shen, and Y. Ide. 2010. Multiple-scaled spatial genetic structures of R36 Fraxinus mandshurica over a riparian-mountain landscape in Northeast China. Conservation R37 Genetics 11:77-87. R38 Isbell, F., V. Calcagno, A. Hector, J. Connolly, W. S. Harpole, P. B. Reich, M. Scherer-Lorenzen, B. Schmid, D. Tilman, J. van Ruijven, A. Weigelt, B. J. Wilsey, E. S. Zavaleta, and M. Loreau. R39 2011. High plant diversity is needed to maintain ecosystem services. Nature In press. Introduction | 25

Jansson, R., C. Nilsson, and B. Renofalt. 2000. Fragmentation of riparian floras in rivers with multiple dams. Ecology 81:899-903. R1 Jansson, R., U. Zinko, D. M. Merritt, and C. Nilsson. 2005. Hydrochory increases riparian plant species R2 richness: a comparison between a free-flowing and a regulated river. Journal of Ecology R3 93:1094-1103. Johansson, M. A., and C. Nilsson. 1993. Hydrochory, population dynamics and distribution of the R4 clonal Ranunculus lingua. Journal of Ecology 81:81-91. R5 Jongejans, E., N. M. Pedatella, K. Shea, O. Skarpaas, and R. Auhl. 2007. Seed release by invasive thistles: The impact of plant and environmental factors. Proceedings of the Royal Society R6 B: Biological Sciences 274:2457-2464. R7 Joosten, H., and D. Clarke. 2002. Wise use of mires and peatlands. Background and principles R8 including a framework for decision-making. International Mire Conservation Group and International Peat Society, Saarijärvi, Finland. R9 Joshi, J., P. Stoll, H. P. Rusterholz, B. Schmid, C. Dolt, and B. Baur. 2006. Small-scale experimental R10 habitat fragmentation reduces colonization rates in species-rich grasslands. Oecologia 148:144-152. R11 Kamm, U., P. Rotach, F. Gugerli, M. Siroky, P. Edwards, and R. Holderegger. 2009. Frequent long- R12 distance gene flow in a rare temperate forest tree (Sorbus domestica) at the landscape R13 scale. Heredity 103:476-482. Kareiva, P., S. Watts, R. McDonald, and T. Boucher. 2007. Domesticated nature: Shaping landscapes R14 and ecosystems for human welfare. Science 316:1866-1869. R15 Katul, G. G., A. Porporato, R. Nathan, M. Siqueira, M. B. Soons, D. Poggi, H. S. Horn, and S. A. Levin. 2005a. Mechanistic analytical models for long-distance seed dispersal by wind. The R16 American Naturalist 166:368-381. R17 Katul, G. G., A. Porporato, R. Nathan, M. Siqueira, M. B. Soons, D. Poggi, H. S. Horn, and S. A. Levin. R18 2005b. Mechanistic analytical models for long-distance seed dispersal by wind. American Naturalist 166:368-381. R19 Kayranli, B., M. Scholz, A. Mustafa, and Ã. Hedmark. 2010. Carbon storage and fluxes within R20 freshwater wetlands: A critical review. Wetlands 30:111-124. Keddy, P. A., L. H. Fraser, A. I. Solomeshch, W. J. Junk, D. R. Campbell, M. T. K. Arroyo, and C. J. R. R21 Alho. 2009. Wet and wonderful: the world’s largest wetlands are conservation priorities. R22 BioScience 59:39-51. R23 Kéry, M., D. Matthies, and H. H. Spillmann. 2000. Reduced fecundity and offspring performance in small populations of the declining grassland plants Primula veris and Gentiana lutea. R24 Journal of Ecology 88:17-30. R25 Klimkowska, A., R. Van Diggelen, J. P. Bakker, and A. P. Grootjans. 2007. Wet meadow restoration in Western Europe: A quantitative assessment of the effectiveness of several techniques. R26 Biological Conservation 140:318-328. R27 Krause, B., H. Culmsee, K. Wesche, E. Bergmeier, and C. Leuschner. 2011. Habitat loss of floodplain R28 meadows in north Germany since the 1950s. Biodiversity and Conservation 20:2347-2364. Kuparinen, A. 2006. Mechanistic models for wind dispersal. Trends in Plant Science 11:296-301. R29 Laman, T. G. 1996. Ficus seed shadows in a Bornean rain forest. Oecologia 107:347-355. R30 Lehouck, V., T. Spanhove, L. Colson, A. Adringa-Davis, N. J. Cordeiro, and L. Lens. 2009. Habitat disturbance reduces seed dispersal of a forest interior tree in a fragmented African cloud R31 forest. Oikos 118:1023-1034. R32 Leng, X., C. J. M. Musters, and G. R. de Snoo. 2009. Restoration of plant diversity on ditch banks: R33 Seed and site limitation in response to agri-environment schemes. Biological Conservation 142:1340-1349. R34 Mälson, K., I. Backéus, and H. Rydin. 2008. Long-term effects of drainage and initial effects of R35 hydrological restoration on rich fen vegetation. Applied Vegetation Science 11:99-106. May, R. M. 2011. Why should we be concerned about loss of biodiversity. Comptes Rendus - Biologies R36 334:346-350. R37 McKenna, M. F., and G. Houle. 2000. Under-saturated distribution of Floerkea proserpinacoides R38 Willd. (Limnanthaceae) at the northern limit of its distribution. Ecoscience 7:466-473. R39 26 | Chapter 1

Merritt, D. M., C. Nilsson, and R. Jansson. 2010. Consequences of propagule dispersal and river R1 fragmentation for riparian plant community diversity and turnover. Ecological Monographs R2 80:609-626. R3 Merritt, D. M., and E. E. Wohl. 2006. Plant dispersal along rivers fragmented by dams. River Research and Applications 22:1-26. R4 Middleton, B. 1999. Wetland restoration, flood pulsing, and disturbance dynamics. John Wiley and R5 Sons, New York, NY, US. Millennium Ecosystem Assessment. 2005. Ecosystems and human well-being. Biodiversity synthesis., R6 World Resources Institute, Washington. R7 Mitsch, W. J., and J. G. Gosselink. 2007. Wetlands. John Wiley and Sons, New York, NY, US. R8 Moggridge, H. L., and A. M. Gurnell. 2010. Hydrological controls on the transport and deposition of plant propagules within riparian zones. River Research and Applications 26:512-527. R9 Moore, D. R. J., and P. A. Keddy. 1988. The relationship between species richness and standing crop R10 in wetlands: the importance of scale. Vegetatio 79:99-106. Moore, D. R. J., P. A. Keddy, C. L. Gaudet, and I. C. Wisheu. 1989. Conservation of wetlands: Do R11 infertile wetlands deserve a higher priority? Biological Conservation 47:203-217. R12 Moore, M. T., and R. Kröger. 2011. Evaluating plant species-specific contributions to nutrient R13 mitigation in drainage ditch mesocosms. Water, Air, and Soil Pollution 217:445-454. Morgan, J. W. 1999. Effects of population size on seed production and germinability in an R14 endangered, fragmented grassland plant. Conservation Biology 13:266-273. R15 Naeem, S., L. J. Thompson, S. P. Lawler, J. H. Lawton, and R. M. Woodfin. 1994. Declining biodiversity can alter the performance of ecosystems. Nature 368:734-737. R16 Nathan, R., G. G. Katul, G. Bohrer, A. Kuparinen, M. B. Soons, S. E. Thompson, A. Trakhtenbrot, and R17 H. S. Horn. 2011. Mechanistic models of seed dispersal by wind. Theoretical Ecology In R18 press:1-20. Nathan, R., G. G. Katul, H. S. Horn, S. M. Thomas, R. Oren, R. Avissar, S. W. Pacala, and S. A. Levin. R19 2002. Mechanisms of long-distance dispersal of seeds by wind. Nature 418:409-413. R20 Nathan, R., U. N. Safriel, and I. Noy-Meir. 2001. Field validation and sensitivity analysis of a mechanistic model for tree seed dispersal by wind. 82:374-388. R21 Neff, K. P., and A. H. Baldwin. 2005. Seed dispersal into wetlands: Techniques and results for a R22 restored tidal freshwater marsh. Wetlands 25:392-404. R23 Nilsson, C., E. Andersson, D. M. Merritt, and M. E. Johansson. 2002. Differences in riparian flora between riverbanks and river lakeshores explained by dispersal traits. Ecology 83:2878- R24 2887. R25 Nilsson, C., R. L. Brown, R. Jansson, and D. M. Merritt. 2010. The role of hydrochory in structuring riparian and Wetland vegetation. Biological Reviews 85:837-858. R26 Oddou-Muratorio, S., and E. K. Klein. 2008. Comparing direct vs. indirect estimates of gene flow R27 within a population of a scattered tree species. Molecular Ecology 17:2743-2754. R28 Opdam, P. 1991. Metapopulation theory and habitat fragmentation: a review of holarctic breeding studies. Landscape Ecology 5:93-106. R29 Pearson, R. G., and T. P. Dawson. 2005. Long-distance plant dispersal and habitat fragmentation: R30 Identifying conservation targets for spatial landscape planning under . Biological Conservation 123:389-401. R31 Pfadenhauer, J., and A. Grootjans. 1999. Wetland restoration in Central Europe: Aims and methods. R32 Applied Vegetation Science 2:95-106. R33 Pollux, B. J. A., A. Luteijn, J. M. Van Groenendael, and N. J. Ouborg. 2009. Gene flow and genetic structure of the aquatic macrophyte Sparganium emersum in a linear unidirectional river. R34 Freshwater Biology 54:64-76. R35 Price, J. S., and J. M. Waddington. 2000. Advances in Canadian wetland hydrology and biogeochemistry. Hydrological Processes 14:1579-1589. R36 Primack, R. B., and S. L. Miao. 1992. Dispersal can limit local plant distribution. Conservation Biology R37 6:513-519. R38 Royal Botanic Gardens. 2010. Plants under pressure a global assessment. The first report of the IUCN Sampled Red List Index for Plants. Kew, UK. R39 Introduction | 27

Rockström, J., W. Steffen, K. Noone, Å. Persson, F. S. Chapin, E. F. Lambin, T. M. Lenton, M. Scheffer, C. Folke, H. J. Schellnhuber, B. Nykvist, C. A. De Wit, T. Hughes, S. Van Der Leeuw, H. Rodhe, R1 S. Sörlin, P. K. Snyder, R. Costanza, U. Svedin, M. Falkenmark, L. Karlberg, R. W. Corell, V. J. R2 Fabry, J. Hansen, B. Walker, D. Liverman, K. Richardson, P. Crutzen, and J. A. Foley. 2009. A R3 safe operating space for humanity. Nature 461:472-475. Sala, O. E., F. S. Chapin Iii, J. J. Armesto, E. Berlow, J. Bloomfield, R. Dirzo, E. Huber-Sanwald, L. F. R4 Huenneke, R. B. Jackson, A. Kinzig, R. Leemans, D. M. Lodge, H. A. Mooney, M. Oesterheld, R5 N. L. Poff, M. T. Sykes, B. H. Walker, M. Walker, and D. H. Wall. 2000. Global biodiversity scenarios for the year 2100. Science 287:1770-1774. R6 Sarneel, J. M. 2010. Colonisation processes in riparian fen vegetation. Dissertation. Utrecht R7 University, Utrecht. R8 Saunders, D. A., R. J. Hobbs, and C. R. Margules. 1991. Biological Consequences Of Ecosystem Fragmentation - A Review. Conservation Biology 5:18-32. R9 Schneider, R. L., and R. R. Sharitz. 1988. Hydrochory And Regeneration In A Bald Cypress Water R10 Tupelo Swamp Forest. Ecology 69:1055-1063. Skarpaas, O., O. E. Stabbetorp, I. Ronning, and T. O. Svennungsen. 2004. How far can a hawk’s beard R11 fly? Measuring and modelling the dispersal of Crepis praemorsa. 92:747-757. R12 Soons, M. B. 2003. Habitat fragmentation and connectivity. Spatial and temporal characteristics of R13 the colonization process in plants. Soons, M. B. 2006. Wind dispersal in freshwater wetlands: Knowledge for conservation and R14 restoration. Applied Vegetation Science 9:271-278. R15 Soons, M. B., G. W. Heil, R. Nathan, and G. G. Katul. 2004. Determinants of long-distance seed dispersal by wind in grasslands. Ecology 85:3056-3068. R16 Soons, M. B., C. Van Der Vlugt, B. Van Lith, G. W. Heil, and M. Klaassen. 2008. Small seed size R17 increases the potential for dispersal of wetland plants by ducks. Journal of Ecology 96:619- R18 627. Stewart, J. D., E. H. Hogg, P. A. Hurdle, K. J. Stadt, P. Tollestrup, and V. J. Lieffers. 1998. Dispersal of R19 white seed in mature aspen stands. Canadian Journal of 76:181-188. R20 Strykstra, R. J., G. L. Verweij, and J. P. Barker. 1997. Seed dispersal by mowing machinery in a Dutch brook valley system. Acta Botanica Neerlandica 46:387-401. R21 Tackenberg, O. 2003. Modeling long-distance dispersal of plant diaspores by wind. Ecological R22 Monographs 73:173-189. R23 Thorsen, M., K. J. M. Dickinson, and P. J. Seddon. 2009. Seed dispersal systems in the New Zealand flora. Perspectives in Plant Ecology, Evolution and Systemetics 11:285-309. R24 Tilman, D., P. B. Reich, and J. M. H. Knops. 2006. Biodiversity and ecosystem stability in a decade-long R25 grassland experiment. Nature 441:629-632. van den Broek, T., R. van Diggelen, and R. Bobbink. 2005. Variation in seed buoyancy of species R26 in wetland ecosystems with different flooding dynamics. Journal of Vegetation Science R27 16:579-586. R28 Van Duren, I. C. 2000. Nutrient limitation in drained and rewetted fen meadows. Rijksuniversiteit Groningen, Groningen. R29 Van Duren, I. C., R. J. Strykstra, A. P. Grootjans, G. N. J. Ter Heerdt, and D. M. Pegtel. 1998. A R30 multidisciplinary evaluation of restoration measures in a degraded Cirsio-Molinietum fen meadow. Applied Vegetation Science 1:115-130. R31 Verhoeven, J. T. A., M. B. Soons, R. Janssen, and N. Omtzigt. 2008. An Operational Landscape Unit R32 approach for identifying key landscape connections in wetland restoration. Journal of R33 Applied Ecology 45:1496-1503. Vitousek, P. M., H. A. Mooney, J. Lubchenco, and J. M. Melillo. 1997. Human domination of Earth’s R34 ecosystems. Science 277:494-499. R35 Vogt, K., L. Rasran, and K. Jensen. 2004. Water-borne seed transport and seed deposition during flooding in a small river-valley in Northern Germany. Flora : Morphologie, Geobotanik, R36 Oekophysiologie 199:377-388. R37 Wassen, M. J., A. Barendregt, P. P. Schot, and B. Beltman. 1990. Dependency of local mesotrophic R38 fens on a regional groundwater flow system in a poldered river plain in the Netherlands. Landscape Ecology 5:21-38. R39 28 | Chapter 1

Wassen, M. J., H. Olde Venterink, E. D. Lapshina, and F. Tanneberger. 2005. Endangered plants R1 persist under phosphorus limitation. Nature 437:547-550. R2 Weston, P. H., A. J. Perkins, and T. J. Entwisle. 2005. More than symbioses: orchid ecology, with R3 examples from the Sydney Region. Cunninghamia 9:1-15. Wheeler, B. D. 1980a. Plant communities of rich-fen systems in England and Wales. I. Introduction, R4 tall sedge and reed communities. Journal of Ecology 68:405-420. Chapter 2 R5 Wheeler, B. D. 1980b. Plant communities of rich-fen systems in England and Wales. II. Communities of calcareous mires. Journal of Ecology 68:405-420. R6 Wheeler, B. D. 1980c. Plant communities of rich-fen systems in England and Wales. III. Fen meadow, R7 fen grassland and fen woordland communities, and contact communities R8 Wichmann, M. C., M. J. Alexander, M. B. Soons, S. Galsworthy, L. Dunne, R. Gould, C. Fairfax, M. Niggemann, R. S. Hails, and J. M. Bullock. 2009. Human-mediated dispersal of seeds over R9 long distances. Proceedings of the Royal Society B: Biological Sciences 276:523-532. R10 Willson, M. F. 1993. Dispersal mode, seed shadows, and colonization patterns. Vegetatio 107- 108:261-280. R11 R12 R13 R14 R15 R16 R17 R18 R19 R20 R21 R22 R23 R24 R25 R26 R27 R28 R29 R30 R31 R32 R33 R34 R35 R36 R37 R38 R39 R1 R2 R3 Chapter 2 R4 R5 R6 R7 R8 the effect of habitat fragmentation and R9 R10 abiotic factors on fen plant occurrence R11 R12 R13 R14 R15 R16 R17 R18 R19 R20 R21 R22 R23 R24 R25 R26 R27 R28 R29 R30 R31 R32 R33 R34 R35 R36 R37 R38 Soomers, H., Karssenberg, D., Verhoeven, J.T.A., Verweij, P.A. and Wassen, M.J. R39 30 | Chapter 2

R1 abstraCt R2 R3 Human landscape modification has led to habitat fragmentation for many species. Habitat R4 fragmentation, leading to isolation, decrease in patch size and increased edge effect, is R5 observed in fen ecosystems that comprise many endangered plant species. However, until now R6 it has remained unclear whether habitat fragmentation per se has a significant additional negative effect on plant species persistence, besides habitat loss and degradation. We R7 investigated the relative effect of isolation, habitat size, and habitat edge compared to the R8 effect of habitat degradation by including both ‘fragmentation variables’ and abiotic variables R9 in best subsets logistic regression analyses for six fen-plant species. For all but one species, R10 besides abiotic variables one or more variables related to fragmentation were included in R11 the regression model. For Carex lasiocarpa, isolation was the most important factor limiting R12 species distribution, while for Juncus subnodulosus and Menyanthes trifoliata, isolation was R13 the second most important factor. The effect of habitat size differed among species and an R14 increasing edge had a negative effect on the occurrence of Carex lasiocarpa and Pedicularis R15 palustris. Our results clearly show that even if abiotic conditions are suitable for certain species, R16 isolation of habitat patches and an increased habitat edge caused by habitat fragmentation R17 affect negatively the viability of characteristic fen plant species. Therefore, it is important not R18 only to improve habitat quality but also to consider spatial characteristics of the habitat of R19 target species when deciding on plant conservation strategies in intensively used landscapes, such as fen areas in Western Europe and North America. R20 R21 R22 R23 R24 R25 R26 R27 R28 R29 R30 R31 R32 R33 R34 R35 R36 R37 R38 R39 Habitat fragmentation and abiotic factors | 31

2.1 IntroduCtIon R1 R2 Over the last decades, human impact on nature has increased (Vitousek et al., 1997). Landscape R3 modification has led to habitat loss and habitat degradation for many species. The former has R4 negative effects on the persistence of species (Tilman et al., 1994) and thus on biodiversity R5 (Fahrig, 2003). The latter, often caused by eutrophication, desiccation and acidification, also R6 decreases species viability (Vitousek et al., 1997). R7 Habitat loss often leads to a fragmented distribution of the remaining habitat patches that R8 are surrounded by a matrix of unsuitable land (Saunders et al., 1991). This causes a decrease in R9 connectivity between the remaining subpopulations and a lower colonization probability of suitable but empty sites. Also, the number of populations and the size of habitat patches, and R10 thus of populations usually decrease as a result of habitat fragmentation (Ouborg et al., 2006). R11 Small, isolated populations are more susceptible to the negative effects of environmental R12 and demographic stochasticity (Leimu et al., 2006). Furthermore, the influence of external R13 conditions in such populations is greater than in large (non-linear) ones (Laurance and Yensen, R14 1991) because of the increasing perimeter:area ratio with increasing fragmentation (edge R15 effect) (Saunders et al., 1991). Last, genetic drift and inbreeding lead to a decrease in genetic R16 diversity in small and isolated populations (Frankham, 2005; Ouborg et al., 2006). For these R17 reasons, fragmentation can have an additional negative effect on species viability, besides the R18 negative effect of habitat loss and habitat degradation per se (Fahrig, 2003). R19 In the literature, proposed plant conservation measures are often limited to restoring abiotic R20 conditions (e.g. Suding et al., 2005; Wassen et al., 2005) and tend to R21 ignore the effect of habitat configuration and dispersal limitation on biodiversity (Ozinga R22 et al., 2009). Conversely, metapopulation models that consider habitat configuration and biotic processes, such as dispersal, often ignore habitat quality (e.g. Helm et al., 2006) by R23 assuming that a habitat patch is either suitable or unsuitable for a species. Furthermore, seed R24 sowing studies that aim at investigating whether seed dispersal limits species distribution by R25 sowing seeds at unoccupied but potentially suitable sites often only consider germination, R26 not establishment of species (Ehrlen et al., 2006). Thereby, these studies ignore the possibility R27 that adult plants are affected by other factors than seedlings, which can lead to erroneous R28 conclusions on dispersal limitation (Ehrlen et al., 2006). Thus, to understand the dynamics R29 of population extinction and (re)colonization in fragmented landscapes, it is necessary to R30 consider the cumulative effects of abiotic and biotic factors on the performance of adult plant R31 species. The relatively negative effect of isolation, decrease in population size, and increase R32 in habitat edge compared to the effect of habitat degradation and loss per se, has remained R33 largely unrevealed until now (but see Pueyo and Alados, 2007). R34 Habitat fragmentation is common in Dutch wetlands and the decline in habitat quality and R35 quantity in them is most evident in nutrient-poor and groundwater-dependent ecosystems, such as species-rich fens (Runhaar et al., 1996). These fens contain many globally endangered R36 plant species and communities which necessitates their conservation and restoration (Lamers R37 et al., 2002). R38 R39 32 | Chapter 2

R1 The aims of this study are first, to determine the current degree of fragmentation of six rich- R2 fen plant species in a Dutch fen area, and second, to disentangle the negative effects of R3 habitat fragmentation from those caused by changes in abiotic site factors on the distribution R4 of these six species. To determine the key factors that explain the presence of these species, R5 a dataset comprising information on site factors and plant species composition was used. R6 This dataset was expanded with four spatial variables that together encompass the different aspects of habitat fragmentation: i) distance to the nearest (other) population, ii) number R7 of populations in the neighbourhood of sampling locations, iii) habitat area, and iv) habitat R8 area: perimeter ratio, representing the edge effect (e.g. Fahrig, 2003; Ouborg et al., 2006; R9 Saunders et al., 1991). These variables were computed with Geographical Information System R10 (GIS) (ESRI, 2006). One of our hypotheses is that these spatial variables contribute significantly R11 to the explanation of species distribution. More specifically, it is hypothesised that the R12 negative effect of dispersal limitation, decreasing habitat area, and increasing habitat edge R13 will be more pronounced for rare species with severely fragmented populations than for more R14 common and less fragmented species (see also Leimu et al., 2006). Another hypothesis is that R15 the effect of habitat fragmentation will be more pronounced for species that only disperse R16 over short distances (see also Ozinga et al., 2005). R17 R18 R19 2.2 metHods R20 2.2.1. Study site R21 The peatlands of the floodplain of the Vecht river in central Netherlands originated in the R22 Holocene era. The peatlands are bordered by the Pleistocene ice pushed hill ridge (0-30 m R23 above msl.) of Het Gooi (5º05’- 5º15’E and 52º07’- 52º20’N (Fig. 2.1)). Rainwater infiltrating in R24 the hill ridge discharges in the fens, leading to nutrient-poor, base-rich conditions. Peat has R25 been excavated since the 12th century, and many lakes originated because of these cutting R26 and dredging activities. The remaining area was drained, and polders have been created since R27 the 19th century. Fragmentation of regional groundwater seepage systems, mainly caused by R28 drinking water abstractions and drainage to the lower-lying polders, has led to a decrease of R29 seepage of base-rich nutrient-poor fresh groundwater that enters the root zone of plants, R30 causing changes in site factors (Schot, 1991). This phenomenon produced a fragmented R31 distribution of the remaining groundwater-dependent ecosystems. Nowadays, the area is R32 used for agriculture and contains some scattered nature reserves (Wassen and Barendregt, 1992). R33 R34 R35 R36 R37 R38 R39 Habitat fragmentation and abiotic factors | 33

R1 R2 R3 R4 R5 R6 R7 R8 R9 R10 R11 R12 R13 R14 R15 R16 R17 R18 R19 R20 R21 R22 R23 R24 R25 R26 R27 Figure 2.1. The Vecht river plain with piezometer locations (point dataset). Inset: location of the R28 study area in the Netherlands. R29 R30 2.2.2. Species descriptions R31 For six fen-plant species differing in abundance at the study site, the degree of fragmentation R32 and the dependency on several abiotic and spatial factors were investigated. Four of these R33 species are included in the Dutch Red List of vascular plant species, i.e. Carex diandra R34 Schrank (category: vulnerable), Carex lasiocarpa Ehrh. (vulnerable), Menyanthes trifoliata L. R35 (nearly threatened) and Pedicularis palustris L. (vulnerable). The two other species, Juncus subnodulosus Schrank and Equisetum fluviatileL., are more common in the Netherlands and R36 also in the Vecht river plain. All considered species except P. palustris are perennial. R37 R38 R39 34 | Chapter 2

R1 C. diandra and C. lasiocarpa are relatively rare small sedge species that have their optimum in R2 the rare, relatively nutrient-poor but species-rich, floating fens (Parvocaricetea class) (Lamers R3 et al., 2002) in the Vecht river plain. R4 P. palustris is a monocarpic hemiparasite with a primarily biennial life cycle (Schmidt and R5 Jensen, 2000). The species mainly occurs in floating (rich) fens and mesotrophic grasslands R6 (Van der Meijden, 1996). M. trifoliata, which is protected by the Flora and Fauna Act in the Netherlands, is a helophyte R7 occurring in shallow waters and on semi-terrestrial fens (Van der Meijden, 1996). This species R8 is more common than P. palustris, C. Diandra, and C. lasiocarpa (Van der Meijden, 1996). R9 J. subnodulosus and E. fluviatileoccur in small sedge fens and along ditches (Van der Meijden, R10 1996). The latter can also occur in shallow water and in fens dominated by large sedges R11 (Lamers et al., 2002). E. fluviatile, which is considered to be a seepage indicator, belongs to R12 the class of the Equisetopsida and produces spores instead of seeds (Van der Meijden, 1996). R13 Dispersal characteristics of all species are given in Table 2.1. Terminal velocity (i.e. falling R14 velocity of seeds in still air (m.s-1)) is a seed trait that is an important determinant of the seed’s R15 ability to disperse via wind (Soons and Heil, 2002) (the lower the terminal velocity, the farther R16 the seed can disperse). Seed buoyancy (days) enables seeds to disperse (over large distances) R17 via water. R18 table 2.1. Mean seed terminal velocity (i.e. falling velocity of seeds in still air (m.s-1)), seed buoyancy, R19 other potential dispersal vectors besides wind and surface water, and the assumed dominant long R20 distance dispersal vector of the studied fen-plant species. R21 species Mean seed Seed buoyancyb Seed dispersal Hypothesised terminal velocity (days) vectors (besides dominant long R22 (m/s)a wind and water)c distance dispersal R23 vectorh R24 Carex diandra 2.6 99 , animal fur Water f R25 Carex lasiocarpa 2.0 49 Ants, animal fur Water Equisetum fluviatiled ?g ? ? Wind R26 Juncus subnodulosus 0.9-1.2e 30 Animal fur Water R27 Menyanthes trifoliata 3.85 119 Animals (internally) Water R28 Pedicularis palustris 3.3 71 Ants Water

R29 aKleyer et al., 2008. bThe number of days after which 50 % of the seeds have sunk in stagnant water R30 (van den Broek et al., 2005). cBouman et al., 2000. dThis species produces spores, not seeds. eThe R31 value is unknown for this species. Values depicted here are based on two Juncus species with seed size, shape and weight comparable to that of J. subnodulosus: J. captitatus (0.9 m/s) & J.compressus R32 (1.2 m/s). fThe value measured in moving water (approximately 0.06 m.s-1) is presented here because R33 of a missing value for stagnant water. gThe Terminal velocity of spores of E. fluviatile is unknown. Nevertheless, it can be assumed to be very low (<0.3 m/s, (Soons, 2006)), due to the extremely small R34 size (approximately 40 µm (Lehmann et al., 1984)). hSpecies were characterised as long distant R35 wind dispersers if terminal velocity < 0.3 (Soons, 2006) and as long distant water dispersers if seed R36 buoyancy > 2 days (Johansson et al., 1996). R37 R38 R39 Habitat fragmentation and abiotic factors | 35

2.2.3. General approach R1 We used point and polygon spatial data to test our hypotheses. The point data comprised R2 species presence/absence data and data on abiotic factors at these points. The polygon data R3 consisted of species distribution maps for each of the six fen species, a nature type map of the R4 study area, and a topographic map of the study area. R5 For the floodplain of the Vecht river, we first quantified the current degree of fragmentation R6 of each species using species distribution maps, and second determined the key factors that R7 explain the presence of these fen species. The latter was realised by relating plant occurrence R8 (presence/absence at points) to spatial variables and to abiotic factors, using logistic regression R9 analyses. The datasets used, and the variables that were computed from these datasets are described R10 in Table 2.2. To create the species distribution and the vegetation type maps, species or R11 vegetation type respectively were mapped per cadastral parcel. R12 R13 2.2.4. Current fragmentation R14 Species distribution patterns were analysed using area covering species distribution maps R15 (see Table 2.2) to estimate the current degree of fragmentation of each species. Thus, we R16 computed 1) the total area of the parcels in which a species is present within the study area, R17 2) the distance from a population to the nearest population of the same species, averaged R18 over all populations, 3) the percentage of 1 x 1 km cells in the study area in which a species R19 is present, 4) the mean number of populations per 1 x 1 km cell, and 5) the total number of R20 populations for all species. R21 Fragmentation was considered to be severe when the total area covered by a certain species R22 was relatively small and the mean distance to the nearest population was relatively large (Fahrig, 2003). Furthermore, one would expect that a highly fragmented (meta-)population R23 comprises a relatively small number of populations distributed over a relatively large number R24 of 1 x 1 km cells. This would cause a low mean number of populations per occupied 1 x 1 R25 km cell. Finally, the number of populations within an area can initially increase during the R26 fragmentation process because existing populations divide, but eventually it will probably R27 decrease because of habitat loss (Fahrig, 2003). R28 Mean population size often decreases with increasing fragmentation (Ouborg et al., 2006). In R29 our data, the size of each population was not registered, but was considered to be similar to R30 the size of the cadastral parcel on which the population was growing, ignoring the possibility R31 that the population did not cover the whole parcel. Therefore, this variable was not included. R32 R33 R34 R35 R36 R37 R38 R39 36 | Chapter 2

R1 2.2.5. Regression analyses R2 The presence of each plant species was related to abiotic factors and spatial variables by R3 applying logistic regression analyses (see Table 2.2). The spatial variables included (Table 2.2) R4 are often mentioned in the literature about habitat fragmentation (e.g. Ewers et al., 2007; R5 Fahrig, 2003; Saunders et al., 1991). The data used as input for these analyses are described R6 below and shown in Table 2.2. R7 2.2.5.1. Datasets used for the regression analyses R8 Point data: abiotic factors and species occurrence R9 Between 1986 and 1992, groundwater samples were taken in the growing season (April- R10 September) from piezometers with a filter depth of 50 cm below the soil surface (n=250) (De R11 Mars, 1996; Wassen and Barendregt, 1992) (Fig. 2.1). Of these samples, Electro-Conductivity R12 + - + - (EC25), pH, groundwater table relative to the peat surface, nutrients (K , NO3 , NH4 , H2PO4 , P) R13 2+ 2+ + - - 2- (mg/l) and major and minor ions (Ca , Mg , Na , Cl , HCO3 , SO4 , Fetot) (mg/l) were measured. R14 At each sampling location, presence or absence of fen species was recorded in a 10 m² plot. R15 Data on the occurrence of C. diandra, C. lasiocarpa, M. trifoliata, P. palustris, J. subnodulosus R16 and E. fluviatile were chosen from this database. R17 R18 Polygon data: spatial variables R19 In addition to the abiotic factors, four different spatial variables that are related in the literature to habitat fragmentation were included in the logistic regression analyses as R20 independent variables (Table 2.2). These spatial variables were 1) number of populations R21 within 500 m, 2) distance to the nearest (other) population (related to isolation), 3) habitat R22 area, and 4) habitat edge. Also, distance from the sampling location to the hill ridge was R23 included as an independent variable. Several different GIS maps (polygon datasets) were R24 used to compute these spatial variables (see Table 2.2). The private nature conservation R25 organisation Natuurmonumenten and the province of Utrecht supplied maps of ‘Nature Types’ R26 (©Vereniging Natuurmonumenten, ‘s-Graveland; Provincie Utrecht, 2003). The classification R27 of Nature Types (see Table 2.2) includes vegetation types, such as quagfen and wet grassland, R28 but also anthropogenic classes, such as road or garden. The spatial variables were constructed R29 using the GIS package ArcGIS 9.1 (ESRI, 2006). R30 For each sampling location, the distance to the hill ridge was computed using an overlay R31 between the sampling location point map and the zero elevation isoline in the Dutch R32 topographic map that represents the edge of the hill ridge.

R33 Number of populations within 500 m and nearest population. R34 The number of populations within 500 m of each sampling location and the Euclidian distance R35 to the centroid of the nearest (other) population from each sampling location were calculated R36 using the area covering species distribution maps (Table 2.2). Polygons in each species R37 distribution map that shared a common edge were merged as they were considered to be R38 one population. For each species, both spatial variables were computed by making an overlay R39 Habitat fragmentation and abiotic factors | 37

between the resulting species distribution polygon map and the map containing the sampling R1 location points (point data). If a sampling location was situated within a population of the R2 considered species, this population was not included in the computed number of nearby R3 populations nor in the computed distance to the nearest population. The distance of 500 R4 m was chosen because this is approximately the maximum distance seeds of the considered R5 species can disperse, when considering wind and water dispersal (Soomers et al., 2010; Soons, R6 2006). R7 R8 Habitat area and area:perimeter. R9 For each species, habitat was mapped by selecting the polygons of those Nature Types in which the species can be found according to Vereniging Natuurmonumenten (©Vereniging R10 Natuurmonumenten, ‘s-Graveland) and Schaminée et al. (1995). Habitat was defined as a R11 continuous patch suitable for the species. Polygons that shared a common edge were merged. R12 Then, for each species, 1) habitat area and 2) habitat area: perimeter ratio of the habitat R13 polygon that contained the sampling location were calculated at each sampling location. R14 When the sampling location was not inside any habitat polygon for a certain species, habitat R15 area and area: perimeter ratio were set to 0 for the species under consideration. R16 R17 2.2.5.2. Analysis R18 To gain insight into the key factors that determine plant presence for each of the considered R19 species, ‘best subsets logistic regression analyses’ were run following the procedure described R20 by King (2003) and using the statistical package SAS version 9.1. Abiotic and spatial variables R21 were independent factors in the models, and presence/absence of plant species was the R22 dependent factor. From all possible models, the model with the lowest Mallow’s Cp value was chosen as the best model (Draper and Smith, 1998). To check for consistency of the results, R23 backward stepwise regression analyses with model selection through a likelihood ratio test R24 were run and the results were compared to those of the best subsets regression. R25 To avoid collinearity or multi-collinearity (De Veaux and Ungar, 1994), from groups of strongly R26 correlating predictor variables (r >0.8) only one was kept. This led to the omission of pH, Mg2+, R27 + - Na , HCO3 and EC25. Because the habitat in which was sampled was relatively homogeneous, R28 the gradient in the environmental data was relatively short and a linear species response R29 to these data was found in an explorative data analysis. Therefore, only linear terms were R30 included in the regression analyses (see Lepš and Šmilauer, 2007). Single imputation was used R31 to estimate missing values (Catellier et al., 2005), preventing loss of power due to a diminished R32 sample size. Model results were compared to those of a backward stepwise regression. R33 Moran’s I values (Diniz-Filho et al., 2003; Legendre and Legendre, 1998), obtained at seven R34 distance intervals, were used to test for spatial autocorrelation in the standardised residuals of R35 the regression models. Positive autocorrelation at small distances would indicate that certain explanatory variables were missing in the analyses (Diniz-Filho et al., 2003). Moran’s I values R36 were calculated using SAM software (Rangel et al., 2006). The statistical significance of these R37 values was assessed using Monte Carlo randomization (200 permutations). R38 R39 38 | Chapter 2

R1 R2 R3 ©Vereniging ©Vereniging

R4 c R5 , using five datasets R6 5 R7

R8 e R9 R10 R11 R12 nalysis 2: (for each species separately) Dependent variable Independent variables (equal for all species) Used to compute two independent variables for each point in the dataset: 1. Habitat area 2. Area: perimeter Used to compute one independent variable for each point in the dataset: Distance to hill ridge (equal for all species). R13 a Regression analyses Used to compute two independent variables for each point in the dataset: 1. Number of populations within 500 m 2. Distance to the nearest population R14 Determining for each species the key factors (spatial R15 e R16 R17 R18 R19 R20 R21 R22 Topografische Dienst, 2005, Topografische d R23

R24 Provincie Utrecht, 2003; Province of Noord-Holland, Unpublished results

R25 b

R26 nalysis 1: a Current fragmentation (for each species separately) Used to compute for each species: area of all parcels in which the species 1. Total is present 2. Mean nearest neighbour distance between populations 3. Percentage of occupied 1x1 km cells 4. Mean number of populations per 1x1 km cell 5. Number of populations R27 a R28 R29 R30 a

R31 d R32 b R33 c atasets d Species presence/absence Abiotic factors Species distribution maps Dutch topographic map (top10) R34 Nature Types map R35

R36 Description of the variables computed for 1) analysis current fragmentation and 2) regression analyses R37

R38 ataset type able 2.2. d Point data Polygon data De Mars, 1996; Wassen and Barendregt, 1992, De Mars, 1996; Wassen a t (rows), of which two were point- and three polygon data. and abiotic variables) that explain the presence of these fen species. R39 Natuurmonumenten, ‘s-Graveland; Provincie Utrecht, 2003, Habitat fragmentation and abiotic factors | 39

2.3 results R1 R2 2.3.1. Quantification of current fragmentation R3 The small sedge species C. diandra and C. lasiocarpa and the herbaceous species P. palustris R4 have a small number of populations (≤50), a total population area less than 100 ha (the total R5 area of the study site is 26 400 ha), and a relatively high mean nearest neighbour distance (>370 R6 m) (Table 2.3). Conversely, the more common species E. fluviatile and J. subnodulosus have a R7 considerably larger number of populations and total population area and a mean nearest R8 neighbour distance below 220 m. Furthermore, the percentage of occupied 1 x 1 km cells is R9 considerably higher for these species than for the above-mentioned species. Intermediate values for these variables were found for M. trifoliata (Table 2.3). R10 R11 table 2.3. Number of current populations of the considered species in the Vecht river plain, total R12 area of the populations, mean nearest neighbour (NNB) distance for the populations, percentage R13 of square kilometre cells in the study area that is occupied by the considered species, and the mean number of populations per occupied 1 x 1 km cell for each of the considered species. R14 Species Number of Total area of Mean NNB Percentage (%) Mean number R15 populations populations distance of 1x1 km cells in of populations R16 (m²) (m) the area that is per occupied occupied 1x1 km cell. R17 Carex diandra 35 673 858 525.7 8.3 1.6 R18 Carex lasiocarpa 50 900 366 378.9 11.7 1.6 R19 Equisetum fluviatile 1107 4 639 995 120.2 49.2 8.5 R20 Juncus 343 3 676 467 218.8 39.0 3.3 subnodulosus R21 Menyanthes 95 1 879 093 338.3 17.0 2.1 R22 trifoliata R23 Pedicularis palustris 41 730 333 491.9 12.5 1.2 R24 R25 2.3.2. Regression analyses: key factors that determine species occurrence R26 Mean and standard deviation values for the independent variables included in the regression R27 analyses are given in the Appendix. R28 Table 2.4 shows the Spearman correlation coefficient for all pairs of abiotic independent R29 2+ + - variables included in the regression analyses. The variables pH, Mg , Na , HCO3 , and EC25 R30 correlated strongly (r>0.8) with one or more other independent variables and were therefore R31 excluded from the regression analysis and not listed in Table 2.4. Correlation coefficients (r) R32 for the remaining abiotic variables are smaller than 0.6 or larger than -0.6 (Table 2.4). R33 R34 R35 R36 R37 R38 R39 40 | Chapter 2

table 2.4. Spearman correlation coefficients for all combinations of abiotic factors. WL: groundwater R1 level below peat surface; other variables: concentrations of nutrients and ions in groundwater. R2 WL Ca Cl Fe K NH4 NO3 PO4 SO4 P R3 WL - 0.134* 0.232** -0.074 0.018 -0.105 0.021 0.027 -0.201** -0.048 R4 Ca - 0.265** 0.084 0.137* -0.123 -0.183** -0.081 0.088 -0.130* R5 Cl - -0.166** 0.365** -0.139* -0.053 0.059 0.030 -0.029 Fe - -0.255** -0.194** -0.396** -0.223** 0.590** 0.139* R6 K - 0.239** 0.261** 0.189** -0.059 0.126* R7 NH4 - 0.323** 0.324** -0.121 0.161* R8 NO3 - 0.543** -0.395** 0.144* R9 PO4 - -0.383** 0.422** SO4 - 0.033 R10 P - R11 R12 *=p<0.05, **=p<0.01 R13 R14 Table 2.5 presents the results of the best model (having the lowest Mallow’s Cp value) of R15 the best-subsets logistic regression analyses for the six species. For all but one species, the R16 resulting regression models comprised both abiotic and spatial variables. Only for C. diandra, R17 the model does not included spatial variables. R18 Species occurrence of all six species was negatively related to water table depth (i.e. wetter R19 conditions were preferred). The presence of C. diandra, J. subnodulosus, E. fluviatile and M. trifoliata was positively related to the Calcium content of the groundwater (hereafter R20 referred to as [Ca]), while [Ca] was negatively, but not significantly related to the presence of R21 C. lasiocarpa. [Cl] was negatively related to both the occurrence of C. diandra and M. trifoliata. R22 A negatively significant relation between species occurrence and the contents of the nutrients R23 K (C. lasiocarpa and E. fluviatile), PO4 (J. subnodulosus, E. fluviatile,and M. trifoliata), NO3 (J. R24 subnodulosus) and NH4 (P. palustris) was found. Conversely, a positively significant relation R25 was found between the occurrence of M. trifoliata and E. fluviatile and both [NH4] and [P]. R26 The occurrence of J. subnodulosus was significantly positive related to the number of R27 populations within 500 m of the sampling location, while the occurrence of both M. trifoliata R28 and C. lasiocarpa was negatively related to the distance from the sampling location to the R29 nearest population. The area: perimeter ratio of the habitat patch was positively related R30 to the occurrence of C. lasiocarpa and P. palustris, which suggests a negative effect of an R31 increasing edge for these species. R32 The presence of J. subnodulosus was significantly positive related to the habitat area, while the occurrence of E. fluviatile and P. palustris were significantly negative related to the habitat R33 area. R34 For all species, backward stepwise regression analyses with model selection through a R35 likelihood ratio test produced results similar to best-subsets regression for all species. The only R36 difference between the two methods was an additionally negative, non-significant (P=0.115) R37 relation between the occurrence of C. diandra and the distance to the hill ridge when running R38 the stepwise regression. R39 Habitat fragmentation and abiotic factors | 41

R1

Wald R2 R3 R4 0.899 2.0347 0.377 -0.001 5.325* NI 0.195 21.258** -0.032 5.501* Pedicularis palustris NI NI -0.93 7.834** NI R5 R6 R7 Wald B R8

NI R9 0.093 7.7934 0.499 -0.003 11.190** NI NI Menyanthes trifoliata NI NI -0.022 2.355 NI R10 R11 R12 Wald B R13 R14 R15 0.310 7.421 0.335 0.001 19.368** NI 0.088 7.947** NI 0.271 4.626* 0.486NI 8.111** -0.446NI 1.911 NI Equisetum fluviatile 4.194 6.317* 4.507 3.133 NI NI -0.307 6.088* NI R16 R17

Wald B R18 R19 R20 0.813 5.0169 0.485 NI NI 0.55 10.874** -0.02 4.694* NI -0.758NI 2.3610.191 NI 17.543** NI NI -1.594 18.763** -.310 3.338 -2.016 10.823** NI Juncus subnodulosus NI NI R21 R22 R23 Wald B R24 R25 0.947 1.1313 0.168 NI NI -0.001 7.197** NI NI 0.051 5.024** NI NI NI NI NI NI Carex lasiocarpa NI -0.268 3.341 NI R26 R27 R28 Wald B R29 R30 0.836 2.2074 NI B NI NI NI NI NI NI NI Carex diandra NI - 0.098 -0.036 16.31** 4.710* -0.064 12.917** -0.079 18.790** -0.068 8.506** -0.111 16.088** 0.037 28.24** 0.037 -0.11 28.24** 2.669 0.013 3.842* 0.013 4.864* 0.027 12.385** NI -0.017 5.79* -0.017 NI 5.79* NI R31 R32 R33 R34 R35

Regression coefficient (B) and Wald statistics (the squared ratio of B to the Standard error) of the predictor variables (abiotic factors (abiotic variables predictor the of error) Standard the to B of ratio squared (the statistics Wald and (B) coefficient Regression R36 R37 (mg/l) (mg/l) (mg/l) (mg/l) 3 4 4 4 R38 able 2.5. able HL Cp-statistic Nagelkerk-R² total model 0.356 NO Distance to hill ridge (m) NI Fe (mg/l) Predictor variables NH Distance to nearest population (m) Area of habitat (hm²) Area: perimeter ratio of NI habitat (m²/m) Number of populations within 500 m Species P (mg/l) SO Water level (cm below Water mowing field) Ca (mg/l) Cl (mg/l) K (mg/l) PO NI=not included, *=p≤0.05, **=p≤0.01 t six of the the presence/absence variables were Dependent each species. model for regression in the best logistic variables) included and spatial considered fen plant species. Nagelkerk-R² values indicating the variance explained by total model, and Hosmer Lemeshow goodness of fit values (HL; <0.05 indicate that the model does not fit data) are presented for each model. R39 42 | Chapter 2

R1 R2 P R3 R4

R5 3.5 km M R6 R7 P R8 R9 R10 3 km M 0.017 0.66 -0.013 0.58 R11

R12 P R13 R14 R15 2.5 km M -0.016 0.4850.05 -0.121 0.015 0.03 -0.083 0.02 R16

R17 P R18 R19 2 km M R20 R21

R22 P R23 R24 1.5 km M 0.077 0.005 0.08 0.005 R25 R26 R27 P R28

R29 1 km M -0.059 0.045 -0.203 0.005 0.047 0.09 -0.0330.014 0.1 0.595 0.059 0.005 0.085 0.845 -0.025 -0.047 0.39 0.04 R30 R31 R32 P R33

R34 Lags 500 m M 0.039 0.10.181 0.005 0.0220.006 0.395 0.775 0.0050.028 -0.031 0.865 0.205 0.2150.035 -0.067 0.024 0.021 0.02 0.1 0.375 0.350.064 -0.035 0.011 0.025 0.05 0.012 0.09 0.585 0.605 0.085 0.015 0.035 0.004 0.52 0.28 0.8 0.021 0.004 0.004 0.26 0.89 0.9 -0.023 0.485 -0.015 0.011 0.535 0.655 R35

R36 lag distances. Bold values are significantly I autocorrelation (M) of the standardised model residuals for each species, different Moran’s R37

R38 <0.05) positively auto-correlated. able 2.6. Carex diandra Carex lasiocarpa Juncus subnodulosus Equisetum fluviatile Menyanthes trifoliata Pedicularis palustris P R39 t ( Habitat fragmentation and abiotic factors | 43

The standardised model residuals for C. lasiocarpa and P. palustris were significantly spatially R1 auto-correlated for the first separation lag (500 m) in a Moran’s I autocorrelation analysis. R2 For M. trifoliata and P. palustris, the standardised residuals were positively spatially auto- R3 correlated from 1000 m to 2000 m and from 2000 m to 2500 m, respectively (Table 2.6). R4 R5 R6 2.4 dIsCussIon R7 R8 We have investigated whether spatial variables related to habitat fragmentation have an R9 additional effect on the distribution of rich-fen plant species, on top of the effect of abiotic site factors. We expected that spatial variables contributed significantly to the explanation of R10 species occurrences. R11 The results of the regression analyses clearly confirm the hypothesis that there is an additional R12 negative effect of habitat fragmentation on the occurrence of fen plant species besides the R13 negative effect of habitat degradation. For all species, except for C. diandra, one or more R14 variables related to fragmentation were included in the best regression model. For C. lasiocarpa R15 (distance to the nearest population) and P. palustris (area: perimeter ratio), the highest Wald R16 values in the regression model were even attributed to a ‘fragmentation variable’. R17 The Morans’ I values demonstrate that in spite of the inclusion of dispersal-related variables R18 in the regression models for three species, some spatial autocorrelation remained present in R19 the model residuals. This might indicate that explanatory variables are missing (Diniz-Filho R20 et al., 2003). Furthermore, spatial autocorrelation in the model residuals could exaggerate R21 the strength of the relations (Whittaker et al., 2007). However, Hawkins et al. (2007) found R22 that, at least for ordinary least squares regression of gridded geographical data, regression coefficients were not seriously affected by the presence of spatial autocorrelation (see also R23 Bini et al., 2009). Thus, the presence of some autocorrelation for three of the species would R24 not change our major conclusion that both abiotic factors and spatial configuration of habitat R25 patches influence fen-plant-species distribution. R26 The results did not support our hypothesis that species for which the populations are already R27 fragmented would be more prone to negative effects of habitat fragmentation. Highly R28 fragmented rare species and less fragmented common species were negatively affected by the R29 effects of habitat fragmentation (Table 2.3 and 2.5). Leimu et al. (2006) report that rare species R30 tend to be more susceptible to the negative effects of habitat fragmentation than common R31 species although their results were not significant. Conversely, Honnay and Jacquemyn R32 (2007) found that the genetic consequences of habitat fragmentation were equally or more R33 pronounced in common than in rare species. To clarify this issue, more research is needed. R34 R35 2.4.1. Dispersal limitation Habitat fragmentation leads to isolation of the remaining habitat patches (Hanski, 1999). R36 Consequently, dispersal capacity of species in fragmented areas can limit species persistence R37 (Ozinga et al., 2009). It can then be assumed that the probability that a species occurs at R38 R39 44 | Chapter 2

R1 a certain suitable site would increase with decreasing distance to the nearest population R2 (i.e. seed source), and with increasing number of populations nearby. This is reflected in the R3 results found for J. subnodulosus, C. Lasiocarpa, and M. trifoliata. For J. subnodulosus, the R4 probability of occurrence increased with increasing number of nearby populations, while R5 for C. lasiocarpa and M. trifoliata the probability of occurrence increased with decreasing R6 distance to the nearest population. These highly significant results indicate that seed dispersal is limiting species distribution of J. subnodulosus, C. lasiocarpa, and M. trifoliata. Similar R7 results were found for a semi-arid area by Pueyo & Alados (2007), who conclude that the R8 re-establishment of the steppe species Lygeum spartum in a fragmented Mediterranean area R9 was most strongly related to the distance to propagule sources. Our results are also congruent R10 with findings of the long-term seed sowing study by Ehrlen et al. (2006), in which it was R11 observed that establishment of forest herbs can occur after sowing in sites unoccupied by R12 adult plants of the same species and in which hence is concluded that these forest herbs were R13 limited by seed dispersal. Nevertheless, no effect related to dispersal limitation was found for R14 the abundant species E. fluviatileand for the rare species C. diandra and P. palustris. It can R15 be expected that for E. fluviatile, distances between suitable habitat patches are too short R16 to impair colonization probability. Conversely, the occurrence of both highly fragmented R17 and rare Red List species C. diandra and P. palustris was not promoted by a close distance to R18 other population or an abundant number of neighbouring populations either. Long-distance R19 dispersal for these species is most likely to be realised via surface water (Table 2.1). However, surface water can only enhance colonization probability or connectivity for hydrochorous (i.e. R20 water dispersing) species if the water bodies connect seed sources with potentially suitable R21 habitat. Ozinga et al. (2009) indicate that the dispersal infrastructure (i.e. the spatial structure R22 of the dispersal vector, e.g. surface water, of a species) for many species in North-western R23 Europe is insufficient. If habitat patches of hydrochorous species are not connected by surface R24 water bodies, seeds of highly fragmented species with few populations and narrow ecological R25 amplitudes, such as C. diandra and P. palustris, will not be able to colonize new habitat patches, R26 even if these are nearby. We expect that this could explain that small distances to a population R27 or many populations within a certain distance were not related to a higher probability of R28 occurrence for C. diandra and P. palustris. These results stress the importance of further R29 research on dispersal mechanisms of plant species in fragmented habitats and the importance R30 of the improvement of the dispersal infrastructure for such species. Thus, although our results R31 reflect that seed dispersal limitation affects negatively the distribution of rich-fen species, R32 additional to the negative effect of habitat degradation, the hypothesis that the effect of habitat fragmentation will be more important for species that only disperse over short R33 distances was not supported. Future research should focus on effective dispersal distances R34 instead of potential Euclidian dispersal distances, considering the dispersal infrastructure of R35 the considered species. R36 R37 R38 R39 Habitat fragmentation and abiotic factors | 45

2.4.2. Edge and area effect R1 In general, the proximity of habitat edges can have a positive, negative or neutral effect on R2 species abundance and on biodiversity inside a given habitat, depending on the edge type and R3 the species’ life history strategies and habitat requirements (see reviews of Ewers and Didham, R4 2005; Saunders et al., 1991). In habitat patches with a small area: perimeter ratio, the effect of R5 the external conditions will be relatively greater, causing a smaller survival probability in these R6 patches if the effect of the edge is negative. R7 In our study area, habitat patches are surrounded by a matrix of unsuitable agricultural land R8 in which rare species with a narrow ecological amplitude, such as C. lasiocarpa and P. palustris, R9 cannot establish. The results for C. lasiocarpa and P. palustris demonstrate that these species are more likely to be present in patches that have a relatively large area: perimeter ratio. This R10 indicates that these species might be occurring more often in round than in linearly shaped R11 landscape elements, which suggests that these species may suffer from the ‘edge effect’ (e.g. R12 Saunders et al., 1991). However, these findings could also be explained by the fact that the R13 two species usually do not occur in linearly shaped landscape elements, such as ditches or ditch R14 banks, because the site factors tend to be unsuitable (i.e. too nutrient rich) for them. R15 It should be noted that it is difficult to disentangle an edge and an area effect, as synergistic R16 interaction between these two variables can determine species occurrence (Ewers et al., 2007). R17 Nevertheless, Smith et al. (2009) state that when trying to evaluate the relative effect of R18 different fragmentation related predictors, such as amount of habitat, mean patch size and R19 edge size, on species abundance, standard multiple regression performs as well as or better R20 than methods that consider collinearity. R21 Habitat area was, as expected, positively related to species occurrence for J. subnodulosus. R22 However, a negative relation between habitat size and species occurrence was found for E. fluviatile and P. palustris. The negative relation found for E. fluviatile might be caused because R23 this species abundantly occurs at line elements with a usually small area, such as ditch banks, R24 in the study site. R25 Collins et al. (2009) found that woody encroachment progressed faster in larger patches, R26 leading to less favourable germination conditions for early-successional species. This effect R27 possibly caused the negative relation between patch size and the persistence of annual early- R28 successional species found by Collins et al. (2009). If woody encroachment also progresses R29 faster in larger patches in our study site, the negative relation between habitat size and R30 species occurrence found for the biannual species P. palustris could possibly be explained by R31 that mechanism. R32 R33 2.4.3. Abiotic factors R34 Generally, the occurrence of most species was negatively related to the contents of one or R35 more nutrients. Furthermore, all species occurred more often at sites with higher water levels and most of the species preferred sites with high calcium content.This agrees with the results R36 of studies by for example van Diggelen et al. (1996) and Sjors and Gunnarsson (2002), who R37 concluded that species of low productive rich-fens are generally persisting at alkaline wet R38 R39 46 | Chapter 2

R1 sites. These conditions are usually present at groundwater discharge sites, or at sites supplied R2 by mineral rich but relatively nutrient poor surface water. Therefore, our results indicate that R3 it is important to enhance discharge of calcium rich groundwater in the study site or to enable R4 the supply of clean, nutrient poor, alkaline surface water to rich fen patches or potential rich R5 fen sites. R6 2.4.4. Implications for nature conservation and further research R7 Our results show that even if abiotic conditions are suitable for certain species, isolation of R8 habitat patches and decreased area: perimeter ratios of these patches negatively influence R9 the viability of the investigated fen plant species. Since for the majority of the fen plant R10 species fragmentation indicators add significantly to the explanation of their occurrences, R11 we conclude that it is important not only to improve habitat quality, but also to connect R12 fragmented populations acknowledging the dispersal capacity and the dominant dispersal R13 vectors of fen-plant species. R14 Furthermore, it would be advisable to restore rich-fen vegetation in such a way that circular R15 rather than linear areas (the latter having relatively more edge) are created and that R16 newly created habitat patches are situated within the dispersal range of the species. When R17 attempting to improve the dispersal infrastructure for hydrochorous species, one should R18 ensure that subpopulations are connected by the same water body and that water borne R19 seeds of (semi)terrestrial species can be deposited at suitable germination sites. The latter could be realised by the creation of shallow ditch banks or lake shores (Soomers et al., 2010). R20 Although our study focused on fens and plant species only, we may generalize the results R21 to other ecosystems and species. In densely populated areas, such as Western Europe and R22 North America, many ecosystems suffer from both deterioration of habitat quality (by for R23 instance diffuse pollution via airborne or waterborne pollutants) as well as from a fragmented R24 distribution of the remaining habitat patches. Intensive land use and drainage has transformed R25 most of the rural area into unsuitable ‘deserts’ for species of forests, nutrient-poor grasslands R26 and lakes, for example. Therefore, our results stress the importance of considering both R27 habitat quality and connectivity in restoration attempts in intensively used landscapes. This R28 requires the development of models which consider both aspects. R29 R30 acknowledgments R31 The authors thank the Province of Utrecht, the Province of Noord-Holland, and R32 Natuurmonumenten for providing data and Hans de Mars for collecting data; Maarten Zeylmans van Emmichoven for his advice on GIS procedures and Rogier Donders for his advice R33 on statistics. R34 R35 R36 R37 R38 R39 Habitat fragmentation and abiotic factors | 47

reFerenCes R1 R2 Bini, L.M., Diniz-Filho, J.A.F., Rangel, T.F.L.V.B., Akre, T.S.B., Albaladejo, R.G., Albuquerque, F.S., R3 Aparicio, A., Araújo, M.B., Baselga, A., Beck, J., Bellocq, M.I., Böhning-Gaese, K., Borges, P.A.V., Castro-Parga, I., Chey, V.K., Chown, S.L., De Marco Jr, P., Dobkin, D.S., Ferrer-Castán, R4 D., Field, R., Filloy, J., Fleishman, E., Gómez, J.F., Hortal, J., Iverson, J.B., Kerr, J.T., Kissling, R5 W.D., Kitching, I.J., León-Cortés, J.L., Lobo, J.M., Montoya, D., Morales-Castilla, I., Moreno, J.C., Oberdorff, T., Olalla-Tárraga, M.A., Pausas, J.G., Qian, H., Rahbek, C., Rodríguez, M.A., R6 Rueda, M., Ruggiero, A., Sackmann, P., Sanders, N.J., Terribile, L.C., Vetaas, O.R., Hawkins, R7 B.A., 2009. Coefficient shifts in geographical ecology: An empirical evaluation of spatial R8 and non-spatial regression. Ecography 32, 193-204. Bouman, F., Boesewinkel, D., Bregman, R., Devente, N., Oostermeijer, G., 2000. Verspreiding van R9 zaden. KNNV Uitgeverij, Utrecht. R10 Catellier, D.J., Hannan, P.J., Murray, D.M., Addy, C.L., Conway, T.L., Yang, S., Rice, J.C., 2005. Imputation of missing data when measuring physical activity by accelerometry. Medicine R11 and Science in Sports and Exercise 37, S555-S562. R12 Collins, C.D., Holt, R.D., Foster, B.L., 2009. Patch size effects on plant species decline in an R13 experimentally fragmented landscape. Ecology 90, 2577-2588. De Mars, H., 1996. Chemical and physical dynamics of fen hydro-ecology, p. 167. Rijksuniversiteit R14 Utrecht, Utrecht. R15 De Veaux, R.D., Ungar, L.H., 1994. Multicollinearity: A tale of two nonparametric regressions., In Selecting models from data: AI and Statistics IV. eds P. Cheeseman, R.W. Oldford, pp. 293- R16 302. Springer-Verlag. New York, NY, US. R17 Diniz-Filho, J.A., Bini, L.M., Hawkins, B.A., 2003. Spatial autocorrelation and red herrings in R18 geographical ecology. Global Ecology and Biogeography 12, 53-64. Draper, N.R., Smith, H., 1998. Applied Regression Analysis, Third edn. John Wiley & Sons, Inc., New R19 York. R20 Ehrlen, J., Münzbergová, Z., Diekmann, M., Eriksson, O., 2006. Long-term assessment of seed limitation in plants: results from an 11-year experiment. Journal of Ecology 94, 1224-1232. R21 ESRI, 2006. ArcGIS 9.1. Environmental Systems Research Institute, , California. R22 Ewers, R.M., Didham, R.K., 2005. Confounding factors in the detection of species responses to R23 habitat fragmentation. Biological Reviews 81, 117-142. Ewers, R.M., Thorpe, S., Didham, R.K., 2007. Synergistic interactions between edge and area effects R24 in a heavily fragmented landscape. Ecology 88, 96-106. R25 Fahrig, L., 2003. Effects of Habitat Fragmentation on Biodiversity. Annu. Rev. Ecol. Evol. Syst. 34, 487-515. R26 Frankham, R., 2005. Genetics and extinction. Biological Conservation 126, 131-140. R27 Hanski, I., 1999. Metapopulation Ecology. Oxford University Press, Oxford. R28 Hawkins, B.A., Diniz-Filho, J.A.F., Mauricio Bini, L., De Marco, P., Blackburn, T.M., 2007. Red herrings revisited: Spatial autocorrelation and parameter estimation in geographical ecology. R29 Ecography 30, 375-384. R30 Helm, A., Hanski, I., Partel, M., 2006. Slow response of plant species richness to habitat loss and fragmentation. Ecology Letters 9, 72-77. R31 Honnay, O., Jacquemyn, H., 2007. Susceptibility of common and rare plant species to the genetic R32 consequences of habitat fragmentation. Conservation Biology 21, 823-831. R33 Johansson, M.E., Nilsson, C., Nilsson, E., 1996. Do rivers function as corridors for plant dispersal? Journal of Vegetation Science 7, 593-598. R34 King, J.E., 2003. Running a Best-Subsets Logistic Regression: An Alternative to Stepwise Methods. R35 Educational and Psychological Measuerement 63, 392-403. Kleyer, M., Bekker, R.M., Knevel, I.C., Bakker, J.P., Thompson, K., Sonnenschein, M., Poschlod, P., R36 van Groenendael.J.M., K., L., Klimesova, J., Klotz, S., Rusch, G.M., Hermy, M., Adriaens, R37 D., Boedeltje, G., Bossuyt, B., Dannemann, A., Endels, P., Götzenberger, L., Hodgson, R38 J.G., Jackel, A-K., Kühn, I., Kunzmann, D., Ozinga, W.A., Römermann, C., Stadler, M., R39 48 | Chapter 2

Schlegelmilch, J., Steendam, H.J., Tackenberg, O., Wilmann, B., Cornelissen, J.H.C., Eriksson, R1 O.,Garnier, E., Peco, B. , 2008. The LEDA Traitbase: A database of life-history traits of the R2 Northwest European flora. Journal of Ecology 96, 1266-1274. R3 Lamers, L.P.M., Smolders, A.J.P., Roelofs, J.G.M., 2002. The restoration of fens in the Netherlands. Hydrobiologia 478, 107-130. R4 Laurance, W.F., Yensen, E., 1991. Predicting the impacts of edge effects in fragmented habitats. R5 Biological Conservation 55, 77-92. Legendre, P., Legendre, L., 1998. Numerical Ecology. Elsevier, Amsterdam. R6 Lehmann, H., Neidhart, H.V., Schlenkermann, G., 1984. Ultrastructural investigations on sporogenesis R7 in Equisetum fluviatile. Protoplasma 123, 38-47. R8 Leimu, R., Mutikainen, P., Koricheva, J., Fischer, M., 2006. How general are positive relationships between plant population size, fitness and genetic variation? Journal of Ecology 94, 942- R9 952. R10 Lepš, J. and Šmilauer, P., 2007. Multivariate Analysis of Ecological Data using CANOCO. Cambridge University Press, Cambridge. R11 Ouborg, N.J., Vergeer, P., Mix, C., 2006. The rough edges of the conservation genetics paradigm for R12 plants. Journal of Ecology 94, 1233-1248. R13 Ozinga, W.A., Römermann, C., Bekker, R.M., Prinzing, A., Tamis, W.L.M., Schaminée, J.H.J., Hennekens, S.M., Thompson, K., Poschlod, P., Kleyer, M., Bakker, J.P., van Groenendael, R14 J.M., 2009. Dispersal failure contributes to plant losses in NW Europe. Ecology Letters 12, R15 66-74. Ozinga, W.A., Schaminee, J.H.J., Bekker, R.M., Bonn, S., Poschlod, P., Tackenberg, O., Bakker, J., van R16 Groenendael, J.M., 2005. Predictability of plant species composition from environmental R17 conditions is constrained by dispersal limitation. Oikos 108, 555-561. R18 Provincie Utrecht, 2003. Handleiding 2003, Ecologisch Onderzoek, onderdeel Flora en Vegetatie. Sector Ecologisch onderzoek en Groene regelgeving, Utrecht. R19 Pueyo, Y., Alados, C.L., 2007. Effects of fragmentation, abiotic factors and land use on vegetation R20 recovery in a semi-arid Mediterranean area. Basic and Applied Ecology 8, 158-170. Rangel, T.F.L.V.B., Diniz-Filho, J.A.F., Bini, L.M., 2006. Towards an integrated computational tool for R21 spatial analysis in macroecology and biogeography. Global Ecology and Biogeography. 15. R22 Runhaar, J., vanGool, C.R., Groen, C.L.G., 1996. Impact of hydrological changes on nature conservation R23 areas in the Netherlands. Biological Conservation 76, 269-276. Saunders, D.A., Hobbs, R.J., Margules, C.R., 1991. Biological Consequences Of Ecosystem R24 Fragmentation - A Review. Conservation Biology 5, 18-32. R25 Schaminée, J.H.J., Weeda, E.J., Westhoff, V., 1995. De Vegetatie van Nederland. Deel 2. Plantengemeenschappen van wateren, moerassen en natte heiden. Opulus Press, Uppsala/ R26 Leiden. R27 Schmidt, K., Jensen, K., 2000. Genetic Structure and AFLP Variation of Remnant Populations in the R28 Rare Plant Pedicularis palustris (Scrophulariaceae) and Its Relation to Population Size and Reproductive Components. American Journal of Botany 87, 678-689 R29 Schot, P.P., 1991. Solute transport by groundwater flow to wetland ecosystems. Utrecht University, R30 Utrecht. Sjors, H., Gunnarsson, U., 2002. Calcium and pH in North and Central Swedish Mire Waters. Journal R31 of Ecology 90, 650-657. R32 Smith, A.C., Koper, N., Francis, C.M., Fahrig, L., 2009. Confronting collinearity: Comparing methods R33 for disentangling the effects of habitat loss and fragmentation. Landscape Ecology 24, 1271-1285. R34 Soomers, H., Winkel, D.N., Du, Y., Wassen, M.J., 2010. The dispersal and deposition R35 of hydrochorous plant seeds in drainage ditches. Freshwater Biology. DOI: 10.1111/j.1365- 2427.2010.02460.x. In press. R36 Soons, M.B., 2006. Wind dispersal in freshwater wetlands: Knowledge for conservation and R37 restoration Applied Vegetation Science 9, 271-278. R38 Soons, M.B., Heil, G.W., 2002. Reduced colonization capacity in fragmented populations of wind- dispersed grassland forbs. Journal of Ecology 90, 1033-1043. R39 Habitat fragmentation and abiotic factors | 49

Suding, K.N., Collins, S.L., Gough, L., Clark, C., Cleland, E.E., Gross, K.L., Milchunas, D.G., Pennings, S., 2005. Functional- and abundance-based mechanisms explain diversity loss due to N R1 fertilization. Proceedings of the National Academy of Sciences of the United States of R2 America 102, 4387-4392 R3 Tilman, D., May, R.M., Lehman, C.L., Nowak, M.A., 1994. Habitat destuction and the extinction debt. Nature 371, 65-66. R4 Topografische Dienst, 2005. Topografische ondergrond (c), Emmen. R5 van den Broek, T., van Diggelen, R., Bobbink, R., 2005. Variation in seed buoyancy of species in wetland ecosystems with different flooding dynamics. Journal of Vegetation Science 16, R6 579-586. R7 Van der Meijden, R., 1996. Heukels’ Flora van Nederland Wolters-Noordhoff bv., Groningen. R8 van Diggelen, R., Molenaar, W.J., Kooijman, A.M., 1996. Vegetation succession in a floating mire in relation to management and hydrology. Journal of Vegetation Science 7, 809-820. R9 Vitousek, P.M., Mooney, H.A., Lubchenco, J., Melillo, J.M., 1997. Human domination of Earth’s R10 ecosystems. Science 277, 494-499. Wassen, M.J., Barendregt, A., 1992. Topographic Position And Water Chemistry Of Fens In A Dutch R11 River Plain. Journal of Vegetation Science 3, 447-456. R12 Wassen, M.J., Olde Venterink, H., Lapshina, E.D., Tanneberger, F., 2005. Endangered plants persist R13 under phosphorus limitation. Nature 437, 547-550. Whittaker, R.J., Nogués-Bravo, D., Araújo, M.B., 2007. Geographical gradients of species richness: A R14 test of the water-energy conjecture of Hawkins et al. (2003) using European data for five R15 taxa. Global Ecology and Biogeography 16, 76-89. R16 R17 R18 R19 R20 R21 R22 R23 R24 R25 R26 R27 R28 R29 R30 R31 R32 R33 R34 R35 R36 R37 R38 R39 50 | Chapter 2

R1 aPPendIx R2 The means and standard deviations for each variable. The ‘fragmentation variables’ are specific for R3 each species. A:P=Area: perimeter ratio of habitat (m²/m), DNP=Distance to nearest population (m), R4 AREA=Area of habitat (m²), NRPOP=Number of other populations within 500 m of the sampling location. Chapter 3 R5 Species Variable Mean sd R6 All Water level (cm below mowing field) 11.14 10.85 R7 All Ca (mg/l) 40.47 27.68 R8 All Cl (mg/l) 33.39 33.87 R9 All K (mg/l) 1.33 1.63 All PO (mg/l) 0.87 1.22 R10 4 All P (mg/l) 0.087 0.116 R11 All SO4 (mg/l) 26.44 21.99 R12 All Fe (mg/l) 3.29 5.29 All NO (mg/l) 0.64 0.81 R13 3 All NH (mg/l) 0.73 1.32 R14 4 All Distance to hill ridge (m) 1759.65 1013.47 R15 Carex diandra A:P 5.24 7.08 R16 Carex diandra DNP 753.16 896.01 R17 Carex diandra AREA 4824.66 8456.47 R18 Carex diandra NRPOP 1.348 1.718 Carex lasiocarpa A:P 5.4186 7.0135 R19 Carex lasiocarpa DNP 839.03 1150.54 R20 Carex lasiocarpa AREA 4572.26 8082.25 R21 Carex lasiocarpa NRPOP 1.98 2.06 R22 Juncus subnodulosus A:P 6.053 6.905 Juncus subnodulosus DNP 214.79 377.90 R23 Juncus subnodulosus AREA 6035.91 10740.41 R24 Juncus subnodulosus NRPOP 7.57 4.02 R25 Equisetum fluviatile A:P 8.772 17.176 R26 Equisetum fluviatile DNP 157.73 113.43 Equisetum fluviatile AREA 103 197.73 455 205.46 R27 Equisetum fluviatile NRPOP 12.696 6.651 R28 Menyanthes trifoliata A:P 7.804 14.121 R29 Menyanthes trifoliata DNP 505.52 641.66 R30 Menyanthes trifoliata AREA 93 159.76 32 3081.98 R31 Menyanthes trifoliata NRPOP 3.52 3.08 Pedicularis palustris A:P 5.044 6.999 R32 Pedicularis palustris DNP 478.08 543.49 R33 Pedicularis palustris AREA 4308.40 8043.89 R34 Pedicularis palustris NRPOP 1.54 1.74 R35

R36 R37 R38 R39 Chapter 3

Factors influencing the seed source and sink functions of a floodplain nature reserve in the netherlands

Soomers, H., Sarneel, J.M., Patberg, W., Verbeek, S.K., Verweij, P.A., Wassen, M.J., and Van Diggelen, R. 2011. Journal of Vegetation Science 22: 445-456. 52 | Chapter 3

R1 abstraCt R2 R3 Question: How do species traits and abiotic factors influence the extent of hydrochorous R4 dispersal into and out of a small floodplain area along a free-flowing river in The Netherlands? R5 R6 Location: The Kappersbult nature reserve (53º07’28”N, 6º37’14”E), which is a floodplain along the Dutch river Drentsche Aa. R7 R8 Methods: Seeds transported by the river were collected in fine mesh nets for 24 consecutive R9 hours once or twice a week for 1 year, upstream and downstream of the studied floodplain. R10 Data on the captured seeds were related to species traits and abiotic factors and species R11 composition in the floodplain. R12 R13 Results: The floodplain functioned both as a seed source and sink. High levels of river water R14 seemed to promote seed transport to or from the floodplain. Seeds of river bank species R15 occurred significantly more often in the river water than expected. Net source species had R16 significantly higher seed production, taller stature and higher seed buoyancy, but lower site R17 elevation than net sink species. Seed weight was significantly higher for sink species than for R18 other species. R19 Conclusion: Our study found that inundation, and therefore more natural river water R20 management, is a prerequisite for seed transport to and from a floodplain. The restoration R21 of target floodplain vegetation may be successful for common species that produce many R22 seeds and grow in proximity to the river. Consequently, it is expected that the probability R23 of restoring vegetation types that occur further from the river, such as wet grasslands, by R24 hydrochorous dispersal is low. R25 R26 Nomenclature: Van der Meijden (2005) R27 R28 R29 R30 R31 R32 R33 R34 R35 R36 R37 R38 R39 Seed source and sink functions of a floodplain | 53

3.1 IntroduCtIon R1 R2 European river basins have been used intensively for centuries. The canalization and regulation R3 of rivers and the reclamation and eutrophication of floodplains have resulted in a decline in R4 water quality and biodiversity (Jensen et al. 2006, Nienhuis et al. 2001). Recently, the focus R5 with respect to river management has changed to a more natural approach, in anticipation R6 of the EU Water Framework Directive (Nienhuis et al. 2002). In addition, the implementation R7 of the Dutch policy directive ‘Room for the River’ (Baan et al. 2004), which is aimed at R8 mitigating the effects of peak discharges of major rivers in the future, requires a shift in river R9 management policy. The forthcoming implementation of these legislative measures requires profound knowledge of the functioning of ecosystems in riverine habitats and provides the R10 opportunity to restore floodplain habitats. R11 Although attempts to restore habitats are often successful, characteristic target species are R12 not necessarily re-established (Grootjans et al. 2002, Jensen et al. 2006). A major constraint R13 in the re-establishment of target plant communities is the limited availability of propagules R14 (Ehrlen et al. 2000, Bischoff 2002, Boedeltje et al. 2003, Ehrlen et al. 2006, Jansson et al. 2007). R15 For instance, Ozinga et al. (2005) concluded from their study that poorer dispersers were R16 under-represented in suitable habitat patches, in contrast to plants with a high capacity for R17 long-distance dispersal. R18 For riparian plant seeds, surface water is an important dispersal vector (e.g. Schneider et al. R19 1988, Nilsson et al. 1991, Goodson et al. 2003, Boedeltje et al. 2004, Gurnell et al. 2006, Soomers R20 et al. 2010) and hydrochory (i.e. water dispersal) is especially important for the restoration of R21 plant communities in river valleys (Rosenthal 2006). Furthermore, Van den Broek et al. (2005) R22 found a positive correlation between the flooding probability of plant communities and the average seed buoyancy in species from these communities. Consequently, the dispersal R23 range of riparian plant species is much greater than when dispersal takes place exclusively R24 by anemochory (i.e. wind dispersal) (Boedeltje et al. 2003). The discharge dynamics of river R25 water, and more specifically, flood pulses (Junk et al. 1989) are considered to be important R26 factors that influence hydrochory (Moggridge et al. 2009). For riverine plant communities, the R27 importance of flood pulses has been stressed by Middleton (1999), Tockner et al. (2000) and R28 Boedeltje et al. (2004). Similarly, Jansson et al. (2005) found that flooding had a positive effect R29 on plant biodiversity along the river. R30 When focusing on hydrochory in floodplains, two different processes must be distinguished; R31 (1) seed inflow to and deposition in a floodplain (seed sink function), and (2) outflow from R32 a floodplain of seeds produced there (seed source function). Both processes are important R33 for the maintenance of populations of riparian plant species that occur in a metapopulation R34 along a river. In this paper, seed source and seed sink functions only refer to seed-related R35 dispersal processes, not to source or sink populations as described within the metapopulation theory (Hanski 1999). R36 R37 R38 R39 54 | Chapter 3

R1 Only a few of the numerous studies that have investigated hydrochory relate species traits R2 to the presence and abundance of species in seed samples that have been captured in a river R3 (Boedeltje et al. 2003, Vogt et al. 2006, Gurnell et al. 2008). Moreover, these studies did not R4 take into account the spatial location of possible source plants, nor did they analyse the R5 relation between river water levels and the transport of seeds. The study reported herein R6 had as its aim investigation of the influence of species traits and abiotic factors on the extent of hydrochorous dispersal into and out of a floodplain area along a free-flowing river in The R7 Netherlands. R8 In contrast to other studies that focus either on diaspore deposition at river margins or R9 floodplains (e.g. Merritt et al. 2006, Vogt et al. 2007) or on the quantification of the flow of R10 propagules through a river (e.g. Boedeltje et al. 2003, Boedeltje et al. 2004), our experimental R11 approach and the spatial configuration of the floodplain allowed us to distinguish between R12 seed inflow and outflow and to evaluate separately the species traits and abiotic factors that R13 influence each process. R14 The main research questions addressed in this study are: (1) does the considered floodplain R15 function as a net seed source or seed sink, and (2) which plant traits and growing location R16 characteristics are related to these two different processes? We examined the species traits R17 seed buoyancy, seed weight, seed-shedding season, plant height and seed production in the R18 floodplain and the abiotic factors river water level, soil elevation and distance to the river of R19 source plant locations. Our study focuses on herbaceous species only. In the study area, except for riparian vegetation, other types of floodplain vegetation are R20 absent upstream of the floodplain under consideration. Therefore, we hypothesize that the R21 floodplain functions as a net source of seeds towards downstream areas for plants that inhabit R22 the studied floodplain. R23 R24 R25 3.2 metHods R26 R27 3.2.1 Site description R28 The 27 ha Kappersbult nature reserve (53º07’28”N, 6º37’14”E) is a single separate floodplain R29 area along the small Dutch river Drentsche Aa (see Figure 3.1). Upstream of the study site, the R30 river is embanked, without presence of floodplains. The river banks consist of sandy vegetated R31 levees. Vegetation in the floodplain ranges from highly productive vegetation types [tall R32 sedges (Caricion gracilis), reed canary grass marshes (Phalaris arundinacea community) and reed manna grass marshes (Glyceria maxima community)] close to the river, through small R33 sedge vegetation (Caricion nigrae), to low productive litter meadow (Junco molinion) and R34 grassland (Calthion palustris) furthest from the river (Bakker 1987, Klimkowska et al. 2009). R35 The average discharge of the river is 1.95 ± 1.85 m³s-¹ (± SD; 1998-1999) and the water level R36 0.51 m to 1.47 m (average 0.65 ± 0.088 m) above mean sea level (a.s.l.) (1998). River water R37 levels do not fluctuate strongly as a consequence of downstream water regulation. R38 R39 Seed source and sink functions of a floodplain | 55

R1 R2 R3 R4 R5 R6 R7 R8 R9 R10 R11 R12 R13 R14 R15 R16 R17 R18 Figure 3.1. The location of the research site in The Netherlands, the distribution of vegetation types and location of the nets within the research site. R19 R20 R21 3.2.2 General approach R22 To quantify the number of seeds transported by the river, seeds were collected in fine mesh nets for 24 consecutive hours once or twice a week, during 1 year, upstream and downstream of the R23 studied floodplain nature reserve. These numbers and the species caught were compared with R24 the estimated numbers of seeds produced in the floodplain, and with the species composition R25 of the floodplain. The sampling design enabled us to calculate the net seed inflow or outflow. R26 We define the net seed inflow, or seed sink function, of the floodplain as the net inflow of R27 seeds from the river into the floodplain (number of seeds captured upstream > number of R28 seeds captured downstream). Similarly, we define the net seed outflow, or source function, of R29 the floodplain as the net outflow of seeds from the floodplain into the river (number of seeds R30 captured downstream > number of seeds captured upstream). However, it should be noted R31 that seeds could also remain in the river channel between the upstream and downstream nets R32 (Gurnell et al. 2007). R33 Species traits, such as seed weight and seed buoyancy, and spatial variables such as the average R34 distance of plant sites to the river, were compared between species for which the floodplain R35 functioned as a (net) seed source (hereafter referred to as ‘(net) source species’) and other species, and also between species for which the floodplain functioned as a (net) sink and other R36 species. Furthermore, the river water level was related to the net outflow of viable seeds that R37 were captured in every 2-wk period. R38 R39 56 | Chapter 3

R1 3.2.3 Seeds transported by the river R2 Trapping of seeds transported by the river R3 Fine mesh nets (mesh size: 150 x 150 µm) fixed to a 53 x 50 cm (53 x 25 cm subsurface) wooden R4 frame with attached wooden floats were used to collect waterborne seeds. Note that the nets R5 only sampled the water column near the water surface. Three nets were placed upstream R6 (south) and three downstream (north) of the nature reserve, equidistant from each other and the riverbanks (see Figure 3.1). At the northern location, the river was 18.8 m wide, and at the R7 southern location, 14.4 m. The two seed capture locations were 1282 m apart. Between June R8 and November 1998, seeds were trapped twice a week by putting nets in the river on Monday R9 and Thursday and taking them out 24 h later. Between December and May 1999, we sampled R10 once a week because we expected lower seed numbers in winter. During the latter period we R11 sampled on Monday, except for 3 instances, when we sampled on Tuesday. R12 R13 Identification of trapped seeds R14 The collected nets were rinsed and the captured seeds were stratified at 4 ºC for at least 1 R15 month under moist conditions. After stratification, samples were sown on wet sterile soil that R16 was covered with a thin layer of sterile sand and placed in a greenhouse at temperatures of 25 R17 ºC during the day and 15 ºC at night. A regime of 12 h light and 12 h darkness was maintained. R18 Seeds that germinated were removed after they had been identified. This procedure was R19 continued until no new seedlings emerged for at least 4 weeks. Not all seedlings could be identified to species level. R20 R21 Calculation of total annual numbers of seeds transported by the river R22 The total number of captured viable seeds was transformed into annual numbers of viable R23 seeds per species transported across the entire river width at the two locations, using equation R24 1. R25 R26 Stot,s,a = Sc,s,a · (tj /tc ) · (wr,a /wn) eq. 1 R27

R28 In equation 1, Stot,s,a represents the annual number of seeds for species s at location a. Sc,s,a

R29 represents the total number of captured (c) seeds for species s at location a , tj the number of

R30 minutes in a year (525600), tc the number of minutes for which these seeds were captured, wr,a the width of the river at location a and w the total width of the three nets (1.59 m). R31 n R32 3.2.4 Species traits and abiotic explanatory variables R33 Timing of seed-shedding and river water level R34 To relate seed transport by the river to the seed-shedding period and river water level, for all R35 species in the river and the floodplain, seed-shedding periods were extracted from a database R36 (Kleyer et al. 2008). Furthermore, The Hunze and Aa regional Water Board provided data on R37 river water levels, measured downstream of the Kappersbult (Fig. 3.2). R38 R39 Seed source and sink functions of a floodplain | 57

Species traits and spatial variables R1 To analyse the potential differences in species traits and spatial variables between (net) source R2 and (net) sink species, data on seed weight (Flynn et al. 2006), minimum plant height (van R3 der Meijden 2005) and seed buoyancy (Boedeltje et al. 2003, van den Broek et al. 2005) were R4 derived from the literature. We used the buoyancy data in Boedeltje et al. (2003). Buoyancy R5 data for additional species were derived from Van den Broek et al. (2005) by standardizing R6 these values to the data set of Boedeltje et al. (2003) using a regression analysis. R7 To gain insight in the species composition and seed production in the floodplain, vegetation R8 relevés (0.9 x 0.9 m) were taken in 10 randomly selected representative subplots in each R9 community. For a detailed description of the methodology used to sample the vegetation and to estimate the seed production, see Klimkowska et al. (2009). We estimated the number of R10 seeds per species in each plot and multiplied the estimated number of seeds per vegetation R11 type with the area of each vegetation type to estimate the number of seeds per species R12 that were produced in the floodplain (Klimkowska et al. 2009). No vegetation relevés were R13 undertaken in the forested patches, and therefore, trees and woody species were excluded R14 from this analysis. R15 Using GIS maps with the spatial location of vegetation types and the location of the river, R16 and the data on estimated numbers of seeds produced per species per vegetation type in the R17 floodplain, the weighted average and weighted average minimum distance from the growing R18 location to the river were calculated for each plant species present in the floodplain. Similarly, R19 the weighted average, weighted average minimum and weighted average maximum soil R20 elevation of the growing location were calculated for each species using an elevation raster R21 map (van Heerd et al. 2000). R22 The average water level of the Drentsche Aa river was 0.65 m a.s.l To investigate the effect of inundation on the number of waterborne seeds, the estimated number of seeds produced R23 at sites that would be inundated when the water level rises to 0.8 m was calculated using R24 an elevation raster map and the gathered information on seed production per species per R25 vegetation type. We chose a height of 0.8 m a.s.l. as the river water level at which overbank R26 flooding starts to occur. Our choice was based on the fact that about 10% of the river R27 bank is lower than this height, assuming significant overbank flooding at this water level. R28 Furthermore, approximately 25% of the floodplain area is below this level. R29 R30 3.2.5 Data analysis R31 The floodplain as a net seed source or sink? R32 We tested all data for normality and homogeneity of variance. If assumptions of parametric R33 statistical tests were violated, we used a non-parametric test. R34 Total numbers and species of seeds that were captured upstream and downstream were R35 compared with species growing in the floodplain. For species that were found in the floodplain and exclusively in the downstream nets (as seeds), the percentage of the seeds produced in R36 the floodplain in 1 year that reaches the river water was assessed using the calculated number R37 of produced seeds in the floodplain and the calculated total number of seeds transported R38 R39 58 | Chapter 3

R1 through the river downstream of the floodplain. The number of seeds trapped upstream was R2 related to the number of seeds trapped downstream, for each species, using a Spearman R3 correlation test. R4 R5 Timing of seed-shedding and river water level R6 To test the hypothesis that more seeds will reach the river water in the seed-shedding season of the considered species, for both the upstream and downstream locations, the average number R7 of viable seeds per month found in the nets in the seed-shedding period of a considered R8 species was compared with the average number of seeds per month in the period in which no R9 seeds are shed by this species (Wilcoxon test). R10 To relate the river water level to number of seeds trapped upstream or downstream and to R11 net seed outflow, 2-week averages of river water level and 2-week total number of captured R12 viable seeds were used. For these analyses, the number of captured viable seeds for the R13 different species was summed in order to derive the total number of captured viable seeds. R14 In contrast to the species-specific dataset, data for this analysis were normally distributed. To R15 investigate the effect of river water level on the net inflow/outflow of viable seeds, a Pearson R16 correlation test was used. R17 R18 Species traits and spatial variables R19 To investigate the effect of species traits and spatial characteristics of their growing location on the net seed inflow or outflow for herbaceous species, four issues were addressed, each R20 with a corresponding analysis. Different selections of species (excluding the woody species R21 Alnus spp, Betula pubescens, Crataegus monogyna, Salix alba and Salix cinerea), were made R22 for the purpose of each analysis, according to the location(s) of seed capture and the presence R23 or absence of the plant species in the floodplain (see Table 3.2). R24 We investigated (1) which species traits and spatial variables determine whether a species that R25 is growing in a floodplain enters the river water; (2a) which species traits and spatial variables R26 determine whether a species that is growing in the floodplain acts as a net seed source species; R27 (2b) which species traits and spatial variables determine whether a species that is growing in R28 the floodplain and interacting with the river acts as a net seed source or sink species; and (3) R29 which species traits determine the inflow of seeds to the floodplain from the river. R30 In each of the analyses, the differences in species traits and spatial variables were tested R31 between the two groups ((net) sink and (net) source species) mentioned in Table 3.2, using a R32 Mann-Whitney U-test. R33 1: Exclusive source analysis R34 Considering analysis 1, a selection of the total data set, including only species that occur in R35 the Kappersbult and were not found in the upstream nets was made. Within this selection, R36 the species traits and spatial variables were compared between species not found in the R37 downstream nets (group a) and species that were found in the downstream nets (group b). It R38 is assumed that only species from group b were able to enter the river water. R39 Seed source and sink functions of a floodplain | 59

2a/2b: Net source/sink analysis R1 The selection for analysis 2a comprised all species that were present in the floodplain, within R2 which species that were more abundant in the downstream nets than upstream (net source R3 species, group a) were compared with the species for which this did not hold (group b). In R4 analysis 2b, the same selection and groups as in analysis 2a was used, except for the exclusion R5 of species that were not found in any net in this analysis. R6 R7 3: Exclusive sink analysis R8 Analysis 3 considered a selection of species that did occur in the upstream nets, but were not R9 found in the floodplain. In this analysis, variables for species not found downstream (‘exclusive sink species’, group a) were compared with those for species found in both nets (group b). R10 Because the species selected in this analysis did not grow in the floodplain, spatial variables R11 were omitted for this analysis. R12 R13 R14 3.3 results R15 R16 3.3.1 The floodplain as a net seed source or sink? R17 Total numbers of seeds and species that were captured upstream and downstream and R18 identified in the floodplain are given in Table 3.1. Table S1 in the Appendix lists the identified R19 species that were found either in the vegetation of the Kappersbult or in at least one of R20 the nets. The computed annual number of viable seeds floating along the total width of R21 the river was rather similar at the downstream and upstream locations (Table 3.1). From the R22 viable seeds caught in the river during the year-round experiment, 76 different species could be identified to species level. As may be seen in Tables 3.1 and S1, 45 species were found in R23 both nets, 18 only upstream and 13 only downstream. Of the 13 species that were found R24 exclusively in the downstream nets, three (Glyceria maxima, Carex nigra and Typha latifolia) R25 occurred in the Kappersbult vegetation (see Table S1). For Carex nigra, 0.00264% and for R26 Glyceria maxima, 0.000022% of the seeds produced in the Kappersbult reached the river water R27 downstream of the area and remained viable. No estimation of seed production was available R28 for Typha latifolia. The most abundant herbaceous species in the upstream nets were Juncus R29 bufonius, Poa trivialis and Juncus effusus, while Epilobium hirsutum and Lycopus europaeus R30 were the most abundant species in the downstream nets (see Appendix). A total of 55.3 % R31 of the captured species were found more often in the upstream than in the downstream R32 nets. There was a significant correlation between the number of seeds per species captured R33 in the upstream location and captured downstream (Spearman correlation coefficient=0.541, R34 P<0.001). Nevertheless, many species exhibited distinct net inflow or outflow rates (see Table R35 S1). R36 R37 R38 R39 60 | Chapter 3

table 3.1. Figures on calculated annual number of seeds and number of species captured in upstream R1 and downstream nets and growing in the Kappersbult. R2 Computed Number of Number of identified Number of identified R3 annual identified species also growing species exclusively at R4 number of species in the Kappersbult the location described seeds in the first column R5 Seeds captured in 111 429 63 28 18 * R6 upstream nets R7 Seeds captured in 103 038 58 25 13 * downstream nets R8 Total seeds captured in 214 467 76 R9 nets R10 Species’ seeds found in 45 22 R11 both nets 6 R12 Species growing in /seeds 25 117·10 63 32 produced in Kappersbult R13 R14 * some species also occur in the Kappersbult R15 R16 3.3.2 Timing of seed-shedding and river water level R17 Figure 3.2 shows that during the seed-shedding season, most of the time river water levels R18 were too low to allow overbank flooding of the floodplain; except for one flooding event R19 that began at the end of October and lasted until the beginning of November. There was no significant difference between the average monthly number of seeds found in R20 the nets during the seed-shedding season of each species and the number of seeds outside R21 the seed fall season, for both the upstream as for the downstream nets (Wilcoxon, P= 0.238 R22 and 0.924 respectively). R23 The 2-week average river water level did not correlate significantly with either the 2-week R24 number of seeds trapped upstream or downstream (Pearson, P=0.558 and P=0.925 respectively), R25 or the net seed outflow (P=0.391). Similarly, the 2-week maximum river water level was not R26 correlated significantly with the number of seeds trapped upstream, downstream or net seed R27 outflow (Pearson,P =0.950, P=0.997 and P=0.933 respectively). However, Fig. 3.2 illustrates R28 that the third peak in river water level, which began at the end of October and lasted until R29 the beginning of November (i.e. towards the end of the seed shedding season), coincides with R30 a clear net inflow of seeds to the floodplain, whereas higher river water levels in February R31 and March coincide with a net outflow of seeds. For lower water levels, the net seed outflow R32 was closer to zero. R33 3.3.3 Species traits and spatial variables R34 Exclusive source analysis R35 Seed production at lower sites of the floodplain (< 0.8 m a.s.l.) was significantly higher for R36 exclusive source species (see Table 3.2, analysis 1), than for species for which no viable seeds R37 entered the river water from the floodplain (Mann-WhitneyU , P=0.011, N=27, table 3.2). R38 R39 Seed source and sink functions of a floodplain | 61

A trend to higher seed production in the floodplain for absolute source species compared R1 to non-hydrochorous species was found (Mann-Whitney U, P=0.051, N=27, table 3.2). Other R2 variables did not differ significantly among the two groups. R3 R4 R5 R6 R7 R8 R9 R10 R11 R12 R13 R14 R15 R16 R17 R18 R19 R20 R21 R22 R23 R24 R25 Figure 3.2. River water level in 1998 (June-December) and 1999 (January-May) of the River Drentsche R26 Aa (line, right y-axis), percentage of species that release seeds in the considered months (bars, left y-axis) and net seed outflow divided by 100 (seeds downstream minus seeds upstream) (black dots, R27 left y-axis; each dot represents the total number of seed outflow in a 2-week period, river-wide). R28 Water level data for December were not available. R29 R30 R31 R32 R33 R34 R35 R36 R37 R38 R39 62 | Chapter 3

R1 R2 R3 R4 =0.011 =0.015 =0.003 =0.013 =0.038 =0.01 =0.003 =0.011

R5 P P P P P P P P R6 Result a>b, a>b, a>b, a>b, ab, a>b, a>b, R7 R8 R9 R10 =All species absent at this location are ○ R11 b. a with group test, comparing group u R12

R13 hitney-

R14 w an

R15 m Variable -Seed production at locations lower than 0.8m -Seed production: -Seed production at locations lower than 0.8m -Minimum plant height -Minimum soil elevation -Buoyancy -Minimum plant height -Seed weight R16 R17 R18 R19 R20

R21 b roup g Species not found in the downstream nets. Species not found more often downstream Species found more upstream than downstream (net sink species) Species found in both nets (hydrochoric species from upstream) R22 R23 R24 R25 R26 R27 =All species found at this location are included in the selection,

R28 ● roup a roup

R29 g Species found in the downstream nets (exclusive source species) Species found downstream more often than upstream (net source species) more often than upstream (net source species) Species not found in the downstream nets (exclusive sink species) R30 =significance value. R31 P * * Species found downstream R32 election FUD ● ○ ● ● ○ ● R33 s R34 R35

R36 seeds found as species of selections for four species, b) of (a and groups two variables between spatial traits and species of Comparison R37

R38 ame of analysis able 3.2. able n 1: Exclusive source 2a: Net source/sink 2b: Net source/sink 3: Exclusive sink † t Bullets nets. D=downstream nets, U=Upstream F=Floodplain, selection: the in considered Locations floodplain. the in plants as and/or river the in refer to criteria for selection of species: included in the selection. *Species not found river at all were excluded from If two criteria are used, both first and second apply to the selection. R39 † Spatial variables were omitted in analysis 3, because these species absent the floodplain Seed source and sink functions of a floodplain | 63

Net source/sink analyses R1 Considering all species growing in the floodplain separately, a Mann-WhitneyU -test yielded a R2 significantly higher estimated number of seeds produced in the nature reserve for net source R3 species (see Table 3.2, analysis 2a) than for net sink and non-hydrochoruous species (P=0.015, R4 N=49). Furthermore, the minimum plant height and number of seeds produced at lower R5 sites were significantly higher for net source species than for net sink and non-hydrochorous R6 species (P=0.013, N=63 and P=0.003, N=49, respectively). There was a trend towards a smaller R7 minimum distance of the species’ location to the river and towards a lower minimum soil R8 elevation for net source species (P=0.052, N=49 and P=0.059, N=49 respectively). This was R9 confirmed by a Chi-square test, which revealed that species that are found at the river bank of the floodplain had a positive net seed outflow significantly more often, and species that are R10 found in the floodplain and not at the river bank significantly less often than expected (Chi- R11 square, p=0.012, N=63). A trend to a higher buoyancy for net source species was found (Mann- R12 Whitney U, P=0.093, N=47). Other variables did not differ significantly among the two groups. R13 For analysis 2b, species that were found in the Kappersbult, but not in one of the nets, were R14 excluded from the selection (see Table 3.2, analysis 2b). For this selection, minimum plant R15 height and seed buoyancy were significantly higher for net source than for net sink species R16 (Mann-Whitney U, P=0.003, N=31 and P=0.01, N=26, respectively). Minimum soil elevation was R17 significantly lower for net source species and there was a trend towards a smaller minimum R18 distance to the river for net source species (P=0.038, N=24 and P=0.081, N=24, respectively). R19 Other variables did not differ significantly among the two groups. R20 R21 Exclusive sink analysis R22 For species not growing in the floodplain, but found in the upstream nets, seed weight was significantly higher for exclusive sink species (see Table 3.2, analysis 3) than for species R23 found in both nets (Mann-Whitney U, P=0.009, N=33). A trend towards a lower buoyancy for R24 absolute sink species was found (Mann-Whitney U, P=0.098, N=17). Minimum plant height did R25 not differ significantly among the two groups. R26 R27 R28 3.4 dIsCussIon R29 R30 3.4.1 Net seed source or sink? R31 Our study considers one relatively small separate floodplain area, situated along a part of R32 an embanked river. Although this setting has limitations for the general applicability of our R33 results, it enables us to distinguish the specific contribution of the floodplain as a seed source R34 for downstream areas as well as the specific function of the floodplain in capturing seeds R35 carried by the river. The results do not support the initial hypothesis that the Kappersbult floodplain functions mainly as a seed source. The floodplain appears to function as either a R36 seed source or as a sink, depending on the species and conditions. However, for at least half R37 of the species in the dataset (55%), the Kappersbult functions as a net seed sink; the evidence R38 R39 64 | Chapter 3

R1 for this claim being that the total number of viable seeds for these species was lower in the R2 river downstream of the floodplain than upstream. Furthermore, the number of species found R3 in the upstream nets was higher than in the downstream nets. R4 The herb species that were most abundant in the downstream nets also occurred in the R5 floodplain vegetation and the upstream nets. Species that were abundant in the nets are R6 common species that are not confined to floodplains, but can also grow on river levees. However, for species that occur in small sedge and wet grassland (Litter meadow and R7 Grassland) vegetation in the Kappersbult, exchange of seeds between floodplain and river R8 seems much less common. Only the small sedge Carex nigra, which produced many seeds R9 in the floodplain, was found in the downstream nets, whereas the small sedgesC. curta, C. R10 panicea and C. aquatilis, and the wet grassland species Silene flos-cuculi and Caltha palustris R11 occurred in neither the downstream nor the upstream nets. These results indicate that, R12 although flooding of the Small sedge, Litter meadow and Grassland vegetation is possible, R13 the effect of hydrochory on the vegetation composition of floodplain species that do not R14 grow at riverbanks is very limited. This is in agreement with the results of Bissels et al. (2004), R15 who conclude that restoring more natural flooding conditions did not result in the recovery of R16 species richness of alluvial grasslands in Germany. For seeds of more distant floodplain species, R17 the coupling of anemochory (primary dispersal) and hydrochory (secondary dispersal) could R18 result in long-distance dispersal. However, if this strategy were very successful in dispersing R19 seeds via river water, we would expect to find a higher number of seeds of these species in the downstream nets than we did. R20 Thirteen species were found exclusively in the downstream nets, but of these only three R21 occurred in the floodplain. This may indicate that the methods used to capture seeds were not R22 as efficient as they might have been, with the result that a number of species were not collected R23 in the upstream nets. Alternatively, these species (a) may be present in the Kappersbult but R24 may have been overlooked by us while mapping species or (b) may be present along the river R25 bank opposite the floodplain in the stretch between the upstream and downstream nets. R26 These various possibilities illustrate the general uncertainties related to studies of seed R27 dispersal. Seed dispersal is a stochastic process and, whatever methods are used for capturing R28 seeds, in most cases only a small proportion of the seeds can be captured. This holds especially R29 for wind dispersal studies, but also applies to our case, in which we could not sample the R30 entire width of the river continuously for a year. Hydraulic variation within the river could R31 prevent uniform distribution of seeds across the channel. This could mean that the annual R32 calculations of seed numbers across the entire river width could be an over- or under- estimation, depending on the exact manifestation of this variation. Moreover, extrapolating a R33 process that is stochastic over time also brings the risk that either rare species are not found, R34 or that rare events/species that are coincidentally sampled are getting an unrealistically high R35 weight as a result of extrapolation. Furthermore, the nets with the small mesh size (used to R36 make sure that even very small seeds were captured) can be blocked by drifting material, as R37 a result of which the effective trapping period may have been shorter than 24 h (Boedeltje R38 et al. 2003). Therefore, we note that caution must be taken when drawing conclusions R39 Seed source and sink functions of a floodplain | 65

from seed capture studies such as ours, because of sampling constraints and stochasticity. In R1 particular, caution should be observed when drawing conclusions from the calculated annual R2 seed numbers; these should be seen as rough estimates rather than precise figures. However, R3 our sampling design was dense compared to other studies; we sampled about 10% of the R4 river width, frequently, for one year, while many other studies only considered samples taken R5 over a much shorter period of time, less frequently (Goodson et al. 2003, Moggridge et al. R6 2009, Moggridge & Gurnell 2010) or even only after one flooding event (Cellot et al. 1998, R7 Andersson et al. 2000, Vogt et al. 2006, 2007). R8 In summary, we conclude that floodplains can function both as seed source and sink, but that R9 source-sink dynamics with accompanying genetic exchange between sub-populations might only play a visible role for common species growing on river banks or dikes. R10 R11 3.4.2 Timing of seed-shedding and river water level R12 In our study, the high water level peak in autumn, shortly after the seed fall season, was related R13 to net seed deposition in the floodplain. However, high water levels in winter (February- R14 March) corresponded to a clear net seed outflow, rather than seed deposition. In relation R15 to the former effect, other studies report a considerable amount of overall seed deposition R16 after flooding (Vogt et al. 2004, 2006). We found no direct evidence in the literature for the R17 latter effect (e.g. source effect). However, Boedeltje et al. (2004) found indirect evidence in R18 the form of a positive relation between mean number of species and diaspores captured in a R19 channel and high water discharge levels. Further, the river flow regime according to the time R20 of the year has been reported to influence hydrochory by, for example, Boedeltje et al. (2004), R21 Gurnell et al. (2006) and Moggridge & Gurnell (2010). R22 We suggest that seed transport to or from a floodplain is initiated by flooding, but that the main direction of the transport (net source or net sink) depends on the time of the year. In R23 herbaceous wetlands, vegetation roughness is probably greater in autumn, just after the seed R24 fall season, than in late winter, when most plants have decayed. Therefore, we hypothesize R25 that during late winter, seeds lying on the soil surface might reach the river more easily R26 after inundation. In autumn, most seeds are captured in the vegetation, whether they were R27 deposited by the river or originated from the vegetation. The importance of vegetation R28 roughness in relation to hydrochory is also mentioned by Pollux et al. (2009). R29 Our results show that flooding events seem to promote both the deposition of seeds in R30 floodplains and the outflow of seeds from the floodplain. Hence, inundation is expected R31 to promote the ability to colonize, and therefore the viability of wetland plant species, by R32 connecting otherwise fragmented wetland patches. This implies that the regulation of river R33 water levels, which results in constant water level without flooding events, can diminish the R34 effectiveness of a river as a dispersal vector. This is supported by the flood pulse concept (Junk R35 et al. 1989, Middleton 1999), which states that the inundation of floodplains by a free-flowing river is a prerequisite for the restoration of riverine biodiversity. We note that the river in our R36 study area was a low-energy river. In contrast, in high-energy river systems, seeds might be R37 mobilized from the river bed within a river reach (Gurnell et al. 2008). Furthermore, in rivers R38 R39 66 | Chapter 3

R1 with higher energy, non-floating seeds could be transferred through the river channel or into R2 the floodplain by lateral turbulent exchanges. Therefore, a near surface sampling design such R3 as ours could miss important seed transfers in high-energy rivers. R4 R5 3.4.3 Species traits and spatial variables R6 Whether the floodplain functioned as a (net) seed source or sink for a certain species was influenced not only by phenology and flooding, but also by species traits and spatial variables. R7 The number of seeds produced in the floodplain, in particular seed production at lower sites, R8 was the main factor promoting (net) seed outflow. This is in line with the results of Boedeltje R9 et al. (2003), who found for riparian species a positive correlation between seed production R10 and the number of seeds trapped in a lowland stream. In addition, Peart (1989) found that R11 the relative abundance of species in seed traps in open grassland patches is determined mainly R12 by seed production. Thus, relatively high seed production is an important condition for seeds R13 reaching the river. However, when we focus on those species that were captured in the river, R14 seed production no longer appears as an important factor (see analysis 2b in Table 3.2). The R15 species groups compared in analysis 2b concern seeds that have reached the river, either from R16 outside the study area or from the Kappersbult. Thus, species with low seed production and R17 therefore a lower variation in seed production in the floodplain were probably excluded. In R18 this analysis, the spatial variables that appear important for net seed outflow are minimum R19 soil elevation and minimum distance to the river. We suggest that the positive influence of low soil elevation on seed source behaviour might be related to the higher incidence of inundation R20 at those sites. Furthermore, lower soil elevations occur at smaller distances to the river. Seeds R21 of species growing at larger distances from the river (and thus higher elevation) might have R22 greater probability of captured by the standing vegetation in the floodplain before reaching R23 the river. Lastly, species growing at low elevation close to the river are wetland species, which R24 are generally adapted to hydrochory (high buoyancy) (see van den Broek et al. 2005). This R25 might also explain why seeds originating from lower elevations are found in the downstream R26 nets more often than seeds originating from higher elevation sites. The trend towards a lower R27 minimum distance to the river for net source species stresses the above-suggested importance R28 of proximity to the river for seed outflow, which is also indicated by the fact that a significantly R29 greater number of the species that grow at the river bank than we had expected were net R30 source species. Important species traits for net seed outflow were minimum plant height and R31 buoyancy. The influence of plant height is probably due to the fact that wind acts as a primary R32 dispersal vector for seed transport to the river, because plant height is known to be a major factor influencing wind dispersal distance (e.g. Soons et al. 2004, Muller-Landau et al. 2008). R33 The significantly higher buoyancy for net source species than for net sink species is in line with R34 the results of Boedeltje et al. (2003), who found that, next to seed production, buoyancy was R35 the most important factor determining hydrochory for riparian species. R36 Analogously to this, exclusive sink species tended to have lower buoyancy than species R37 captured in both nets that were not found in the floodplain. This confirms the finding of R38 Chang et al. (2008) that seeds with very low buoyancy are retained in larger numbers by the R39 Seed source and sink functions of a floodplain | 67

vegetation than seeds with higher buoyancy. On the other hand, high buoyancy will increase R1 dispersal distance and thereby the probability that distant patches will be colonized (Pollux R2 et al. 2009). R3 The seeds of exclusive sink species were found to be significantly heavier than those of species R4 that occur in both nets and not in the floodplain, which indicates that heavier seeds are more R5 likely to be deposited in the floodplain. This might be related to the fact that heavier seeds R6 are often also relatively large. Schneider and Sharitz (1988) indeed found that vegetation R7 trapped larger water tupelo fruits more efficiently than it trapped smaller bald cypress seeds. R8 It should be noted that vegetative propagules were not taken into account in the current R9 study. Boedeltje et al. (2003) found that 95.8% of the propagules captured in a lowland stream had a vegetative origin. However, 87.1% of the seeds of trapped (semi-)terrestrial species R10 were generative diaspores (Boedeltje et al. 2003). Given that we were mainly interested in the R11 source-sink dynamics of a floodplain, which implies a focus on (semi-)terrestrial species, we R12 think that it is unlikely that the omission of vegetative propagules had an important effect R13 on the results. R14 In summary, high seed production appears to be a major factor increasing the probability R15 of seed outflow. However, a large number of seeds will not be available for the usually rare R16 target species. Therefore, in addition to seed production, buoyancy, the distance of the R17 growing location from, and its elevation in relation to, the river are expected to be the main R18 factors determining species’ ability to disperse via river water. This implies that it is unlikely R19 that sources of floodplain species that are distant from the river, for instance, low-productive R20 wet grasslands or litter meadows, will be successful in dispersing seed via river water. Further, R21 such grasslands do not have high potential as a sink area, because a lower frequency of R22 inundation of these sites reduces the chance that target seeds will be deposited at locations that are suitable for their germination. Thus, once suitable abiotic conditions are restored, R23 restoration of the biodiversity of floodplain communities that is facilitated by hydrochory can R24 be successful for communities that are located in floodplains adjacent to and connected to the R25 river, provided that the river water is not regulated. R26 R27 acknowledgements R28 The authors thank The Hunze and Aa regional Water Board for providing data on river R29 water level and discharge. Jos Verhoeven and two anonymous reviewers provided valuable R30 comments that improved the manuscript. R31 R32 R33 R34 R35 R36 R37 R38 R39 68 | Chapter 3

R1 reFerenCes R2 R3 Andersson, E., Nilsson, C. & Johansson, M.E. 2000. Plant dispersal in boreal rivers and its relation to the diversity of riparian flora.Journal of Biogeography 27: 1095-1106. R4 Baan, P.J.A. & Klijn, F. 2004. Flood Risk Perception and Implications for Flood Risk Management in R5 the Netherlands. International journal of river basin management 2: 1-10. Bakker, J.P., Brouwer, C., Van Den Hof, L., Jansen, A. 1987. Vegetational Succession, Management R6 and Hydrology in a Brookland (The Netherlands). Acta Botanica Neerlandica 36: 39-58. R7 Bischoff, A. 2002. Dispersal and establishment of floodplain grassland species as limiting factors in R8 restoration. Biological Conservation 104: 25-33 Bissels, S., Holzel, N., Donath, T.W. & Otte, A. 2004. Evaluation of restoration success in alluvial R9 grasslands under contrasting flooding regimes. Biological Conservation 118: 641-650. R10 Boedeltje, G., Bakker, J.P., Bekker, R.M., Van Groenendael, J.M. & Soesbergen, M. 2003. Plant dispersal in a lowland stream in relation to occurrence and three specific life-history traits R11 of the species in the species pool. Journal of Ecology 91: 855-866. R12 Boedeltje, G., Bakker, J.P., Ten Brinke, A., Van Groenendael, J.M. & Soesbergen, M. 2004. Dispersal R13 phenology of hydrochorous plants in relation to discharge, seed release time and buoyancy of seeds: the flood pulse concept supported.Journal of Ecology 92: 786-796. R14 Cellot, B., Mouillot, F. & Henry, C.P. 1998. Flood drift and propagule bank of aquatic macrophytes in R15 a riverine wetland. Journal of Vegetation Science 9 631-640. Chang, E.R., Veeneklaas, R.M., Buitenwerf, R., Bakker, J.P. & Bouma, T.J. 2008. To move or not to R16 move: Determinants of seed retention in a tidal marsh. Functional Ecology 22: 720-727. R17 Ehrlen, J. & Eriksson, O. 2000. Dispersal limitation and patch occupancy in forest herbs. Ecology 81: R18 1667-1674. Ehrlen, J., Munzbergova, Z., Diekmann, M. & Eriksson, O. 2006. Long-term assessment of seed R19 limitation in plants: results from an 11-year experiment. Journal of Ecology 94: 1224-1232. R20 Flynn, S., Turner, R.M. & Stuppy, W.H. 2006. Seed information database. Royal Botanic Gardens, Kew. Available at: http://data.kew.org/sid/sidsearch.html. Accessed 1 February 2008. R21 Goodson, J.M., Gurnell, A.M., Angold, P.G. & Morrissey, I.P. 2003. Evidence for hydrochory and R22 the deposition of viable seeds within winter flow-deposited sediments: The River Dove, R23 Derbyshire, UK. River Research and Applications 19: 317-334. Grootjans, A.P., Bakker, J.P., Jansen, A.J.M. & Kemmers, R.H. 2002. Restoration of brook valley R24 meadows in the Netherlands. Hydrobiologia 478: 149-170. R25 Gurnell, A., Goodson, J., Thompson, K., Clifford, N. & Armitage, P. 2007. The river-bed: a dynamic store for plant propagules? Earth Surface Processes and Landforms 32: 1257-1272. R26 Gurnell, A.M., Boitsidis, A.J., Thompson, K. & Clifford, N.J. 2006. Seed bank, seed dispersal and R27 vegetation cover: Colonization along a newly-created river channel. Journal of Vegetation R28 Science 17: 665-674. Gurnell, A, Thompson, K., Goodson, J. & Moggridge, H. 2008. Propagule deposition along river R29 margins: Linking hydrology and ecology. Journal of Ecology 96: 553-565. R30 Hanski, I. 1999. Metapopulation Ecology. Oxford University Press, Oxford. Jansson, R., Nilsson, C. & Malmqvist, B. 2007. Restoring freshwater ecosystems in riverine landscapes: R31 the roles of connectivity and recovery processes. Freshwater Biology 52: 589-596. R32 Jansson, R., Zinko, U., Merritt, D.M. & Nilsson, C. 2005. Hydrochory increases riparian plant species R33 richness: a comparison between a free-flowing and a regulated river. Journal of Ecology 93: 1094-1103. R34 Jensen, K., Trepel, M., Merritt, D. & Rosenthal, G. 2006. Restoration ecology of river valleys. Basic and R35 Applied Ecology 7: 383-387. Junk, W.J., Bayley, P.B. & Sparks, R.E. 1989. The floodplain concept in river systems. In: Dodge, R36 D.P. (ed.) Proceedings of the International Large River Symposium (LARS), pp. 110-127. R37 Department of Fisheries and Oceans, Ottawa, Honey Harbour, Ontario, Canada. R38 R39 Seed source and sink functions of a floodplain | 69

Kleyer, M., Bekker, R.M., Knevel, I.C., Bakker, J.P., Thompson, K., Sonnenschein, M., Poschlod, P. & van Groenendael.J.M., Klimeš, L., Klimesova, J., Klotz, S., Rusch, G.M., Hermy, M., R1 Adriaens, D., Boedeltje, G., Bossuyt, B., Dannemann, A., Endels, P., Götzenberger, L., R2 Hodgson, J.G., Jackel, A-K., Kühn, I., Kunzmann, D., Ozinga, W.A., Römermann, C., Stadler, R3 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 R4 of the Northwest European flora.Journal of Ecology 96: 1266-1274. R5 Klimkowska, A., van Diggelen, R., den Held, S., Brienen, R., Verbeek, S. & Vegelin, K. 2009. Seed production in fens and fen meadows along a disturbance gradient. Applied Vegetation R6 Science 12: 304-315. R7 Merritt, D.M. & Wohl, E.E. 2006. Plant dispersal along rivers fragmented by dams. River Research R8 and Applications 22: 1-26. Middleton, B.A. 1999. Wetland Restoration. Flood Pulsing and Disturbance Dynamics. John Wiley R9 and Sons, Inc., New York. R10 Moggridge, H.L. & Gurnell, A.M. 2010. Hydrological Controls on the Transport and Deposition of Plant Propagules within Ripirian Zones. River Research and Applications 26: 512-527. R11 Moggridge, H.L., Gurnell, A.M. & Mountford, J.O. 2009. Propagule input, transport and deposition in R12 riparian environments: the importance of connectivity for diversity. Journal of Vegetation R13 Science 20: 465-474. Muller-Landau, H.C., Wright, S.J., Calderón, O., Condit, R. & Hubbell, S.P. 2008. Interspecific variation R14 in primary seed dispersal in a tropical forest. Journal of Ecology 96: 653-667. R15 Nienhuis, P.H., Buijse, A.D., Leuven, R., Smits, A.J.M., de Nooij, R.J.W. & Samborska, E.M. 2002. Ecological rehabilitation of the lowland basin of the river Rhine (NW Europe). R16 Hydrobiologia 478: 53-72. R17 Nienhuis, P.H. & Leuven, R. 2001. River restoration and flood protection: controversy or synergism? R18 Hydrobiologia 444: 85-99. Nilsson, C., Gardfjell, M. & Grelsson, G. 1991. Importance Of Hydrochory In Structuring Plant- R19 Communities Along Rivers. Canadian Journal of Botany 69: 2631-2633. R20 Ozinga, W.A., Schaminee, J.H.J., Bekker, R.M., Bonn, S., Poschlod, P., Tackenberg, O., Bakker, J. & van Groenendael, J.M. 2005. Predictability of plant species composition from environmental R21 conditions is constrained by dispersal limitation. Oikos 108: 555-561. R22 Peart, D.R. 1989. Species Interactions in a Successional Grassland.1. Seed Rain and Seedling R23 Recruitment. Journal of Ecology 77: 236-251. Pollux, B.J.A., Verbruggen, E., van Groenendael, J.M. & Ouborg, N.J. 2009. Intraspecific variation R24 of seed floating ability inSparganium emersum suggests a bimodal dispersal strategy. R25 Aquatic Botany 90: 199-203 Rosenthal, G. 2006. Restoration of wet grasslands - Effects of seed dispersal, persistence and R26 abundance on plant species recruitment. Basic and Applied Ecology 7: 409-421. R27 Schneider, R.L. & Sharitz, R.R. 1988. Hydrochory and Regeneration in a Bald Cypress Water Tupelo R28 Swamp Forest. Ecology 69: 1055-1063. Soomers, H., Winkel, D.N., Du, Y. & Wassen, M.J. 2010. The dispersal and deposition of hydrochorous R29 plant seeds in drainage ditches. Freshwater Biology 55: 2032-2046. R30 Soons, M.B., Heil, G.W., Nathan, R. & Katul, G.G. 2004. Determinants of long-distance seed dispersal by wind in grasslands. Ecology 85: 3056-3068. R31 Tockner, K., Malard, F. & Ward, J.V. 2000. An extension of the flood pulse concept.Hydrological R32 Processes 14: 2861-2883. R33 van den Broek, T., van Diggelen, R. & Bobbink, R. 2005. Variation in seed buoyancy of species in wetland ecosystems with different flooding dynamics.Journal of Vegetation Science 16: R34 579-586. R35 van der Meijden, R. 2005. Heukels’ Flora van Nederland. 23. Wolters-Noordhoff bv, Groningen. van Heerd, R.M., Kuijlaars, E.A.C., Teeuw, M.P. & van ‘t Zand, R.J. 2000. Productspecificatie AHN R36 2000. MDTGM2000, 13. Rijkswaterstaat, Delft. R37 Vogt, K., Rasran, L. & Jensen, K. 2006. Seed deposition in drift lines during an extreme flooding R38 event - Evidence for hydrochorous dispersal? Basic and Applied Ecology 7: 422-432. R39 70 | Chapter 3

Vogt, K., Rasran, L. & Jensen, K. 2007. Seed deposition in drift lines: Opportunity or hazard for R1 species establishment? Aquatic Botany 86: 385-392. R2 Vogt, K., Rasran, L. & Jensen, K. 2004. Water-borne seed transport and seed deposition during R3 flooding in a small river-valley in Northern Germany.Flora: Morphologie, Geobotanik, Oekophysiologie 199: 377-388. R4 R5 R6 R7 R8 R9 R10 R11 R12 R13 R14 R15 R16 R17 R18 R19 R20 R21 R22 R23 R24 R25 R26 R27 R28 R29 R30 R31 R32 R33 R34 R35 R36 R37 R38 R39 Seed source and sink functions of a floodplain | 71

aPPendIx R1 R2 table s1. Plant species for which (I) seeds were captured in the River Drentsche Aa and/or (II) which R3 occur in the Kappersbult floodplain (last column). The second and third columns indicate numbers of seeds caught downstream and upstream respectively, computed to annual figures. R4 Species Total number Total number Occurring in Occurring R5 of seeds of seeds per vegetation in the river R6 per year year upstream Kappersbult? bank of the downstream 1=yes, 0=no Kappersbult? R7 1=yes, 0=no R8 Acorus calamus 0.00 0.00 1 1 R9 Aegopodium podagraria 0.00 171.29 0 0 R10 Agrostis canina 732.59 359.41 1 1 Agrostis capillaris 76.47 217.18 1 0 R11 Agrostis stolonifera 808.17 1937.46 1 0 R12 Alisma plantago-aquatica 0.00 102.78 0 0 R13 Alnus 7844.46 9886.42 mv 0 R14 Alopecurus geniculatus 0.00 263.67 0 0 R15 Alopecurus pratensis 0.00 0.00 1 1 Angelica sylvestris 0.00 0.00 1 1 R16 Anthoxanthum odoratum 0.00 79.53 1 0 R17 Anthriscus sylvestris 0.00 0.00 1 1 R18 Atriplex prostrata 0.00 168.24 0 0 R19 Berula erecta 550.14 479.01 0 0 Betula pubescens 8558.59 5839.29 mv 0 R20 Bidens cernua 76.47 229.41 0 0 R21 Calamagrostis canescens 134.14 374.09 1 1 R22 Caltha palustris 0.00 0.00 1 0 R23 Cardamine pratensis 67.29 742.07 1 0 Carex acuta 0.00 0.00 1 0 R24 Carex aquatilis 0.00 0.00 1 0 R25 Carex curta 0.00 0.00 1 0 R26 Carex disticha 670.08 263.67 0 0 R27 Carex elata 0.00 0.00 1 1 R28 Carex nigra 4078.94 0.00 1 0 Carex panicea 0.00 0.00 1 0 R29 Carex pseudocyperus 3152.12 0.00 0 0 R30 Carex riparia 268.29 0.00 0 0 R31 Cicuta virosa 622.03 298.24 1 1 R32 Cirsium palustre 157.53 290.59 1 1 Comarum palustre 0.00 0.00 1 0 R33 Conyza canadensis 157.53 162.42 0 0 R34 Crataegus monogyna 0.00 171.29 mv 0 R35 Dactylus glomerata 203.41 0.00 0 0 R36 Deschampsia cespitosa 0.00 0.00 1 0 Eleocharis palustris 0.00 0.00 1 0 R37 Epilobium ciliatum 249.29 355.74 0 0 R38 R39 72 | Chapter 3

R1 Species Total number Total number Occurring in Occurring of seeds of seeds per vegetation in the river R2 per year year upstream Kappersbult? bank of the R3 downstream 1=yes, 0=no Kappersbult? R4 1=yes, 0=no Epilobium hirsutum 14314.18 5868.96 1 1 R5 Epilobium obscurum 134.14 0.00 0 0 R6 Epilobium tetragonum 210.62 1542.56 0 0 R7 Eriophorum angustifolium 0.00 0.00 1 0 R8 Eupatorium cannabinum 875.91 357.88 1 1 R9 Festuca pratensis 0.00 58.12 1 0 Festuca rubra 0.00 171.29 1 0 R10 Filipendula ulmaria 0.00 0.00 1 1 R11 Galium aparine 0.00 52.00 0 0 R12 Galium palustre 67.29 1490.56 1 0 R13 Glyceria fluitans 76.47 59.65 1 0 Glyceria maxima 97.88 0.00 1 1 R14 Gnaphalium uliginosum 1576.78 3116.02 0 0 R15 Heracleum sphondylium 76.47 0.00 0 0 R16 Holcus lanatus 2730.00 3927.53 1 1 R17 Hypochaeris radicata 0.00 79.53 0 0 Iris pseudacorus 0.00 0.00 1 1 R18 Juncus acutiflorus 0.00 205.55 1 0 R19 Juncus articulatus 0.00 0.00 1 0 R20 Juncus bufonius 1975.91 13977.60 0 0 R21 Juncus conglomeratus 0.00 0.00 1 0 R22 Juncus effusus 5007.65 7220.35 1 0 Lycopus europaeus 5365.79 2261.08 1 1 R23 Lythrum salicaria 939.01 3494.40 0 0 R24 Mentha aquatica 2483.23 433.74 1 1 R25 Mentha arvensis 0.00 52.00 1 1 R26 Myosotis laxa subsp. cespitosa 316.14 117.76 1 1 Myosotis scorpioides 134.14 688.54 1 1 R27 Myosoton aquaticum 81.06 0.00 0 0 R28 Pedicularis palustris 0.00 0.00 1 0 R29 Persicaria amphibia 0.00 0.00 1 1 R30 Persicaria hydropiper 0.00 0.00 1 1 Peucedanum palustre 139.18 65.76 1 1 R31 Phalaris arundinacea 4081.62 182.31 1 1 R32 Phleum pratense 134.14 0.00 0 0 R33 Phragmites australis 0.00 0.00 1 1 R34 Plantago lanceolata 0.00 0.00 1 0 Plantago major 0.00 205.55 0 0 R35 Poa palustris 0.00 0.00 1 1 R36 Poa pratensis 0.00 102.78 0 0 R37 Poa trivialis 3521.55 12686.78 1 0 R38 Ranunculus acris 0.00 0.00 1 0 R39 Seed source and sink functions of a floodplain | 73

Species Total number Total number Occurring in Occurring R1 of seeds of seeds per vegetation in the river per year year upstream Kappersbult? bank of the R2 downstream 1=yes, 0=no Kappersbult? R3 1=yes, 0=no R4 Ranunculus flammula 0.00 102.78 1 0 R5 Ranunculus repens 277.91 1189.88 1 0 Ranunculus sceleratus 1125.65 762.56 0 0 R6 Rhinantus angustifolius 0.00 0.00 1 0 R7 Rorippa amphibia 0.00 308.33 0 0 R8 Rorippa palustris 4367.82 4749.74 0 0 R9 Rorippa sylvestris 324.24 598.00 0 0 Rumex actetosa 1205.18 893.18 1 0 R10 Rumex hydrolapathum 1560.00 521.53 0 0 R11 Rumex obtusifolius 0.00 52.00 0 0 R12 Sagina procumbens 519.56 265.20 0 0 R13 Salix alba 804.87 1027.76 mv 0 R14 Salix cinerea 134.14 308.33 mv 0 Salvia pratensis 105.53 0.00 0 0 R15 Scutellaria galericulata 134.14 0.00 0 0 R16 Silene flos-cuculi 0.00 0.00 1 0 R17 Solidago canadensis 0.00 0.00 1 1 R18 Sonchus asper 1341.45 469.22 0 0 Stachys palustris 0.00 0.00 1 0 R19 Stellaria palustris 0.00 0.00 1 0 R20 Stellaria uliginosa 715.96 0.00 0 0 R21 Succisa pratensis 0.00 0.00 1 0 R22 officinale 268.29 102.78 0 0 Typha angustifolia 939.01 205.55 0 0 R23 Typha latifolia 319.65 0.00 1 1 R24 Urtica dioica 536.58 5344.38 0 0 R25 Veronica beccabunga 472.59 162.42 0 0 R26 Veronica scutellata 0.00 0.00 1 0 Viola palustris 0.00 47.41 0 0 R27 R28 mv=missing value R29 R30 R31 R32 R33 R34 R35 R36 R37 R38 R39 Chapter 4 Chapter 4

the dispersal and deposition of hydrochorous plant seeds in drainage ditches

Soomers, H., Winkel, D.N., Du, Y., and Wassen, M.J. 2010. Freshwater Biology 55: 2032-2046. 76 | Chapter 4

R1 abstraCt R2 R3 1. Surface water is an important dispersal vector for wetland plant species. However, most R4 previous studies on hydrochory (i.e. water dispersal) have focused on ecosystems with R5 relatively rapid water flow. Therefore, there is a need to study such dispersal in slow-flowing R6 or stagnant water bodies, such as drainage ditches, which might act as dispersal corridors between habitat patches. R7 2. To gain insight into the mechanisms by which seeds are transported in drainage ditches, R8 the effect of the velocity of wind and water on the rate of transport of floating seeds of R9 three wetland species (Carex pseudocyperus L., Iris pseudacorus L. and Sparganium erectum R10 L.) was investigated. Furthermore, in release and retrace experiments with painted Carex R11 pseudocyperus seeds, a number of factors potentially determining the probability of seed R12 deposition were investigated. R13 3. Net wind speed was found to be the main factor determining the rate at which seeds are R14 transported in drainage ditches. No relation between water flow at mid-depth in the ditches R15 and seed transport was found. Wind speed and flow at the water surface were positively R16 related. The effect of wind speed on the rate of transport of floating seeds was greater for R17 Sparganium erectum seeds, because a greater ratio of their volume protrudes from the water, R18 than for Carex pseudocyperus and Iris pseudacorus seeds. R19 4. The principal factors that determine seed deposition were aquatic plant cover, ditch slope and indentations in the ditch bank. Seeds changed direction if the wind direction changed, or R20 if there was a bend in the ditch. The final pattern of deposition was related to mean net wind R21 speed. Mean transport distance after 2 days varied between 34 and 451 m. R22 5. Unlike in rivers, seed transport in ditches was determined by wind speed and direction, R23 enabling multidirectional seed dispersal. We conclude that in slow flowing waters wind is a R24 more important driver for hydrochorous seed transport than the flow of water. This sheds a R25 new light on hydrochory, and has important consequences for the management of otherwise R26 fragmented wetland remnants. R27 R28 R29 R30 R31 R32 R33 R34 R35 R36 R37 R38 R39 Dispersal and deposition of hydrochorous plant seeds | 77

4.1 IntroduCtIon R1 R2 Seed dispersal is an important process determining plant distribution patterns and the genetic R3 diversity of populations (Wright, 1943; Venable & Brown, 1993; Levin et al., 2003). Dispersal R4 limitation caused by the fragmentation of habitats leads to reduced genetic exchange between R5 subpopulations and a smaller probability that empty but suitable sites will be colonized. As a R6 result, there is an overall reduction in the size and viability of population networks (Wilcove, R7 McLellan & Dobson, 1986; Saunders, Hobbs & Margules, 1991; Soons & Heil, 2002). Knowledge R8 of mechanisms that determine the range of dispersal of plant seeds is of vital importance for R9 determining ways of restoring biodiversity successfully (Middleton, van Diggelen & Jensen, 2006). R10 In wetlands, surface water is an important vector for dispersal (Merritt & Wohl, 2002; Boedeltje R11 et al., 2004; Neff & Baldwin, 2005). For instance, Jansson et al. (2005) observed that hydrochory R12 (i.e. plant dispersal by surface water) increased plant species richness in riverine ecosystems. R13 Furthermore, the study by Schneider & Sharitz (1988) in a swamp forest showed that, for R14 cypress seeds (Taxodium distichum L.), hydrochory was more common than wind dispersal. R15 Research on hydrochory has focussed on ecosystems in which water flow is relatively rapid, R16 such as rivers, streams and estuaries. In such ecosystems, water flow determines the velocity R17 of seed transport. Several studies have shown that seeds or seed mimics may be carried over R18 a distance of several kilometres (Griffith & Forseth, 2002) to more than a hundred kilometres R19 (Andersson, Nilsson & Johansson, 2000). R20 In Western Europe and North America, however, the significance of flow as a vector for seed R21 dispersal of floodplain plant species has decreased as rivers have been regulated to reduce R22 the risk of flooding. Such regulation often has the consequence that the lateral connectivity between a river and its floodplain or valley is broken (Tockner & Stanford, 2002; Trepel & R23 Kluge, 2002; McCartney & de la Hera, 2004). Moreover, flood plains have been reclaimed for R24 agricultural purposes, with the result that connectivity has diminished further. At the same R25 time, numerous drainage ditches have been dug in reclaimed farmland (Moss, 1983; Borger, R26 1992; Bootsma, 2000). The banks of these ditches serve as important refuges for many wetland R27 plant species (Blomqvist et al., 2006). Drainage ditches may also function as habitat corridors R28 between isolated fen nature reserves (Milsom et al., 2004). Therefore, it seems feasible that R29 hydrochorous dispersal in these cultivated areas no longer takes place via the river, but rather R30 via this network of man-made drainage ditches in which water flows much more slowly than R31 in rivers. Thus, Bulle, van Groenedael & Jurgens (1994) related potential dispersal distances of R32 helophytes to the spatial location of subpopulations of these species along drainage ditches R33 in the Netherlands, and concluded that hydrochorous dispersal via ditches can contribute to R34 the connectivity between the subpopulations. R35 Because of the relatively low rate of flow in drainage ditches, the results of previous studies of hydrochory in rapidly flowing water cannot simply be extrapolated to these areas. It is thus R36 necessary to study hydrochory in areas in which the water flows slowly or is stagnant. To our R37 knowledge, only three studies have been published on hydrochory in slow-flowing drainage R38 R39 78 | Chapter 4

R1 ditches (Bulle et al., 1994; Gornall, Hollingsworth & Preston, 1998; Beltman, Van Den Broek & R2 Vergeer, 2005), of which only Beltman et al. (2005) directly measured seed dispersal. However, R3 none of these studies revealed the mechanisms that determine hydrochorous seed dispersal in R4 slow flowing to stagnant waters. Given that wind shear stress on water influences the speed R5 and direction of surface currents in lakes (Laval et al., 2003; Stevens, Lawrence & Hamblin, R6 2004), it can be expected that wind speed and direction also influence hydrochorous dispersal in drainage ditches. However, neither Bulle et al. (1994) nor Beltman et al. (2005) related wind R7 speed to seed transport. R8 Besides the actual transport of floating seeds, seed deposition at sites suitable for R9 germination also determines the successful exchange of propagules between fragmented R10 wetland populations (Levine & Murrell, 2003; Chang et al., 2008). There is no agreement in R11 the literature about factors that determine seed deposition. In free-flowing rivers and tidal R12 marshes, seed deposition is promoted by inundation (e.g. Middleton, 2000; Merritt & Wohl, R13 2002; Goodson et al., 2003; Vogt, Rasran & Jensen, 2006; Chang, Veeneklaas & Bakker, 2007). R14 Furthermore, emergent obstacles, such as trees and logs, have been found to increase the R15 density of deposited seeds (Schneider & Sharitz, 1988). However, Andersson et al. (2000) found R16 a significant relation between the number of deposited seed mimics and rapid current, but R17 not between seed mimics and the percentage cover of trees and shrubs. R18 Clearly, there is insufficient knowledge of how wind and currents influence the transport of R19 seeds in slowly flowing waters. Also, not much is known about the factors that determine the deposition of seeds along the banks of slowly flowing water bodies. Since harvesting of R20 vegetation from drainage ditches and ditch banks is common, management has potentially R21 important implications for transport and deposition processes. Therefore, we investigated R22 the transport and deposition of hydrochorous seeds in drainage ditches in a semi-natural fen R23 area in the Netherlands. Firstly, we examined the relative contribution of wind and water R24 flow to the rate of transport of floating seeds. In addition to the indirect effect of wind R25 shear stress on the movement of floating seeds, via surface flow, we expected wind to drive R26 hydrochorous seed movement directly. If so, the effect of wind should have a greater effect R27 on the movement of seeds which protrude from the water than for those which float lower R28 in the water. Here we attempted to disentangle the direct effect of such seed ‘sailing’ from R29 that of wind shear stress. We hypothesized that: 1) wind speed and direction are the main R30 factors determining seed transport in these ditches, and 2) the effect of wind on the rate of R31 transport of floating seeds increases as the fraction of the seed protruding from the water R32 increases. Secondly, we investigated what characteristics of ditches and ditch banks determine seed deposition. R33 R34 R35 R36 R37 R38 R39 Dispersal and deposition of hydrochorous plant seeds | 79

4.2 metHods R1 R2 4.2.1 Study site R3 The floodplain of the Vecht River (Netherlands) is a lowland peat area north of Utrecht and R4 southeast of Amsterdam (between 52°19’N - 5°06’E and 52°07’N - 5°10’E). Anthropogenic R5 land use in the area began in the Middle Ages by cutting peat and draining the remaining R6 land to make it suitable for agriculture. The area consists of polders (i.e. low-lying land, R7 reclaimed from a water body or peatland and protected by dikes), each with its own artificially R8 controlled surface water system, created in the 19th century. For this purpose, an intensive R9 network of drainage ditches was designed (Wassen et al., 1990). Currently, the area comprises agricultural pastures interconnected by numerous ditches, lakes and turf ponds (i.e. water R10 bodies resulting from the excavation of peat). Although there are some nature reserves, the R11 remaining species-rich fen vegetation is currently deteriorating, due to structural changes in R12 hydrology, agricultural intensification and nutrient enrichment (Wassen et al., 1990; Wassen et R13 al., 1996). The hydrological connection between the River Vecht and the valley has been lost, R14 but the drainage ditches and turf ponds that dominate the landscape might allow dispersal R15 of wetland plant seeds. Carex pseudocyperus L., Iris pseudacorus L. and Sparganium erectum R16 L. are common on ditch banks and in shallow water. Within the Vecht River floodplain, Polder R17 Westbroek (Fig. 4.1) was chosen as the study site. Details of the vegetation, land use and R18 hydrology in this area are given by Wassen et al. (1990) and Verhoeven et al. (1988). The R19 network of ditches in this polder is interconnected by covered culverts (i.e. pipes below a field R20 that connect ditches). The experiments to investigate the relative contribution of wind and R21 water to the floating velocity of seeds were performed in ditch 4 (Fig. 4.1). The experiments R22 on seed deposition were performed in ditches 1-4 (Fig. 4.1). The general characteristics of the four ditches are given in Table 4.1. R23 R24 table 4.1. Orientation, length from culvert to culvert, approximate mean width and depth, R25 and general shape of the four research ditches. se-nw=southeast-northwest orientation, sw- R26 ne=southwest-northeast orientation R27 Ditch number 1 2 3 4 R28 First part Second part R29 Orientation se-nw sw-ne sw-ne sw-ne sw-ne R30 Length (m) 100 830 758 1037 810 R31 Width (m) 4.5 2 3 2 2 Depth (m) 0.50 0.35 0.40 0.35 0.40 R32 Shape one 90º bend straight some bends* straight R33 R34 *Seeds were not transported by water movement to the bend in the ditch R35 R36 R37 R38 R39 80 | Chapter 4

R1 R2 Forest Pasture R3 Water R4 Buildings R5 Drainage canals Experiment ditches R6 3

R7 2 R8

R9 4 R10 R11 1 the Netherlands R12 R13 Amsterdam Polder Westbroek R14 Utrecht R15 R16

R17 50 km R18 200 metres 7380 R19 Figure 4.1 Location of Polder Westbroek in the Netherlands and the four research ditches within R20 this polder. R21 R22 4.2.2 Seed description R23 In the experiments, the achenes of C. pseudocyperus, the seeds of I. pseudacorus and the fruits R24 of S. erectum were used. Hereafter, the propagules of all three species will be called ‘seeds’. R25 The seeds of these three wetland species differ in size, weight and the fraction of the volume R26 that protrudes from the water. Characteristics of the seeds are shown in Fig. 4.2. R27 R28 R29 R30 Water R31 7485 R32 Species Carex pseudocyperus Iris pseudacorus Sparganium erectum R33 Length (mm) 5 8.5 10 R34 Width (mm) 1.4 7.5 6 R35 Dry Mass (mg) 0.73 49 16.7 R36 Figure 4.2. Approximate mean size (Cappers, Bekker & Jans, 2006) and dry mass (Liu et al., 2008) of R37 seeds of C. pseudocyperus, I. pseudacorus and S. erectum. The shape and the fraction of the seed that protrudes from the surface when floating are illustrated in the upper part of the figure. The R38 seed images of C. pseudocyperus and S. erectum are from the Groningen Institute of Archeology R39 (Cappers et al., 2006). Dispersal and deposition of hydrochorous plant seeds | 81

4.2.3 The relative contribution of wind speed and water velocity to seed transport rate R1 We investigated mean water velocity (averaged over 30 seconds) at two different depths and R2 the possible influence of net mean wind speed and water flow on the current at the water R3 surface in ditch 4. Water flow was measured 1) 1 cm below the water surface (hereafter called R4 ‘surface current’), and 2) at mid-depth halfway across the ditch (approximating mean water R5 velocity; Herschy, 1999). These measurements were made with a SonTec FlowTracker (SonTek, R6 2001). During each water flow measurement, mean wind speed and direction were measured R7 using an anemoscope and a wind vane with degree scale, respectively. R8 We also assessed (in ditch 4) the effect of water velocity at the two depths and net mean R9 wind speed (averaged over 5 seconds) on the floating speed of Carex pseudocyperus seeds. For this purpose, a wooden frame (0.75 x 0.75 m) was used. Floats at each corner of the frame R10 prevented it from touching the water and thereby influencing flow. For each measurement, R11 a seed was released from one side of the frame, and the time taken to be transported to the R12 other side was recorded. Simultaneously, mean wind speed and direction and mean water R13 velocity at the two depths were measured. Similar experiments, but without measuring R14 velocity at mid-depth (due to time constraints), were performed with seeds of Iris pseudacorus R15 and Sparganium erectum. R16 R17 4.2.4 Factors determining seed deposition R18 To investigate seed deposition, painted seeds of Carex pseudocyperus were released in ditches R19 1-3 (April-June 2007) and 4 (May-June 2008). Prior to painting, all seeds were dried in an oven R20 for 3 h at 80°C to prevent later germination in the field (see Vogtet al., 2004). To determine R21 the effect of painting and heating on the buoyancy of C. pseudocyperus seeds, the buoyancy R22 of 1) painted, 2) heated, 3) painted and heated and 4) untreated seeds was measured for 4 days. For each of the four classes, 10 buckets were filled with tap water and 50 seeds were R23 placed in each bucket. Every day, the sunken seeds were counted after stirring for 20 s to R24 reduce the influence of surface tension (Danvind & Nilsson, 1997). For the first three days, all R25 seeds remained buoyant. On the fourth day, three seeds of the painted and heated class, and R26 two seeds of the painted class sank, whereas all the seeds of the other two classes remained R27 buoyant, although differences between the classes were not significant (Kruskal-Wallis, R28 P=0.228). Because painted and heated seeds started to sink after 4 days, the duration of the R29 deposition experiments did not exceed 3 days. R30 To determine the effect of ditch-bank morphology and vegetation on seed deposition, both R31 banks of the ditches were divided into 2 m sections (Fig. 4.3). In each section, the slope of R32 the bank, the aquatic vegetation cover, the type of vegetation at the bank and the shape R33 of the bank were recorded. Four slope classes were distinguished: 0-30º (shallow), 31-45º, R34 46-60º and 61-90º (steep). Aquatic vegetation cover in the ditch, from the bank up to 0.5 m R35 from the bank, was categorized according to Daubenmire’s cover classes (<5 %, 5-25 %, 26- 50%, 51-75%, 76-95% and 96-100 %) (Daubenmire, 1968). The aquatic vegetation cover was R36 measured every week because it increased over the course of the experiments. Submerged R37 plants were not taken into account, as they do not trap floating seeds. The shape of the bank R38 R39 82 | Chapter 4

R1 was recorded as straight, or including an indentation, a bulge, or an inner or outer bend. For R2 the vegetation structure on the banks, pasture and carr forest were distinguished. R3 In each of the 12 release experiments (two in ditch 1, three in ditch 2, two in ditch 3 and five R4 in ditch 4), 1000 Carex pseudocyperus seeds were released at the release point (Fig. 4.3), and R5 relocated after 3 h, one day and two days. At these times, the number of seeds in each section R6 was counted. Furthermore, for every section we recorded whether the wind was blowing in the direction of the ditch bank, away from the bank or along the ditch. Data on the mean daily R7 wind velocity were taken from the online database of the Royal Netherlands Meteorological R8 Institute (KNMI, 2008). R9 R10 R11 R12 R13 R14 R15 R16 R17 R18 R19 R20

R21 Figure 4.3. Example of sections along ditch banks and the positive and negative directions of R22 transport. Sections were marked in both the positive and negative directions along the ditches. R23

R24 4.2.5 Statistical analyses R25 Wind, water flow and seed transport R26 The relation between net wind speed and surface current was determined using a Spearman R27 rank correlation test. R28 To determine the relative contribution of surface current and velocity at mid-depth, as well R29 as net wind speed, to the rate of transport of floating seeds, a multiple linear regression R30 analysis, with seed speed as the dependent variable, was performed. The net wind speed was R31 the speed along (x-direction) the ditch (Fig. 4.3), computed from vector calculations using R32 the measured mean wind speed and the wind direction. To avoid multicollinearity, we tested whether independent variables were strongly correlated. If this was the case, one of the R33 correlating variables was omitted. The remaining independent variables were entered into R34 the regression model. R35 We assumed that, if wind also drives hydrochorous seed transport directly, the wind should R36 have a greater effect on the movement of seeds which protrude from the water than for R37 those which float lower in the water, and thus the ‘seed transport rate/surface water velocity’ R38 ratio of the former should be greater. Therefore, we hypothesized that the effect of wind R39 Dispersal and deposition of hydrochorous plant seeds | 83

on seed transport rate would increase as the fraction of the seed protruding from the water R1 increased. This hypothesis was tested by analysing with an ANOVA (plus Tukey post-hoc test) R2 whether there was a significant difference in the ratio of the seed floating velocity to the R3 surface current velocity between the small Carex pseudocyperus seeds, the intermediate Iris R4 pseudacorus seeds and the relatively large Sparganium erectum seeds. Note that for practical R5 reasons, we only used one species in each of the three categories (hereafter called ‘size R6 categories’). R7 R8 Seed deposition R9 For the seed deposition experiment, we tested whether there was a significant difference in the number of seeds deposited between the different classes (hereafter called groups) of the R10 explanatory variables (see ‘Factors determining seed deposition’). To correct for the decrease R11 in the number of seeds available to deposit as the distance from the release point increases R12 (Andersson et al., 2000), the number of seeds per section, expressed as the percentage of the R13 seeds not yet retrieved upstream, was calculated from Equation 1 and used as the dependent R14 variable. R15 R16

S=(ns /( nrtot - nrc ))100 eq. 1 R17 R18 where S is the number of seeds deposited in a 2 m section (as a percentage of those entering R19 the section), n is the number of seeds found in the section, n is the total number of retrieved s rtot R20 seeds and n the number of seeds retrieved upstream of the section. ‘Upstream’ was defined rc R21 as in the direction of the release location of the seeds. R22 Because these data were not normally distributed and variances of the groups were not equal, a nonparametric analysis of variance (Kruskal-Wallis test, pairwise Mann-Whitney U R23 (MWU) tests) was performed. Bonferroni corrections were used to correct for these multiple R24 comparisons. The last two sections in which seeds were present were omitted from the analysis R25 because the calculation of percentages relative to the denominator of Eqn 1 would reveal R26 percentages close to 100% for those sections. Cases in which the wind direction changed after R27 one or two days were also omitted, because then S could no longer be calculated. R28 The mean distance of the deposited seeds was calculated, and the mean distance 24 h after R29 release was plotted versus the mean wind speed on the release day and on the day after R30 release for each of the 10 experiments (in total) executed in the straight ditches 2, 3 and 4 R31 (Fig. 4.1). R32 R33 R34 R35 R36 R37 R38 R39 84 | Chapter 4

R1 4.3 results R2 R3 4.3.1 Relative contribution of wind speed and water flow velocity to seed transport rate R4 The surface current was positively correlated with net mean wind speed (Spearman’s R5 coefficient, P=0.025, r=0.230). At mid-depth, in 92% of the cases the water was flowing in R6 the opposite direction than at the surface, whereas the surface current always followed the direction of the wind. R7 R8 For all three species, there was a significant positive correlation between mean wind speed and R9 seed velocity (Pearson correlation. Carex pseudocyperus: R²=0.451, P<0.001, N=64; Sparganium R10 erectum: R²=0.313, P=0.001, N=32; Iris pseudacorus: R²=0.323, P=0.001, N=30) (Fig. 4.4). R11 R12 R13 R14 R15 R16 R17 R18 R19 R20

R21 Figure 4.4. Relationships between ‘net mean wind speed’ (m s-1) and ‘rate of transport of floating R22 seeds’ (m s-1) for Carex pseudocyperus (R²=0.451, P<0.001, regression line: y=0.0256x + 0.0232), Iris R23 pseudacorus (R²=0.323, P=0.001, regression: y=0.0107x + 0.0399) and Sparganium erectum (R²=0.313, P=0.001, regression: y=0.0206x + 0.0398). R24 R25 Multiple regression analysis revealed a significant effect of both surface current velocity R26 and mean wind speed on the velocity of floating seeds of Sparganium erectum and Carex R27 pseudocyperus, whereas surface current velocity had no effect on the transport of seeds of Iris R28 pseudacorus (Table 4.2). R29 R30 For all three species, the direction of seed transport was always similar to the net wind R31 direction. For all species, the velocity of floating seeds was less than the mean wind speed R32 (Fig. 4.4), but greater than surface current velocity (Fig. 4.5). R33 R34 R35 R36 R37 R38 R39 Dispersal and deposition of hydrochorous plant seeds | 85

table 4.2. Regression coefficients (B) and significance valuesP ( ) of the predictor variables ‘net mean wind speed’ and ‘surface current velocity’ of the multiple linear regression model with ‘rate R1 of transport of floating seeds’ as the dependent variable, and the proportion of variation in the R2 dependent variable explained by the full regression model (R²). Analyses were performed for seeds R3 of Carex pseudocyperus (N=30), Iris pseudacorus (N=31) and Sparganium erectum (N=33). The predictor variables were poorly correlated (r=0.230) R4 Dependent Rate of transport of Carex Rate of transport of Iris Rate of transport of R5 variable pseudocyperus seeds pseudacorus seeds Sparganium erectum R6 seeds R7 Independent B P B P B P variable R8 Net mean wind 0.026 <0.001 0.011 0.002 0.012 0.050 R9 speed R10 Surface current 1.253 0.001 0.061 0.870 1.122 0.006 R11 velocity R12 R² 0.648 0.324 0.468 R13 R14 R15 R16 6.00 A A B R17 R18 R19 5.00 R20 R21

4.00 R22 R23 R24

3.00 R25 ) per surface current velocity (m/s) −1 R26 R27 2.00 R28 R29 R30 1.00

Seed floating velocity (m s R31 R32 R33 0.00 Carex pseudocyperus Iris pseudacorus Sparganium erectum Seed species R34 R35 Figure 4.5. Mean value and standard deviation for the ratio ‘seed floating velocity/surface current R36 velocity’ for Carex pseudocyperus, Iris pseudacorus and Sparganium erectum (Anova (Tukey post- hoc), F=14.34, P<0.001). B differs significantly from A. R37 R38 R39 86 | Chapter 4

R1 For Carex pseudocyperus seeds, the water velocity at mid-depth was measured simultaneously R2 with wind speed, surface current and seed velocity (N=24). Adding this variable as a potential R3 predictor variable did not change the results; velocity at mid-depth was not significantly R4 related to seed velocity. R5 There was no relationship between the velocity of water flow at mid-depth and the rate R6 of transport of Carex pseudocyperus seeds (Fig. 4.6). Quadrant I of the figure is empty. This represents cases in which both seed transport and water flow at mid-depth would be in a R7 south-eastern direction (positive, see Fig. 4.3). In quadrants II and IV, the directions of seed R8 transport and water flow at mid-depth were opposite to each other (47 cases). Observations R9 in quadrant III represent cases in which seeds were transported and water flowed in a north- R10 westerly direction (11 cases). R11 R12 0.1 R13 R14 0.08 IV I R15 0.06 )

R16 −1 0.04 R17 R18 0.02 R19 0

R20 −0.02 R21 −0.04

R22 Seed transport velocity (m s R23 −0.06

R24 −0.08 III II R25 −0.1 R26 −0.015 −0.01 −0.005 0 0.005 0.01 0.015 Water flow (m s−1) at mid−depth R27 R28 Figure 4.6. The velocity (m s-1) of water flow at mid-depth plotted versus the transport velocity (m -1s ) of Carex pseudocyperus seeds. Positive values indicate movement in a southerly direction, negative R29 values in a northerly direction. R30 R31 R32 For S. erectum seeds, the fraction of the propagule protruding from the water is greater than for C. pseudocyperus and I. pseudacorus seeds (Fig. 4.2). To test the hypothesis that seeds with R33 a greater fraction of the volume protruding are more influenced by wind speed, we compared R34 the ratios of the velocity of floating seeds to the surface current velocity for the three species. R35 This ratio was significantly higher for S. erectum than for C. pseudocyperus and I. pseudacorus R36 (ANOVA, F=14.34, P<0.001, Tukey post-hoc) (Fig. 4.5). No significant difference in this ratio R37 was found between C. pseudocyperus and I. pseudacorus (P=0.585), although the mean ratio R38 was slightly higher for the larger I. pseudacorus seeds (Fig. 4.5). R39 Dispersal and deposition of hydrochorous plant seeds | 87

4.3.2 Factors determining seed deposition R1 Overall, the corrected number of seeds deposited per section (S) (see Eqn 1) was significantly R2 lower for sections with less aquatic plant cover (Table 4.3) (Kruskal-Wallis, P<0.0033). It was R3 only between the sections with the most plant cover that no significant difference in the R4 number of seeds deposited was found (Table 4.3). R5 Twenty-four and 48 h after release, corrected numbers of deposited seeds were found to R6 be significantly higher in sections with shallow ditch banks than in steep sections (Table 4.3) R7 (Kruskal-Wallis, P<0.0083). This pattern was less clear 3 h after release. R8 The corrected number of seeds in sections that contained a bulge was significantly smaller R9 than in sections with an indentation (P<0.0083). In straight sections, the corrected number of seeds deposited was lower than in sections with an indentation (MWU, 24 h after release, R10 P<0.0083; 48 h after release, P<0.0083), but higher than in sections with a bulge (P<0.005). R11 On the day of release, the corrected number of seeds deposited on the lee side was significantly R12 greater than in sections situated in the windward direction (MWU, P<0.001). Furthermore, R13 significantly more seeds were found 24 h after release where the ditch bank was covered with R14 carr forest than in sections with pasture vegetation at the ditch bank (MWU, P<0.001). R15 R16 Table 4.4 gives the mean, minimum and maximum distances of retraced deposited seeds from R17 the release point for every experiment. The mean number of retraced seeds was 407.4 (±188.9) R18 on day 1, 217.5 (± 136.1) on day 2 and 146.5 (± 121.2) on day 3. For five of the 12 experiments, R19 the maximum distance that seeds were transported after 24 and 48 h equals the total length R20 of the ditch, measured from the release point. At the end of the ditches, these channels were R21 interconnected by culverts. Yet we never saw seeds being transported through these culverts R22 and never found any on the other side. Table 4.4 shows that for experiments 1, 2, 8, 9, 10 and 11, the mean distance that seeds had R23 been transported after 3 h exceeded 80% of the mean distance of transport after 48 h. In R24 experiments 3, 5, 7 and 12, more than 60% of the mean distance of transport after 48 h had R25 already been reached after 3 h. In one experiment (experiment 6), the seeds were transported R26 back when the wind direction changed after 1 day. Ditch 1 makes a 90-degree bend after R27 100 m. In the first experiment in this ditch (experiment 1), many seeds were deposited in this R28 bend on the day of release but were transported further, around the corner, when the wind R29 direction changed after that day. For experiment 12, the wind direction was opposite to the R30 -1 surface current on day 3. That day, the wind speed was low (0.32 m s ). In this case, since R31 the direction of seed transport was not recorded, it could not be related to the direction of R32 wind and water. However, wind direction was similar to that on the previous two days, and R33 the seeds did not seem to be transported back (see mean and minimum distance, Table 4.4, R34 experiment 12). R35 Figure 4.7 gives the relation between mean distance of seed deposition after 24 h and the mean net wind speed on the day of release and 1 day later. The direction of seed transport prior R36 to deposition was always similar to the net wind direction on the day of release. Experiments 1 R37 and 2 (see Table 4.4) were excluded from Fig. 4.7 as the ditch had a bend in it, and experiment R38 6 was excluded because the transport direction of the seeds changed after one day. R39 88 | Chapter 4

table 4.3. (continuing on the next page) Factors determining seed deposition: differences between R1 pairs of groups of independent variables in the corrected number of seeds per section. Significance R2 values (P) of Kruskal-Wallis tests (KW) for each independent variable and Mann-Whitney U (MWU) R3 tests for each pair of groups within the independent variables, with number of seeds per section (as the percentage of seeds not yet retrieved upstream) as the dependent variable, are presented for R4 the periods 3 h, 24 h and 48 h after release. In the column MWU/KW, the pairs of groups for which R5 the MWU test was performed are given. The column Results indicates which group is significantly smaller, larger or equal to the other group. ≈: no statistically significant difference between the R6 groups. Explanation of groups: Aquatic plant cover: 1=0-5%, 2=5-25%, 3=26-50%, 4=51-75%, 5=76- R7 95%, 6=96-100%. For MWU tests: P<0.0033 indicates a statistically significant difference (Bonferroni R8 corrected). Slope of the bank: 1=0-30º, 2=31-45º, 3=46-60º, 4=61-90º, Significance: P<0.0083. Shape of the bank: 1=indentation, 2=bulge, 3=straight, 4=inner bend, 5=outer bend. Significance: P<0.005 R9 (five classes)/P<0.0083 (four classes). Wind direction: 0=parallel to ditch, 1=in direction of bank, R10 2=away from the bank. Significance: P<0.017. Vegetation structure at the ditch bank: 1=pasture, 2=carr forest. Significance:P <0.05 R11 R12 R13 R14 R15 R16 R17 R18 R19 R20 R21 R22 R23 R24 R25 R26 R27 R28 R29 R30 R31 R32 R33 R34 R35 R36 R37 R38 R39 Dispersal and deposition of hydrochorous plant seeds | 89

3 h after release 24 h after release 48 h after release R1 Independent MWU/ P Results P Results P Results variable KW R2 Aquatic plant cover 1-2 <0.001 1<2 0.004 1≈2 <0.012 1≈2 R3 0-50 cm from ditch 1-3 <0.001 1<3 <0.001 1<3 <0.001 1<3 R4 bank 1-4 <0.001 1<4 <0.001 1<4 <0.001 1<4 R5 1-5 <0.001 1<5 <0.001 1<5 <0.001 1<5 1-6 <0.001 1<6 <0.001 1<6 <0.001 1<6 R6 2-3 <0.001 2<3 <0.001 2<3 0.106 2≈3 R7 2-4 <0.001 2<4 <0.001 2<4 <0.001 2<4 2-5 <0.001 2<5 <0.001 2<5 <0.001 2<5 R8 2-6 0.002 2<6 <0.001 2<6 <0.001 2<6 R9 3-4 0.002 3<4 0.001 3<4 <0.001 3<4 R10 3-5 <0.001 3<5 <0.001 3<5 <0.001 3<5 3-6 0.189 3≈6 <0.001 3<6 <0.001 3<6 R11 4-5 0.451 4≈5 0.080 4≈5 0.027 4≈5 R12 4-6 0.626 4≈6 0.003 4<6 0.074 4≈6 5-6 0.314 5≈6 0.107 5≈6 0.935 5≈6 R13 KW <0.001 <0.001 <0.001 R14 Slope of ditch bank 1-2 0.551 1≈2 0.793 1≈2 0.049 1≈2 R15 1-3 0.098 1≈3 0.007 1>3 <0.001 1>3 R16 1-4 0.032 1≈4 <0.001 1>4 <0.001 1>4 2-3 <0.001 2>3 <0.001 2>3 0.022 2≈3 R17 2-4 <0.001 2>4 <0.001 2>4 <0.001 2>4 R18 3-4 0.200 3≈4 <0.001 3>4 <0.001 3>4 KW <0.001 <0.001 <0.001 R19 Shape of ditch bank 1-2 0.001 1>2 <0.001 1>2 <0.001 1>2 R20 1-3 0.469 1≈3 <0.001 1>3 <0.001 1>3 R21 1-4 0.691 1≈4 0.588 1≈4 0.550 1≈4 R22 1-5 0.036 1≈5 - - - - 2-3 <0.001 2<3 0.005 2<3 0.001 2<3 R23 2-4 0.713 2≈4 0.793 2≈4 0.769 2≈4 R24 2-5 0.002 2<5 - - - - 3-4 0.786 3≈4 0.724 3≈4 0.682 3≈4 R25 3-5 0.020 3≈5 - - - - R26 4-5 0.153 4≈5 - - - - R27 KW 0.012 <0.001 <0.001 Wind direction 0-1 - † † R28 relative to ditch 0-2 - R29 bank 1-2 <0.001 1<2 R30 KW <0.001 R31 Vegetation structure 1-2 0.560 1≈2 <0.001 1<2 0.089 1≈2 at the ditch bank R32 R33 - No cases for one of the two groups, † The effect of wind direction was tested only for the day of release because the wind direction changed during the week R34 R35 R36 R37 R38 R39 90 | Chapter 4

R1 R2 R3 3 R4

R5 2 R6 R7

R8 1 R9 R10 3 R11 R12 2 mv -2 58 mv 62 R13 2 2 108 124 118 R14 R15 Minimum distance (m) Maximum distance (m) 1

R16 .

R17 A minus sign indicates that those seeds were deposited north of R18 ‡ 3 R19 R20

R21 2 The first number is the length from release point to culvert south of

R22 † R23 Mean distance (m) 88 88 83 20 22 24 100 152 152 29 mv 34 4 155 251 253 66 60 60 234 434 432 133 451 451 76 451 451 216 451 451 76 77 97 22 24 26 212 202 176 149 10 47 98 -10 ‡ -10 ‡ 152 98 156 35 40 52 0 273 296 306 57 61 61 405 405 405 398 400 404 107 101 101 405 405 405 373 384 375 113 271 301 387 395 393 372 391 395 89 117 185 403 405 405 R24 259 378 391 41 63 117 405 405 405 R25 R26 Day 1 R27 R28 R29 R30 Max. ditch distance 100/830 * 88/830 * 576 451 533 220/817 † 192 405 405 405 405 R31 405 R32 R33 number 1 1 2 2 2 3 3 4 4 4 4 R34 4 R35

R36 Mean, minimum and maximum distances of retraced deposited seeds, measured from the release point, for ditch 1 to 4 on day (3 R37 R38 able 4.4. Trial numberTrial Ditch 1 2 3 4 5 6 7 8 9 10 11 12 t the from the distance refers to distance ditch value. Max. mv=missing release). (48 h after day 3 and after release) (24 h day 2 h after release), release point to the culvert in direction of seed transport for that experiment. Ditch number identifies ditch which experiment was performed (see Fig. 4.1). R39 * The first number is the length of southeast-northwest stretch ditch from release point to bend, while second is the length of southwest-northeast stretch ditch. release point, which was the transport direction on day 1. The second number is length from point to culvert north of release point, because the directions of wind and transport changed on day 2. the release point, while on day 1 all seeds were deposited south of point Dispersal and deposition of hydrochorous plant seeds | 91

500 R1 R2 450 R3

400 R4 R5 350 R6 R7 300 R8

250 R9 R10 200 R11 R12 150

Average distance (m) 24 hours after release R13

100 R14 R2=0.644 R15 50 R16 R17 0 0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 R18 Net wind velocity (m s−1), averaged over day 1 and 2 R19 Figure 4.7. Mean distance of traced seeds 1 day after release, related to the net mean wind speed for the release day (day 1) and 1 day after release (day 2). Regression: y=280.62*ln(x) + 72.592 R20 R21 R22 4.4 dIsCussIon R23

Seed dispersal is important for the maintenance and enhancement of plant biodiversity in R24 fragmented habitats (e.g. Saunders et al., 1991; Primack & Miao, 1992; Soons et al., 2005; R25 Ozinga et al., 2009). The focus of our study was on hydrochorous seed dispersal via drainage R26 ditches, since ditches can serve as dispersal corridors for plant species with buoyant seeds R27 growing near or in ditches. Ditch banks can serve as important refuges for many wetland R28 plant species (Blomqvist et al., 2006). However, species richness at ditch banks is declining and, R29 despite restoration efforts, remnant populations are increasingly isolated (Blomqvist et al., R30 2003), threatening biodiversity. R31 Previous studies on hydrochory have focussed on ecosystems where water flow is rapid such as R32 rivers, streams and estuaries (e.g. Merritt & Wohl, 2002; Vogt, Rasran & Jensen, 2004; Gurnell R33 et al., 2008). In such ecosystems, water flow determines the speed and direction of seed R34 transport. R35 Our first hypothesis was upheld in that wind is the predominant determinant of the speed and direction of hydrochorous seed transport in drainage ditches. For all species, rate and R36 direction of seed transport were positively related to wind speed and direction, and not to R37 water flow at mid-depth (C. pseudocyperus). R38 R39 92 | Chapter 4

R1 It is known that surface current determines the rate of transport of floating seeds (Erftemeijer R2 et al., 2008) and that surface current is affected by wind shear stress (Laval et al., 2003; Stocker R3 & Imberger, 2003; Stevens et al., 2004). Disentangling the indirect effect of wind on the R4 surface current (via wind shear stress) and a potential direct effect of wind (‘sailing effect’) R5 on hydrochorous seeds is complicated, because wind speed and surface current are positively R6 related. In our study, the rate of transport of floating seeds was always greater than the surface current. Furthermore, seed velocity per unit surface current was higher for the larger R7 Sparganium erectum seeds, with a greater fraction of the seed volume protruding from the R8 water, than for the smaller and flat Carex pseudocyperus and Iris pseudacorus seeds. These R9 two results provide evidence that not only wind shear stress, but also a direct wind effect R10 (sailing effect), cause the movement of hydrochorous plant seeds in systems with very slow R11 water flow. R12 As already noted in the Methods section, we only used one species in each of the three size R13 categories. Strictly speaking, this does not allow for rigorously testing the hypothesis that the R14 effect of wind on seed transport rate would increase as the fraction of the seed protruding R15 from the water increased. To gain more insight into which seed traits determine seed transport R16 rate in slow flowing ditches, future research should focus on variation in weight, volume R17 and seed shape among a larger number of species. Nevertheless, this study is the first that R18 attempts to separate the relative importance of water and wind velocity for hydrochorous R19 seed dispersal in slow flowing ditches. Furthermore, we recommend further research into instances where the flow rate of water at mid-depth is higher (owing to pumping by water R20 managers after heavy rainfall) to determine whether there is a threshold above which water R21 flow counteracts the influence of wind in seed transport. If this is found not to be the case R22 it is unlikely that introducing measures that affect the flow of water will increase potential R23 dispersal distances. Further, it should be noted that when wind speed is very low, it is possible R24 that hydrochorous seeds are not transported by wind but by water. R25 The correlation between the rate of seed transport and wind speed does not necessarily imply R26 a relation between seed transport distance and net wind speed. For wind-dispersed seeds, R27 the positive relation between wind speed and dispersal distance is obvious (e.g. Skarpaas et R28 al., 2004; Soons et al., 2004). However, for hydrochorous seeds, it is possible that ditch bank R29 characteristics and ditch traits determine the deposition pattern of seeds, independent of the R30 speed at which the seeds are transported. Levine & Murrell (2003) stated that seed deposition R31 cannot, in general, be easily related to dispersal kernels (i.e. a probability distribution, usually R32 skewed, describing the spatial distribution of deposited seeds), but is better explained by seed-trapping elements in the landscape. For instance, Johansson & Nilsson (1993) found a R33 disproportionately high number of tagged ramets in bends and obstacles along a Swedish R34 river. Similarly, Schneider & Sharitz (1988) revealed that emergent substrata increased the R35 density of deposited seeds in a swamp forest. These results are in line with our findings that R36 the presence of abundant stands of aquatic plants and forested ditch banks increased the R37 probability that seeds will be deposited. Nevertheless, mean wind speed within 24 h after R38 release was positively correlated to the mean deposition distance of seeds 1 day after release. R39 Dispersal and deposition of hydrochorous plant seeds | 93

Thus, although seed-trapping landscape elements explain some of the deposition, wind speed R1 also determined the distance over which seeds are transported. R2 There was a greater probability of seed deposition 3 h after release in sections that were R3 not exposed to the wind than in other sections. Field observations showed that, when wind R4 direction was perpendicular to the long dimension of the ditch, the high banks on the wind- R5 exposed side of the ditch seemed to reverse the wind direction close to the water surface, R6 thereby transporting seeds to the opposite side of the ditch. Indentations in the bank were R7 found to promote seed deposition, while bulges decreased the probability that seeds would R8 be deposited. In summary, net wind speed determined the speed of seed transport. The R9 ultimate seed rain in our experiment, however, was determined by the combination of wind speed and direction with landscape elements. R10 In non-riverine ecosystems with drainage ditches, we found the effect of wind direction on R11 seed dispersal to differ markedly from the unidirectional hydrochorous dispersal in rivers. R12 The fact that wind direction and speed affect seed dispersal in slow flowing water bodies R13 provides extra opportunities for the restoration of ecosystems. Restored sites situated in all R14 directions from source sites potentially receive seeds from them because wind direction varies. R15 Nevertheless, it should be taken into account that there is often a predominant wind direction R16 that will influence most dispersal events. R17 In river systems, seed morphology may influence seed deposition because the size of seeds R18 affects the probability that they will be trapped by obstacles (Schneider & Sharitz, 1988). It R19 may be expected that this effect is also present in drainage ditches and that the additional R20 ‘sailing effect’ also influences the variation in the final deposition pattern of different plant R21 species seeds in such systems. Hence, we recommend that experiments be conducted that R22 are designed to reveal the effect of seed size on patterns of seed deposition and the spatial distribution of vegetation. R23 The mean distance of seed transport in our experiment ranged from 34 to 451 m, the R24 maximum dispersal distance being limited by culverts. These distances were in the same R25 order of magnitude as those found by Beltman et al. (2005) for water dispersing Carex elata R26 All. seeds in a comparable ecosystem. Although our experiments only lasted three days, it R27 is unlikely that longer experiments would have resulted in much longer dispersal distances, R28 considering that deposition distances did not differ substantially between day 2 and day 3 (see R29 Table 4.4). In rivers, seeds can be transported over considerably greater distances (Andersson R30 et al., 2000; Griffith & Forseth, 2002). Considering the dispersal distances found in the study R31 by Beltman et al. (2005) and in our study, and given the current configuration of ditches, R32 culverts and ponds in the floodplain of the Vecht River, it is unlikely that seed exchange would R33 occur between polders by hydrochorous dispersal. However, a general problem in dispersal R34 experiments is the difficulty of measuring long-distance dispersal (Skarpaas et al., 2004), R35 which may occur due to extreme events such as storms. Furthermore, besides hydrochory, zoochory (e.g. seed dispersal by waterfowl or fish) can cause long-distance dispersal in R36 aquatic plant species (Pollux, Santamaria & Ouborg, 2005). Therefore, it should be noted that R37 the distances found in our experiment might be lower than the actual dispersal distances. R38 R39 94 | Chapter 4

R1 The dispersal range found in drainage ditches in our study is greater than when dispersal R2 of unplumed seeds takes place exclusively by anemochory (i.e. wind dispersal) (Boedeltje et R3 al., 2003; Soons et al., 2004). This stresses the importance of hydrochory for the connectivity R4 between wetland populations. In that light, it is important to maintain sufficient surface water R5 infrastructures that enable hydrochory. Ozinga et al. (2009) found an over-representation of R6 water-dispersed species among declining species, and concluded that a degraded dispersal infrastructure explains plant diversity losses. In both our study and the study by Beltman et R7 al. (2005), culverts limited seed transport. We found seeds at the end of a ditch where it was R8 connected to another one via a culvert. Yet they were never transported through the culvert R9 because aquatic plants were abundant in front of the culverts and partly obstructed the R10 passage. Thus, we expect that hydrochorous dispersal in areas with drainage ditches can be R11 promoted by creating wider passages between water bodies, thereby improving the dispersal R12 infrastructure for hydrochorous species. Dispersal over longer distances could thus improve R13 the connectivity between remnants of native vegetation in a matrix of agricultural landscape R14 [see also the review in Trakhtenbrot et al. (2005)] and thereby reduce the negative effects of R15 fragmentation, as reviewed by Saunders et al. (1991). R16 For many of the deposition experiments, most of the final distance was covered within 24 h. R17 However, if the wind turned after 1 day, the seeds could be dislodged and transported in the R18 opposite direction (Table 4.4, experiment 6). A changing wind direction also enabled seed R19 transport around the corner of a curved ditch (Table 4.4, experiment 1). This indicates that connecting parallel ditches by ditches perpendicular to the main orientation of the existing R20 ditch may promote dispersal over greater distances than would be reached by anemochory R21 alone (see also Beltman et al. (2005)). R22 Factors that promote seed deposition might prevent seeds from being transported over long R23 distances, and vice versa. If the original vegetation is to be restored at created wetlands or sites R24 otherwise intended for the re-establishment of a natural vegetation, then the characteristics R25 of the ditch banks at these sites should be suitable for seed deposition. Seed deposition at R26 such sites can be promoted by a shallow bank slope, irregular banks (indentations) and a high R27 abundance of aquatic plants. However, an abundance of aquatic plants or helophytes can R28 block the transportation of diaspores of target species. Therefore, it should be noted that, R29 with ongoing vegetation succession, dispersal success can decrease. Clearing of ditches that R30 are situated between source populations and habitats to be recolonised can promote the seed R31 transport process between such areas. R32 Because ditch banks serve as refuges for many wetland species, it is important to improve dispersal between remnant wetland populations to slow down and ultimately reverse the R33 decline in species richness. Currently, drainage ditches are mainly situated parallel to one R34 another, without an abundant network of cross-cutting connections between ditches. Our R35 results indicate that, for optimal seed transport between otherwise fragmented wetland R36 remnants, it is beneficial to design connections between drainage ditches in all directions, R37 to maintain an open landscape (which promotes the influence of wind) and to clear ditches R38 regularly. R39 Dispersal and deposition of hydrochorous plant seeds | 95

acknowledgments R1 The authors thank the following people: Jolanda Serné for helping with painting the seeds; R2 Kjell ‘t Hoen for helping with the figures; Pita Verweij and Marieke Schouten for assistance R3 with the field work; Bert van Dijk (Staatsbosbeheer) for giving permission to perform the R4 experiments in Polder Westbroek; and Pieter Filius (Water Board ‘Velt en Vecht’) for giving R5 permission to use their SonTek FlowTracker. We also thank two anonymous referees and Pita R6 Verweij for their comments on an earlier version of the manuscript. R7 R8 R9 R10 R11 R12 R13 R14 R15 R16 R17 R18 R19 R20 R21 R22 R23 R24 R25 R26 R27 R28 R29 R30 R31 R32 R33 R34 R35 R36 R37 R38 R39 96 | Chapter 4

R1 reFerenCes R2 R3 Andersson, E., Nilsson, C., & Johansson, M.E. (2000) Plant dispersal in boreal rivers and its relation to the diversity of riparian flora.Journal of Biogeography, 27, 1095-1106. R4 Beltman, B., Van Den Broek, T., & Vergeer, P. (2005) The limited success of peat pond restoration R5 [Het beperkte success van laagveenrestauratie]. Landschap, 4, 173-179. Blomqvist, M.M., Tamis, W.L.M., Bakker, J.P., & van der Meijden, E. (2006) Seed and (micro)site R6 limitation in ditch banks: Germination, establishment and survival under different R7 management regimes. Journal for Nature Conservation, 14, 16-33. R8 Blomqvist, M.M., Vos, P., Klinkhamer, P.G.L., & Ter Keurs, W.J. (2003) Declining plant species richness of grassland ditch banks - A problem of colonisation or extinction? Biological Conservation, R9 109, 391-406 R10 Boedeltje G., Bakker J.P., Bekker R.M., van Groenendael J.M. & Soesbergen M. (2003) Plant dispersal in a lowland stream in relation to occurrence and three specific life-history traits of the R11 species in the species pool. Journal of Ecology, 91, 855-866 R12 Boedeltje, G., Bakker, J.P., Ten Brinke, A., Van Groenendael, J.M., & Soesbergen, M. (2004) Dispersal R13 phenology of hydrochorous plants in relation to discharge, seed release time and buoyancy of seeds: the flood pulse concept supported.Journal of Ecology, 92, 786-796. R14 Bootsma, M.C. (2000) Stress and recovery in wetland ecosystems, PhD thesis, Utrecht University, R15 Utrecht. Borger, G.J. (1992). Draining, digging, dredging; the creation of a new landscape in the peat areas R16 of the low countries. In Fens and Bogs in the Netherlands: Vegetation, History, Nutrient R17 Dynamics and Conservation. (ed J.T.A. Verhoeven), pp. 131-171. Kluwer Academic R18 Publishers, Dordrecht. Bulle, M., van Groenendael, J.M., & Jurgens, C.R. (1994) Helofyten in het Wageningse Binnenveld. R19 Voorkomen in relatie tot dispersiekenmerken. [Helophytes in “het Wageningse R20 Binnenveld”]. Landschap, 11, 19-28. Cappers, R.T.J., Bekker, R.M. & Jans, J.E.A. (2006) Digitale Zadenatlas van Nederland (Digital Seed R21 Atlas of the Netherlands). Barkhuis Publishing & Groningen University Library, Groningen R22 (www.seedatlas.nl). R23 Chang, E.R., Veeneklaas, R.M., & Bakker, J.P. (2007) Seed dynamics linked to variability in movement of tidal water. Journal of Vegetation Science, 18, 253-262. R24 Chang, E.R., Veeneklaas, R.M., Buitenwerf, R., Bakker, J.P., & Bouma, T.J. (2008) To move or not to R25 move: Determinants of seed retention in a tidal marsh. Functional Ecology, 22, 720-727. Danvind, M. & Nilsson, C. (1997) Seed floating ability and distribution of alpine plants along a R26 northern Swedish river. Journal of Vegetation Science, 8, 271-276. R27 Daubenmire, R.F. (1968) Plant Communities: a Textbook of Plant Synecology Harper & Row, New R28 York. Erftemeijer, P.L.A., Van Beek, J.K.L., Ochieng, C.A., Jager, Z., & Los, H.J. (2008) Eelgrass seed dispersal R29 via floating generative shoots in the Dutch Wadden Sea: A model approach. Marine R30 Ecology Progress Series, 358, 115-124. Goodson, J.M., Gurnell, A.M., Angold, P.G., & Morrissey, I.P. (2003) Evidence for hydrochory and R31 the deposition of viable seeds within winter flow-deposited sediments: The River Dove, R32 Derbyshire, UK. River Research and Applications, 19, 317-334. R33 Gornall, R.J., Hollingsworth, P.M., & Preston, C.D. (1998) Evidence for spatial structure and directional gene flow in a population of an aquatic plant, Potamogeton coloratusHeredity , 80, 414- R34 421 R35 Griffith, A.B. & Forseth, I.N. (2002) Primary and secondary seed dispersal of a rare, tidal wetland annual, Aeschynomene virginica. Wetlands, 22, 696-704. R36 Gurnell, A., Thompson, K., Goodson, G., & Moggridge, H. (2008) Propagule deposition along river R37 margins: linking hydrology and ecology. Journal of Ecology, 96, 553-565. R38 Herschy, R.W. (1999). Flow Measurement. In Hydrometry. Principles and Practices (ed R.W. Herschy). pp. 9-84. John Wiley & Sons, Chichester. R39 Dispersal and deposition of hydrochorous plant seeds | 97

Jansson, R., Zinko, U., Merritt, D.M., & Nilsson, C. (2005) Hydrochory increases riparian plant species richness: a comparison between a free-flowing and a regulated river.Journal of Ecology, R1 93, 1094-1103. R2 Johansson, M.A. & Nilsson, C. (1993) Hydrochory, population dynamics and distribution of the clonal R3 aquatic plant Ranunculus lingua. Journal of Ecology, 81, 81-91. KNMI (2008) Daggegevens van het weer in Nederland. Station de Bilt. KNMI, De Bilt (www.knmi.nl). R4 Laval, B., Imberger, J., Hodges, B.R., & Stocker, R. (2003) Modeling circulation in lakes: Spatial and R5 temporal variations. Limnology and Oceanography, 48, 983-994. Levin, S.A., Muller-Landau, H.C., Nathan, R., & Chave, J. (2003) The ecology and evolution of seed R6 dispersal: a theoretical perspective. Annual Review of Ecology, Evolution, and Systematics, R7 34, 575-604. R8 Levine, J.M. & Murrell, D.J. (2003) The Community-Level Consequences of Seed Dispersal Patterns. Annual Review of Ecology, Evolution, and Systematics, 34, 549-574. R9 Liu, K., Eastwood, R.J., Flynn, S., Turner, R.M., and Stuppy, W.H. (2008) Seed Information Database R10 (release 7.1, May 2008) http://www.kew.org/data/sid McCartney, M.P. & de la Hera, A. (2004) Hydrological assessment for wetland conservation at Wicken R11 Fen. Wetlands Ecology and Management, 12, 189-204. R12 Merritt, D.M. & Wohl, E.E. (2002) Processes governing hydrochory along rivers: Hydraulics, hydrology, R13 and dispersal phenology. Ecological Applications, 12, 1071-1087. Middleton, B. (2000) Hydrochory, seed banks and regeneration dynamics along the landscape R14 boundaries of a forested wetland. Plant Ecology, 146 169-184. R15 Middleton, B., van Diggelen, R., & Jensen, K. (2006) Seed dispersal in fens. Applied Vegetation Science, 9, 279-284. R16 Milsom, T.P., Sherwood, A.J., Rose, S.C., Town, S.J., & Runham, S.R. (2004) Dynamics and management R17 of plant communities in ditches bordering arable fenland in eastern England. Agriculture, R18 Ecosystems & Environment, 103, 85-99. Moss, B. (1983) The Norfolk Broadland: experiments in the restoration of a complex wetland. R19 Biological Reviews, 58, 521-561. R20 Neff, K.P. & Baldwin, A.H. (2005) Seed dispersal into wetlands: Techniques and results for a restored tidal freshwater marsh. Wetlands, 25, 392-404. R21 Ozinga, W.A., Römermann, C., Bekker, R.M., Prinzing, A., Tamis, W.L.M., Schaminée, J.H.J., R22 Hennekens, S.M., Thompson, K., Poschlod, P., Kleyer, M., Bakker, J.P., & van Groenendael, R23 J.M. (2009) Dispersal failure contributes to plant losses in NW Europe. Ecology Letters, 12, 66-74. R24 Pollux, B.J.A., Santamaria, L., & Ouborg, N.J. (2005) Differences in endozoochorous dispersal R25 between aquatic plant species, with reference to plant population persistence in rivers. Freshwater Biology, 50, 232-242. R26 Primack, R.B. & Miao, S.L. (1992) Dispersal can limit local plant distribution. Conservation Biology, R27 6, 513-519 R28 Saunders, D.A., Hobbs, R.J., & Margules, C.R. (1991) Biological Consequences of Ecosystem Fragmentation - A Review. Conservation Biology, 5, 18-32. R29 Schneider, R.L. & Sharitz, R.R. (1988) Hydrochory and regeneration in a bald cypress-water tupelo R30 swamp forest. Ecology, 69, 1055-1063. Skarpaas, O., Stabbetorp, O.E., Ronning, I., & Svennungsen, T.O. (2004) How far can a hawk’s beard R31 fly? Measuring and modelling the dispersal of Crepis praemorsa. Journal of Ecology, 92, R32 747-757. R33 SonTek (2001) FlowTracker Handheld ADV Technical Documentation, SonTek/YSI, Inc., San Diego, CA. R34 Soons, M.B. & Heil, G.W. (2002) Reduced colonization capacity in fragmented populations of wind- R35 dispersed grassland forbs. Journal of Ecology, 90, 1033-1043. Soons, M.B., Heil, G.W., Nathan, R., & Katul, G.G. (2004) Determinants of long-distance seed dispersal R36 by wind in grasslands. Ecology, 85, 3056-3068. R37 Soons, M.B., Messelink J.H., Jongejans E., & Heil, G.W. (2005) Habitat fragmentation reduces R38 grassland connectivity for both short-distance and long-distance wind-dispersed forbs. Journal of Ecology, 93, 1214-1225. R39 98 | Chapter 4

Stevens, C.L., Lawrence, G.A., & Hamblin, P.F. (2004) Horizontal dispersion in the surface layer of a R1 long narrow lake. Journal of Environmental Engineering and Science, 3, 413-417. R2 Stocker, R. & Imberger, J. (2003) Horizontal transport and dispersion in the surface layer of a medium- R3 sized lake. Limnology and Oceanography, 48, 971-982. Tockner, K. & Stanford, J.A. (2002) Riverine flood plains: present state and future trends. R4 Environmental Conservation, 29, 308-330. Chapter 5 R5 Trakhtenbrot, A., Nathan, R., Perry, G., Richardson, D.M., & 2005 (2005) The importance of long- distance dispersal in biodiversity conservation. Diversity and Distributions, 11, 173-181. R6 Trepel, M. & Kluge, W. (2002) Ecohydrological characterisation of a degenerated valley peatland in R7 northern Germany for use in restoration. Journal for Nature Conservation, 10, 155-169. R8 Venable, D.L. & Brown, J.S. (1993) The population-dynamic function of seed dispersal. Vegetatio, 108, 31-55. R9 Verhoeven, J.T.A., Koerselman, W., & Beltman, B. (1988). The vegetation of fens in relation to their R10 hydrology and nutrient dynamics: a case study. In Vegetation of Inland Waters (ed J.J. Symoens), pp. 249-282 Kluwer Acadmic Publishers, Dordrecht. R11 Vogt, K., Rasran, L., & Jensen, K. (2004) Water-borne seed transport and seed deposition during R12 flooding in a small river-valley in Northern Germany.Flora: Morphologie, Geobotanik, R13 Oekophysiologie, 199, 377-388. Vogt, K., Rasran, L., & Jensen, K. (2006) Seed deposition in drift lines during an extreme flooding R14 event - Evidence for hydrochorous dispersal? Basic and Applied Ecology, 7, 422-432. R15 Wassen, M.J., Barendregt, A., Schot, P.P., & Beltman, B. (1990) Dependency of local mesotrophic fens on a regional groundwater flow system in a poldered river plain in the Netherlands. R16 Landscape Ecology, 5, 21-38. R17 Wassen, M.J., Van Diggelen, R., Wolejko, L., & Verhoeven, J.T.A. (1996) A comparison of fens in R18 natural and artificial landscapes.Vegetatio , 126, 5-26. Wilcove, D.S., McLellan, C.H., & Dobson, A.P. (1986). Habitat fragmentation in the temperate zone. R19 In Conservation Biology. The Science of Scarcity and Diversity. (ed M.E. Soulé), pp. 237-256. R20 Sinauer Associates, Sunderland, Massachusetts. Wright, S. (1943) Isolation by distance. Genetics, 28, 114-38. R21 R22 R23 R24 R25 R26 R27 R28 R29 R30 R31 R32 R33 R34 R35 R36 R37 R38 R39 Chapter 5

linking habitat suitability and seed dispersal models in order to analyse the effectiveness of hydrological fen restoration strategies

van Loon, A.H., Soomers, H., Schot, P.P., Bierkens, M.F.P., Griffioen, J., and Wassen, M.J. 2011. Biological Conservation 144: 1025-1035. 100 | Chapter 5

R1 abstraCt R2 R3 The effectiveness of measures targeted at the restoration of populations of endangered species R4 in anthropogenically dominated regions is often limited by a combination of insufficient R5 restoration of habitat quality and dispersal failure. Therefore, the joint prediction of suitable R6 habitat and seed dispersal in dependency of management actions is required for effective nature management. Here we demonstrate an approach, which links a habitat suitability and R7 a seed dispersal model. The linked model describes potential species distribution as a function R8 of current species distribution, species-specific dispersal traits, the number of successful R9 dispersal events, dispersal infrastructure and habitat configuration, the last two being related R10 to water management actions. We demonstrate the applicability of the model in a strategy R11 analysis of hydrological restoration measures for a large fen area in which still numerous R12 endangered plant species grow. R13 R14 With the aid of the linked model, we were able to optimise the spatial planning of restoration R15 measures, taking into account both the constraints of water management practices on R16 abiotic restoration and the effects of habitat fragmentation on dispersal. Moreover, we R17 could demonstrate that stand-alone habitat suitability models, which assume unlimited R18 dispersal, may considerably overestimate restoration prospects. For these reasons, we R19 conclude that linked habitat suitability and dispersal models can provide useful insights into spatially differentiated potentials and constraints of nature restoration measures targeted R20 at the sustainable conservation of endangered plant populations whose habitats have R21 been deteriorated due to undesirable effects of land and water management on abiotic R22 conditions. These insights may contribute to the design of cost-effective nature restoration R23 and conservation measures. R24 R25 R26 R27 R28 R29 R30 R31 R32 R33 R34 R35 R36 R37 R38 R39 Linking habitat suitability and seed dispersal models | 101

5.1 IntroduCtIon R1 R2 The sustainable conservation of endangered plant species and the restoration of their habitats R3 were agreed upon internationally in the early 1990s (Millennium Ecosystem Assessment, 2005). R4 A recent analysis of Northwest Europe, however, showed only modest progress in the reduction R5 of the negative effects that habitat loss and fragmentation have on the viability of endangered R6 vascular plant species (Ozinga et al., 2009). Habitat fragmentation, i.e., the breaking apart R7 of large contiguous habitat patches into multiple smaller ones, is disadvantageous for both R8 plant and animal populations, because it results in reduced population sizes, increased edge- R9 effect and reduced habitat connectivity (Ewers and Didham, 2005). Habitat fragmentation, in particular reduced habitat connectivity, may also hinder plants from re-establishing at R10 successfully restored habitat patches due to the limited success of seed dispersal by wind R11 (Soons et al., 2005), water (Boedeltje et al., 2003; Soomers et al., 2010), animal migration R12 (Soons et al., 2008), or anthropogenic activities (Klimkowska et al., 2007). For these reasons, R13 nature restoration measures should aim at the restoration of habitats in such a way that R14 habitat patches can be reconnected (Hooftman et al., 2004; Ozinga et al., 2009). R15 R16 In restoration ecology, habitat suitability and meta-population models can be helpful for R17 predicting changes in species distribution in response to abiotic restoration scenarios. Habitat R18 suitability models describe the probability of plant species occurrence by using tables or R19 regression methods in which habitat conditions serve as the explaining variables and plant R20 species as the dependent variables (Olde Venterink and Wassen, 1997; Schröder et al., 2008; R21 Segurado and Araújo, 2004). When the explaining input variables are derived from physical or R22 chemical speciation models, the effectiveness of abiotic restoration measures can be predicted (Van Ek et al., 2000). However, habitat suitability models usually overestimate species R23 distribution because they assume that dispersal is unlimited; i.e., they ignore the effects of R24 limited dispersal on species distribution in fragmented landscapes (Ozinga et al., 2005). R25 R26 Meta-population models, in contrast, describe population dynamics and viability by solving R27 the dynamic equilibrium between dispersal and local extinction mathematically, using (among R28 others) habitat maps as the input variables (Fagan and Lutscher, 2006; Hanski, 2008; Purves R29 and Dushoff, 2005). Since these habitat maps are derived from field observations and not R30 from physical or chemical speciation models, meta-population models lack the flexibility to R31 predict the response of ecosystems to abiotic restoration measures. R32 R33 To overcome the constraints of both modelling approaches, recent studies have linked habitat R34 suitability and dispersal models in order to evaluate the effects of climate change scenarios R35 on the potential distribution of tree species (Engler and Guisan, 2009; Iverson et al., 2004; Scheller and Mladenoff, 2008). To our knowledge, however, linked habitat suitability and seed R36 dispersal models have not yet been used in restoration ecology for analysing the effectiveness R37 of abiotic restoration measures. Such linkage effectively combines knowledge about physical R38 R39 102 | Chapter 5

R1 or chemical processes determining habitat suitability with seed dispersal processes determining R2 the probability of establishment of target species. It may therefore be helpful for optimising R3 abiotic restoration measures in terms of the restoration of both suitable habitats and sufficient R4 connectivity between habitat patches for viable plant populations. This is particularly true for R5 populations of endangered plant or animal species who´s habitats have been deteriorated R6 due to undesirable effects of land and water management, like increased levels of pesticides (Dormann et al., 2007) and nutrients (Kardol et al., 2008) in abandoned agricultural fields, R7 or altered flooding frequencies (Beumer et al., 2008) and reduced groundwater supply R8 (Barendregt et al., 1995) in managed wetlands. R9

R10 This paper presents a linked habitat suitability and seed dispersal model that predicts potential R11 species distribution as a function of current species distribution, species-specific dispersal R12 traits, dispersal infrastructure and habitat configuration, the last two being related to water R13 management actions. The habitat suitability part of the model is used to assess the regional R14 habitat configuration of fen plant species, while the seed dispersal part of the model is used R15 to identify the habitat patches that are accessible to fen plant species via natural dispersal. R16 The objective of this study was to demonstrate that linked habitat suitability and dispersal R17 models are a powerful tool to optimise the design and spatial planning of nature restoration R18 and conservation measures. R19 R20 5.2 tHe Case R21 R22 The applicability of linked habitat suitability and seed dispersal models in restoration ecology is R23 demonstrated by a strategy analysis of hydrological restoration measures for a c. 150 km2 large R24 low-productive fen area in The Netherlands. We define low-productive fens as all vegetations R25 belonging to the alliance Caricion davallianae, i.e., species-rich fens with open low-growing R26 sedge vegetation (Schaminée et al., 1995). We focus on low-productive fens, because their R27 restoration is, like many other ecosystems, often constrained by both insufficient restoration R28 of habitat quality and dispersal limitation. Another reason to focus on low-productive fens R29 is that the restoration of their habitat strongly depends on the feasibility of hydrological R30 restoration measures (Boomer and Bedford, 2008; Van Wirdum, 1991). This is because many R31 fen plants are typical for sites with a low-nutrient availability (Bedford et al., 1999), a near- R32 neutral pH (Sjörs and Gunnarsson, 2002), and shallow groundwater levels. These site factors are usually associated with a supply of alkaline, nutrient-poor groundwater (Boomer and R33 Bedford, 2008; Almendinger and Leete, 1998; Boyer and Wheeler, 1989). R34 R35 Given the negative effects of water management on the groundwater supply of fens (Van R36 Loon et al., 2009c), substantial changes in water management actions are needed for the R37 sustainable restoration of low-productive fens in anthropogenically dominated regions. R38 R39 Linking habitat suitability and seed dispersal models | 103

Our study region is situated in the Vechtstreek area in The Netherlands (52o10’N and 5o10’E; R1 Fig. 5.1) and covers an area of c. 35 km2. This study region was selected, because (1) remnant R2 populations of low-productive fen plant species persist in fragmented nature reserves, and R3 (2) intensive water management of the study region and its surroundings constrains the R4 availability of fen habitat by reducing the availability of groundwater for fen plants. R5 R6 Originally, the Vechtstreek area was a vast lowland mire that comprised, among others, R7 vast low-productive fens. These fens were provided with groundwater that was laterally R8 redistributed from rather narrow exfiltration zones at the upstream fen margins by lateral flow R9 through the fen root-zone, i.e., throughflow (Van Loon et al., 2009b and c). Since the Middle Ages, the Vechtstreek area was reclaimed through the installation of drainage networks. R10 Subsequent land subsidence has resulted in the current ground surface levels being below sea R11 level (Fig. 5.1a). This has resulted in a complex spatial configuration of exfiltration zones, and R12 exfiltrating groundwater has become increasingly intercepted by drainage networks instead R13 of being laterally redistributed by throughflow (Van Loon et al., 2009a and c). Particularly the R14 latter is thought to contribute to the fragmented configuration of habitat remnants within a R15 matrix unsuitable for fen plants. R16 R17 Nowadays, the Vechtstreek area comprises agricultural fields, urban areas, lakes and fen R18 reserves. Fen reserves are comprised of floating fens and terrestrial fens. Floating fens are R19 early-successional stages of terrestrial ecosystems that prevail in ponds with minimal wave R20 action and receive sufficient groundwater to mediate minerotrophic (groundwater-like), R21 nutrient-poor conditions (Schaminée et al., 1995). Terrestrial fens with a nutrient-poor, organic R22 topsoil can have a species composition comparable to that of floating fens, provided that they receive groundwater. Both types of low-productive fens can be maintained for up to several R23 decades if succession is slowed by mowing (Van Diggelen et al., 1996). R24 R25 Water levels in the Vechtstreek area are intensively controlled to accommodate multiple land R26 uses. Intensively drained water management districts, called polders, have been established R27 for this purpose. The surface water levels of these polders can be controlled independently R28 from each other. During dry summer periods, surface water is supplied from the River Vecht R29 via the ditch network to the polders in order to irrigate crops and nature reserves. Due to R30 pollution, this supply water is unsuitable for fen plants. During wet periods, superfluous water R31 is drained from the polders into the River Vecht. R32 R33 The western boundary of the study area is formed by Lake Loosdrecht and the deep agricultural R34 polders Horstermeer and Bethune (Fig. 5.1). The lake level is maintained at -1 m asl (above R35 sea level ) by means of active water management. The deep agricultural polders Horstermeer and Bethune are reclaimed lakes with surface elevations ranging from -3 to -4 m asl. Due to R36 their low topographic position, both polders drain large amounts of groundwater and surface R37 water from the river valley (Schot and Molenaar, 1992; Wassen et al., 1990). Because of the R38 R39 104 | Chapter 5

R1 intensive agricultural activities in these two polders, nutrients have accumulated in the soils, R2 making them unsuitable for low-productive fen plants. R3 R4 A R5 R6 R7 R8 R9 R10 R11 R12 R13 R14 R15 R16 R17 R18 R19 R20 R21 R22 R23 R24 R25 R26 R27 R28 R29 R30 R31 R32 R33 R34 R35 R36 R37 R38 R39 Linking habitat suitability and seed dispersal models | 105

B R1 R2 R3 R4 R5 R6 R7 R8 R9 R10 R11 R12 R13 R14 R15 R16 R17 R18 R19 R20 R21 R22 R23 R24 R25 R26 R27 R28 R29 R30 Figure 5.1. Topographic features of the Gooi- and Vechtstreek area (The Netherlands). (A) (left page) Surface elevation and spatial aspects of the restoration strategies, (B) Impression of the network of R31 drainage ditches that provides an infrastructure for hydrochorous dispersal. R32 R33 The study area is bordered to the east by the ice-pushed ridge Het Gooi. This ridge consists of R34 elongated sandy hills that have a high permeability. The surface elevation of the ridge ranges R35 from 0 to 30 m asl. Because of the ridge’s relatively high topographic position, groundwater flows from the ice-pushed ridge towards the river valley (Van Loon et al., 2009c). Numerous R36 wells abstract in total 15 million m3 groundwater per year for the production of drinking R37 water, which reduces the supply of groundwater to the fen reserves in the river valley. R38 R39 106 | Chapter 5

R1 Groundwater flows through unconsolidated, sandy aquifers (Van de Meene et al., 1988; for R2 a cross-section see Van Loon et al., 2009c). The hydrological base of the study area consists of R3 early Pleistocene clays of marine origin at -250 to -150 m asl. Discontinuous resistance layers R4 consisting of fluvial clays intercalate the aquifers laterally. The ice-pushed ridge consists of R5 coarse sands that are partly intercalated with sloping clay sheets in the east. A semi-confining R6 peat layer with a thickness of 0.8 to 1.0 m is present in the river valley. Peat is not present in the ice-pushed ridge. R7 R8 R9 5.3 metHods R10

R11 5.3.1 General approach R12 We linked habitat suitability and seed dispersal models in order to predict the potential R13 distribution of fen plant species as a function of water management actions and the number R14 of successive and successful dispersal events. As many low-productive fen plants usually grow R15 at sites that are supplied with groundwater, we defined potential fen habitat as permanently R16 wet, groundwater-fed sites with a morphology that allows for the development of a low- R17 productive fen. By this definition, potential fen habitat consists of (1) ponds that receive R18 exfiltrating groundwater and (2) terrestrial areas that are usually groundwater saturated R19 because of groundwater surplus either by exfiltration or by throughflow. We assumed that drainage ditches do not provide habitat for fen plants because the development of a R20 floating fen is prevented by removal of aquatic plants from the ditches. Because of the above R21 mentioned hydrological controls on fen habitat, we used a groundwater model to predict R22 habitat suitability and, thereby, the configuration of potential fen habitat. The seed dispersal R23 model was used to identify which habitat patches can be potentially colonised by target R24 species via seed dispersal from existing populations. R25 R26 The linked models were used to analyse the effectiveness of hydrological restoration strategies R27 for increasing the current viability of endangered fen plant species in the Vechtstreek area. R28 For this purpose, we considered species viability to increase with (1) an increasing total R29 area of subpopulations, (2) increasing subpopulation sizes and (3) decreasing edge effect R30 on subpopulations, i.e., an increasing area with regard to edge-length of subpopulations. R31 These general indicators of habitat loss and fragmentation (Fahrig, 2003) were determined R32 by modelling the potential distribution of Carex diandra Schrank, a small sedge species. C. diandra was selected as the focal species, because (1) it is a characteristic species of the low- R33 productive fen plant association Caricion davallianea (Schaminée et al., 1995) considered here, R34 (2) it is one of the fen species that is greatly affected by water management practices due to R35 its preference for minerotrophic, wet habitat patches (Van Wirdum, 1991; Wassen et al., 1992; R36 Wheeler and Shaw, 1995), (3) it persists in only 34 highly fragmented remnant populations R37 across the study area (Wassen et al., 1990), and (4) sufficient data are available to compile an R38 area-covering, parcel-scale species distribution map (unpublished data Provinces of Utrecht R39 Linking habitat suitability and seed dispersal models | 107

and North Holland). Note that species-specific dispersal traits, like seed size, weight and shape, R1 of many other small sedges that dominate in low-productive fens are comparable to that of R2 C. diandra (Kleyer et al, 2008; Cappers et al., 2006; Van den Broek et al., 2005). Therefore, we R3 expect that our analysis of the effectiveness of hydrological restoration strategies by using C. R4 diandra as the focal species is representative for a range of characteristic low-productive fen R5 species. R6 R7 The hydrological restoration strategies that were considered are (1) Strategy I: inundation of R8 the deep agricultural polders Horstermeer and Bethune in order to redirect part of the 100 × R9 103 m3/d groundwater that currently discharges into these polders towards the fen surface; (2) Strategy II: closure of all abstraction wells in the river valley and the ridge in order to redirect R10 part of the 40 × 103 m3/d groundwater that is currently intercepted by wells towards the fen R11 surface; (3) Strategy III: elimination of all drainage elements from the upstream fen margins, R12 i.e., a 0.5–2 km wide belt adjacent to the ridge; (4) Strategy IV: elimination of all drainage R13 elements across the study area (both Strategies II and IV aim at reducing diffuse groundwater R14 losses by drainage); and (5) Strategy V: all measures mentioned in the Strategies I–IV. Details R15 of these strategies are provided in Figure 5.1. R16 R17 5.3.2 Groundwater modelling R18 Zones of groundwater supply were modelled during three successive stages. Firstly, we R19 modelled groundwater flow for each restoration strategy using 3-dimensional, steady-state R20 groundwater models based on the MODFLOW-1988 code (McDonald and Harbaugh, 1988). R21 This model was constructed at a 25 × 25 m resolution and it includes a highly conductive root- R22 zone layer to enable the modelling of throughflow of exfiltrated groundwater cf. Van Loon et al. (2009b). Further details about the model structure, data sources used for parameter R23 estimation and model performance are provided in Appendix A. Next, we determined R24 throughflow patterns of exfiltrated groundwater using a MODPATH particle tracking analysis R25 (Pollock, 1994). Finally, the modelled zones with groundwater supply (by direct exfiltration R26 or by throughflow) were down-scaled to a resolution of 2.5 × 2.5 m by assigning exfiltration R27 fluxes to surface water elements within the model cell, or by distributing them evenly over R28 the fen surface in the absence of surface water elements. Wet habitat patches were identified R29 from modelled groundwater tables. We defined wet as having a groundwater table exceeding R30 the lower border of the root zone layer, i.e., 0,3 m below ground surface. The restoration R31 strategies were simulated by adjusting the groundwater model according to the related R32 changes in hydrological characteristics. R33 R34 5.3.3 Seed dispersal modelling R35 Both anemochorous and hydrochorous seed dispersal of C. diandra were modelled for each restoration strategy using a 2-dimensional stepping-stone model that considered a successive R36 number of dispersal stages (Fig. 5.2). In the first dispersal stage, the seed shadows of the R37 current populations of C. diandra were calculated and new populations were assumed to be R38 R39 108 | Chapter 5

R1 established at habitat patches in the seed shadow. The second and following dispersal stages R2 repeated this procedure, with seed shadows being calculated from each new population. New R3 dispersal stages were initiated until seed dispersal no longer resulted in the establishment of R4 new populations. We analysed the configuration of the populations for each dispersal stage, R5 i.e., after successive dispersal events. R6 Anemochory R7 Anemochory is the dispersal of seeds by wind. Like other Carex species, C. diandra has R8 plumeless seeds that are not specifically adapted to long distance anemochory. Nevertheless, R9 plumeless seeds can disperse over distances varying from several metres to several kilometres, R10 depending mainly on wind velocity and species-specific dispersal traits, i.e., release height and, R11 particularly, seed terminal velocity (Soons, 2006). We derived the release height of C. Diandra R12 from Van der Meijden, 1996. The terminal velocity of the seeds of C. diandra was determined R13 using an experimental setup conform Soons and Heil (2002). Using this setup, seed terminal R14 velocity was calculated using the gravitational acceleration, drop height and drop time. R15 Drop time was measured for 20 seeds (cf. Soons and Heil, 2002) randomly selected from 5 R16 populations situated in the study area. We established an average terminal velocity of 2.58 m/s R17 (standard deviation: 0.47). Seeds with a terminal velocity greater than 2 m/s generally do not R18 exceed distances of several metres when dispersing via wind (Soons, 2006). More specifically, R19 the 95% percentile dispersal distance of plants that have dispersal traits comparable to that of C. diandra does not exceed 10 m, even if the wind speed approaches that of severe storms R20 (Soons et al., 2004). For this reason, we set the maximum dispersal distance via wind at 10 m. R21 We analysed the duration of each wind direction during the dispersal period of C. diandra R22 for the period 1981-2000 AD at station De Bilt (KNMI, 2010). From this analysis it R23 appeared that wind direction was variable with a maximum frequency of only 10 % from R24 south-west direction. Therefore, we assumed that the seeds could disperse in any direction R25 from a population. This assumption implies that habitat patches could become colonized by R26 C. diandra via anemochorous dispersal if they were situated within a distance of 10 m from a R27 population. Because we also assumed secure establishment of C. Diandra in habitat patches R28 where seeds arrive, a habitat patch will become entirely colonised as soon as it overlaps with R29 a seed shadow of a nearby population. R30 R31 R32 R33 R34 R35 R36 R37 R38 R39 Linking habitat suitability and seed dispersal models | 109

R1 R2 R3 R4 R5 R6 R7 R8 R9 R10 R11 R12 R13 R14 R15 Figure 5.2. Conceptual model of integrated anemochorous-hydrochorous dispersal of C. diandra R16 for three dispersal stages. Dispersal is a function of the current species distribution, species-specific dispersal traits, the dispersal infrastructure (i.e., the surface water system), and the configuration R17 of habitat patches. Anemochorous dispersal distances of C. diandra ranged from 0 to 10 m, while R18 hydrochorous dispersal distances ranged from 0 to 500 m. Anemochorous and hydrochorous dispersal was allowed in any direction. New dispersal stages were initiated until seed dispersal no R19 longer resulted in the establishment of new populations, implying that time is not limiting the R20 success of seed dispersal. R21 R22 Hydrochory Hydrochory is the dispersal of seeds or (parts of) plants by water. Like many other Carex R23 species, C. diandra is characterised by a relatively high seed buoyancy of ca. 100 days (Van R24 den Broek et al., 2005), i.e., 50% of the seeds still float in stagnant water after ca. 100 days. R25 This potentially enables long-distance dispersal via the surface water system. However, R26 dispersal barriers, such as aquatic plants or helophytes, rather than seed buoyancy seem to R27 limit hydrochorous dispersal distances in anthropogenically drained fens in The Netherlands R28 (Soomers et al., 2010). In these fens, dispersal distances of at most 500 m along surface water R29 elements have been observed for seeds of Carex species (Beltman et al., 2005; Soomers et R30 al., 2010). Because wind stress on the water surface, not water flow, is the principle driver of R31 hydrochorous seed dispersal via surface water elements in stagnant or slow-flowing water R32 (Soomers et al., 2010), we assumed that seeds could be dispersed over a distance of 500 m R33 in any direction via surface water elements. We further assumed that seeds could neither R34 pass culverts (Soomers et al., 2010) nor disperse from one polder to the other, because the R35 polders were not directly connected via surface water elements. The dispersal infrastructure for hydrochory, i.e., the configuration of the surface water system, was derived from a high- R36 resolution topographical map (see example in Fig. 5.1b), gridded to a resolution of 2.5 × 2.5 R37 m, and combined with the modelled water levels to identify inundated areas. R38 R39 110 | Chapter 5

R1 5.4 results R2 R3 5.4.1 Current condition R4 The groundwater model for the current condition showed that most of the groundwater that R5 exfiltrates into the river valley is intercepted by the deep agricultural polders and agricultural R6 drainage ditches outside the deep polders (Table 5.1) and does not become available for fen plants as a result. This hinders the establishment of contiguous fen habitat patches (Fig. 5.3a). R7 By consequence, the success of anemochorous dispersal of C. diandra is limited, as indicated by R8 the small number of subpopulations that can potentially establish via this dispersal mechanism R9 (Fig. 5.4a). The effect of habitat fragmentation on the hydrochorous dispersal of C. diandra R10 is less severe, even though only half the total amount of habitat is potentially accessible R11 for C. Diandra via hydrochorous dispersal (Fig. 5.4a). Colonisation of the majority of these R12 potentially accessible habitat patches is possible within five successive dispersal events (Fig. R13 5.5). Due to limited dispersal and the small sizes of the fen habitat patches, the populations R14 of C. diandra have a rather low viability as indicated by the small number, area, sizes and area R15 to edge-length ratios of the modelled subpopulations that can potentially establish (Fig. 5.4). R16 Note that the habitat patches that are inaccessible to fen plants, even after an infinite number R17 of successive anemochorous and hydrochorous dispersal events, are diverged across the study R18 area, with the majority of them being situated in three zones, one in the south, one in the R19 centre and one in the north-east of the study area R20 5.4.2 Inundation of deep agricultural polders (Strategy I) R21 Inundation of the deep agricultural polders Horstermeer and Bethune (Strategy I) would R22 redirect most of the groundwater that currently discharges into the deep polders towards the R23 drainage ditches across the river valley (Table 5.1). As drainage ditches do not provide suitable R24 habitat for low-productive fen plants, this strategy would only restore few small fen habitat R25 patches in the direct vicinity of the polders (Fig. 5.3b). In the north, these new habitat patches R26 could be colonised by C. diandra via hydrochorous dispersal. In the south, these new habitat R27 patches could not be colonised by C. diandra, even after an infinite number of dispersal R28 events. Overall, the dispersal success of C. Diandra is not improved as indicated by the minor R29 increase in population area for successive dispersal events, which is comparable to that for R30 the current condition (Fig. 5.5). As a result, only minor changes in the spatial characteristics R31 of potential subpopulations of C. diandra compared to the current condition were modelled R32 (Fig. 5.4). Thus, inundation of the deep agricultural polders, particularly polder Bethune in the south, would only have minor effects on the viability of C. diandra. R33 R34 R35 R36 R37 R38 R39 Linking habitat suitability and seed dispersal models | 111 a a a a R1 Total 368 262 209 204 328 249 R2 R3 2 2 2

10 13 22 R4 Terrestrial Terrestrial areas R5 R6 R7 R8 23 30 29 60 22 26 R9 /d) Potential fen habitat Shallow lakes and inundated areas 3 R10 m 3 R11 R12 R13 88 243 122 113 204 215 R14 Drainage ditches and deep lakes R15 R16 R17 R18 9 6 79 100 100 R19 Unsuitable matrix Deep agricultural polders Modelled exfiltration fluxes (× 10 R20 R21 R22 R23 R24 R25 R26 R27 R28 R29 R30 R31 R32 R33 ux decreases compared to the current condition. This is caused by a reduction of volume surface water that recharges fl R34 R35 ltration Modelled groundwater exfiltration fluxes into the unsuitable matrix and potential fen habitat patches for current condition R36 fi R37 R38 able 5.1. Current condition or restoration strategy Strategy II: Closure of abstraction wells Strategy III: Drainage elements eliminated from upstream fen margins 100 Drainage elements eliminated across river valley Strategy IV: All measures implemented Strategy V: Current condition Strategy I: Inundation of deep agricultural polders Total ex Total a the aquifer under river valley in response to a decreased head gradient (Strategies I and V) or interrupted ability supply alien surface water to polders (Strategies III, IV and V). t and five hydrological fen restoration strategies. The unsuitable matrix consists of deep agricultural polders, anthropogenic drainage ditches deep lakes. Potential fen habitat consists of shallow lakes, inundated areas and terrestrial areas. R39 112 | Chapter 5

R1 5.4.3 Closure of abstraction wells (Strategy II) R2 By closing the abstraction wells (Strategy II), most of the groundwater that is currently R3 directed towards the abstraction wells would be redirected towards the drainage ditches R4 across the river valley (Table 5.1). Compared to Strategy I, this strategy would restore even less R5 fen habitat, leaving the fen habitat configuration largely unaltered compared to the current R6 condition (Fig. 5.3c). Also, as the dispersal infrastructure for C. diandra is not improved, the habitat patches which are currently inaccessible to fen plants are still inaccessible under R7 this strategy (Fig. 5.4), and the expansion of the meta-population occurs at a rate that is R8 comparable to that under the current condition (Fig. 5.5). The viability of C. diandra is thus R9 not improved (Fig. 5.4) by this strategy. R10

R11 5.4.4 Drainage elements eliminated from upstream fen margins (Strategy III) R12 The elimination of drainage elements from the upstream fen margins (Strategy III) would R13 cause a decrease in the total groundwater exfiltration flux into the study area compared to R14 the current condition (Table 5.1). This decrease relates to the interrupted ability to supply alien R15 surface water to the upstream polders, which causes a reduction in the volume of surface water R16 that recharges the aquifer under the river valley. However, a substantial part of the volume of R17 groundwater that is currently intercepted by drainage ditches and deep lakes would become R18 redirected towards shallow lakes and terrestrial areas (Table 5.1). As a result, superfluous R19 groundwater can be laterally redistributed by throughflow (Fig. 5.3d), which would restore fen habitat throughout the upstream fen margins. In the south, these new habitat patches R20 can become colonised by C. diandra via anemochorous dispersal. As anemochorous dispersal R21 distances are small, the colonisation of new habitat can be severely delayed as indicated by R22 the minor increase in population area for successive dispersal events (Fig. 5.5). Nevertheless, R23 this would eventually result in increased number, area, sizes and area to edge-length ratios R24 of potential subpopulations (Fig. 5.4). Note that hydrochorous dispersal has become less R25 effective, because its dispersal infrastructure is lost when the upstream drainage ditches are R26 eliminated and only small areas have become inundated (Fig. 5.4). R27 R28 5.4.5 Drainage elements eliminated across valley (Strategy IV) R29 Compared to Strategy III, the elimination of the drainage elements across the river valley R30 (Strategy IV) would cause a further decrease in the total groundwater exfiltration flux into R31 the study area (Table 5.1). This relates to the interrupted ability to supply alien surface water R32 to the study area. Nevertheless, the volume of groundwater that is available for fen plants would increase compared to Strategy III, because of the reduced interception of groundwater R33 by drainage ditches and the deep agricultural polders (Table 5.1). Moreover, the area and R34 contiguity of habitat patches is further increased by enhanced throughflow at the upstream R35 fen margins and the development of groundwater-fed ponds near the centre of the study R36 area (Fig. 5.3e). This increases the success of both anemochorous and hydrochorous dispersal, R37 which results in a more effective colonisation of new habitat (Fig. 5.5). As a result, the total R38 area, sizes and area to edge-length ratios of subpopulations of C. diandra can increase R39 Linking habitat suitability and seed dispersal models | 113

substantially after each successive dispersal event (Fig. 5.4 and 5.5). Note that the effectiveness R1 of this restoration strategy is spatially differentiated, because restored habitat patches at the R2 north-eastern and southern part of the study area remain inaccessible to C. Diandra, and R3 about half the amount of potentially accessible habitat can only be colonised by C. Diandra R4 after 25 successive dispersal events or more (Fig. 5.5). R5 R6 5.4.6 All measures implemented (Strategy V) R7 Finally, Strategy V, which consists of all of the measures discussed in Strategies I–IV, would R8 cause an even further increase of the availability of groundwater for fen plants compared to R9 Strategy IV (Table 5.1, Fig. 5.3f). The larger volume of groundwater that is redirected towards the ground surface in the study area causes an expansion of both the groundwater-fed ponds R10 and the throughflow zone (Fig. 5.3f). As a result, contiguous habitat patches can establish R11 that are easily colonised via anemochorous and hydrochorous dispersal (Fig. 5.5), except for R12 the patches that become established in the north-east and south of the study area. This would R13 result in a further increase in the viability of C. diandra compared to the other strategies (Fig. R14 5.4). However, a large number of successive dispersal events (>25) are required before the R15 meta-population of C. Diandra can reach its maximum extent (Fig. 5.5). R16 R17 R18 R19 R20 R21 R22 R23 R24 R25 R26 R27 R28 R29 R30 R31 R32 R33 R34 R35 R36 R37 R38 R39 114 | Chapter 5

R1 a. Current condition b. Strategy I c. Strategy II R2 R3 R4 R5 R6 R7 R8 R9 R10 R11 R12 R13 R14 R15 d. Strategy III e. Strategy IV f. Strategy V R16 R17 R18 R19 R20 R21 R22 R23 R24 R25 R26 R27 R28 R29 R30 R31 Figure 5.3. Configuration of potential fen habitat provided by groundwater exfiltration into R32 terrestrial areas, groundwater exfiltration into ponds and inundated areas and throughflow of R33 exfiltrated groundwater. (a) Current condition. (b) Strategy I: Inundation of deep agricultural R34 polders. (c) Strategy II: Closure of abstraction wells. (d) Strategy III: Draining elements eliminated from upstream fen margins. (e) Strategy IV: Draining elements eliminated across the river valley. (f) R35 Strategy V: All measures implemented. See Fig. 5.1 for details of the restoration strategies. Note R36 that fen habitat under the current condition, strategy I and strategy II (a, b and c) exclusively consists of shallow lakes that receive exfiltrating groundwater, while fen habitat under strategy III, IV and R37 V (d, e and f) increasingly consists of terrestrial areas that receive groundwater by exfiltration or R38 by throughflow of exfiltrated groundwater. Besides, semi-aquatic fen habitat becomes increasingly R39 associated to inundated areas instead of to turf ponds from strategy III through V, because exfiltration zones will become reallocated towards the upstream fen margins, where water levels will rise to above the ground surface. Linking habitat suitability and seed dispersal models | 115

R1 R2 R3 R4 R5 R6 R7 R8 R9 R10 R11 R12 R13 R14 R15 R16 R17 R18 R19 R20

Figure 5.4. Spatial characteristics of potential subpopulations of C. diandra calculated using R21 anemochorous, hydrochorous and coupled anemochorous-hydrochorous dispersal models for the R22 current condition and for the restoration strategies. (a) Number of subpopulations. (b) Total area of R23 subpopulations. (c) Size of largest subpopulation. (d) Largest area to edge-length ratio (A:P ratio) of subpopulations. R24 The coupled anemochorous-hydrochorous dispersal model integrates the effects of both dispersal R25 mechanisms on the colonisation of abandoned fen habitat patches (see Fig. 5.3). The x-axis refers to the restoration strategies I–V (see Table 5.1), including the current condition (denoted by C). Note R26 that the number of subpopulations for the hydrochorous dispersal model exceeds the number of R27 subpopulations for the coupled anemochorous-hydrochorous dispersal model (Fig. 5.4a), because R28 the small subpopulations that can establish via solely hydrochorous dispersal are merged together via subsequent anemochorous dispersal. R29 R30 R31 R32 R33 R34 R35 R36 R37 R38 R39 116 | Chapter 5

R1 25 Current condition R2 Inundation of deep agricultural polders Closure of abstraction wells R3 Drainage elements eliminated from upstream fen margins 20 Drainage elements eliminated across river valley R4 All measures implemented R5

R6 15 R7 R8 R9 10 R10 R11 5 R12 R13

R14 Total area of subpopulations Carex diandra after dispersal event x (km2) 0 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 >25 R15 Dispersal event R16 Figure 5.5. Area of sub-populations of C. Diandra plotted against the number of successive anemochorous and hydrochorous dispersal events for the current condition and the restoration R17 strategies I-V (see Table 5.1). Note the delayed, but gradual colonisation of vast habitat patches via R18 anemochorous dispersal under restoration strategy III, and the efficient colonisation of vast habitat R19 patches via hydrochorous dispersal across inundated areas under restoration strategies IV and V. R20 R21 5.5 dIsCussIon R22

R23 5.5.1 Methodological approach R24 In this study, we used a linked habitat suitability and seed dispersal model in order to analyse R25 spatially differentiated potentials and constraints of a number of hydrological restoration R26 strategies targeted at the counteraction of habitat loss and fragmentation of low-productive R27 fens. Compared to stand-alone habitat suitability models, the added value of the linked R28 approach is that the effects of restoration strategies are not quantified by changes in R29 potentially suitable habitat, but by changes in potentially colonisable habitat. Potentially R30 colonisable habitat is a more relevant indicator of the restoration prospects of fragmented R31 ecosystems, because habitat fragmentation reduces the dispersal success of plants (Soons et R32 al., 2005) and may consequently hinder or delay the colonisation of habitat patches by target species (Bischoff, 2002; Ozinga et al., 2005). Our model showed that, for each strategy, 20–35% R33 of the potentially suitable habitat could not be colonized by C. diandra, and that the delay of R34 colonisation widely differs among the strategies, depending on the spatial configuration of R35 both the newly established habitat patches and the dispersal infrastructure for hydrochorous R36 dispersal. This indicates that habitat suitability models that ignore the effects of limited R37 dispersal on colonisation may considerably overestimate the prospects for fen restoration. For R38 this reason, we believe the presented modelling approach may serve as an example for future R39 attempts at improving the predictive ability of habitat suitability models. We particularly Linking habitat suitability and seed dispersal models | 117

advocate our modelling approach for the analysis of the effectiveness of habitat restoration R1 measures that affect both the habitat configuration and dispersal infrastructure of species R2 that have a limited dispersal ability. R3 R4 Habitat suitability models have been employed in a number of former studies in order to R5 predict ecosystem responses to hydrological restoration measures (Olde Venterink and R6 Wassen, 1997; Van Ek et al., 2000). These studies considered habitat suitability on an interval or R7 continuous scale in order to quantify the probability of plant species occurrence as a function R8 of multiple predictors, like ion concentrations in shallow groundwater, soil characteristics and R9 management actions. In the current study, we simplified the habitat suitability concept by considering fen habitat to be either suitable or unsuitable. Moreover, we ignored the effects R10 of non-hydrological management actions, like mowing (Fojt and Harding, 1995), and soil- R11 chemical processes (Lamers et al., 1998) on habitat suitability. Despite of these simplifications R12 we could identify spatially differentiated potentials for the re-establishment of low-productive R13 fen plants by restoration of their habitat at sites that are accessible via seed dispersal within R14 a certain number of successive dispersal events. However, in order to further specify these R15 potentials a more complete analysis of fen habitat suitability under the restoration strategies R16 is required. R17 R18 Numerous studies have used dispersal models to analyse population dynamics (Engler and R19 Guisan, 2009; Scheller and Mladenoff, 2008) or species viability (Soons et al., 2005) in relation R20 to climate change or habitat fragmentation. These studies used fitted or physically based R21 dispersal kernels to model seed deposition as a function of the distance to the parent plant. R22 In the present study, we used uniform dispersal kernels for anemochorous and hydrochorous dispersal that were based on empirically established dispersal distances. A sensitivity R23 analysis demonstrated that dispersal distance settings had limited to considerable effect on R24 the predicted meta-population area in time (see Table 5.2). Given this sensitivity and the R25 assumption of secure establishment of target species at a habitat patch once seeds have R26 arrived, our modelling results have no absolute value. Nevertheless, the sensitivity analysis R27 confirmed the applicability of our dispersal model to the analysis of regional-scale hydrological R28 fen restoration measures. This because the sensitivity of the modelled meta-population area R29 was small compared to the change in meta-population area under each restoration strategy, R30 except for the strategies that had minor effects on population viability (Table 5.2). However, R31 more realistic plant dispersal models are required for a sound analysis of the effects of habitat R32 restoration measures on species distribution on a local scale. R33 R34 Long-distance dispersal by animals (Soons et al., 2008) or man (Klimkowska et al., 2007) may R35 substantially reduce the colonisation delay of potentially accessible habitat. Long-distance dispersal may also make distant and apparently isolated habitat patches accessible to target R36 species. As long-distance dispersal events are usually rare and unpredictable, and their effects R37 on dispersal are highly uncertain, we could not incorporate long-distance dispersal in our R38 model. R39 118 | Chapter 5 (-) 3

R1 4 - - R2 ∞ R3 R4 4 - - / 0.0 0.0 /

R5 5 R6 R7

R8 4 - 1 R9 ). R10 R11 Current, default 4 (%) Robustness of the linked model 2 ∞ R12 – A R13 7 0.0 / 0.0 / -0.1 / +0.1 0.0 /

R14 scen X, default R15 4 / +12. 5

R16 ) / (A default R17

R18 scen X, 4 R19 – A 1

R20 Sensitivity of the linked model R21 scen X,sensitivity R22 ]·100 %. ∞ ) R23 2 (km

R24 1

R25 5 R26 current/scen X, default of the default dispersal distance for anemochory or hydrochory, and the second value corresponds value second the and hydrochory, or anemochory for distance dispersal default the of R27 ) / A 1 calculated for each scenario using default dispersal settings and sensitivity robustness of the linked 0.461.91 1.73 5.04 7.64 16.10 -2.2 / +4.3 -3.1 / +3.1 -6.9 / +5.2 -17.9 -0.3 / +1.2 -0.3 / +0.6 -5.9 / +4.4 0.0 / 0.461.91 1.732.50 5.04 7.64 7.07 16.10 -34.8 / +52.2 -21.5 / +18.3 22.26 -27.2 / +6.9 -11.1 / +2.2 -24.0 / +21.2 -6.9 / +0.5 -13.7 / +3.0 -0.1 / 0.0 -5.0 / +7.5 0.0 / -23.2 / +5.9 -0.1 / +0.1 0.0 / -0.1 / 0.0 0.0 / 0.0 / 0.0 0.44 1.73 2.11 -2.3 / +2.3 -6.4 / +8.1 -1.4 / +0.9 -1.0 / +1.0 -5.4 / +6.9 -1.5 / +1.0 0.44 1.73 2.11 -36.4 / +54.5 -28.3 / +6.9 -27.0 / +1.4 -15.6 / +23.5 -24.2 / +5.9 -28.2 / +1.5 0.57 2.10 2.57 -1.8 / +1.8 -5.2 / +8.6 -4.7 / +1.2 -0.1 / +0.1 -0.2 / +0.4 -0.2 / +0.1 R28 2.50 7.070.57 22.26 2.10 -2.8 / +2.4 -16.3 / +12.6 2.57 0.0 / -38.6 / +45.6 -27.6 / +9.5 0.0 / -26.5 / +1.6 -0.1 / 0.0 -1.2 / +1.4 -1.2 / +0.4 0.0 / -1.2 / +0.1 Population area for default dispersal settings R29 decrease

R30 current/scen X, default

R31 – A Carex diandra R32 sensitivity R33 R34 of the default dispersal distance anemochory or hydrochory. Scenario III Scenario IV Scenario III Scenario IV Scenario V Scenario II Number of successive dispersal events Scenario II Current conditionScenario I 0.43 1.71 2.09 0.0 / +2.3 -6.4 / +8.2 -1.4 / +1.0 Scenario V Scenario I Current condition 0.43 1.71 2.09 -34.9 / +55.8 -28.7 / +7.0 -26.8 / +1.4 -

R35 current/scen X, R36 increase R37 Population area of R38 The first values correspond to a 50 % 50 a to correspond values first The Default dispersal settings were 10 m for anemochory and 500 m for hydrochory. Default dispersal settings were 10 m for anemochory and 500 hydrochory. Defined as the difference between the meta-population areas that were calculated using sensitivity and default dispersal distance settings relative Defined as the difference able 5.2. Change in anemochorous dispersal settings Change in hydrochorous dispersal settings Defined as the ratio of (1) the difference between the meta-population areas that were calculated using sensitivity and default dispersal distance Defined as the ratio of (1) difference settings and (2) the difference between the meta-population area calculated for each restoration strategy and current settings and (2) the difference situation using default dispersal settings; Sensitivity / Change = (A 4 t model for anemochorous and hydrochorous dispersal distance settings. Details about the restoration strategies are provided in Fig. 5.1. 1 2 3 to a 50 % R39 to the meta-population area that was calculated using default dispersal distance settings; Sensitivity = [(A Linking habitat suitability and seed dispersal models | 119 (-) 3

4 Nevertheless, we think that our modelling approach is useful for the identification of spatially R1 - - ∞ differentiated potentials for nature restoration. This because the restoration of habitat R2 patches that are accessible via frequently available dispersal vectors should, in our opinion, be R3 assigned priority over the restoration of distant habitat patches that are exclusively accessible R4 4 - - / 0.0 0.0 /

to target species via rarely available long-distance dispersal vectors. 5 R5 R6 A general constraint of habitat suitability and meta-population models is the lack of reference R7 data to evaluate predicted responses to external changes. For this reason, species distribution

4 R8 - 1

). models can only be evaluated using independent (or re-sampled) reference data for the R9 current condition (Guisan and Thuiller, 2005). In our case, we could evaluate the linked habitat suitability and seed dispersal model by comparing the modelled realised habitat for the R10 R11 Current, default current condition with the actual distribution of C. diandra derived from presence/absence 4 (%) Robustness of the linked model 2 ∞ – A data projected on a cadastral map. According to our analysis, 20 (i.e., 59%) of the 34 parcels R12 inhabited by C. diandra could be explained by the linked model. This limited fit between R13 7 0.0 / 0.0 / -0.1 / +0.1 0.0 /

predictions and observations may be related to our method for downscaling the modelled R14 scen X, default zones with groundwater supply by assigning exfiltrating groundwater to drainage elements R15 4 / +12. 5

) / (A within model cells. As a result, floating fens bordering drainage ditches may not have been R16

default identified as suitable habitat, because the entire volume of exfiltrating groundwater was R17 assigned to the drainage ditches, while part of this groundwater is, in reality, directed to the scen X, R18 floating fens. Note that succession may soon cause the loss of fen habitat in the absence of 4 – A R19 1 open water bordering floating fens (Verhoeven and Bobbink, 2001). If this happens, 10 of

Sensitivity of the linked model R20 the 14 parcels inhabited by C. diandra but not explained by the linked model, will render R21 unsuitable conditions for fen plants in the nearby future. scen X,sensitivity ]·100 %. R22 ∞ ) 2 5.5.2 Model results R23 (km 1 The strategy analysis indicated that both the inundation of polders and the removal of R24

5 abstraction wells would cause an increase in the volume of groundwater that is directed R25 current/scen X, default towards the area where fen plants grow. However, most of this groundwater is intercepted by R26 of the default dispersal distance for anemochory or hydrochory, and the second value corresponds value second the and hydrochory, or anemochory for distance dispersal default the of ) / A drainage ditches, which do not provide habitat for low-productive fen plants. For this reason, R27 1 calculated for each scenario using default dispersal settings and sensitivity robustness of the linked 0.461.91 1.73 5.04 7.64 16.10 -2.2 / +4.3 -3.1 / +3.1 -6.9 / +5.2 -17.9 -0.3 / +1.2 -0.3 / +0.6 -5.9 / +4.4 0.0 / 0.461.91 1.732.50 5.04 7.64 7.07 16.10 -34.8 / +52.2 -21.5 / +18.3 22.26 -27.2 / +6.9 -11.1 / +2.2 -24.0 / +21.2 -6.9 / +0.5 -13.7 / +3.0 -0.1 / 0.0 -5.0 / +7.5 0.0 / -23.2 / +5.9 -0.1 / +0.1 0.0 / -0.1 / 0.0 0.0 / 0.0 / 0.0 0.44 1.73 2.11 -2.3 / +2.3 -6.4 / +8.1 -1.4 / +0.9 -1.0 / +1.0 -5.4 / +6.9 -1.5 / +1.0 0.44 1.73 2.11 -36.4 / +54.5 -28.3 / +6.9 -27.0 / +1.4 -15.6 / +23.5 -24.2 / +5.9 -28.2 / +1.5 0.57 2.10 2.57 -1.8 / +1.8 -5.2 / +8.6 -4.7 / +1.2 -0.1 / +0.1 -0.2 / +0.4 -0.2 / +0.1 2.50 7.070.57 22.26 2.10 -2.8 / +2.4 -16.3 / +12.6 2.57 0.0 / -38.6 / +45.6 -27.6 / +9.5 0.0 / -26.5 / +1.6 -0.1 / 0.0 -1.2 / +1.4 -1.2 / +0.4 0.0 / -1.2 / +0.1 only small and fragmented habitat patches are restored and the unsuitable matrix is largely R28 Population area for default dispersal settings decrease unaltered. As a result, most of the restored habitat patches are inaccessible to fen plants and R29 the colonisation delay of accessible habitat is comparable to that under the current condition. current/scen X, default R30 Water management actions thus constrain the restoration of low-productive fens under this – A R31 Carex diandra strategy, because of both limited restoration of fen habitat and minor improvement of the R32 sensitivity

dispersal ability of target species. Note that this strategy may contribute to improving the R33 quality of surface water, as the increased availability of groundwater in the fen area provides R34 of the default dispersal distance anemochory or hydrochory. opportunities to reduce the amount of polluted surface water that is supplied to polders Scenario III Scenario IV Scenario III Scenario IV Scenario V Scenario II Number of successive dispersal events Scenario II Current conditionScenario I 0.43 1.71 2.09 0.0 / +2.3 -6.4 / +8.2 -1.4 / +1.0 Scenario V Scenario I Current condition 0.43 1.71 2.09 -34.9 / +55.8 -28.7 / +7.0 -26.8 / +1.4 - current/scen X, R35 during the growing season. This could be particularly beneficial for aquatic species, but also for R36 increase the regeneration of floating fens that are currently polluted by alien surface water (Lamers et Population area of al., 1998). These effects can be analysed in terms of population viability using a methodology R37 comparable to the one presented here and by using additional surface water quality models. R38 The first values correspond to a 50 % 50 a to correspond values first The Default dispersal settings were 10 m for anemochory and 500 m for hydrochory. Default dispersal settings were 10 m for anemochory and 500 hydrochory. Defined as the difference between the meta-population areas that were calculated using sensitivity and default dispersal distance settings relative Defined as the difference able 5.2. Change in anemochorous dispersal settings Change in hydrochorous dispersal settings Defined as the ratio of (1) the difference between the meta-population areas that were calculated using sensitivity and default dispersal distance Defined as the ratio of (1) difference settings and (2) the difference between the meta-population area calculated for each restoration strategy and current settings and (2) the difference situation using default dispersal settings; Sensitivity / Change = (A 4 t model for anemochorous and hydrochorous dispersal distance settings. Details about the restoration strategies are provided in Fig. 5.1. 1 2 3 to a 50 % to the meta-population area that was calculated using default dispersal distance settings; Sensitivity = [(A R39 120 | Chapter 5

R1 Measures that include the elimination of drainage elements were effective in increasing the R2 volume of groundwater that is available for fen plants. These kind of measures also support R3 a more efficient supply of groundwater to fen root-zones by throughflow as found in natural R4 fens (Van Loon et al., 2009c). This leads to the establishment of contiguous habitat patches at R5 the upstream fen margins, even if drainage networks are maintained near the centre of the R6 fens and groundwater continues to be abstracted through wells. Although a rather effective infrastructure for hydrochorous dispersal is lost with the elimination of drainage ditches R7 (Soomers et al., 2010), dispersal is not necessarily a bottleneck for successful fen restoration. R8 This is particularly true at sites where contiguous habitat patches have established near R9 existing populations of target species, because the majority of these patches can be easily R10 colonised via anemochorous dispersal. Given the small transport distances of seeds dispersed R11 by wind, the time lag between restoration and colonization of habitat patches is considerable R12 in the absence of inundated areas that provide an infrastructure for hydrochorous dispersal R13 (see Fig. 5.5). For this reason, local measures may be needed in order to further enhance R14 the colonisation of habitat patches by target species. These measures may consist of the R15 maintenance of a number of drainage ditches in order to provide a dispersal corridor for fen R16 plants. As this measure may be disadvantageous for abiotic restoration of low-productive fens R17 (Van Loon et al., 2009a), we also advocate the joint use of habitat suitability and dispersal R18 models for the optimisation of such local measures using population viability and the time to R19 colonisation as indicators of restoration potentials. R20 R21 5.6 ConClusIon and ImPlICatIons For Fen restoratIon R22

R23 This paper demonstrates the applicability of linked habitat suitability and seed dispersal R24 models in restoration ecology by applying such a model in a strategy analysis of hydrological R25 restoration measures targeted at the sustainable restoration of a large low-productive fen in R26 The Netherlands. Hydrological restoration of this fen type is often constrained by the presence R27 of anthropogenic land uses that have completely opposite demands for water management R28 than are desirable for abiotic fen restoration. Likewise, abiotic fen restoration of sites that R29 have a low probability for re-colonisation by target species is not cost-effective. With the aid R30 of the linked model, we were able to optimise the spatial planning of hydrological restoration R31 measures, taking into account both the constraints of water management practices on R32 abiotic restoration and the effects of habitat fragmentation on dispersal. Moreover, we demonstrated that stand-alone habitat suitability models, which assume unlimited dispersal, R33 may considerably overestimate restoration prospects. For these reasons, the presented R34 modelling approach may serve as a basis for future attempts to improve the predictive ability R35 of habitat suitability models. R36 R37 R38 R39 Linking habitat suitability and seed dispersal models | 121

We conclude that linked habitat suitability and dispersal models can provide useful insights R1 into spatially differentiated potentials and constraints of nature restoration measures. This is R2 particularly true for measures targeted at the sustainable conservation of endangered plant R3 populations who´s habitats have been reduced due to undesirable effects of land and water R4 management on abiotic conditions. The joint use of habitat suitability and dispersal models R5 may therefore contribute to spatially optimized and cost-effective nature restoration, and R6 to conserve populations of endangered species in areas under reclamation by smart spatial R7 planning of the reclaimed areas. R8 R9 acknowledgments The authors thank the Provincie Utrecht and Provincie Noord Holland for providing databases R10 of the presence of C. diandra across the Vechtstreek area, Laura Cobb for proof-reading the R11 manuscript and three anonymous reviewers for their constructive suggestions to improve the R12 manuscript. R13 R14 R15 R16 R17 R18 R19 R20 R21 R22 R23 R24 R25 R26 R27 R28 R29 R30 R31 R32 R33 R34 R35 R36 R37 R38 R39 122 | Chapter 5

R1 reFerenCes R2 R3 Almendinger, J.E., Leete, J.H., 1998. Peat characteristics and groundwater geochemistry of calcareous fens in the Minnesota River Basin, U.S.A. Biogeochemistry 43, 17-41. R4 Barendregt, A., Wassen, M.J., Schot, P.P., 1995. Hydrological systems behind a nature reserve, the R5 major problem in wetland conservation of Naardermeer (The Netherlands). Biological Conservation 72, 393-405. R6 Bedford, B.L., Walbridge, M.R., Aldous, A., 1999. Patterns in nutrient availability and plant diversity R7 of temperate North American wetlands. Ecology 80, 2151-2169. R8 Beltman, B., Van den Broek, T., Vergeer, P., 2005. Het beperkte success van laagveen restauratie. Landschap 22, 173-179. R9 Beumer, V., Van Wirum, G., Beltman, B., Griffioen, J., Grootjans, A.P., Verhoeven, J.T.A., 2008. R10 Geochemistry and flooding as determining factors of plant species composition in Dutch winter-flooded riverine grasslands. Science of The Total Environment 25, 70-81. R11 Bischoff, A., 2002. Dispersal and establishment of floodplain grassland species as limiting factors in R12 restoration. Biological Conservation 104, 25-33. R13 Boedeltje, G., Bakker, J.P., Bekker, R.M., Van Groenendael, J.M., Soesbergen, M., 2003. Plant dispersal in a lowland stream in relation to occurrence and three specific life-history traits of the R14 species in the species pool. Journal of Ecology 91, 855-866. R15 Boomer, K.M.B., Bedford, B.L., 2008. Groundwater-induced redox-gradients control soil properties and phosphorus availability across four headwater wetlands, New York, USA. R16 Biogeochemistry 90, 259-274. R17 Boyer, M.L.H., Wheeler, B.D., 1989. Vegetation patterns in spring-fed calcareous fens, calcite R18 precipitation and constraints on fertility. Journal of Ecology 77, 597-609. Cappers, R.T.J., Bekker, R.M., Jans, J.E.A., 2006. Digitale Zadenatlas van Nederland (Digital Seed R19 Atlas of the Netherlands). Barkhuis Publishing & Groningen University Library, Groningen R20 (www.seedatlas.nl). Engler, R., Guisan, A., 2009. MigClim: Predicting plant distribution and dispersal in a changing R21 climate. Diversity and Distributions 15, 590-601. R22 Dormann,C.F., Schweiger, O., Augenstein, I., Bailey, D., Billeter, R., De Blust, G., DeFilippi, R., Frenzel, R23 M., Hendrickx, F., Herzog, F., Klotz, S., Liira, J., Maelfait, J, Schmidt, T., Speelmans, M., Van Wingerden, W.K.R.E., Zobel, M., 2007. Effects of landscape structure and land-use R24 intensity on similarity of plant and animal communities. Global Ecology and Biogeography R25 16, 774-787. Ewers, R.M., Didham, R.K., 2005. Confounding factors in the detection of species responses to habitat R26 fragmentation. Biological Reviews of the Cambridge Philosophical Society 81, 117-142. R27 Fagan, W.F., Lutscher, F., 2006. Average dispersal success: linking home range, dispersal and R28 metapopulation dynamics to reserve design. Ecological Applications 16, 820-828. Fahrig, L., 2003. Effects of habitat fragmentation on biodiversity. Annual Review of Ecology, R29 Evolution and Systematics 34, 487-515. R30 Fojt, W., Harding, M., 1995. Thirty years of change in the vegetation communities of three valley mires in Suffolk, England. Journal of Applied Ecology 32, 561-577. R31 Guisan, A., Thuiller, W., 2005. Predicting species distribution: offering more than simple habitat R32 models. Ecology Letters 8, 993-1009. R33 Hanski, I., 2008. Spatial patterns of coexistence of competing species in patchy habitat. Theoretical Ecology 1, 29-43. R34 Hooftman, D.A.P., Billeter, R.C., Schmid, B., Diemer, M., 2004. Genetic effects of habitat fragmentation R35 on common species of Swiss fen meadows. Conservation Biology 18, 1043-1051. Iverson, L.R., Schartz, M.W., Prasad, A.M., 2004. Potential colonization of newly available tree- R36 species habitat under climate change: an analysis for five eastern US species. Landscape R37 Ecology 19, 787-799. R38 Kardol, P., Van der Wal, A., Bezemer, T.M., De Boer, W., Duyts, H., Holtkamp, R., Van der Putten, W.H., 2008. Restoration of species-rich grasslands on ex-arable land: Seed addition outweighs R39 soil fertility reduction. Biological Conservation 141, 2208-2217. Linking habitat suitability and seed dispersal models | 123

Kleyer M., Bekker, R.M., Knevel, I.C., 2008. The LEDA Traitbase: A database of life-history traits of the Northwest European flora. Journal of Ecology 96, 1266-1274. R1 Klimkowska, A., Van Diggelen, R., Bakker, J.P., Grootjans, A.P., 2007. Wet meadow restoration in R2 Western Europe: A quantitative assessment of the effectiveness of several techniques. R3 Biological Conservation 140, 318-328. KNMI, 2010. Frequentietabellen windrichting in graden. http://www.knmi.nl/klimatologie/ R4 frequentietabellen/uur_freq.cgi. R5 Lamers, L.P.M., Tomassen, H.B.M., Roelofs, J.G.M., 1998. Sulfate-induced eutrophication and phytotoxicity in freshwater wetlands. Environmental Science and Technology 32, 199-205. R6 McDonald, M.G., Harbaugh, A.W., 1988. A modular three-dimensional finite-difference groundwater R7 flow model. United States Government Printing Office, Washington. R8 Millennium Ecosystem Assessment, 2005. Ecosystems and human well-being: Wetlands and water synthesis. World Resources Institute, Washington. R9 Olde Venterink, H., Wassen, M.J., 1997. A comparison of six models predicting vegetation response R10 to hydrological habitat change. Ecological Modelling 101, 347-361. Ozinga, W.A., Romermann, C., Bekker, R.M., Prinzing, A., Tamis, W.L.M., Schaminee, J.H.J., R11 Hennekens, S.M., Thompson, K., Poschlod, P., Kleyer, M., Bakker, J.P., Van Groenendael, R12 J.M., 2009. Dispersal failure contributes to plant losses in NW Europe. Ecology Letters 12, R13 66-74. Ozinga, W.A., Schaminée, J.H.J., Bekker, R.M., Bonn, S., Poschlod, P., Tackenberger, O., Bakker, J., Van R14 Groenendael, J.M., 2005. Predictability of plant species composition from environmental R15 conditions is constrained by dispersal limitation. Oikos 108, 555-561. Pollock, D.W., 1994. User’s Guide for MODPATH/MODPATH-PLOT, version 3: A particle tracking R16 post-processing package for MODFLOW, the U.S. Geological Survey finite-difference R17 groundwater flow model. U.S. Geological Survey, Reston, Virginia. R18 Purves, D.W., Dushoff, J., 2005. Directed seed dispersal and metapopulation response to habitat loss and disturbance: application to Eichhornia paniculata. Journal of Ecology 93, 658-669. R19 Schaminée, J.H.J., Weeda, E.J., Westhoff, V., 1995. Deel 2. Plantengemeenschappen van wateren, R20 moerassen en natte heiden. Opulus Press, Leiden. Scheller, R.M., Mladenoff, D.J., 2008. Simulated effects of climate change, fragmentation and R21 inter-specific competition on tree species migration in northern Wisconsin, USA. Climate R22 Research 36, 191-202. R23 Schot, P.P., Molenaar, A., 1992. Regional changes in groundwater flow patterns and effects on groundwater composition. Journal of Hydrology 130, 151-170. R24 Schröder, B., Rudner, M., Biedermann, R., Kögl, H., Kleyer, M., 2008. A landscape model for R25 quantifying the trade-off between conservation needs and economic constraints in the management of a semi-natural grassland community. Biological conservation 141, 719- R26 732. R27 Segurado, P., Araújo, M.B., 2004. An evaluation of methods for modelling species distribution. R28 Journal of Biogeography 31, 1555-1568. Sjörs, H., Gunnarsson, U., 2002. Calcium and pH in North and Central Swedish mire waters. Journal R29 of Ecology 90, 650-657. R30 Soomers, H., Winkel, D.N., Du, Y., Wassen, M.J., 2010. The dispersal and deposition of hydrochorous plant seeds in drainage ditches. Freshwater Biology, accepted for publication (doi:10.1111/ R31 j.1365-2427.2010.02460.x). R32 Soons, M.B., 2006. Wind dispersal in freshwater wetlands: knowledge for conservation and R33 restoration. Applied Vegetation Science 9, 271-278. Soons, M.B., Heil, G.W., 2002. Reduced colonization capacity in fragmented populations of wind- R34 dispersed grassland forbs. Journal of Ecology 90, 1033-1043. R35 Soons, M.B., Messelink, J.H., Jongejans, E., Heil, G.W., 2005. Habitat fragmentation reduces grassland connectivity for both sort-distance and long-distance wind-dispersed forbs. Journal of R36 Ecology 93, 1214-1225. R37 Soons, M.B., Nathan, R., Katul, G.G., 2004. Human effects on long-distance wind dispersal and R38 colonization by grassland plants. Ecology 85, 3069-3079. R39 124 | Chapter 5

Soons, M.B., Van der Vlugt, C., Van Lith, B., Heil, G.W., Klaassen, M., 2008. Small seed size increases R1 the potential for dispersal of wetland plants by ducks. Journal of Ecology 96, 619-627. R2 Van de Meene, E.A., Van Meerkerk, M., Van der Staay, J., 1988. Toelichting bij de geologische kaart R3 van Nederland 1 : 50.000. Blad Utrecht Oost (31O). Rijks Geologische Dienst, Haarlem. Van den Broek, T., Van Diggelen, R., Bobbink, R., 2005. Variation in seed buoyancy of species in R4 wetland ecosystems with different flooding dynamics. Journal of Vegetation Science 16, R5 579-586. Van der Meijden, R. 1996. Heukels’Flora van Nederland. Tweeëntwintigste druk. Wolters-Noordhoff, R6 Groningen. R7 Van Diggelen, R., Molenaar, W.J., Kooijman, A.M., 1996. Vegetation succession in a floating mire in R8 relation to management and hydrology. Journal of Vegetation Science 7, 809-820. Van Ek, R., Witte, J.M., Runhaar, H., Klijn, F., 2000. Ecological effects of water management in The R9 Netherlands: the model DEMNAT. Ecological Engineering 16, 127-141. R10 Van Loon, A.H., Schot, P.P., Bierkens, M.F.P., Griffioen, J., Wassen, M.J., 2009a. Local and regional impact of anthropogenic drainage on fen contiguity. Hydrology and Earth System Sciences R11 13, 1837-1848. R12 Van Loon, A.H., Schot, P.P., Griffioen, J., Bierkens, M.F.P., Batelaan, O., Wassen, M.J., 2009b. R13 Throughflow as a determining factor for habitat contiguity in a near-natural fen. Journal of Hydrology, in press. R14 Van Loon, A.H., Schot, P.P., Griffioen, J., Bierkens, M.F.P., Wassen, M.J., 2009c. Palaeo-hydrological R15 reconstruction of a managed fen area in The Netherlands. Journal of Hydrology, in press. Van Wirdum, G., 1991. Vegetation and hydrology of floating rich fens. PhD-thesis, University of R16 Amsterdam, Amsterdam. R17 Verhoeven, J.T.A., Bobbink, R., 2001. Plant diversity of fen landscapes in The Netherlands, In: R18 Biodiversity in wetlands: assessment function and conservation. eds B. Gopal, W.J. Junk, J.A. Davis, pp. 65-87. Backhuys Publishers, Leiden. R19 Wassen, M.J., Barendregt, A., 1992. Topographic position and water chemistry of fens in a Dutch R20 river plain. Journal of Vegetation Science 3, 447-456. Wassen, M.J., Barendregt, A., Palczynski, A., De Smidt, J.T., De Mars, H., 1992. Hydro-ecological R21 analysis of the Biebrza mire (Poland). Wetlands Ecology and Management 2, 119-134. R22 Wassen, M.J., Barendregt, A., Schot, P.P., Beltman, B., 1990. Dependency of local mesotrophic fens R23 on a regional groundwater flow system in a poldered river plain in the Netherlands. Landscape Ecology 5, 21-38. R24 Wheeler, B.D., Shaw, S.C., 1995. Restoration of damaged peatlands, vol. 1. HMSO, London. R25 R26 R27 R28 R29 R30 R31 R32 R33 R34 R35 R36 R37 R38 R39 Linking habitat suitability and seed dispersal models | 125

aPPendIx R1 R2 appendix a: Structure, data sources and performance of the groundwater model for the R3 current condition. R4 R5 a1 Introduction R6 This appendix provides additional information about the groundwater model of the R7 Vechtstreek area and its surroundings. Firstly, we outline the model structure and data R8 sources used for parameter estimation. Then, we discuss the performance of the model after R9 calibration. R10 a2 model structure and data sources R11 The groundwater model is based on the Modflow-code (McDonald and Harbaugh, 1988). The R12 model consists of seven confined model layers, each of which is discretized into 25 × 25 m R13 grid cells. The upper model layer represents the fen root zone of 0.3 m thickness. This layer is R14 highly conductive in cells with modelled groundwater levels exceeding its bottom elevation, R15 whereas this layer is inactive in cells with modelled groundwater levels below its bottom R16 elevation. The other six model layers represent aquifers or aquitards, who’s spatial dimensions R17 are derived from the geological stratigraphy of the catchment of the Vechtstreek area. Each R18 layer is represented by transmissivities for horizontal flow and leakance factors for vertical R19 flow. Values of these parameters are derived after Vernes et al. (2005). At the ice-pushed R20 ridge, anisotropy factors derived after Gehrels (1995) are used to model the horizontally R21 variable transmissivity of the ice-pushed ridge. R22

The model is bordered to the south by the River Rhine, to the west by the River Vecht and to R23 the North by Lake IJsselmeer. The eastern model boundary is defined at sufficient distance R24 from the Vechtstreek area so that no influence on groundwater flow in and towards the study R25 area was exerted. All model boundaries are represented by no-flow boundary conditions, R26 assuming that groundwater flow near the model boundaries is directed towards the surface. R27 R28 The Modflow Drain-package is used to model topographic control of groundwater levels R29 by surface runoff after complete inundation of topographic depressions. For this purpose, R30 highly conductive drains were placed across the model domain. Their elevations were set R31 equal to the ground surface elevation at the most downstream grid cell of each topographic R32 depression. This elevation was established by a gradient analysis of a high-resolution digital R33 elevation model (Van Heerd et al., 2000). R34 R35 The Modflow River-package is used to model groundwater-surface water interactions. Surface water levels in polders (i.e., water management districts) were set equal to local target levels R36 set by the water board, whereas those at the ice-pushed ridge were established from a high- R37 resolution digital elevation model (Van Heerd et al., 2000). Drainage was allowed throughout R38 R39 126 | Chapter 5

R1 the model domain, whereas infiltration of surface water was only allowed in the polders. R2 River conductances were set proportional to the density of surface water elements, which R3 were derived from topographical maps. R4 Chapter 6 R5 The Modflow Recharge-package is used to model groundwater recharge. Recharge fluxes were R6 set equal to the precipitation surplus that was established from a time-series of precipitation and Makkink reference evapotranspiration for the period 2000-2005, and using crop factors R7 and interception factors derived from Gehrels (1995) and Spieksma (1995). Land cover was R8 derived from land use maps. R9

R10 The Modflow Well-package is used to model groundwater abstractions. Spatially variable R11 abstraction rates were derived from databases that included both permanent and temporal R12 abstractions for the period 2000-2005. R13 The model represents steady-state groundwater flow, meaning that it is representative R14 for long-term mean-annual conditions. In The Netherlands, these conditions are similar to R15 conditions prevailing in spring (beginning of April). R16 R17 a3 Calibration and model performance R18 The groundwater model is calibrated for transmissivities of all layers, except for the root zone R19 layer, river conductances and groundwater recharge. For this purpose, a reference data set consisting of mean-annual heads for the period 2000-2005 was established using 659 time- R20 series of observed heads across the model domain. After calibration, the difference between R21 modelled and reference heads was less than 0.2 m for 47 % of the reference cells and less R22 than 0.5 m for 75 % of the reference cells. Deviations exceeding 0.5 m relate either to the R23 presence of abstraction wells, or to heterogeneous clay sheets in the eastern part of the ice- R24 pushed ridge. R25 R26 a4 references R27 Gehrels, J.C., 1995. Niet-stationaire grondwater modellering van de Veluwe. Een studie naar de invloed van grondwater winningen, inpoldering en verloofing op de grondwaterstand R28 sinds 1951. PhD thesis, Free University Amsterdam, Amsterdam. R29 R30 McDonald, M.G., and Harbaugh, A.W., 1988. A modular three-dimensional finite-difference groundwater flow model. Techniques of water-resources investigations of the United R31 States Geological Survey. United States Government Printing Office, Washington. R32 Spieksma, J.F.M., Dolman, A.J., and Schouwenaars, J.M., 1995. De parametrisatie van de verdamping R33 van natuurterreinen in hydrologische modellen. R34 R35 Van Heerd, R.M., Kuijlaars, E.A.C., Teeuw, M.P., and Van ‘t Zand, R.J., 2000. Productispecifictatie AHN 2000. MDTGM2000.13, Rijkswaterstaat, Delft. R36 R37 Vernes, R.W., Van Doorn, T.H.M., Bierkens, M.F.P., Van Gessel, S.F., and De Heer, E., 2005. Van gidslaag naar hydrogeologische eenheid. Toelichting op de totstandkoming van de dataset REGIS R38 II. NITG 05-038-B, TNO-NITG (Dutch Institute of Applied Geoscience), Utrecht R39 Chapter 6 wind and water dispersal of wetland plants across fragmented landscapes

Soomers, H., Karssenberg, D., Soons, M.B., Verweij, P.A., Verhoeven, J.T.A., and Wassen, M.J. 128 | Chapter 6

R1 abstraCt R2 R3 Over the last decades, landscape modification has led to habitat fragmentation for many R4 species. Therefore, knowledge on mechanisms that determine the dispersal range of plant R5 seeds is of vital importance for determining promising ways to restore biodiversity. In Western R6 Europe, drainage ditches are ubiquitous in former wetlands that are reclaimed and now used as farmland. The banks of these ditches serve as refuges for many wetland plant species. The R7 ditches may also function as dispersal corridors between fragmented populations of riparian R8 species. Apart from dispersal via water through these ditches, also dispersal by wind plays a R9 role in dispersing seeds of wetland species. Until now, no spatially explicit models exist in which R10 both wind and water dispersal are included. We developed a spatial wind- and water dispersal R11 model for agricultural areas with drainage ditches. We assess the relative importance of wind R12 and water as dispersal vectors for two wetland plant species differing in seed characteristics, R13 and for different landscape configurations. The simulations show that water dispersal distances R14 for typical wind dispersers are similar to those of typical water dispersers. 90th percentile water R15 dispersal distances were between 100 m and >1000 m, depending on ditch roughness and R16 obstructions, after wind and water dispersal. Wind dispersal distance was nil for the typical R17 water disperser, whereas 90th percentile wind dispersal distance was 28 m for the typical wind R18 disperser. Density or direction of the ditch network did not seem to influence simulated water R19 dispersal distances substantially, whereas roughness of the ditch and obstructions in the ditch strongly limited seed dispersal distances. Our study suggests that in a modern agricultural R20 landscape with a network of ditches, water is an important dispersal vector for both typical R21 wind and typical water dispersers. From a biodiversity restoration perspective, it is important R22 that riparian populations and suitable sites are connected by a network of surface water and R23 that obstructions in this network are avoided as much as possible. R24 R25 R26 R27 R28 R29 R30 R31 R32 R33 R34 R35 R36 R37 R38 R39 Wind and water dispersal of wetland plants | 129

6.1 IntroduCtIon R1 R2 Habitat destruction and habitat fragmentation have reduced large, continuous natural areas R3 to smaller and less connected habitat patches worldwide. These patches contain smaller and R4 more isolated populations of plants and animals, which often experience an increased risk R5 of extinction (Fahrig 2003, Fischer and Lindenmayer 2007). In fragmented, isolated habitats, R6 seed dispersal can become limiting for the regional survival of species (Ozinga et al. 2009). R7 The colonisation of new habitats and gene flow between populations usually decreases with R8 fragmentation (Nathan 2001, Hanski 2005), which is therefore considered a major threat to R9 biodiversity (Fahrig 2003, Fischer and Lindenmayer 2007). Agriculture is a widespread driver of fragmentation since c. 40 % of the world’s land surface R10 is covered by agricultural fields such as croplands and pastures (Foley et al. 2005) and areas R11 of intensive agriculture are inhospitable for wild flora and fauna (Donald and Evans 2006). R12 Still, populations of wild species can persist in fragmented habitat patches surrounded by R13 an agricultural matrix (Donald and Evans 2006). For plants, exchange between fragmented R14 habitats takes place by seed dispersal via wind, water, animals, and human activity. Until now, R15 plant dispersal research has mainly focused on quantitatively understanding seed dispersal R16 via wind (anemochory) (e.g. Tackenberg 2003, Soons et al. 2004, Katul et al. 2005, Nathan et R17 al. 2011). Recent studies, however, have shown the importance of surface water as a dispersal R18 vector for riparian plant species (i.e. hydrochory; (Boedeltje et al. 2003, Goodson et al. 2003, R19 Boedeltje 2005, Jansson et al. 2005, Gurnell et al. 2006). R20 In relatively wet landscapes or poorly drained soils, land drainage systems are often required R21 for profitable agricultural production (Vaughan et al. 2007). Therefore, agricultural landscapes R22 are often dominated by numerous drainage- or irrigation ditches (Moss 1983, Borger 1992, Bootsma 2000). These types of agricultural landscapes with an extensive network of ditches R23 can for instance be found in the Mississippi River Basin, USA (Moore and Kröger 2011, Kröger R24 et al. in press), in the UK (Moss 1983, Davies et al. 2008), Germany (Kahle et al. 2008), the R25 Netherlands (Blomqvist et al. 2003) and other parts of Europe (e.g. Pelacani et al. 2008, R26 Maljanen et al. 2010). Although agricultural fields are unsuitable as habitat for most wild R27 plants, the banks of these drainage ditches may serve as refuges for wetland plant species R28 (e.g. Bunce and Hallam 1993, Blomqvist et al. 2006). However, habitat quality varies because it R29 is often negatively affected by herbicides and fertilizers applied to adjacent modern farming R30 systems. As a result, many plant species growing at ditch banks exhibit spatially fragmented R31 populations (Geertsema and Sprangers 2002). Consequently, ditches may function as dispersal R32 corridors between subpopulations of hydrochorous wetland plants growing at ditch banks R33 (Geertsema et al. 2002, Milsom et al. 2004, Soomers et al. 2010). R34 Many riparian wetland plant species possess highly buoyant seeds that can float several weeks R35 till several months (van den Broek et al. 2005), enabling them to disperse effectively by surface water. However, water can only transport seeds to areas that are connected to the source R36 area by surface water. Wind, on the other hand, can transport seeds to a wide range of sites R37 distributed all over the landscape (Soons 2006). Many wetland plant seeds can be dispersed R38 R39 130 | Chapter 6

R1 both by surface water and by wind (Bouman et al. 2000, Middleton et al. 2006, Soons 2006). R2 For semi-terrestrial or terrestrial wetland species that do not grow directly along the water R3 body, wind dispersal will often be the first dispersal stage, enabling seeds to enter the water. R4 Therefore, when investigating dispersal of wetland plant seeds in agricultural landscapes, R5 both anemochory and hydrochory should be taken into account. R6 Seed dispersal distances by wind are investigated in many studies, either by experimental research or by (mechanistic) modelling (e.g. Nathan et al. 2001, Soons and Heil 2002, R7 Tackenberg 2003, Skarpaas et al. 2004, Katul et al. 2005, Nathan et al. 2011). However, R8 dispersal distances by water are not well known and hydrochory studies focus mostly on seed R9 transport through rivers. To our knowledge, only two studies (Beltman and van den Broek R10 2006, Soomers et al. 2010) have experimentally determined hydrochorous dispersal ranges in R11 drainage ditches. Moreover, spatially explicit process-based models predicting seed dispersal R12 via water have not yet been developed. We found only three studies that have attempted to R13 model hydrochorous dispersal (Campbell et al. 2002, Levine 2003, Groves et al. 2009). None R14 of the models, however, are spatially and temporally explicit or are applicable in standing or R15 slow flowing waters such as ditches, in which hydrochorous dispersal is driven by wind shear R16 stress (Soomers et al. 2010). R17 It is known that a decline in availability of dispersal vectors in a landscape affects species R18 loss and that this effect is species- and vector-specific (Ozinga et al. 2009). However, it is not R19 clear what portion of a species’ dispersal kernel is determined by wind dispersal and what by water dispersal, and how this changes in different landscapes and for different species R20 with differing dispersal traits. Seed traits such as terminal velocity (i.e. seed falling velocity R21 in still air) or buoyancy are often used to assess whether species are potentially capable of R22 long distance dispersal via wind or water (Soons 2006, Ozinga et al. 2009, Thomson et al. R23 2010). However, this results in rough estimations of maximum dispersal distances only, and R24 does not tell us anything about the spatial seed deposition pattern and the contribution of R25 different dispersal vectors to this dispersal pattern in real landscapes. Thus, to understand R26 metapopulation dynamics and spread and persistence of species in fragmented landscapes, R27 more knowledge on spatial dispersal patterns in different landscapes is needed. R28 The aim of this study was to analyze how dispersal of plant seeds via both wind and via water R29 contributes to the deposition patterns of seeds across agricultural landscapes throughout a R30 year. For this purpose we developed a temporally and spatially explicit, process–based, coupled R31 anemochory-hydrochory model. To parameterize and validate the model, we performed R32 seed release-and-retrace field experiments and seed mimic experiments. Although animals and humans may also disperse seeds of wetland plant species (Strykstra et al. 1997, Soons et R33 al. 2008) we do not include them as dispersal vectors in our study because their effects on R34 dispersal are highly unpredictable. R35 We investigated for two species, one representative for typical wind dispersers and the other R36 for typical water dispersers, how their deposition patterns differ in several landscape matrices. R37 Using the anemochory-hydrochory model, we address the following research questions: (i) R38 what is the relative contribution of wind dispersal and water dispersal to seed dispersal R39 Wind and water dispersal of wetland plants | 131

distances in landscapes with different configurations of drainage ditches, and (ii) to what R1 extent do system characteristics of the landscape (f.i. ditch direction, -density, -roughness, R2 obstructions) determine the dispersal distances of seeds? R3 Our results provide new insights into the relative importance of wind and water dispersal in R4 agricultural landscapes and how different landscape characteristics and features determine R5 deposition patterns of seeds across these landscapes. R6

R7 R8 6.2 metHods R9

6.2.1 Study system R10 We simulated anemochorous and hydrochorous dispersal of two wetland species in an R11 agricultural landscape consisting of meadows or agricultural fields interspaced by a network R12 of drainage ditches. The ditches in such systems are linear and narrow (<3m wide) and R13 usually parallel to each other, and are connected to each other and to larger canals by some R14 perpendicular ditches or canals, together forming a network for discharging the excess R15 water from precipitation and groundwater. The ditch network is also used to irrigate the R16 area with alien surface water in periods of water deficit (Van Loon et al. 2009). The water in R17 the ditches is slow-flowing to stagnant which is essentially different from rivers and streams. R18 Because of these low flow rates, both wind and water dispersal in these ecosystems are wind- R19 driven (Sarneel 2010, Soomers et al. 2010); wind shear stress on the water’s surface drives the R20 transport of hydrochorous seeds. R21 In nature reserves or at sites with agri-environmental schemes (Donald and Evans 2006) the R22 ditches and ditch banks together form ecosystems that support species-rich aquatic, semi- aquatic and terrestrial vegetation such as marsh marigold meadow, reed land and sedge R23 marsh (Van Strien et al. 1989, Leng et al. 2011). R24 For our case study, we selected two tall helophytes, Carex pseudocyperus and Phragmites R25 australis. In agricultural areas in peat districts, both species are common at ditch banks or in R26 shallow standing water. Carex pseudocyperus represents a group of riparian species with clear R27 seed adaptations for hydrochorous dispersal, whereas Phragmites australis represents a group R28 of species that is primarily adapted to anemochorous dispersal. Other examples of such species R29 are given in Table 6.1. R30 R31 R32 R33 R34 R35 R36 R37 R38 R39 132 | Chapter 6

R1 and R2 R3 - - 9-10 - 6-8 9-3 8-10 7-10 10 7-11 - - R4 - -

R5 Carex acutiformis, R6 ‡

R7 § 360000 360000 395 2078 1680 515 R8 Seed number SP R9 R10 1.5 0.7 0.7 2.5 0.88 1.15 23660 0.9 0.85 2466 R11 1.83 30062 R12 R13 2.3* 1.5 >210 1.35 2988 67 109 240 32 121 >112* 0.96 972 44* 112 223 36.5* 0.75 5000 233 R14 >210 1.05 248 R15 R16 obtusifolius Rumex acetosela, Rumex acetosa, Rumex R17 R18 R19 Value estimated by averaging data of similar species: Value § 0.11 3.00 0.14 2.69 2.00 0.17 2.26 2.70 0.19 2.43 4.28 0.26 2.12 R20 3.07 R21 R22 R23 R24 R25 R26 R27 Typha latifolia Typha Rumex hydrolapathum Typha angustifolia Typha Carex pseudocyperus Lycopus europaeus Lycopus Phragmites australis Carex paniculata Lysimachia vulgaris Lysimachia Epilobium hirsutum Carex acuta Iris pseudacorus Representative speciesCirsium arvense velocity Terminal FP50 RH Angelica archangelica R28 Carex riparia R29

R30 species: similar of data averaging by estimated Value ‡ R31 , R32 R33 R34 Carex riparia Carex R35 and

R36 Dispersal categories in wetland plants (group A and B), listing representative species of each category with their seed terminal Two R37 R38 able 6.1. Category Group A: Wind-dispersers. Low terminal velocity (≤0.3 m/s), low to intermediate FP50 (< 70 d), high RH (≥ 0.7 m) Group B: Water-dispersers. High terminal velocity (≥2.0 m/s), high FP50 (>100 d), RH (≥ 0.7 m). Rumex crispus. SP: seed shedding period (months, from-to). Source: BioPop (2011), -: missing value. Carex acuta Carex t and seed shedding period (SP). velocities, seed buoyancy (FP50), release height (RH), number, R39 velocity: average terminal velocity (m/s) (i.e. seed falling in still air), Source: Kleyer et al. (2008). Terminal Boedeltje *Source: (2005). et al. Broek den van Source: water, stagnant in floating still is seeds the 50% of which after days of number the FP50: et al. (2003). RH: average seed release height (m), Source: Kleyer et al. (2008). Seed number: number of seeds per shoot, Source: Kleyer et al. (2008), Wind and water dispersal of wetland plants | 133

6.2.2 Model structure R1 We developed a dynamic, spatially explicit coupled anemochory-hydrochory model. The R2 model predicts dispersal densities of wind and water dispersed seeds in a 2 x 2 km landscape R3 surrounding a central source population of wetland plants. A modelling landscape consists of R4 a raster of 1000 x 1000 cells, of 2 x 2 m each. The central source population covers 135 cells. R5 All source plants, which are representative for riparian species, are located along a ditch and R6 not in the agricultural fields. One million seeds are shed from each source-cell throughout R7 the seed shedding period. Primary dispersal takes place by wind; seeds that are blown into R8 a ditch after wind dispersal disperse further by surface water. During transport by surface R9 water, seeds may get captured and may become mobile again until they finally  sink  if they                     st     R10          are not captured permanently. The model uses a time step of one day and runs from July 1   th             R11    until April 30 the next year. The end date was chosen because it was assumed that seeds     R12 germinate in spring (Bewley 1997), and by consequence are then not transported  any  further.            The seed shedding season was assumed to be  from  July 1st to the  31 st of August  with an    R13    assumed uniform seed shedding distribution throughout this 62 day period.  Wind and  water  R14                      dispersal velocity is determined by wind speed  and wind  shear stress  on the  water  surface,     R15                          respectively, and therefore varies from day to day according to meteorological data. Outside           R16  the seed shedding season, water dispersal of seeds that were already in the water continues R17 to takes place. The model was developed using the PCRaster Python modelling framework R18  (Karssenberg et al. 2007). R19     R20 6.2.3 Anemochory module   R21 To simulate wind dispersal distances for the two model species, the mechanistic Markov chain  R22 model for Synthetic Turbulance Generation, adjusted for grassland ecosystems  (hereafter:        STG model) (Soons et al. 2004), was used. The  STG model  is a coupled  Eulerian-Lagrangian          R23                    stochastic dispersion model, which assumes that the  change  in position  and  velocity  of a seed   R24           R25   is described by a Markov chain process. It realistically simulates fluctuations in both horizontal and vertical wind velocities, which results in an accurate prediction of dispersal distances by R26 wind, as has been shown for grasslands and forests. Details on this model are given by Soons R27 et al. (2004) and Nathan et al. (2002). Here we will restrict ourselves to describing   how  we used   R28                        the model in the coupled anemochory–hydrochory model.        R29        As running the STG model for each individual seed  in our simulations   is not  feasible,   we used    R30   the STG model to calculate dispersal distance frequency distributions for 11 average daily R31  horizontal wind speed classes (class width 1 m/s). These frequency distributions are used as R32  input to the coupled anemochory-hydrochory model. For each average daily wind speed class,    R33 the frequency distribution was calculated by simulating wind dispersal of 104 seeds. For each        R34       seed a horizontal wind speed was drawn from a  frequency distribution  of  10-minute  wind       R35  speeds of the considered average daily wind speed class (KNMI 2011). Furthermore, for each            R36 seed, terminal velocity v (m/s) was drawn from a normal distribution    ,   term      and seed release height h (m) was drawn from a uniform distribution    R37 0       R38                   R39        

    134 | Chapter 6

R1 The coupled anemochory-hydrochory model retrieves daily average wind speed and wind R2 direction for each time step of a day. All seeds that are shed at the time step are distributed R3 in the direction of the wind at the time step, and according to the frequency distribution of R4 dispersal distances belonging to the daily average wind speed of the time step. We used daily R5 wind data for 2009, location De Bilt, the Netherlands. This year was selected because it is R6 representative for the years 2000-2010 (KNMI 2011).

R7 6.2.4 Hydrochory module R8  Wind dispersed seeds that end up in ‘land cells’ remain there, whereas those that are blown R9                into ditch cells are consecutively transported via the network of ditches. Ditch cells into which R10              wind dispersed seeds are blown are hereafter called source water cells. R11 As a full individual based modelling approach, where processes are described at the level of R12 individual seeds, is not feasible at the scale of interest, we track homogeneous packages of  R13 seeds. For each time step and each grid cell, a package is created containing those seeds in the  R14 grid cell at the start of the time step that have the same residence time in the water, i.e. time R15 in days  since the  seeds  were  blown  into  the  ditch cell.  Homogeneity   of  residence   time  within R16 a package is required   as sinking  of seeds  in  a package    is modelled    as a  function  of  residence   R17 time. Up to the end of the time step, seeds in a package are transported in the downwind  direction through the network of ditches, with a speed v that is variable between ditch cells R18             t,i     i = 1,2,..,n. The value of v is calculated for each time step t and cell i as a function of wind R19    t,i           velocity at t (see Parameterisation). During this transport process, a proportion c of the seeds R20 transporting through it is captured in each cell. The seed fluxF (number of seeds per day R21 passing a location) of the package over each time step is modelled as: R22   R23     eq.1 R24   R25 in which x is downwind distance along the ditch (in unit cell length of 2 m) and c is capture R26 proportion   per  cell.               R27   R28 After seed transport   and capture,  a proportion   p of the captured   seeds remain  permanently  in R29 the cell in which  they  were  captured,  whereas   the rest  of the captured   seeds  become  mobile               R30 again. Subsequently, ratio rt,ts of the seeds that are mobile sink, whereas the remaining mobile seeds are transported  further  the  next  time  step. These  processes  are described   in  appendix A. R31  R32  6.2.5 Parametrisation R33  Anemochory R34 The STG model was parameterised using input values given in Table 6.2. Vegetation R35 characteristics were chosen representing a reed land shoreline community in a fen meadow              R36 nature reserve. Vegetation leaf area index (LAI) values were adapted from Hirose and Werger R37             (1995) and  vegetation  height  was measured   at (and  averaged  for)  6 representative   ditch R38 banks in fen meadow nature reserve the Westbroekse Zodden in the Netherlands. Average R39                          

  Wind and water dispersal of wetland plants | 135

seed release height was taken from Van der Meijden (1996). Terminal velocity was determined R1 (Soons, unpublished) using an experimental setup conforming to Soons and Heil (2002). For R2 this purpose, drop time was measured for 10 seeds randomly selected from four different R3 populations situated in central Netherlands. R4 R5 table 6.2: Parameter values for wind dispersal module. R6 C. pseudocyperus P. australis R7 N 1 000 000 c R8 Number of source cells 135 LAI (m2/m2) 3.16 R9 h (m) 1.17 R10 h (minimum-maximum) (m) 0.5-1.0 1.0-3.0 0 R11 d 0.61* R12 z0 0.059* -1 R13 vterm (mean/minimum/sd) (ms ) 2.16/0.51/0.823 0.15/0.11/0.036 R14 N : Number of seeds shed per cell per season. LAI: leaf area index. h: vegetation height. h : seed c 0 R15 release height. d: zero-plane displacement height. z0: momentum roughness length. b: terminal velocity. Values that were used for both species are placed in the middle.*values are calculated by R16 the STG model, based on LAI and h. R17 R18 Hydrochory R19 Seed speed and direction are significantly positively correlated to wind speed and direction in R20 ditches and no relation between water flow at mid-depth of a ditch and seed transport speed R21 and direction was found by Soomers et al. (2010). Therefore, seed transport direction for each R22 time step is determined by the average wind direction at the corresponding days in 2009 (KNMI 2011). At ditch junctions, the transport direction most similar to the wind direction is R23 chosen. R24 R25 R26 Seed transport speed per time step per ditch cell (vt,i) is calculated with use of a regression equation relating seed transport speed to net wind speed NWt at 1 m height, given in Soomers R27 et al. (2010). Wind speed at 1 m height is derived from observed wind speed at 20 m height R28 following Monteith (1973) (Table 6.3). The net wind speed in the length direction of the ditch R29 is calculated from wind speed using vector calculations (see Table 6.3). R30 R31 The values of the parameters p, c and r (see Appendix A) are determined in five seed release t,ts R32 and retrace experiments and 2 long term seed mimic experiments. In the seed release and R33 retrace experiments, 1000 seeds were released and retraced in an 810 m long ditch divided R34 into 2-meter sections (cf. Soomers et al. 2010 and Appendix B1). The long term seed mimic R35 experiments implied that the location of 250 individually marked seed mimics, released in a 1037 m. long ditch in a fen meadow reserve (de Westbroekse Zodden) in the Netherlands, was R36 registered every week during nine, respectively 16 weeks. Equations resulting from analysing R37 these experiments, and other parameter values for the hydrochory module, are given in R38 R39 136 | Chapter 6

R1 Table 6.3. Details on the experiments, statistical analyses and parameterisation are given in R2 Appendix B1-4. R3 R4 table 6.3.   Parameter values, and  (regression)  equations  used  to determine   parameter  values, for the water  dispersal  module.                              R5              C. pseudocyperus  P. australis R6 v (m/s)      0.0256NW  + 0.0232     t,i    t   R7        R8 NWt (m/s)                      R9                          R10                        (m/s)                             R11                 R12  c    ((1-α)*β)/10000       R13                 R14                                α        R15                             R16          β      R17          R18  c 0.96 culvert  p   R19  0.168   rt,ts a / [100 - a ((t - ts)-e) + a] R20    a  0.61  0.90 R21    e (days) 33.1  0.0 R22         v= seed transport speed at time step t for cell i, NW =net wind speed (m/s, in the direction of R23 t,i  t                     the ditch) at 1 m height, at time step t and cell i, WDt=daily average wind direction at time step t                  R24 (North= 0º)  (KNMI  2011),  DD=direction of  the  ditch  at  cell  i,  =horizontal wind speed  at time       i                         step t, 1 m height (Monteith 1973),   = horizontal wind speed at time step t, 20 m height (KNMI R25        2011), d =zero-plane   displacement height:     0.61,  z =momentum      roughness length:   0.059,  c=capture      0    R26 probability    at time step t for  cell i, α=zero  model   output of Zero-inflated   regression   (see Appendix    B1 for explanation), H=helophyte abundance (%), F=fine floating material (present: 1, absent: 0), R27                    M=coarse  floating material  (present: 1, absent: 0), β=count model output  of Zero-inflated  regression                       R28 (see Appendix   B1), c  =capture   probability   at  ‘culvert-cell’    in scenario   SW-HD-H-C.     Value  derived     culvert              R29 from experiments,     see    Appendix    B3,   p=permanent capture   probability    per cell.  Value    derived from                      experiments.   See  Appendix    B2,  r =sink   probability     for  time step  t and ‘starting   time    step’ ts (i.e. t,ts   R30                       time step at which the considered seeds entered the water), a=regression coefficient, derived from                          R31                                                data points from van  den Broek  et al.  (2005),  e =number  of  days  after  which   floating  seeds start                         R32 sinking, derived from data points from van  den Broek  et al.  (2005)  (See  Appendix   B4  for  explanation                  of a and e). R33               R34    R35  R36 R37 R38 R39

       

  Wind and water dispersal of wetland plants | 137

6.2.6 Validation R1 Seed release and retrace experiments (described in appendix B1 and in Soomers et al. 2010), R2 were used to validate the hydrochory module. Three out of eight randomly chosen repetitions R3 of the experiment were kept separate, to use them for model validation. We measured (or R4 obtained from databases) in one ditch the variables that drive seed transport via water in R5 the model. These variables were obtained for each experiment section (i.e. a 2 meter long R6 part of the ditch) which corresponds with one model cell. In the model, 1000 seeds were R7 released in the ditch at the same location as in the experiments. The model was run with the R8 settings of each experiment for 2 time steps. The percentage of released seeds located in R9 each model cell after 2 time steps of hydrochorous dispersal (48 hours) was compared with the percentage of retrieved seeds in the experiment-sections 48 hours after release. To assess R10 model performance, seed percentages in the model cells were related to seed percentages R11 in the experiment sections using a Spearman correlation test for all experiments together. R12 Furthermore, cumulative seed percentages for both the model and experimental results were R13 plotted against distance from release point for the three experiments separately. Details on R14 the validation procedure and experiments used for validation are given in Appendix C1. R15 Validation of the wind module (STG model) was performed and described by Soons et al. R16 (2004). R17 R18 6.2.7 Sensitivity analyses R19 To asses the sensitivity of the hydrochorous model component to changes in parameter values, R20 each parameter of the hydrochorous module was changed with plus and minus 50% and R21 90%, and the percentage change in median dispersal distance as a result of these parameter R22 changes was calculated. Sensitivity analysis for the wind module (STG model) was performed by Soons et al. (2004). R23 R24 6.2.8 Scenarios R25 We created four modelling landscapes representative for Dutch agricultural landscapes, in R26 which we varied 1) ditch direction and 2) ditch density in the landscape (Figure 6.1). In the R27 first two modelling landscapes (Figure 6.1 a and c), the majority of the ditches are oriented R28 from southwest towards northeast (along the main wind direction in the Netherlands), and R29 the ditch density is high respectively low (coded as scenario SW-HD and SW-LD, respectively). R30 In the third and the fourth modelling landscape, the majority of the ditches are oriented R31 perpendicular to the direction used in the first two landscapes, and also have a high and R32 low density of ditches, respectively (Figure 6.1 b and d, coded as scenario NW-HD and NW- R33 LD, respectively). In the high density landscapes, the distance between parallel ditches is 40 R34 meters whereas in the low density landscapes this is 300 meters. In agricultural areas, drainage R35 ditches are often connected with each other by ditches perpendicular to the majority of the ditches, to be able to drain excess water out of the area. Therefore, all modelling landscapes R36 include 3 ditches that are oriented perpendicular to the majority of the ditches. R37 R38 R39 138 | Chapter 6

R1 These four modelling landscapes were each run for both Carex pseudocyperus and Phragmites R2 australis and for two scenarios of ditch roughness: 1) roughness scenarios in which helophyte R3 abundance along the bank (0-50 cm from the ditch bank) was 25,6% per cell (the average R4 abundance in our experiments) and where no other obstructions were present in the ditch R5 (hereafter called the highway scenario. Final code-letter: H. Example code: SW-HD-H), and R6 2) roughness scenarios in which helophyte abundance was 50%, and fine floating material was present in the cells (hereafter called increased roughness scenario. Final code-letters: IR. R7 Example code: SW-HD-IR). See Appendix B1 for details on the roughness scenarios. R8 Furthermore, a scenario with 2 randomly placed culverts (i.e. pipes below a field that connect R9 ditches in case a road or a passage for cattle runs across a ditch) at each ditch-junction directly R10 around the population and 1 randomly placed culvert at each other ditch-junction was run R11 for both modelling species for the SW-HD landscape (code: SW-HD-H-C). See Appendix B3 for R12 details on the culvert scenario. R13 R14 R15 R16 R17 R18 R19 R20 R21 R22 R23 R24 R25 R26 R27 R28 R29 R30 R31 R32 R33 R34 R35 R36 R37 R38 R39 Wind and water dispersal of wetland plants | 139

R1 R2 R3 R4 R5 R6 R7 R8 R9 R10 R11 R12 R13 R14 R15 R16 R17 R18 R19 R20 R21 R22 R23 R24 R25 R26 R27 R28 R29 R30 R31 R32 Figure 6.1. The four modelling landscapes, and satellite images of examples of four Dutch landscapes R33 dominated by agriculture on which the modelling landscapes were based: a) the majority of the R34 ditches oriented southwest-northeast and a high ditch density (Polder Achttienhoven, N:52°09’42, E:5°08’38), b) the majority of the ditches oriented northwest-southeast and a high ditch density R35 (Polder Middelblok, N:51°59’01, E:4°41’58), c) the majority of the ditches oriented southwest- R36 northeast and a low ditch density (Southern Flevoland, N:52°24’25, E:5°25’03), and d) the majority R37 of the ditches oriented northwest-southeast and a low ditch density (Noordoostpolder, N:52°47’05, E:4°59’21). Satellite images derived from Google Earth. R38 R39 140 | Chapter 6

R1 6.3 results R2 R3 6.3.1 Validation R4 For three repetitions of the seed release-retrace experiment, the observed cumulative seed R5 percentages per distance after 48 hours were plotted versus the modelled percentages (see R6 Appendix C2; Figure C2.1). An overall correlation coefficient of 0.512 (Spearman correlation, P < 0.01) was found when relating the percentage of seeds in cells in the model output to the R7 observed percentage of seeds in experiment-sections. R8 The STG model, used in the wind module of our coupled model, performed well in simulating R9 realistic dispersal distances. The regression coefficient of a linear regression model relating R10 the median distance of simulated distances of the STG model to the measured distance per R11 dispersal event was close to 1 (0.92). Furthermore, of all measured dispersal events, 92% are R12 within the range simulated for them by the STG model (Soons et al. 2004). R13 R14 6.3.2 Scenario results R15 Water dispersal contributed clearly to the total dispersal kernel for both species (Table 6.4- R16 6.5). Even for the typical wind disperser, P. australis, the simulated 90th-percentile dispersal R17 distances after hydrochory were at least more than 3.5 times larger than the 90th-percentile R18 dispersal distances after wind dispersal only (Table 6.5). R19 Maximum simulated wind dispersal distance was only 3 m for C. pseudocyperus (median and 90th-percentile distances less than 2 m) whereas seeds of P. australis dispersed up to more R20 than 1000 m (out of the model area) (median: 6m, 90th-percentile: 28 m) (Table 6.4-6.5, Figure R21 6.2). Consequently, the percentage of the model area at which wind dispersed seeds were R22 deposited differed greatly between the two species; for P. australis this percentage was 623 R23 times higher than for C. pseudocyperus (Table 6.4-6.5). Nevertheless, even for the typical wind R24 disperser P. australis a large majority of seeds were, after wind dispersal, deposited close to the R25 source population, in contrast to seeds that had been wind and water dispersed (Figure 6.2 R26 and 6.4). The contribution of water dispersal to the total number of simulated seeds deposited R27 at distances more than 100 m from the seed source was much higher than the contribution of R28 wind dispersal, even for the typical wind disperser P. australis (Appendix D, Figure D1). R29 For P. australis, eighteen percent of the modelling area was occupied by wind dispersed seeds R30 after 305 time steps, which was at least 3.5 times higher than the percentage of the area at R31 which water dispersed seeds (that had been previously dispersed by wind) had deposited. R32 This was also true for landscapes with a dense ditch network; most of the simulated seeds landed on land. For C. pseudocyperus, however, the percentage of model-cells in which wind R33 dispersed seeds were present after deposition was much lower than that for water dispersed R34 seeds, because wind dispersal distances for this species were extremely short. R35 As we assumed the plants of the modelled source-population to grow at the ditch-edge and R36 because simulated wind dispersal distances were nil for C. pseudocyperus, results show that R37 more than 99% of its seeds entered the water. For P. australis, a higher percentage of seeds R38 landed at ‘land-cells’ after wind dispersal and thus did not enter the water. Therefore, the R39 Wind and water dispersal of wetland plants | 141

percentage of C. pseudocyperus seeds that were deposited at the ditch bank after water R1 dispersal was structurally higher compared to P. australis. (Table 6.4-6.5). Nevertheless, the R2 percentage of the area covered by seeds deposited after water dispersal (excluding the seeds R3 that only dispersed by wind) was similar for both species. Although buoyancy of P. australis R4 seeds is much lower than for C. pseudocyperus seeds, simulated water dispersal distances and R5 spatial patterns were also similar for the two species (Table 6.4-6.5, Figure 6.2). R6 For both species, simulated hydrochorous dispersal distances were highest for the landscape R7 with a high ditch density and a ditch orientation parallel to the main wind direction (SW-HD). R8 Nevertheless, spatial patterns did not differ markedly between the four landscapes differing R9 in ditch orientation and density (Figure 6.4). Orientation of ditches perpendicular to the main wind direction did not hinder seeds to disperse in the main wind direction. The seeds even R10 reached the ditches furthest away from the source population (Figure 6.4). R11 Increasing ditch roughness or placing culverts in the ditch and thereby increasing seed capture R12 probability has a much higher impact on the modelled dispersal distances than the effect of R13 spatial orientation or density of ditches (Table 6.4-6.5). The limiting effect of culverts on the R14 dispersal distance was more pronounced for C. pseudocyperus than for P. australis (Table 6.4 R15 and 5; 90th percentile water dispersal distances), because P. australis seeds are able to enter R16 the water behind a culvert due to their higher wind dispersal ability than C. pseudocyperus. R17 The percentage of released seeds that were deposited along the ditch bank inside the R18 modelling area and thus potentially could germinate (assuming that ‘land-cells’ are agricultural R19 fields and thus not suitable for germination) is higher for the scenarios with increased ditch R20 roughness than for the highway scenarios, because for the latter scenarios most hydrochorous R21 seeds dispersed out of the modelling area (Table 6.4-6.5, Figure 6.3a versus 6.3b). R22 R23 R24 R25 R26 R27 R28 R29 R30 R31 R32 R33 R34 R35 R36 R37 R38 R39 142 | Chapter 6

R1 R2 R3 R4 R5 R6 99.9 99.9 99.9 99.9 92.1 23.1 18.0 18.5 16.7 % of total number of released seeds deposited in ditch R7 R8 R9 R10 R11 R12 0.03 0.03 0.03 0.03 0.03 0.03 0.03 0.03 0.03 R13 Percentage of map- area covered by wind dispersing seeds (%)* R14 R15 R16 R17 R18 0.9 5.2 0.9 5.1 5.2 0.9 4.2 0.8 4.1 R19 Percentage of map-area covered by deposited hydrochorous seeds (%)* R20 R21 R22 from distances as Euclidian given are distances Dispersal scenarios. 9 different for R23 R24 R25 <2/<2 <2/<2 <2/<2 <2/<2 <2/<2 <2/<2 <2/<2 <2/<2 <2/<2 50th/90th percentile of transport distance after wind dispersal (m) R26 R27 were considered to be occupied. Carex pseudocyperus, Carex

R28 -45 R29 R30 R31 2/102 3/134 <2/104 48/307 4/827 >1000/>1000 >1000/>1000 >1000/>1000 >1000/>1000 R32 50th/90th percentile of transport distance of all hydrochorous seeds after wind and water dispersal (m) R33 ‡ R34 R35 R36 . Model results after 305 days for 305 days after results . Model R37 R38 able 6.4 able NW-LD-IR NW-HD-IR SW-LD-IR SW-HD-IR SW-HD-H-C NW-LD-H NW-HD-H SW-LD-H SW-HD-H Landscape and ditch roughness scenario Scenario codes: SW: majority of ditches oriented southwest-northeast, NW: majority of ditches oriented northwest-southeast, HD: high ditch majority of ditches oriented southwest-northeast, NW: Scenario codes: SW: density. LD:low ditch density, H=’highway scenario’, IR=increased roughness in ditch, C=culvert. These codes are combined to get the full scenario LD:low ditch density, density. codes as given in the first table column. For explanation of scenarios, see Figure 6.1 and subsection Scenarios. *Cells at which seed numbers were > 10 t the source. R39 ‡ Wind and water dispersal of wetland plants | 143

R1 § R2 R3 Note: a large

§ R4 R5 R6 99.9 13.36 99.9 15.47 99.9 13.35 99.9 19.01 92.1 16.66 23.1 3.18 18.0 2.83 18.5 2.54 16.7 % of total number of released seeds deposited in ditch 3.25 % of total number of released seeds deposited in ditch R7 R8 R9 R10 R11 R12 0.03 18.7 0.03 18.7 0.03 18.7 0.03 18.7 0.03 18.7 0.03 18.7 0.03 18.7 0.03 18.7 0.03 Percentage of map- area covered by wind dispersing seeds (%)* 18.7 Percentage of map- area covered by wind dispersing seeds (%)* were considered to be occupied. R13 -45 R14 R15 R16 R17 R18 0.9 0.9 5.2 5.2 0.9 0.9 5.1 5.2 5.2 5.2 0.9 0.8 4.2 4.9 0.8 0.8 4.1 Percentage of map-area covered by deposited hydrochorous seeds (%)* 4.8 Percentage of map-area covered by deposited hydrochorous seeds (%)* R19 R20 R21 for 9 different scenarios. Dispersal distances are given as Euclidian distances from distances as Euclidian given are distances Dispersal scenarios. 9 different for for 9 different scenarios. Dispersal distances are given as Euclidian distances. for 9 different R22 R23 R24 R25 <2/<2 6/28 <2/<2 6/28 <2/<2 6/28 <2/<2 6/28 <2/<2 6/28 <2/<2 6/28 <2/<2 6/28 <2/<2 6/28 <2/<2 50th/90th percentile of transport distance after wind dispersal (m) 6/28 50th/90th percentile of transport distance after wind dispersal (m) R26 R27 Phragmites australis, were considered to be occupied. Carex pseudocyperus, Carex

-45 R28 R29 R30 R31 2/102 2/102 3/134 10/140 <2/104 <2/101 48/307 57/310 4/827 4/>1000 >1000/>1000 >1000/>1000 >1000/>1000 >1000/>1000 >1000/>1000 >1000/>1000 >1000/>1000 50th/90th percentile of transport distance of all hydrochorous seeds after wind and water dispersal (m) >1000/>1000 50th/90th percentile of transport distance of all hydrochorous seeds after wind and water dispersal (m) R32 R33 ‡ R34 R35 R36 . Model results after 305 days for . Model results after 305 days for 305 days after results . Model R37 R38 able 6.5 able 6.4 able NW-LD-IR NW-LD-IR NW-HD-IR NW-HD-IR SW-LD-IR SW-LD-IR SW-HD-IR SW-HD-IR SW-HD-H-C SW-HD-H-C NW-LD-H NW-LD-H NW-HD-H NW-HD-H SW-LD-H SW-LD-H SW-HD-H Landscape and ditch roughness scenario SW-HD-H Landscape and ditch roughness scenario Scenario codes: SW: majority of ditches oriented southwest-northeast, NW: majority of ditches oriented northwest-southeast, HD: high ditch majority of ditches oriented southwest-northeast, NW: Scenario codes: SW: For explanation of scenario codes, see Table 6.4. * Cells at which seed numbers were > 10 For explanation of scenario codes, see Table density. LD:low ditch density, H=’highway scenario’, IR=increased roughness in ditch, C=culvert. These codes are combined to get the full scenario LD:low ditch density, density. codes as given in the first table column. For explanation of scenarios, see Figure 6.1 and subsection Scenarios. *Cells at which seed numbers were > 10 t t the source. ‡ percentage of seeds (approximately 80%) is deposited on land by wind dispersal. R39 144 | Chapter 6

R1 R2 R3 R4 R5 R6 R7 R8 R9 R10 R11 R12 R13 R14 R15 R16 R17 R18 R19 R20 R21 R22 R23 R24 R25 R26 R27 R28 2 R29 Figure 6.2. Simulated seed dispersal patterns [number of seeds per cell (4m )] for C. pseudocyperus (upper panels A and B) and P. australis (lower panels C and D). Left (A and C): seed distribution after R30 wind dispersal. Right (B and D): deposited hydrochorous seeds for the scenario with the majority R31 of the ditches oriented in southwest-northeastern direction, high ditch density, and increased ditch roughness (SW-HD-IR). R32 R33 R34 R35 R36 R37 R38 R39 Wind and water dispersal of wetland plants | 145

R1 R2 R3 R4 R5 R6 R7 R8 R9 R10 R11 R12 R13 R14 R15 R16 R17 R18 R19 R20 R21 R22 R23 R24 R25 R26 R27 R28 R29 R30 R31 R32 R33 R34 R35 R36 Figure 6.3. Simulated seed dispersal patterns of deposited hydrochorous seeds for Phragmites R37 australis, SW-HD landscapes. A) Highway scenario, B) Increased roughness scenario, C) Culvert R38 scenario R39 146 | Chapter 6

R1 R2 R3 R4 R5 R6 R7 R8 R9 R10 R11 R12 R13 R14 R15 R16 R17 R18 R19 R20 R21 R22 R23 R24 R25 R26 R27 R28 Figure 6.4. Simulated seed dispersal patterns of deposited hydrochorous seeds of Phragmites R29 australis. Highway scenarios, all four landscapes. R30 R31 6.3.3 Sensitivity analysis R32 Results of the sensitivity analyses for C. pseudocyperus demonstrate that the model is not sensitive to changes in seed transport speed and seed buoyancy (Table 6.6). The results show R33 that seed capture probability and permanent seed capture probability are most important in R34 determining hydrochorous seed dispersal distance for C. pseudocyperus. R35 Sensitivity analysis for the wind module (STG model) was performed by Soons et al. (2004). If R36 all four grassland study-species used by Soons et al. (2004) are analysed together, the median R37 distances simulated by the STG model were most sensitive to changes in seed release height, R38 followed by horizontal wind speed and terminal velocity. R39 Wind and water dispersal of wetland plants | 147

table 6.6. Median dispersal distances (m) by water for a 50% and 90% increase and decrease in each parameter value compared to the default settings (i.e. SW-HD-IR scenario for Carex pseudocyperus). R1 Parameter Default Parameter Parameter Parameter Parameter R2 settings value +50% value -50% value +90% value -90% R3 Seed speed, v 48.1 48.1 48.1 48.1 48.1 t,i R4 Capture probability, p 48.1 14.1 159.6 2.0 1080.1 t,i R5 Permanent capture, c 48.1 19.8 133.5 7.7 648 R6 Sinking, rt,ts 48.1 48.1 48.1 48.1 48.1 R7 R8 R9 6.4 dIsCussIon R10

Our results show that hydrochory contributes considerably to the dispersal of wetland plant R11 seeds across the agricultural landscape considered here, for both wind and water dispersal R12 specialists. In contrast, wind dispersal distances were, as expected, nil for water dispersal R13 specialists with high terminal velocity (> 2.0 m/s). Also for wind dispersal specialists, wind R14 dispersal distances were limited; 90% of the wind dispersed seeds remained within only 30 R15 m of the seed source. Seeds of both species’ types that ended up in the water, however, R16 could disperse effectively via the network of ditches in the agricultural landscapes, thereby R17 multiplying their original wind dispersal distance at least 3.5 times (90-percentile distances). R18 Note, however, that for the wind dispersal specialist, a smaller percentage of seeds end up in R19 the water because part of the seeds are blown over the ditches by wind dispersal and deposit R20 in the agricultural fields, where they can not establish. R21 The importance of hydrochory compared to anemochory was also stressed for a forested R22 floodplain in the USA by Schnieder and Sharitz (1988), who found that 10-100 times as many cypress seeds and tupelo fruits were transported into each plot by water as by wind R23 dispersal alone. Similarly, Jansson et al (2005) showed that along a Swedish river, 36-58% more R24 species colonised flooded plots (subjected to hydrochory) than unflooded plots that were R25 only subjected to anemochory and zoochory. These and our results show that hydrochorous R26 dispersal has great potential to increase riparian plant diversity. R27 Remarkably, our results show that, after entering the surface water, seeds of wind dispersal R28 specialists dispersed equally far by water as seeds of water dispersal specialists did. Note that R29 seeds of many species that are adapted to wind dispersal, not to water dispersal, also float R30 when they enter the surface water (Nilsson et al. 2010). Seeds with very low terminal velocity R31 (<0.2 m/s), and thus high ability for long distance wind dispersal, such as our wind dispersal R32 representative Phragmites australis, often have wings or hairs (Bouman et al. 2000, Nilsson R33 et al. 2010). These structures also increase their ability to remain on the water by surface R34 tension (Nilsson et al. 2010). Although the seeds of such species rarely float as long as seeds R35 of species typically adapted to hydrochory, our results show that seed buoyancy is not limiting hydrochorous dispersal distances for such wind dispersers on the scale of our study (up to 4 R36 km2). Pollux et al. (2009) demonstrated a trade-off between germination ability and floating R37 ability for Sparganium emersum (a typical water disperser). This species showed a negative R38 R39 148 | Chapter 6

R1 intra-specific correlation between seed buoyancy and germination success. If this generally R2 holds for hydrochorous dispersers, this would mean that investing in traits that increase R3 seed buoyancy is a disadvantage in the establishment phase. Moreover, if this is true, this R4 would mean that in rural landscapes investing in seed buoyancy is not an advantageous trait R5 anymore, since increased seed buoyancy apparently does not lead to a higher probability R6 of ending up in suitable sites, whereas it goes together with decreased germinative power. We hypothesize that investing in enhanced floating capacity only matters in more natural R7 conditions. Plants investing in buoyancy have evolved in natural floodplains in which the R8 hydrological dynamics are very different from the dynamics in managed agricultural areas. R9 When growing in a floodplain, seeds that are shed in summer may fall close to the source R10 plant on moist soil, but hydrochory will only be possible from autumn or winter on, when the R11 floodplain inundates (Chormanski et al. 2011). If seeds have lost buoyancy by then, they will R12 stay close to the source plant and will have to compete for resources with it after germination, R13 whereas hydrochorous dispersal would enable them to deposit and germinate at empty but R14 suitable patches further away. Although our results did not show clear differences in distance R15 or spatial pattern between the typical wind disperser and the typical water disperser, it should R16 be noted that we do not know the faith of the seeds leaving the modelling area, since the local R17 spatial scale of our study (up to 4 km2) does not allow extrapolating to larger spatial scales. R18 Possibly, extreme long-distance hydrochorous dispersal events can only take place for highly R19 buoyant seeds and not for less buoyant seeds (such as the species in our typical wind disperser group; Table 6.1). Because long-distance dispersal events are rare and highly stochastic, their R20 importance for plant distribution and population viability is not well understood and has R21 been debated (Nathan 2006). R22 Besides the differences between wind and water dispersal, we investigated the effects R23 of ditch orientation, -density and ‘-roughness’ on hydrochorous dispersal distances and R24 patterns. The orientation of the ditches relative to the main wind direction did not seem to R25 be very important: hydrochorous dispersal is only slightly more efficient in landscapes with R26 southwest-northeast oriented ditch patterns, which is parallel to the main wind direction R27 in the Netherlands. As long as ditches are connected to each other by some perpendicular R28 ditches, a seemingly unfavourable main orientation of ditches does not largely limit dispersal R29 distance. A decrease in ditch density did not clearly affect dispersal distances or patterns either. R30 However, changing helophytic plant abundances or other obstructions in the ditches greatly R31 affected model results. For the so-called ‘highway scenarios’ the majority of the hydrochorous R32 seeds left the modelling area, whereas for the scenarios with increased ditch roughness by plant material and for the scenarios with culverts at ditch crossings at least 90% of the seeds R33 that had entered the water stayed within the modelling area. The huge difference in the R34 dispersal distances predicted by the highway scenario on the one hand and the increased R35 roughness- and culvert scenarios on the other hand show that cleaning ditches or constructing R36 alternative designs for culverts (f.i. wider diameter, bridges) would probably greatly magnify R37 realised hydrochorous dispersal distances. Considering our results, these measures would have R38 a higher positive effect on dispersal capacity than digging a denser network of extra ditches R39 or creating extra short-cuts between ditches. Wind and water dispersal of wetland plants | 149

It should be noted, that temporal and permanent seed capture probability parameters were R1 not species specific in the hydrochorous dispersal model we developed. Seed size or shape R2 could affect deposition probability (see: Schneider and Sharitz 1988), and therefore dispersal R3 experiments with different seed types should be performed to improve parameterisation of R4 water dispersal models. Furthermore, considering the sensitivity of our model to permanent R5 capture probability, a reliable way to assess this parameter should be found. To assess R6 permanent capture probability, it is necessary to mark seeds individually. However, marking R7 seeds individually is difficult for real seeds, and seed mimics do not necessarily behave as real R8 seeds. Also, retracing real (usually small) seeds during a long period of time (preferably one R9 year) is difficult, if not impossible. Alternatively, population genetic techniques could be used to calibrate and/or validate dispersal models (Ouborg et al. 1999) over a large temporal and R10 spatial scale. R11 Summarising, more field data, collected over large temporal and spatial scale, is needed R12 to reliably parameterise models such as ours. However, model validation showed that the R13 hydrochory module underestimated water dispersal distances (see Appendix C2), meaning R14 that the conclusion that dispersal via water ensures considerably larger dispersal distances R15 than dispersal via wind for both wind and water dispersal specialists is valid despite of model R16 uncertainty. R17 Additionally, we would like to note that we assumed similar seed numbers for both model R18 species. In reality, seed numbers greatly differ between species, and therefore one should R19 keep in mind that species with high seed numbers will have higher probability for successful R20 dispersal to a patch at a certain distance from the seed source than species comprising seeds R21 with similar dispersal characteristics but producing less seeds. Given the sensitivity of our R22 model to the seed capture parameters and the limited extent of the experiments used to determine these parameter values, combined with the mediocre fit between observed and R23 modelled data, conclusions on dispersal distances should be drawn carefully and focus on R24 relative differences between scenarios, species or dispersal mechanisms and on general spatial R25 patterns rather than absolute values. Nevertheless, Geertsema (2005), who investigated R26 colonisation and extinction events for riparian plant populations at ditch banks along arable R27 fields, found that most colonisation events took place within 50 m of the nearest conspecific R28 population, but that colonisation distances exceeding 200 m also occurred. These figures are in R29 the same order of magnitude as the results of our increased roughness and culvert scenarios, R30 after dispersal by wind and water. This may indicate that the model we developed generates R31 realistic seed deposition patterns. For the so-called highway scenarios (scenarios with low R32 ditch-roughness) most of the seeds dispersed out of the modelling area (travel distances of R33 more than one kilometre), which is further than those found by Geertsema (2005). R34 Despite of the apparent potential for long distance dispersal of hydrochorously dispersing R35 seeds, Ozinga et al. (2009) stress that species with adaptations for water dispersal are overrepresented among declining species. This is probably caused by the lack of connectivity R36 between riparian habitats via water nowadays, caused by the regulation of rivers by dams and R37 sluices. These obstructions are known to prevent effective hydrochorous dispersal via rivers; R38 free flowing rivers enable more efficient hydrochorous dispersal than artificially controlled R39 150 | Chapter 6

R1 rivers (e.g. Andersson et al. 2000, Jansson et al. 2000, Merritt and Wohl 2006). Similarly, our R2 results show that culverts greatly reduce water dispersal distances in agricultural landscapes, R3 especially for water dispersal specialists, which are not able to overcome these barriers R4 by initial long distance wind dispersal. These results stress the importance of free flowing R5 connected water bodies for population connectivity and colonisation and thus for habitat R6 restoration, especially for species with seeds that are mostly dependent on surface water as a dispersal vector, considering their very limited wind dispersal ability. It is known that many R7 plant seeds can also be dispersed by animals (zoochory; (Clausen et al. 2002, Soons et al. 2008, R8 Brochet et al. 2010) or humans (Strykstra et al. 1997, Wichmann et al. 2009). Although such R9 dispersal events are highly stochastic and difficult to track, it will be important to include them R10 in future dispersal studies, in order to assess the relative importance of such dispersal events. R11 Besides the transport process, another aspect important for effective seed exchange between R12 populations or for colonisation is that seeds also need to deposit at places suitable for R13 germination. The most favourable measure for seed exchange and colonisation would be R14 to create ‘dispersal highways’ at places where habitat quality is unsuitable for germination R15 or establishment anyway, and increase deposition probability at suitable places. The latter R16 can for instance be realised by a less regular ditch cleaning regime, enabling helophytes to R17 grow along the ditch banks, or creating seed ‘landing strips’ consisting of ditch banks with a R18 physical structure suitable for catching seeds (see: Soomers at al. 2010). R19 To summarize, our study suggests that in a modern agricultural landscape with a network of ditches in which hydrochory is driven by wind, water is an important dispersal vector for both R20 typical wind and typical water dispersers, whereas wind as a dispersal vector is only relevant R21 for typical wind dispersers. Via wind seeds can disperse in any direction whereas dispersal R22 via water is restricted to the spatial lay-out of the surface water infrastructure. Nevertheless, R23 dispersal distances by water surpass those by wind, even for typical wind dispersing species. R24 R25 acknowledgments R26 We would like to thank Cees Wesseling for his support during model development. Furthermore, R27 Florus Sibma and Saskia van de Venne are acknowledged for doing field experiments. R28 R29 R30 R31 R32 R33 R34 R35 R36 R37 R38 R39 Wind and water dispersal of wetland plants | 151

reFerenCes R1 R2 Andersson, E., C. Nilsson, and M. E. Johansson. 2000. Effects of river fragmentation on plant dispersal R3 and riparian flora. River Research and Applications 16:83-89. Atkins, D. C., and R. J. Gallop. 2007. Rethinking how family researchers model infrequent outcomes: R4 a tutorial on count regression and zero-inflated models. Journal of Familiy Psychology R5 21:726-735. Beltman, B., and T. van den Broek. 2006. Het beperkte succes van Laagveenrestauratie. Landschap. R6 Bewley, J. D. 1997. Seed germination and dormancy. Plant Cell 9:1055-1066. R7 Biopop. 2011. Biopop. Biologische Parameter des Populationsmanagements bei Pflanzen. R8 Blomqvist, M. M., W. L. M. Tamis, J. P. Bakker, and E. van der Meijden. 2006. Seed and (micro) site limitation in ditch banks: Germination, establishment and survival under different R9 management regimes. Journal for Nature Conservation 14:16-33. R10 Blomqvist, M. M., P. Vos, P. G. L. Klinkhamer, and W. J. Ter Keurs. 2003. Declining plant species richness of grassland ditch banks - A problem of colonisation or extinction? Biological R11 Conservation 109:391-406. R12 Boedeltje, G. 2005. The role of dispersal, propagule banks and abiotic conditions in the establishment R13 of aquatic vegetation. Radboud Universiteit Nijmegen, Nijmegen. Boedeltje, G., J. P. Bakker, R. M. Bekker, J. M. Van Groenendael, and M. Soesbergen. 2003. Plant R14 dispersal in a lowland stream in relation to occurrence and three specific life-history traits R15 of the species in the species pool. Journal of Ecology 91:855-866. Bootsma, M. C. 2000. Stress and recovery in wetland ecosystems. Universiteit Utrecht, Utrecht. R16 Borger, G. J. 1992. Draining, digging, dredging; the creation of a new landscape in the peat areas R17 of the low countries. Pages 131–171 in J. T. A. Verhoeven, editor. Fens and Bogs in the R18 Netherlands: Vegetation, History, Nutrient Dynamics and Conservation. Kluwer Academic Publishers, Dordrecht. R19 Bouman, F., D. Boesewinkel, R. Bregman, N. Devente, and G. Oostermeijer. 2000. Verspreiding van R20 zaden. KNNV Uitgeverij, Utrecht. Brochet, A. L., M. Guillemain, H. Fritz, M. Gauthier-Clerc, and A. J. Green. 2010. Plant dispersal R21 by teal (Anas crecca) in the Camargue: Duck guts are more important than their feet. R22 Freshwater Biology 55:1262-1273. R23 Bunce, R. G. H., and C. J. Hallam. 1993. The ecological significance of linear features in agricultural landscapes in Britain. Pages 11-19 in R. G. H. Bunce, L. Ryskowski, and M. G. Paoletti, R24 editors. Landscape Ecology and Agroecosystems. Lewis Publishers, Boca Raton. R25 Campbell, G. S., P. G. Blackwell, and F. I. Woodward. 2002. Can landscape-scale characteristics be used to predict plant invasions along rivers? Journal of Biogeography 29:535-543. R26 Chormanski, J., T. Okruszko, S. Ignar, O. Batelaan, K. T. Rebel, and M. J. Wassen. 2011. Flood mapping R27 with remote sensing and hydrochemistry: A new method to distinguish the origin of flood R28 water during floods. Ecological Engineering 37:1334-1349. Clausen, P., B. A. Nolet, A. D. Fox, and M. Klaassen. 2002. Long-distance endozoochorous dispersal R29 of submerged macrophyte seeds by migratory waterbirds in northern Europe - A critical R30 review of possibilities and limitations. Acta Oecologica 23:191-203. Davies, B. R., J. Biggs, P. J. Williams, J. T. Lee, and S. Thompson. 2008. A comparison of the catchment R31 sizes of rivers, streams, ponds, ditches and lakes: Implications for protecting aquatic R32 biodiversity in an agricultural landscape. Hydrobiologia 597:7-17. R33 Donald, P. F., and A. D. Evans. 2006. Habitat connectivity and matrix restoration: The wider implications of agri-environment schemes. Journal of Applied Ecology 43:209-218. R34 Fahrig, L. 2003. Effects of Habitat Fragmentation on Biodiversity. Pages 487-515 Annual Review of R35 Ecology, Evolution, and Systematics. Fischer, J., and D. B. Lindenmayer. 2007. Landscape modification and habitat fragmentation: A R36 synthesis. Global Ecology and Biogeography 16:265-280. R37 R38 R39 152 | Chapter 6

Foley, J. A., R. DeFries, G. P. Asner, C. Barford, G. Bonan, S. R. Carpenter, F. S. Chapin, M. T. Coe, G. C. R1 Daily, H. K. Gibbs, J. H. Helkowski, T. Holloway, E. A. Howard, C. J. Kucharik, C. Monfreda, R2 J. A. Patz, I. C. Prentice, N. Ramankutty, and P. K. Snyder. 2005. Global consequences of R3 land use. Science 309:570-574. Geertsema, W. 2005. Spatial dynamics of plant species in an agricultural landscape in the Netherlands. R4 Plant Ecology 178:237-247. R5 Geertsema, W., P. Opdam, and M. J. Kropff. 2002. Plant strategies and agricultural landscapes: survival in spatially and temporally fragmented habitat. Landscape Ecology 17:263-279. R6 Geertsema, W., and J. T. C. M. Sprangers. 2002. Plant distribution patterns related to species R7 characteristics and spatial and temporal habitat heterogeneity in a network of ditch R8 banks. Plant Ecology 162:91-108. Goodson, J. M., A. M. Gurnell, P. G. Angold, and I. P. Morrissey. 2003. Evidence for hydrochory and R9 the deposition of viable seeds within winter flow-deposited sediments: The River Dove, R10 Derbyshire, UK. River Research and Applications 19:317-334. Groves, J. H., D. G. Williams, P. Caley, R. H. Norris, and G. Caitcheon. 2009. Modelling of floating seed R11 dispersal in a fluvial environment. River Research and Applications 25:582-592. R12 Gurnell, A. M., A. J. Boitsidis, K. Thompson, and N. J. Clifford. 2006. Seed bank, seed dispersal and R13 vegetation cover: Colonization along a newly-created river channel. Journal of Vegetation Science 17:665-674. R14 Hanski, I. 2005. Landscape fragmentation, biodiversity loss and the societal response - The longterm R15 consequences of our use of natural resources may be surprising and unpleasant. EMBO reports 6:388-392. R16 Hirose, T., and M. J. Werger. 1995. Canopy structure and photon flux partitioning among species in R17 a herbaceous plant community. Ecology 76:466-474. R18 Jansson, R., C. Nilsson, and B. Renofalt. 2000. Fragmentation of riparian floras in rivers with multiple dams. Ecology 81:899-903. R19 Jansson, R., U. Zinko, D. M. Merritt, and C. Nilsson. 2005. Hydrochory increases riparian plant species R20 richness: a comparison between a free-flowing and a regulated river. Journal of Ecology 93:1094-1103. R21 Kahle, P., B. Tiemeyer, B. Eichler-Löbermann, and B. Lertnartz. 2008. Phosphorus losses from an R22 agriculturally used lowland catchment in North-Eastern Germany. Phosphoraustrag aus R23 Einem Landwirtschaftlich Genutzten Einzugsgebiet des Nordostdeutschen Tieflandes 98:10-15. R24 Karssenberg, D., K. de Jong, and J. van der Kwast. 2007. Modelling landscape dynamics with Python. R25 International Journal of Geographical Information Science 21:483-495. Katul, G. G., A. Porporato, R. Nathan, M. Siqueira, M. B. Soons, D. Poggi, H. S. Horn, and S. A. R26 Levin. 2005. Mechanistic analytical models for long-distance seed dispersal by wind. The R27 American Naturalist 166:368-381. R28 Kleyer, M, R. M. Bekker, I.C. Knevel, J.P. Bakker, K. Thompson, M. Sonnenschein, P. Poschlod, J.M. van Groenendael, L. Klimes, J. Klimesová, S. Klotz, G.M. Rusch, M. Hermy, D. Adriaens, R29 G. Boedeltje, B. Bossuyt, A. Dannemann, P. Endels, L. Götzenberger, J.G. Hodgson, A-K. R30 Jackel, I. Kühn, D. Kunzmann, W.A. Ozinga, C. Römermann, M. Stadler, J. Schlegelmilch, H.J. Steendam, O. Tackenberg, B. Wilmann, J.H.C. Cornelissen, O. Eriksson, E. Garnier, B. R31 Peco. B. 2008. The LEDA Traitbase: A database of life-history traits of Northwest European R32 flora. Journal of Ecology 96:1266-1274. R33 KNMI. 2011. Klimatologie. Informatie over Het Weer in het Verleden KNMI. Kröger, R., M. T. Moore, J. L. Farris, and M. Gopalan. in press. Evidence for the Use of Low-Grade R34 Weirs in Drainage Ditches to Improve Nutrient Reductions from Agriculture. Water, Air R35 and Soil Pollution in press:1-12. Leng, X., C. J. M. Musters, and G. R. de Snoo. 2011. Effects of mowing date on the opportunities of R36 seed dispersal of ditch bank plant species under different management regimes. Journal R37 for Nature Conservation 19:166-174. R38 Levine, J. M. 2003. A patch modeling approach to the community-level consequences of directional dispersal. Ecology 84:1215-1224. R39 Wind and water dispersal of wetland plants | 153

Maljanen, M., B. D. Sigurdsson, J. Guömundsson, H. Öskarsson, J. T. Huttunen, and P. J. Martikainen. 2010. Greenhouse gas balances of managed peatlands in the Nordic countries present R1 knowledge and gaps. Biogeosciences 7:2711-2738. R2 Merritt, D. M., and E. E. Wohl. 2006. Plant dispersal along rivers fragmented by dams. River Research R3 and Applications 22:1-26. Middleton, B., R. Van Diggelen, and K. Jensen. 2006. Seed dispersal in fens. Applied Vegetation R4 Science 9:279-284. R5 Milsom, T. P., A. J. Sherwood, S. C. Rose, S. J. Town, and S. R. Runham. 2004. Dynamics and management of plant communities in ditches bordering arable fenland in eastern England. Agriculture, R6 Ecosystems & Environment 103:85–99. R7 Monteith, J. L. 1973. Principles of environmental physics. Edward Arnold, London, UK. R8 Moore, M. T., and R. Kröger. 2011. Evaluating plant species-specific contributions to nutrient mitigation in drainage ditch mesocosms. Water, Air, and Soil Pollution 217:445-454. R9 Moss, B. 1983. The Norfolk Broadland: experiments in the restoration of a complex wetland. R10 Biological Reviews 58:521–561. Nathan, R. 2001. Dispersal biogeography. Pages 127-152 in S. A. Levin, editor. Encyclopedia of R11 Biodiversity. Academic Press, New York. R12 Nathan, R. 2006. Long-distance dispersal of plants. Science 313:786-788. R13 Nathan, R., H. S. Horn, J. Chave, and S. A. Levin. 2001. Mechanistic models for tree seed dispersal by wind in dense forests and open landscapes.in D. J. Levey, W. R. Silva, and M. Galetti, editors. R14 Seed dispersal and frugivory: Ecology, Evolution and Conservation. CAB International R15 Press, Oxfordshire. Nathan, R., G. G. Katul, G. Bohrer, A. Kuparinen, M. B. Soons, S. E. Thompson, A. Trakhtenbrot, and R16 H. S. Horn. 2011. Mechanistic models of seed dispersal by wind. Theoretical Ecology In R17 press:1-20. R18 Nathan, R., G. G. Katul, H. S. Horn, S. M. Thomas, R. Oren, R. Avissar, S. W. Pacala, and S. A. Levin. 2002. Mechanisms of long-distance dispersal of seeds by wind. Nature 418:409-413. R19 Nilsson, C., R. L. Brown, R. Jansson, and D. M. Merritt. 2010. The role of hydrochory in structuring R20 riparian and Wetland vegetation. Biological Reviews 85:837-858. Ouborg, N. J., Y. Piquot, and J. M. Van Groenendael. 1999. Population genetics, molecular markers R21 and the study of dispersal in plants. Journal of Ecology 87:551-568. R22 Ozinga, W. A., C. Römermann, R. M. Bekker, A. Prinzing, W. L. M. Tamis, J. H. J. Schaminée, S. M. R23 Hennekens, K. Thompson, P. Poschlod, M. Kleyer, J. P. Bakker, and J. M. Van Groenendael. 2009. Dispersal failure contributes to plant losses in NW Europe. Ecology Letters 12:66-74. R24 Pelacani, S., M. Märker, and G. Rodolfi. 2008. Simulation of soil erosion and deposition in a R25 changing land use: A modelling approach to implement the support practice factor. Geomorphology 99:329-340. R26 Pollux, B. J. A., E. Verbruggen, J. M. Van Groenendael, and N. J. Ouborg. 2009. Intraspecific variation R27 of seed floating ability in Sparganium emersum suggests a bimodal dispersal strategy. R28 Aquatic Botany 90:199-203. Sarneel, J. M. 2010. Colonisation processes in riparian fen vegetation. Utrecht University, Utrecht. R29 Schneider, R. L., and R. R. Sharitz. 1988. Hydrochory And Regeneration In A Bald Cypress Water R30 Tupelo Swamp Forest. Ecology 69:1055-1063. Skarpaas, O., O. E. Stabbetorp, I. Ronning, and T. O. Svennungsen. 2004. How far can a hawk’s beard R31 fly? Measuring and modelling the dispersal of Crepis praemorsa. 92:747-757. R32 Soomers, H., D. N. Winkel, Y. Du, and M. J. Wassen. 2010. The dispersal and deposition of hydrochorous R33 plant seeds in drainage ditches. Freshwater Biology 55:2032-2046. Soons, M. B. 2006. Wind dispersal in freshwater wetlands: Knowledge for conservation and R34 restoration. Applied Vegetation Science 9:271-278. R35 Soons, M. B., and G. W. Heil. 2002. Reduced colonization capacity in fragmented populations of wind-dispersed grassland forbs. Journal of Ecology 90:1033-1043. R36 Soons, M. B., G. W. Heil, R. Nathan, and G. G. Katul. 2004. Determinants of long-distance seed R37 dispersal by wind in grasslands. Ecology 85:3056-3068. R38 R39 154 | Chapter 6

Soons, M. B., C. Van Der Vlugt, B. Van Lith, G. W. Heil, and M. Klaassen. 2008. Small seed size R1 increases the potential for dispersal of wetland plants by ducks. Journal of Ecology 96:619- R2 627. R3 Strykstra, R. J., G. L. Verweij, and J. P. Barker. 1997. Seed dispersal by mowing machinery in a Dutch brook valley system. Acta Botanica Neerlandica 46:387-401. R4 Tackenberg, O. 2003. Modeling long-distance dispersal of plant diaspores by wind. Ecological R5 Monographs 73:173-189. Thomson, F. J., A. T. Moles, T. D. Auld, D. Ramp, S. Ren, and R. T. Kingsford. 2010. Chasing the R6 unknown: Predicting seed dispersal mechanisms from plant traits. Journal of Ecology R7 98:1310-1318. R8 van den Broek, T., R. van Diggelen, and R. Bobbink. 2005. Variation in seed buoyancy of species in wetland ecosystems with different flooding dynamics. Journal of Vegetation Science R9 16:579-586. R10 Van der Meijden, R. 1996. Heukels’ Flora van Nederland. Wolters-Noordhoff, Groningen. Van Loon, A. H., P. P. Schot, M. F. P. Bierkens, J. Griffioen, and M. J. Wassen. 2009. Local and regional R11 impact of anthropogenic drainage on fen contiguity. Hydrology and Earth System Sciences R12 13:1837-1848. R13 Van Strien, A. J., J. Van Der Linden, T. C. P. Melman, and M. A. W. Noordervliet. 1989. Factors affecting the vegetation of ditch banks in peat areas in the western Netherlands. Journal of Applied R14 Ecology 26:989-1004. R15 Vaughan, R. E., B. A. Needelman, P. J. A. Kleinman, and A. L. Allen. 2007. Vertical distribution of phosphorus in agricultural drainage ditch soils. Journal of Environmental Quality 36:1895- R16 1903. R17 Wichmann, M. C., M. J. Alexander, M. B. Soons, S. Galsworthy, L. Dunne, R. Gould, C. Fairfax, M. R18 Niggemann, R. S. Hails, and J. M. Bullock. 2009. Human-mediated dispersal of seeds over long distances. Proceedings of the Royal Society B: Biological Sciences 276:523-532. R19 R20 R21 R22 R23 R24 R25 R26 R27 R28 R29 R30 R31 R32 R33 R34 R35 R36 R37 R38 R39 Wind and water dispersal of wetland plants | 155

aPPendICes R1   R2   appendix a: Model processes R3                       R4       eq. A.1             R5    R6 In equation A.1, the number of mobile seeds S , in source water cell i at time step t is t,i           R7 given by the number of seeds dispersed by wind into cell i at time step t ( F ,, itwind ) plus the         R8 additional number of mobile seeds in cell i at time step t (which depends on the number of R9 hydrochorous seeds entering the cell at time step t-1). The latter seeds originate from wind dispersal at previous time steps. These can be calculated by multiplying the number of mobile R10  seeds entering cell i from an upstream water cell at time step t-1 (i.e. F ) with the capture R11 t-1,i  R12 probability c, resulting  in the number   of captured  seeds  in  cell i  at time step  t-1. The  number  of    captured seeds is then multiplied  with  1 minus the proportion   of captured  seeds  that  remain    R13  permanently in cell i (p) to get the number of seeds in cell i that are mobile again. Then, R14  mobile seeds are multiplied with 1 minus the proportion of mobile seeds that sink (rt-1,ts) to R15 obtain the number of mobile seeds (that were already in the water at time step t-1) in cell i at R16  (the beginning of) time  step  t. The  proportion of  seeds  that  sink is dependent    on  the residence     R17                             time of the seeds (i.e. time since seeds entered the water: t minus ‘time starting’ ts). Therefore, R18 mobile seeds in cell i at t-1 should   be calculated  for  all seed  packages (see  section hydrochory         R19 module) separately, and then added up, resulting in the total number of mobile seeds in cell   R20 i at the end of t-1 and thus in the beginning of time step t (   )                       R21  (except those originating  from  wind  dispersal  at time   step t).                  R22   R23  R24 R25 R26 R27 R28 R29 R30 R31 R32 R33 R34 R35 R36 R37 R38 R39

    156 | Chapter 6    R1 appendix b: Parameterisation   R2   R3 B1: Capture   probability   c          R4 To parameterise   the  proportion   c of  seeds  that remains   in a water  cell i during transport  through   R5 a cell, we used the seed release and retrace experimental design described in Soomers et al. R6 (2010). Five out of eight experiments, executed in 2008 and 2009 in an 810 m. long ditch in the Westbroekse Zodden (the Netherlands, N:52°9’43”, E:5°7’1”), were used for parameterisation, R7  three for validation of the model. For these experiments, 1000 painted seeds were released in R8  the ditch that was divided into sections of 2 meters each, which corresponds to the cell size in R9 the model. For parameterisation, the number of captured seeds in each section, 3 hours after R10 release was registered and transformed to the percentage of the total number of retraced R11 seeds.  We  used the  location  of  the captured   seeds three  hours  after  release because  this time  R12 span  allowed  the  seeds to reach  the end of the ditch,  given  the speed  with which  they were  R13 transported. Furthermore, the explanatory variables helophyte cover 0-50 cm from the bank, R14 the slope  of the  bank, the  wind speed  vector perpendicular  to the  ditch  and  the presence or   R15 absence  of 1) floating coarse  plant material (mowing remnants/aquatic    plants)  and 2) fine   R16 material  such as algae,  duckweed  or an ‘oil’  film on top of the water caused by  iron-oxidising               R17 bacteria, were  registered.    Because  of excess zero’s  in the dependent   variable (i.e.  percentage              R18 of captured  seeds  in a  section,  multiplied  with  100), a  zero-inflated  Poisson  regression  analysis                        R19 was used to investigate the relation between seed percentage and explanatory variables. Zero-inflated    models  are two-component       mixture  models, in  which a  point  mass  at  zero is  R20 combined   with  a  count  distribution    (Atkins    and Gallop    2007).  Only  independent   variables   that  R21            were significant   in both regression  model  components   were selected  as explanatory  variables   R22              in the final  model. This procedure resulted   in the  following  regression    equations (eq. B1.1  and   R23              B1.2) for the zero-component α and the count-component β of the zero-inflated regression  R24  model: R25            R26 α=  eq. B1.1         R27    R28       β=  eq. B1.2 R29      R30     R31   R32 c=((1-α)*β)/10000        eq. B1.3         R33 from which the seed proportion that remains in a cell when transporting through it, and thus R34 the parameter value for capture probability, c (eq. B1.3), can be calculated. H is the percentage R35 of helophyte cover in a section, 0-50 cm from the ditch bank, F is the presence (1) or absence R36 (0) of fine organic   material and M the  presence   (1) or  absence  (0)  of coarse  organic  material.   R37 The resulting  regression  model  was significantly    better  than  the 0-model  P( <0.001,  Likelyhood     R38 Ratio Test). The denominator of equation B1.3 is 10000 because the dependent variable in the R39                                                

    Wind and water dispersal of wetland plants | 157

regression model was the percentage of seeds that were captured in a cell times 100, whereas R1 the parameter value for c is the ratio of seeds that are captured in a cell. R2 For model validation, a special modelling landscape, in which every model cell corresponded R3 to a ditch section of the experiment ditch, was created. To calculate c for each model cell of R4 this model validation landscape for each of the three model validation runs, values for the R5 independent variables, such as measured in the ditch section corresponding to the considered R6 model cell during the three experiments that were kept separate, where used as input for R7 these eq. 5 and 6. R8 For the real model simulations, for which the modelling landscapes visualised in Figure 6.1 R9 were used, two scenarios of ditch roughness were distinguished: 1) roughness scenarios in which helophyte abundance along the bank (0-50 cm from the ditch bank) was 25,6% per cell R10 (the average abundance in our experiments) and where no other obstructions were present R11 in the ditch (called the highway scenario, H), and 2) roughness scenarios in which helophyte R12 abundance was 50%, and fine floating material was present in the cells (called the increased R13 roughness scenario, IR). For both scenarios, helophyte abundance was assumed to be 5% in R14 winter (1st of November-31st of March). Filling in these values for the independent variables in R15 equations B1.1-B1.3 results in a capture probability c of 0.002 for the highway scenarios in the R16 summer season and 0.0007 in the winter season and in a capture probability c of 0.078 in the R17 increased roughness scenarios during the summer season and of 0.016 in the winter season. R18 In these scenarios, c was assumed to be equal for each cell (in contrast to the validation runs). R19

R20 B2: Permanent capture probability p R21 After seed transport and capture, part of the captured seeds remain permanently in the cell in R22 which they were captured, whereas the rest of the captured seeds becomes mobile again. To parameterise the proportion of captured seeds that are deposited permanently, a long term R23 seed mimic experiment was executed as follows. R24 Two hundred and fifty individually marked flat, round, cork seed mimics with a diameter R25 of 4 cm were released in each of two ditches in the Westbroekse Zodden (the Netherlands, R26 N:52°9’43”, E:5°7’1”). A pilot study in which both seed mimics and Carex pseudocyperus R27 seeds were released and retraced in a ditch divided into 10-m sections, revealed that R28 seed deposition one day and three days after release was well represented by seed mimic R29 deposition (Spearman correlation, r=0.933, P<0.001 and r=0.963, P<0.001 respectively). Every R30 week, the xy coordinates of each individual mimic that could be retraced was registered. R31 Thereby, it could be determined whether seed mimics had moved over the week. Seed mimics R32 in ditch 1 could be followed during 9 weeks, after which the ditch was cleaned, and mimics R33 in ditch 2 could be followed during 16 weeks. In both ditches, part of the mimics did not R34 move anymore after they had reached their first deposition location the day after release, R35 which confirms the assumption that part of the captured seeds stay at their capture location permanently. Unfortunately it is almost impossible to execute such an experiment for a whole R36 year. Therefore, we had to assume that mimics that had not moved for several weeks would R37 not become mobile again. Because only 12 mimics of ditch 2 could be retraced every week, R38 R39 158 | Chapter 6

R1 we only used the data retrieved in ditch 1 to determine the percentage of seeds that are R2 deposited permanently after capture (56 mimics). For week 1-5, we calculated the percentage R3 of the mimics captured in that week that were still at the same location at the last observation  R4 day (after 9 weeks). We used a solution of a negative exponential growth equation (eq. B2.2),  to determine the ratio of the captured seeds per day that remains at this location until the R5   R6 end of the experiment.    R7                eq. B2.1 R8        R9                R10           eq. B2.2 R11                        R12 Equation B2.1  represents  the  negative  exponential  growth  equation  for  which  equation  B2.2   R13 is the solution for continuous time. Z(t) refers to the number of seed mimics still floating after   R14 7 days (t=7), Z is the initial number  of floating seed mimics in a certain experiment week and t 0   R15 is the number  of days after  which  the  location  of the  mimics was  registered  again  (7). For each  week (week  1-5), the proportion   of captured  seeds  that are captured   permanently  each  time R16              R17 step, p, was calculated using equation B2.3, which follows from equation B2.2.  R18      R19     eq. B2.3            R20     The weighted average of these proportions that were calculated for experiment week 1-5 R21   was used as the parameter value for the proportion of captured seeds that are deposited R22  permanently each time step (p=0.168 (SD: 0.11)).  R23     R24 B3: Capture probability of culverts c  culvert R25   An experiment was performed to determine the effect of culverts on seed transport. We  R26 released 200 painted C. pseudocyperus seeds in front of 3 different culverts in the Westbroekse               R27 Zodden and placed a net behind the culvert. This experiment was executed fifteen times in                             R28 total. On average, only 4% (SD: 4.9) of the seeds had been transported through the culvert one day after release. Therefore, the ratio c of seeds captured in or before a culvert when R29 culvert    R30 transporting through it was set to 0.96.     R31  B4: Sinking ratio r  R32  t,ts  Data of Van den Broek et al. (2005), who measured seed buoyancy of seeds of wetland species R33               during 210 days, were used to determine the proportion of floating seeds that sunk each day, R34                             for both species. Van den Broek et al. determined for both of our model species the days at R35  which 10%, 25%, 50%, 75% and 90% of the seeds had sunk. We used these data points to R36  determine a linear regression line relating percentage of sunken seeds to number of days  R37 after release. When the intercept was positive, the regression line was forced through the  R38  origin to avoid a situation in the hydrochory module in which seeds start sinking before they               R39                                     Wind and water dispersal of wetland plants | 159

are actually released. The data points of Van den Broek et al. (2005), and the fitted regression R1 lines are plotted in Figure B4.1. Equations for the regression lines are given in the caption of R2 Figure B4.1. R3 R4 R5 R6 R7 R8 R9 R10 R11 R12 R13 Figure b4.1. Percentage of sunken seeds plotted versus the number of days after release, for Carex R14 pseudocyperus and Phragmites australis. Data points (diamond symbols connected by dashed lines) taken from Van den Broek et al. (2005), for stagnant water. Equation of the fitted linear regression R15 line (solid lines) for C. pseudocyperus: y=0.61x-20.1, R2=0.99. Equation of the fitted linear regression R16 2 line, forced through the origin, for P. australis: y=0.90x, R =0.63 R17 R18 The regression equation given in Figure B4.1 can be written as R19 R20 S=a(t-ts) + i eq. B4.1 R21 R22 for both species, in which S is the percentage of sunken seeds, a is the regression coefficient (a=0.61 for C. pseudocyperus, a=0.90 for P. australis), t is the time step, ts is the time step at R23 which the considered seeds were shed (thus: t-ts is the number of days after release of the R24 considered seeds, x), and i is the intercept (i=-20.1 for C. pseudocyperus and i=0 for P. australis) R25 of the regression lines. R26 R27 To determine each time step the ratio of the still floating seeds that sink, equation B4.2, that R28 follows from equation B4.1, is used in the hydrochory module: R29 R30 r = a / [100 - a ((t - ts)-e) + a] eq. B4.2 t,ts R31 R32 r in equation B4.2 is the ratio of the seeds that are still floating that sink at time step t and t,ts R33 were shed into the water at time step ts, e represents the number of days after release at R34 which seeds start sinking. For both species, e is calculated by calculating the value for t-ts in R35 equation B4.1 for which S equals 0 (e=33.1 for C. pseudocyperus, e=0 for P. australis). If t-ts

R1 appendix C: model validations R2 R3 C1: Details on model validation procedure R4 Seed release and retrace experiments (described in appendix B1 and in Soomers et al. 2010), R5 were used to validate the hydrochory module. Three out of eight randomly chosen repetitions R6 of the experiment described in appendix B1 were kept separate, to use them for model validation. We measured (or obtained from databases) in one ditch the variables that drive R7 seed transport via water in the model. These variables were obtained for each experiment R8 section (i.e. a 2 meter long part of the ditch) which corresponds with one model cell. R9 The variables were wind speed and direction on the days at which the experiments were R10 executed (read from the KNMI database (KNMI, XXXX)), helophyte density along the bank, R11 presence of fine floating material (such as algae, a bacterial film, or duckweed), presence of R12 coarse floating material such as mowing remnants, presence of a culvert, and seed buoyancy. R13 Capture probability c varied per model cell, according to helophyte density and presence of R14 fine or coarse floating material (see equations in Appendix B1). In the model, 1000 seeds were R15 released in the ditch at the same location as in the experiments. Because in the experiments R16 seeds were released at 10 am, the average daily wind speed and direction used in the model R17 was calculated from 10 am that day until 10 am the next day. R18 Because we never observed seeds in the nets, which were placed at the end of the ditch for R19 five out of the eight experiments, we could assume that seeds we did not find back were not lost because they had been transported over longer distances than we could trace. Therefore, R20 we assumed that the proportion of retrieved seeds in a cell of the total number of retrieved R21 seeds would be the same as the proportion of retrieved seeds of the total number of released R22 seeds if all seeds had been retrieved. R23 The model was run with the settings of each experiment [i.e. wind speed of the days of the R24 experiment, helophyte density etc. for each model cell (corresponding to a ditch-section) at R25 the days of the experiment] for 2 time steps. The percentage of released seeds located in R26 each model cell after 2 time steps of hydrochorous dispersal (48 hours) was compared with R27 the percentage of retrieved seeds in the experiment-sections 48 hours after release. To assess R28 model performance, seed percentages in the model cells were related to seed percentages R29 in the experiment sections using a Spearman correlation test for all experiments together. R30 Furthermore, cumulative seed percentages for both the model and experimental results were R31 plotted against distance from release point for the three experiments separately. R32 R33 R34 R35 R36 R37 R38 R39 Wind and water dispersal of wetland plants | 161

C2: model validation: observed versus simulated distances R1 R2 R3 R4 R5 R6 R7 R8 R9 R10 R11 R12 R13 R14 R15 R16 R17 R18 R19 R20

R21 R22 R23 R24 R25 R26 R27 R28 R29 R30 R31 R32 Figure C2.1. Cumulative percentages of released seeds plotted versus transport distance for three independent release experiments (solid lines) with C. pseudocyperus seeds and for the model R33 simulations (dashed lines), parameterised with the settings of the corresponding experiments [wind R34 speed and direction of the days of the experiment, helophyte cover and presence or absence of fine or coarse organic material per experiment section (sections correspond with model cells), buoyancy R35 of 100% (according to Appendix B4: C. pseudocyperus seeds start sinking at day 33), permanent R36 capture probability according to Appendix B2)]. Observed and simulated percentages per distance R37 two days after release are plotted. R38 R39 162 | Chapter 6

R1 appendix d: model simulation results R2 R3 R4 Chapter 7 R5 R6 R7 R8 R9 R10 R11 R12 R13 R14 R15 R16 R17 Figure d.1. The contribution of each dispersal mode (1) wind or 2) water after entering via wind for R18 scenario SW-HD-IR) to the total number of Phragmites australis seeds that deposited at a certain R19 distance class from the centre of the seed source. SW-HD-IR: majority of the ditches southwest- R20 northeast oriented, high ditch density, increased roughness (see Appendix B1 and Figure 6.1). R21 R22 R23 R24 R25 R26 R27 R28 R29 R30 R31 R32 R33 R34 R35 R36 R37 R38 R39 Chapter 7

synthesis 164 | Chapter 7

R1 syntHesIs R2 R3 The aim of this thesis was to shed light on the contribution of hydrochory to seed dispersal in R4 fragmented freshwater wetlands impacted by human activity, with special attention for the R5 effect of landscape configuration and characteristics of the water body on realised dispersal R6 distances. R7 Conclusions with respect to the research questions formulated in chapter 1 (Introduction) are R8 given in table 7.1. Furthermore, in this chapter, these results are synthesised and discussed in R9 the context of existing literature, and directions for future research and implications of the R10 results for wetland management are outlined. R11 R12 As explained in chapter 1, human land use change has led to habitat fragmentation of R13 areas all over the world. The resulting spatially discontinuous habitat patches may contain R14 metapopulations, consisting of a number of spatially separated sub-populations that interact R15 with each other through dispersal to a greater or lesser extent. Local extinction may lead R16 to empty patches that may be colonised again after successful dispersal and establishment. R17 Metapopulations can exhibit different structures, depending on the landscape in which they R18 are located. In Figure 7.1, three possible metapopulation structures are given. These may be R19 found in uplands (Figure 7.1 a), riparian vegetation along a river (Figure 7.1 b), and riparian vegetation along ditches (Figure 7.1 c) are given. R20 R21 R22 R23 R24 R25 R26 R27 R28 R29 R30 R31 R32 R33 R34 R35 R36 R37 R38 R39 Synthesis | 165

R1 R2 R3 R4 R5 R6 R7 R8 R9 R10 R11 R12 R13 R14 R15 R16 R17 R18 R19 R20 R21 R22 R23 R24 R25 R26 R27 R28 Figure 7.1. Possible metapopulation structures in landscapes dominated by agriculture of a) R29 vegetation in upland habitat, b) riparian vegetation along a river and c) ditch bank vegetation. R30 R31 R32 R33 R34 R35 R36 R37 R38 R39 166 | Chapter 7

R1 R2 R3 Chapter 2 3 3 4 4 5 R4 R5 R6 R7 R8 R9 R10 R11 R12 R13 R14 R15 R16 R17 R18 R19 R20 R21 R22

R23 Answer explaining in important most were content calcium and level water -Ground occurrence. on isolation- (and thus seed dispersal) and/or edge effects -Additionally, species occurrence were highly significant for 5 out of 6 species. between the amount of Both a seed sink and source. No clear difference seed inflow and outflow was found. -Main seed flow out of the floodplain was related to high production, tall plant stature and high seed buoyancy low elevation of growing location. -Seed flow into the floodplain was related to high seed weight. -Seeds of riverbank species occurred significantly more often in the river water than expected. -High river water levels seemed to promote seed transport from the into the floodplain or from river. -Net wind speed was the main factor determining rate at which hydrochorous seeds were transported. seed transport speed mid-depth and flow at water relation between -No was found. of wind on the rate transport floating -A direct and an indirect effect seeds were found: not only transported by wind shear stress on was reflected The latter effect the water surface, but also by wind directly. in the fact that seeds for which ratio of volume protruding from water was greater were transported relatively faster. Plant cover of emergent aquatic plants along the bank (helophythes), shallow ditch bank slope and indentations in the promoted seed deposition. if seed dispersal is neglected and thereby implicitly assumed unlimited Yes; considerably is distribution species potential modelled, realistically of instead overestimated. Therefore, biodiversity management measures and policy should not be based on habitat suitability models lacking realistic dispersal modules. R24 R25 R26 R27 R28 R29 R30 R31 R32 R33 R34 R35 Research question fen plant six characteristic of occurrence species in a Dutch semi-natural fen area? net seed source or sink? characteristics are related to seed inflow into the floodplain and to seed outflow? of transport of rate the to flow water and floating seeds in drainage ditches? banks determine seed deposition? improved by the explicit modelling of dispersal? R36 Answers to the research questions formulated in chapter 1 for chapters 2-6. 1 What are the key-factors that explain the 2 Does the considered floodplain function as a 3 Which plant traits and growing location 4 What are the relative contributions of wind 5 What characteristics of ditches and ditch 6 Are species distribution models considerably R37

R38 number Question R39 7.1. Table Synthesis | 167

R1 R2 6 6

R3 th ). R4 R5 R6 R7 R8

Carex pseudocyperus R9 R10 R11 R12 R13 percentile<2m for

th R14 R15 R16 R17 ) was simulated to be 28 m. (Appendix D, Figure D1). R18 R19 R20 P. australis P. R21

percentile wind dispersal distance for the typical wind disperser tested disperser wind typical the for distance dispersal wind percentile R22 th Phragmites australis percentile between 100 m and >1000 m, depending on ditch roughness obstructions, after wind and water dispersal). -The contribution of water dispersal to the total number simulated seeds deposited at distances more than 100 m from the seed source was much higher than the contribution of wind dispersal, even for typical disperser Model simulations suggest that: -Density or direction of the ditch network do not seem to influence water dispersal distances substantially. -Roughness of the ditch and obstructions in strongly limit hydrochorous seed dispersal distances. -Modelled dispersal distances of seeds that end up in the water were at least 3.5 times longer after hydrochorous dispersal compared to wind for typical wind dispersers, and even 50 times longer or more distance only, for typical water dispersers. terminal high with dispersers water typical for distances dispersal -Wind velocity were simulated to be nil (90 -90 ( dispersal distances for typical wind dispersers were simulated to -Water be similar to those of typical water dispersers, within polder scale (90 R23 R24 R25 R26 R27 R28 R29 R30 R31 R32 R33 R34 R35 the landscape (f.i. ditch direction, -density, the landscape (f.i. ditch direction, -density, determine the -roughness, obstructions) dispersal distances of seeds? and water dispersal to total seed distances in landscapes with different configurations of drainage ditches? R36 8 To what extent do system characteristics of 7 What is the relative contribution of wind- R37 R38 R39 168 | Chapter 7

R1 7.1 Fragmentation R2 To investigate the effect of habitat fragmentation on freshwater wetland species, data of six R3 characteristic species of small sedge vegetation in a fen area situated in the Holocene part of R4 the Netherlands (the Vechtstreek) were analysed (chapter 2). Small sedge fens are species-rich R5 ecosystems (f.i. belonging to the alliance Caricion davallianae (Hooftman and Diemer 2002)), R6 which are threatened vegetation types in Europe (Boeye et al. 1996, Hooftman et al. 2004) and for which the Netherlands has an international responsibility. This is reflected in the fact that R7 most areas in which this type of vegetation is present are part of the Natura 2000 network in R8 the Netherlands (Ministerie van Economische Zaken, Landbouw en Innovatie 2011). R9 The results showed that fragmentation has a clear additional effect on the distribution of R10 these characteristic fen species (Table 7.1: question 1. chapter 2). On a regional scale, the R11 negative effect of fragmentation was detectable, besides the effect of abiotic factors on R12 species occurrence (chapter 2). Similar results were found for arid ecosystems by Pueyo et al. R13 (2007) and for forests by Verheyen and Hermy (2001). The negative effects of isolation and R14 increased habitat edge were clear and univocal for several species (chapter 2). The fact that R15 variables related to isolation had a negative effect on species occurrence of Carex lasiocarpa, R16 Juncus subnodulosus and Menyanthes trifoliata suggests that at least these species are R17 dispersal limited, which probably means that many other wetland species I did not test may R18 be dispersal limited as well. Many seed sowing experiment studies also found proof for seed R19 limitation in plant species (e.g. Primack and Miao 1992, McKenna and Houle 2000, Ehrlen et al. 2006). Leng et al. (2009) found that plant diversity at ditch banks in agricultural areas R20 decreased significantly with distance to source populations in nature areas, indicating that R21 seed limitation also plays a role for ditch bank vegetation (see also (Leng et al. 2010). R22 At the community level, both seed dispersal limitation and ecological filters have been proposed R23 as determinants for biodiversity (e.g. Grime 1979, Lord and Lee 2001, Xiong et al. 2003). R24 Ecological filters may be biotic (f.i. competition) or abiotic (f.i. nutrient or water availability). R25 Myers and Harms (2009) conducted a meta-analysis of 62 experiments that investigated how R26 seed supply interacted with ecological filters in determining local plant species richness. They R27 conclude that species richness at the study sites was clearly dispersal limited, and that seed R28 arrival and ecological filters interactively determine species richness in a variety of ecosystems, R29 among which wetlands. Evidence for dispersal limitation at the population level in a site, such R30 as presented in chapter 2, does not necessarily mean that a positive relationship between seed R31 arrival and species richness will be found, because immigrants might outcompete resident R32 species. However, considering the results of Meyers and Harms (2009) and the fact that the species that were found to be dispersal limited in our study (see above) are not tall, fast R33 growing species that are expected to become dominant, it may be expected that increased R34 seed arrival will lead to higher local plant species richness in freshwater wetlands such as the R35 area investigated in chapter 2. Conversely, ongoing habitat fragmentation and corresponding R36 further isolation of populations might decrease species richness in freshwater wetlands. R37 R38 R39 Synthesis | 169

Ground water level and calcium content of the groundwater in the root zone are abiotic R1 filters indicated as important for the occurrence of all the 6 characteristic fen species (chapter R2 2). These conditions are met with the supply of alkaline, nutrient-poor groundwater (Wassen R3 et al. 1996). In chapter 5, the potential distribution of one of the rich-fen species analysed in R4 chapter 2 (Carex diandra) was predicted as a function of water management actions that take R5 these requirements into account, considering seed dispersal by wind and water. The water R6 management actions considered in chapter 5 aim at increasing the total habitat area for this R7 species by a larger supply of alkaline, nutrient-poor groundwater to the rootzone. That study R8 emphasised the importance of coupling a dispersal model to habitat suitability models (Table R9 7.1: question 6). Stand-alone habitat suitability models, that neglect dispersal and implicitly assume dispersal to be unlimited, considerably overestimate potential species distribution. R10 To use financial and natural resources as efficiently as possible, it is advisable that realistic R11 dispersal models will be added to habitat suitability models, to optimise spatial planning of R12 restoration measures. R13 The negative effect of habitat fragmentation should also be kept in mind in programs aimed R14 at conservation of existing wetland vegetation. It should be realised that isolated patches R15 result in lower viability than connected ones, even if site factors are suitable. Incorporating R16 realistic dispersal modules in metapopulation models can improve model results in assessments R17 of metapopulation viability in wetlands. Higgins & Cain (2002) showed that results of simple R18 patch-occupancy metapopulation models differed greatly from metapopulation models R19 in which more realistic assumptions on dispersal and local population dynamics were R20 incorporated. R21 R22 7.2 Hydrochory In order to investigate the importance of hydrochory in freshwater wetlands, an innovative R23 combination of experimental (chapter 3 and 4) and modelling techniques (chapter 6) was R24 used in this thesis. In agricultural areas dissected by ditches, of which the banks may serve as R25 habitat for characteristic wetland species at some places (Bunce and Hallam 1993, Blomqvist R26 et al. 2006), hydrochory was found to be an important dispersal mode enabling long distance R27 dispersal (chapter 4 and 6 (Table 7.1: question 7)) ((Cain et al. 2000, Tackenberg 2003), for R28 instance, defined long distance dispersal of plants as dispersal over 100 m). Even in these R29 slow flowing to stagnant ditches, not only the extremes, but a substantial part of water- R30 dispersing seeds can be expected to become transported over more than several hundreds of R31 meters in such ecosystems (chapters 4 and 6). Although it could be argued that dispersal over R32 hundreds of meters is not yet long distance dispersal, our results contradict the earlier idea R33 in the literature that most seeds become dispersed only zero to a few tens of meters (Howe R34 and Smallwood 1982, Cain et al. 2000), and that long distance dispersal events are unusual R35 (Nathan 2006). Our results suggest that even for wetland species specifically adapted to wind dispersal, R36 water dispersal is an effective dispersal mode. Moreover, on a ‘polder scale’, seeds of these R37 species that enter the water may disperse equally far via water as typical water dispersers R38 R39 170 | Chapter 7

R1 do (chapter 6), which was much further than the distance they travelled by wind dispersal. R2 These results are in line with the results of Higgins et al. (2003), who calculated from data of R3 Fridriksson (1975) that the number of species arriving at Surtsey island by water dispersal was R4 approximately four times higher than the number of species arriving by animal dispersal, and R5 approximately seven times higher than those arriving by wind dispersal. When the number of R6 species arriving by a certain dispersal vector was inferred from seed morphology, the number of species arriving by water was erroneously estimated to be half of those arriving by wind R7 and similar to the number of species arriving by animals. Our results and those of Higgins et R8 al. (2003) suggest that it is not advisable to draw conclusions on dispersal mechanisms and R9 distances of species based on seed morphology only. R10 At distances more than 100 m from the seed source, even for wind dispersal specialists the R11 number of seeds of riparian species deposited by water dispersal is likely to be higher than the R12 number of seeds dispersed by wind only and deposited on land (chapter 6, Table 7.1 question R13 7). This result is caused by the fact that also for wind dispersal specialists most seeds deposit R14 very close to the seed source when dispersing by wind, while those seeds that fall in water R15 rather than on land may be effectively carried by water dispersal. Our results imply that the R16 importance of hydrochorous dispersal for connectivity within metapopulations of riparian R17 species in a fragmented landscape may be greater than the importance of wind dispersal, R18 which is almost exclusively local. R19 Long distance dispersal has been recognised as disproportionately important for colonisation and metapopulation dynamics (He et al. 2004, Nathan 2006). Our results suggest that the R20 contribution of hydrochory to long distance dispersal might be relatively large, even in R21 slow flowing waters, and thus that hydrochory may be disproportionately important for R22 metapopulation dynamics in areas with surface water. In line with my results, different studies R23 found that water dispersal is important for colonisation in ecosystems with surface water R24 flowing faster than in drainage ditches (e.g. tidal marshes and river floodplains). Neff and R25 Baldwin (2005) found significantly more seeds in water dispersing seed traps than in equally R26 sized wind dispersing seed traps in a tidal marsh. Also species numbers were higher in water R27 traps than in wind traps. Merritt et al. (2010) tested the relative importance of wind and R28 water dispersal in the colonisation of bare habitat along a river through controlling both R29 dispersal sources with exclosures. Species richness during the initial colonisation years was R30 significantly higher for plots receiving only water dispersed seeds than for those receiving R31 only wind dispersed seeds. Similarly, Schneider and Sharitz (1988) found that approximately R32 10-100 times as many propagules of their two focal species were transported into a forested floodplain by water as by wind only. Additionally, hydrochorous dispersal distances of more R33 than 2 km were found in rivers, also indicating the importance of hydrochory for long distance R34 dispersal (e.g. Andersson et al. 2000, Griffith and Forseth 2002). These and my results (Table R35 7.1: question 7) suggest that hydrochory is an important dispersal mode affecting species R36 composition and metapopulation dynamics in different types of wetlands. However, note R37 that, logically, hydrochorous dispersal can only be of importance for plants for which the R38 seeds are able to enter surface water. The restricted wind dispersal distances found in chapter R39 Synthesis | 171

6, and the results from chapter 3 from which it became clear that mostly species situated R1 directly along the river or situated at sites subject to flooding could disperse seeds via the river R2 water (Table 7.1: question 3), suggest that hydrochory might almost exclusively be important R3 for riparian species as a dispersal vector. R4 R5 Considering the literature on hydrochory, and the results of chapter 6, different types of R6 metapopulation dynamics can be expected in different landscapes (Figure 7.2). In uplands R7 without surface water, hydrochory does not play a role (Fig. 7.2 A). In fragmented landscapes R8 with rivers – e.g. a river landscape where intensive agriculture is the dominant land use type R9 except for a number of riparian wetlands scattered along the river – two situations may occur. Hydrochory prevails for riparian or aquatic species and most seed transport events will R10 be unidirectional, leading to higher downstream biodiversity and genetic diversity within R11 populations (Figure 7.2 B). Evidence for such dynamics was found by for instance Pollux et R12 al. (2009) and Levine (2001). However, in the other situation zoochory and anemochory may R13 be equally effective as hydrochory (Fig. 7.2 C). Such a situation could for instance exist if R14 downstream hydrochorous dispersal is hindered by river fragmentation by dams. In that case R15 no downstream increase in biodiversity and genetic diversity will be found (see: (Jansson et R16 al. 2000, Imbert and Lefèvre 2003, Hu et al. 2010) (Figure 7.2 C). In agricultural areas without R17 rivers but with a network of drainage ditches, the dispersal within a metapopulation of ditch R18 bank vegetation will be multidirectional as well (chapter 4 and 6), probably leading to a more R19 evenly distributed biodiversity and genetic variation pattern than in riparian systems along R20 rivers for which hydrochory prevails (Figure 7.2 D). However, it should be noted that, although R21 multidirectional, hydrochorous dispersal in drainage ditches will be mainly in the direction R22 of the main wind direction (see Figures 6.4-6.6, chapter 6). Sarneel (2010), who investigated hydrochorous seed dispersal in ponds, also found that over the seasons, most seeds were R23 deposited at downwind banks. R24 R25 R26 R27 R28 R29 R30 R31 R32 R33 R34 R35 R36 R37 R38 R39 172 | Chapter 7

R1 R2 R3 R4 R5 R6 R7 R8 R9 R10 R11 R12 R13 R14 R15 R16 R17 R18 R19 R20 R21 R22 R23 R24 R25 R26 R27 R28 R29 R30 R31 R32 R33 Figure 7.2. Different types of metapopulation dynamics in different landscapes. A) In uplands R34 without surface water, hydrochory does not play a role. B) In fragmented landscapes with rivers R35 in which hydrochory prevails, most seed transport events will be unidirectional, leading to higher R36 downstream biodiversity and genetic diversity within populations. C) In river systems in which hydrochory, and zoochory and anemochory are equally effective, no downstream increase in R37 biodiversity and genetic diversity will be found. D) In agricultural areas with a network of drainage R38 ditches, the dispersal within a metapopulation of ditch bank vegetation will be multidirectional, probably leading to a more evenly distributed biodiversity and genetic variation pattern than in R39 riparian systems along rivers for which hydrochory prevails. Synthesis | 173

High seed buoyancy seemed to increase the probability that floodplain species disperse their R1 seeds via river water (chapter 3, Table 7.1: question 3), whereas model simulations suggested R2 that buoyancy is not very important in determining dispersal distances of seeds of riparian R3 species dispersing through drainage ditches (chapter 6). This apparent contradiction might R4 be caused by the expected delay in dispersal in floodplains: flooding seems to promote seed R5 exchange between the river and the floodplain (chapter 3 of this thesis, (Middleton 2000, R6 Boedeltje et al. 2004), but flooding is mostly restricted to the end of autumn until spring R7 whereas the peak of the seed shedding season is in summer/beginning of the autumn. Seeds R8 that are shedded in summer may fall close to the source plant on moist floodplain soil, but R9 hydrochory will only be possible from autumn or winter on, when the floodplain inundates (Chormanski et al. 2011). Relatively short floating seeds might have lost buoyancy by then R10 and consequently are not transported to the main channel when the river water recedes. R11 For riparian species along ditches, on the other hand, the timing is probably different; in R12 most cases, seeds might either end up in the water body immediately, or not at all. Many R13 characteristic species of for instance (rich) fens, wet meadows and reed beds have seeds that R14 float for more than a week (Boedeltje et al. 2003, van den Broek et al. 2005), even when R15 seed morphology suggests that they are adapted to dispersal by wind. Chances are high that R16 such seeds, if they enter the water, are deposited in the ditch bank before they would sink R17 (chapter 6). In line with this, Goodson et al. (2003) claim, in an experimental study on seed and R18 sediment deposition along a river, that the different hydrodynamic properties of various seed R19 species may be more important than seed buoyancy in determining hydrochorous dispersal R20 distance and deposition patterns. Probably only the extreme dispersal events (falling outside R21 the modelling area used in chapter 6) are influenced by seed buoyancy. Also in wider linear R22 water bodies or in lakes, where deposition probability might be lower because of the smaller edge/area ratio of the water body, buoyancy might be more important. R23 R24 Summarising, our results suggest that habitat fragmentation per se may indeed have negative R25 effects on species occurrence in freshwater wetlands, and that hydrochorous dispersal might R26 play an important role in the connectivity between fragments for riparian species. Therefore, R27 improving the infrastructure of linear aquatic systems for hydrochrorous seeds could increase R28 metapopulation viability of riparian species, wetland biodiversity, and the success of wetland R29 restoration projects. R30 R31 7.3 Consequences for wetland restoration and conservation R32 Our results show that in wetland restoration and conservation projects, the negative effect of R33 habitat fragmentation and especially of dispersal limitation should be kept in mind (chapter R34 2). It is advisable to create new nature reserves in the proximity of existing ones, to enable R35 gene flow and colonisation between and from occupied patches. Even when seeds of target species would be supplied artificially to reduce the necessity of natural colonisation, viability R36 of an isolated patch may be lower than a patch connected to others through propagule R37 dispersal (Ouborg et al. 2006). The optimal spatial location of new patches or the viability R38 R39 174 | Chapter 7

R1 of metapopulations could be assessed using habitat suitability models incorporating realistic R2 dispersal models of both anemochorous and hydrochorous (and ideally also zoochorous) R3 dispersal (chapter 5 and 6). Instead of considering Euclidian distance only, the focus should R4 be on roughness and configuration of dispersal corridors, dispersal characteristics of target R5 species, and location of existing populations, apart from local abiotic conditions. The results in R6 this thesis show that water bodies, such as ditches and rivers, can be efficient dispersal corridors for hydrochorous seeds (chapter 3, 4 and 6). This is supported by many other studies that R7 conclude that dispersal is important for colonisation and vegetation composition in riparian R8 ecosystems (e.g. Boedeltje et al. 2004, Jansson et al. 2005, Neff and Baldwin 2005, Merritt et R9 al. 2010, Nilsson et al. 2010). The national ecological network and green-blue veining (i.e. a R10 network of semi-natural terrestrial (green) or aquatic (blue) landscape elements such as ditches, R11 ditch banks and hedgerows in rural, agricultural areas) are concepts in Dutch Nature policy R12 that aim at connecting nature areas (Grashof-Bokdam and Meeuwsen 2005, Grashof-Bokdam R13 and Van Langevelde 2005). The results in this thesis that show the importance of ditches R14 as dispersal corridors for characteristic riparian wetland species underline the importance of R15 such networks for biodiversity. However, although the green-blue veining network comprises R16 corridors connecting nature areas and linear habitat elements for wild species in otherwise R17 hostile agricultural land, this network should not be considered as an alternative for nature R18 areas, but as an addition. R19 Obviously, reducing the effects of fragmentation by constructing dispersal corridors or stepping stones will not have a positive effect on species viability if abiotic circumstances are R20 not suitable in newly connected patches. On the other hand, increasing habitat suitability in R21 a patch will not lead to colonisation if seed dispersal is limiting. R22 Ground water level and calcium content of the groundwater in the root zone are factors R23 indicated as important for the occurrence of all the 6 characteristic fen species considered in R24 chapter 2, and are thereby conditions that probably should have the highest priority when R25 aiming at restoring or conserving abiotic conditions for rich-fens, besides keeping nutrient R26 availability low (Wassen et al. 1996, Bedford et al. 1999). R27 When degraded, desiccated fen reserves in areas dominated by agriculture are to be restored, R28 eliminating drainage ditches might raise the groundwater table and thereby improve local R29 conditions for target species (chapter 2). However, such measures will likely isolate the R30 reserve from nearby areas more than several tens of metres away from the reserve (chapter R31 6), because the eliminated drainage ditches no longer function as hydrochorous dispersal R32 corridors for riparian species. Although model results (chapter 5) suggest that elimination of drainage ditches might have a high potential to increase the total area of rich fens, this R33 measure will (without artificial seed supply, and assuming depleted seed banks (e.g. Valko et R34 al. 2011) only be effective when applied for a large area, creating vast continuous fen habitat R35 that can be colonised gradually, over years, by a source of target species via wind dispersal (see R36 chapter 5). However, such large scale nature restoration projects at the expense of agricultural R37 land are unfortunately not feasible in most western-European countries. Results of chapter 6 R38 show that hydrochorous seeds are almost as effectively dispersed through a landscape with R39 Synthesis | 175

a low density of ditches as through a landscape with a high density of ditches. Therefore, R1 when aiming at restoring rich-fen vegetation, it could be recommended to eliminate enough R2 ditches to restore groundwater flux towards the area, but keep few shallow ditches in tact R3 along the area, to enable seeds to enter the water to disperse to other habitat patches. When R4 aiming for optimal connectivity between nature areas, or between sub populations of riparian R5 plant species at ditch banks, it will be advisable to replace culverts in ditches by, for instance, R6 bridges where possible, so that seeds can disperse unhindered through the dispersal corridors R7 (chapter 6). R8 R9 7.4 Recommendations for further research In this thesis, we investigated hydrochorous dispersal via a river and via slow flowing drainage R10 ditches. When considering the literature on river ecosystems, it is striking that, to my knowledge, R11 no spatially explicit process based models simulating hydrochororous dispersal through a river R12 exist until now, although much experimental research on hydrochory via rivers is done (e.g. R13 Boedeltje et al. 2003, Goodson et al. 2003, Boedeltje et al. 2004, Gurnell et al. 2007) and many R14 hydrological models have been developed for rivers (e.g. Andersen et al. 2001, Shabalova et R15 al. 2003, Yang et al. 2004, Nauta et al. 2005). It will be valuable to develop models such as R16 the dispersal model presented in chapter 6, applicable to river systems. If such models will be R17 developed, it will be necessary to implement details on river hydraulics, as hydrochorous seed R18 dispersal in rivers is water current driven, in contrast to hydrochorous dispersal in stagnant R19 water bodies. Goodson et al. (2003) conclude from their field experiments that sedimentation R20 and seed deposition in river floodplains are closely related processes. Therefore, a next step R21 in research on hydrochory in river systems could be to adapt hydrological transport models, in R22 which sedimentation can be simulated, in such a way that they are able to model hydrochorous seed dispersal and deposition in river systems, and to use these models to answer questions R23 related to the effect of hydrochorous dispersal on metapopulation dynamics and vegetation R24 composition of riparian communities. R25 Although we investigated hydrochory in a (chapter 3), the focus in this thesis R26 was mostly on landscapes with slow flowing to stagnant ditches. In chapter 5 of this thesis, R27 we presented a species distribution model, comprising of a habitat suitability module and R28 a dispersal module, for optimising spatial planning of restoration measures. The results of R29 the model simulations show that species distribution models lacking such dispersal modules R30 would considerably overestimate restoration prospects. However, simulation of dispersal in R31 this linked approach (chapter 5) was based on simple assumptions. In chapter 6, an innovative, R32 more realistic, process based spatially explicit wind- and water dispersal model for slow flowing R33 to stagnant waters was presented. In the future, the latter type of models could be linked to R34 habitat suitability or metapopulation models, to realistically predict restoration prospects and R35 metapopulation viability. Before such dispersal models can be used for predictions of absolute dispersal distances, model parameterisation and validation should be improved. Improvement R36 of parameterisation and validation of such models could be realised by, for instance, flume R37 experiments in which water and wind speed (using a fan with different wind speed settings) R38 R39 176 | Chapter 7

R1 and direction can be varied. In this way, the mechanisms playing a role in hydrochorous R2 dispersal in slow flowing waters can be studied in more detail under controlled conditions and R3 for different types of seeds. Furthermore, it is important that experiments on larger temporal R4 and spatial scale, similar to those used in the model simulations, are executed. Results of the R5 sensitivity analyses in chapter 6 show that it is especially important to pay attention to proper R6 parameterisation of the deposition process of hydrochorous seeds. To know whether seeds or seed mimics are deposited permanently, and if not where they end up after remobilisation, it R7 is necessary to be able to recognise and track them individually. In animal ecology, GPS-based R8 radiotelemetry techniques are widely used to track the movement of animals. Technological R9 innovation causes an ongoing decrease in size of the devices (Cagnacci et al.), and water proof R10 GPS devices are nowadays available (Schofield et al. 2007). In future, technological innovation R11 might enable the use of such techniques to track small seed mimics in surface water. R12 R13 A different approach for validation of hydrochorous dispersal models over a larger scale R14 could for instance be realised with the use of population genetic techniques. By genotyping R15 individuals from different populations in a metapopulation using (chloroplast) microsatellite R16 markers, assessments of gene flow between the populations can be made (e.g. Ouborg et R17 al. 1999, Garcia et al. 2007, Barluenga et al. 2011) and related to gene flow predicted with a R18 dispersal model. However, when using such a technique to validate dispersal models, it should R19 be considered that genetic variation within and between populations is not only the result of gene flow but also of germination and establishment. R20

R21 R22 R23 R24 R25 R26 R27 R28 R29 R30 R31 R32 R33 R34 R35 R36 R37 R38 R39 Synthesis | 177

reFerenCes R1 R2 Andersen, J., J. C. Refsgaard, and K. H. Jensen. 2001. Distributed hydrological modelling of the R3 Senegal River Basin - Model construction and validation. Journal of Hydrology 247:200- 214. R4 Andersson, E., C. Nilsson, and M. E. Johansson. 2000. Plant dispersal in boreal rivers and its relation R5 to the diversity of riparian flora. Journal of Biogeography 27:1095-1106. Barluenga, M., F. Austerlitz, J. A. Elzinga, S. Teixeira, J. Goudet, and G. Bernasconi. 2011. Fine-scale R6 spatial genetic structure and gene dispersal in Silene latifolia. Heredity 106:13-24. R7 Bedford, B. L., M. R. Walbridge, and A. Aldous. 1999. Patterns in nutrient availability and plant R8 diversity of temperate North American wetlands. Ecology 80:2151-2169. Blomqvist, M. M., W. L. M. Tamis, J. P. Bakker, and E. van der Meijden. 2006. Seed and (micro) R9 site limitation in ditch banks: Germination, establishment and survival under different R10 management regimes. Journal for Nature Conservation 14:16-33. Boedeltje, G., J. P. Bakker, R. M. Bekker, J. M. Van Groenendael, and M. Soesbergen. 2003. Plant R11 dispersal in a lowland stream in relation to occurrence and three specific life-history traits R12 of the species in the species pool. Journal of Ecology 91:855-866. R13 Boedeltje, G., J. P. Bakker, A. Ten Brinke, J. M. Van Groenendael, and M. Soesbergen. 2004. Dispersal phenology of hydrochorous plants in relation to discharge, seed release time and buoyancy R14 of seeds: the flood pulse concept supported. 92:786-796. R15 Boeye, D., V. Van Haesebroeck, B. Verhagen, B. Delbaere, M. Hens, and R. F. Verheyen. 1996. A local rich fen fed by calcareous seepage from an artificial river water infiltration system. R16 Vegetatio 126:51-58. R17 Bunce, R. G. H., and C. J. Hallam. 1993. The ecological significance of linear features in agricultural R18 landscapes in Britain. Pages 11-19 in R. G. H. Bunce, L. Ryskowski, and M. G. Paoletti, editors. Landscape Ecology and Agroecosystems. Lewis Publishers, Boca Raton. R19 Cagnacci, F., L. Boitani, R. A. Powell, and M. S. Boyce. Challenges and opportunities of using GPS- R20 based location data in animal ecology. Philosophical transactions of the Royal Society of London. Series B, Biological sciences 365:2155. R21 Cain, M. L., B. G. Milligan, and A. E. Strand. 2000. Long-distance seed dispersal in plant populations. R22 American journal of Botany 87:1217-1227. R23 Chormanski, J., T. Okruszko, S. Ignar, O. Batelaan, K. T. Rebel, and M. J. Wassen. 2011. Flood mapping with remote sensing and hydrochemistry: A new method to distinguish the origin of flood R24 water during floods. Ecological Engineering 37:1334-1349. R25 Ehrlen, J., Z. Munzbergova, M. Diekmann, and O. Eriksson. 2006. Long-term assessment of seed limitation in plants: results from an 11-year experiment. Journal of Ecology 94:1224-1232. R26 Fridriksson, S. 1975. Surtsey: evolution of life on a volcanic island. Butterworths, London, UK. R27 Garcia, C., P. Jordano, and J. A. Godoy. 2007. Contemporary pollen and seed dispersal in a Prunus R28 mahaleb population: Patterns in distance and direction. Molecular Ecology 16:1947-1955. Goodson, J. M., A. M. Gurnell, P. G. Angold, and I. P. Morrissey. 2003. Evidence for hydrochory and R29 the deposition of viable seeds within winter flow-deposited sediments: The River Dove, R30 Derbyshire, UK. River Research and Applications 19:317-334. Grashof-Bokdam, C., and H. Meeuwsen. 2005. Biodiversity: Maintenance and restoration by assessing R31 green blue veining. Biodiversiteit in agrarisch gebied: Behoud en herstel door sturing in R32 groenblauwe dooradering 2:93-101. R33 Grashof-Bokdam, C. J., and F. Van Langevelde. 2005. Green veining: Landscape determinants of biodiversity in European agricultural landscapes. Landscape Ecology 20:417-439. R34 Griffith, A. B., and I. N. Forseth. 2002. Primary and secondary seed dispersal of a rare, tidal wetland R35 annual, Aeschynomene virginica. 22:696-704. Grime, J. P. 1979. Plant Strategies and Vegetation Processes. John Wiley & Sons, Chichester. R36 Gurnell, A., J. Goodson, K. Thompson, N. Clifford, and P. Armitage. 2007. The river-bed: a dynamic R37 store for plant propagules? Earth Surface Processes and Landforms 32:1257-1272. R38 R39 178 | Chapter 7

He, T., S. L. Krauss, B. B. Lamont, B. P. Miller, and N. J. Enright. 2004. Long-distance seed dispersal in R1 a metapopulation of Banksia hookeriana inferred from a population allocation analysis of R2 amplified fragment length polymorphism data. Molecular Ecology 13:1099-1109. R3 Higgins, S. I., and M. L. Cain. 2002. Spatially realistic plant metapopulation models and the colonization-competition trade-off. Journal of Ecology 90:616-626. R4 Higgins, S. I., R. Nathan, and M. L. Cain. 2003. Are long-distance dispersal events in plants usually R5 caused by nonstandard means of dispersal? Ecology 84:1945-1956. Hooftman, D. A. P., R. C. Billeter, B. Schmid, and M. Diemer. 2004. Genetic effects of habitat R6 fragmentation on common species of Swiss fen meadows. Conservation Biology 18:1041- R7 1051. R8 Hooftman, D. A. P., and M. Diemer. 2002. Effects of small habitat size and isolation on the population structure of common wetland species. Plant Biology 4:720-728. R9 Howe, F., and J. Smallwood. 1982. Ecology of seed dispersal. Annual review of ecology and R10 systematics. Volume 13:201-228. Hu, L. J., K. Uchiyama, H. L. Shen, and Y. Ide. 2010. Multiple-scaled spatial genetic structures of R11 Fraxinus mandshurica over a riparian-mountain landscape in Northeast China. Conservation R12 Genetics 11:77-87. R13 Imbert, E., and F. Lefèvre. 2003. Dispersal and gene flow of Populus nigra (Salicaceae) along a dynamic river system. Journal of Ecology 91:447-456. R14 Jansson, R., C. Nilsson, and B. Renofalt. 2000. Fragmentation of riparian floras in rivers with multiple R15 dams. Ecology 81:899-903. Jansson, R., U. Zinko, D. M. Merritt, and C. Nilsson. 2005. Hydrochory increases riparian plant species R16 richness: a comparison between a free-flowing and a regulated river. Journal of Ecology R17 93:1094-1103. R18 Leng, X., C. J. M. Musters, and G. R. de Snoo. 2009. Restoration of plant diversity on ditch banks: Seed and site limitation in response to agri-environment schemes. Biological Conservation R19 142:1340-1349. R20 Leng, X., C. J. M. Musters, and G. R. De Snoo. 2010. Spatial variation in ditch bank plant species composition at the regional level: The role of environment and dispersal. Journal of R21 Vegetation Science 21:868-875. R22 Levine, J. M. 2001. Local interactions, dispersal, and native and exotic plant diversity along a R23 California stream. Oikos 95:397-408. Lord, L. A., and T. D. Lee. 2001. Interactions of local and regional processes: Species richness in R24 tussock sedge communities. Ecology 82:313-318. R25 McKenna, M. F., and G. Houle. 2000. Under-saturated distribution of Floerkea proserpinacoides Willd. (Limnanthaceae) at the northern limit of its distribution. Ecoscience 7:466-473. R26 Merritt, D. M., C. Nilsson, and R. Jansson. 2010. Consequences of propagule dispersal and river R27 fragmentation for riparian plant community diversity and turnover. Ecological Monographs R28 80:609-626. Middleton, B. 2000. Hydrochory, seed banks, and regeneration dynamics along the landscape R29 boundaries of a forested wetland. Plant Ecology 146:169-184. R30 Ministerie van Economische Zaken, Landbouw en Innovatie. 2011. Natura 2000. url: http://www. rijksoverheid.nl/onderwerpen/natuur/natura-2000. Consulted at 09-09-2011. R31 Myers, J. A., and K. E. Harms. 2009. Seed arrival, ecological filters, and plant species richness: A meta- R32 analysis. Ecology Letters 12:1250-1260. R33 Nathan, R. 2006. Long-distance dispersal of plants. Science 313:786-788. Nauta, A. B., J. Bielecka, and E. P. Querner. 2005. Hydrological model of the Lower Biebrza Basin. R34 Using the model as a management tool., Alterra, Wageningen. R35 Neff, K. P., and A. H. Baldwin. 2005. Seed dispersal into wetlands: Techniques and results for a restored tidal freshwater marsh. Wetlands 25:392-404. R36 Nilsson, C., R. L. Brown, R. Jansson, and D. M. Merritt. 2010. The role of hydrochory in structuring R37 riparian and Wetland vegetation. Biological Reviews 85:837-858. R38 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. R39 Synthesis | 179

Ouborg, N. J., P. Vergeer, and C. Mix. 2006. The rough edges of the conservation genetics paradigm for plants. Journal of Ecology 94:1233-1248. R1 Pollux, B. J. A., A. Luteijn, J. M. Van Groenendael, and N. J. Ouborg. 2009. Gene flow and genetic R2 structure of the aquatic macrophyte Sparganium emersum in a linear unidirectional river. R3 Freshwater Biology 54:64-76. Primack, R. B., and S. L. Miao. 1992. Dispersal can limit local plant distribution. Conservation Biology R4 6:513-519. R5 Pueyo, Y., and C. L. Alados. 2007. Effects of fragmentation, abiotic factors and land use on vegetation recovery in a semi-arid Mediterranean area. Basic and Applied Ecology 8:158-170. R6 Sarneel, J. M. 2010. Colonisation processes in riparian fen vegetation. Utrecht University, Utrecht. R7 Schneider, R. L., and R. R. Sharitz. 1988. Hydrochory And Regeneration In A Bald Cypress Water R8 Tupelo Swamp Forest. Ecology 69:1055-1063. Schofield, G., C. M. Bishop, G. MacLean, P. Brown, M. Baker, K. A. Katselidis, P. Dimopoulos, J. D. R9 Pantis, and G. C. Hays. 2007. Novel GPS tracking of sea turtles as a tool for conservation R10 management. Journal of Experimental Marine Biology and Ecology 347:58-68. Shabalova, M. V., W. P. A. van Deursen, and T. A. Buishand. 2003. Assessing future discharge of the R11 river Rhine using regional climate model integrations and a hydrological model. Climate R12 Research 23:233-246. R13 Tackenberg, O. 2003. Modeling long-distance dispersal of plant diaspores by wind. Ecological Monographs 73:173-189. R14 Valko, O., P. Török, B. Tóthmérész, and G. Matus. Restoration Potential in Seed Banks of Acidic R15 Fen and Dry-Mesophilous Meadows: Can Restoration Be Based on Local Seed Banks? Restoration Ecology 19:9-15. R16 van den Broek, T., R. van Diggelen, and R. Bobbink. 2005. Variation in seed buoyancy of species R17 in wetland ecosystems with different flooding dynamics. Journal of Vegetation Science R18 16:579-586. Verheyen, K., and M. Hermy. 2001. The relative importance of dispersal limitation of vascular plants R19 in secondary forest succession in Muizen Forest, Belgium. Journal of Ecology 89:829-840. R20 Wassen, M. J., R. Van Diggelen, L. Wolejko, and J. T. A. Verhoeven. 1996. A comparison of fens in natural and artificial landscapes. Vegetatio 126:5-26. R21 Xiong, S., M. E. Johansson, F. M. R. Hughes, A. Hayes, K. S. Richards, and C. Nilsson. 2003. Interactive R22 effects of soil moisture, vegetation canopy, plant litter and seed addition on plant diversity R23 in a wetland community. Journal of Ecology 91:976-986. Yang, D., T. Koike, and H. Tanizawa. 2004. Application of a distributed hydrological model and R24 weather radar observations for flood management in the upper Tone River of Japan. R25 Hydrological Processes 18:3119-3132. R26 R27 R28 R29 R30 R31 R32 R33 R34 R35 R36 R37 R38 R39 R1 R2 R3 R4 Abstract R5 R6 R7 R8 R9 R10 R11 R12 R13 R14 R15 R16 R17 R18 R19 R20 R21 R22 R23 R24 R25 R26 R27 R28 R29 R30 R31 R32 R33 R34 R35 R36 R37 R38 R39 Abstract 182 | Abstract

R1 R2 R3 R4 R5 R6 R7 R8 R9 R10 R11 R12 R13 R14 R15 R16 R17 R18 R19 R20 R21 R22 R23 R24 R25 R26 R27 R28 R29 R30 R31 R32 R33 R34 R35 R36 R37 R38 R39 Abstract | 183

Conversion of natural ecosystems into agricultural land or urban areas has led to habitat R1 loss and degradation for many plant species. Additionally, land use change typically results R2 in a fragmented distribution of the remaining habitat. Habitat fragmentation is defined as R3 a process during which a large expanse of habitat is transformed into a number of smaller R4 patches of smaller area, isolated from each other by a matrix of habitats unlike the original. R5 Habitat fragmentation leads to a decreased total habitat area and generally to a decreased R6 local population size of plant species. Furthermore, habitat fragmentation generally results R7 in an increased edge:area ratio (the edge-effect) and in increased isolation between local R8 populations. As a result, fragmented populations may suffer from reduced genetic variation, R9 increased influence of demographic and environmental stochasticity, and increased influence of the ‘hostile’ matrix (chapter 1). All these processes may increase the extinction risk of local R10 populations of plant species. R11 Population isolation is one of the processes of habitat fragmentation that lead to, amongst R12 others, reduced genetic variation. To ensure connectivity between fragmented local R13 populations, effective seed dispersal is of vital importance. Knowledge on dispersal capacity R14 of plant species in a range of fragmented habitats differing in size and spatial arrangement R15 of patches and quality of the matrix is thus important for understanding (meta)population R16 dynamics and for spatial optimisation of restoration measures. Only few studies have R17 addressed plant seed dispersal in relation to landscape patterns. R18 Plant seeds can be dispersed by different vectors: wind (anemochory), surface water R19 (hydrochory), animals (zoochory), humans (anthropochory), or by the plant itself (autochory). R20 Some plant seeds are specifically adapted to one type of dispersal, whereas others may be R21 dispersed by several vectors. R22 Although many studies have shown the importance of hydrochory for plant population and community patterns, and thus for biodiversity, the precise mechanisms determining R23 hydrochorous dispersal capacity of plant seeds in different landscapes are underexposed in R24 the literature. R25 The aim of this thesis is to shed light on the contribution of hydrochory to seed dispersal in R26 freshwater wetlands impacted by human activity and on the effect of landscape configuration R27 and -characteristics of the water body on realised dispersal distances. R28 R29 To check whether habitat fragmentation may indeed negatively influence species distribution R30 in freshwater wetlands, in chapter 2, the relative effect of isolation, habitat size and habitat R31 edge compared to the effect of habitat quality on plant occurrence was investigated for 6 R32 characteristic herbaceous fen plant species. For this purpose, large datasets on habitat quality R33 and species distribution in a Dutch semi-natural fen area were used. R34 For all but one species, besides abiotic variables one or more variables related to fragmentation R35 were shown to be important in determining species occurrence. For Carex lasiocarpa, isolation was the most important factor limiting species distribution, while for Juncus subnodulosus R36 and Menyanthes trifoliata, isolation was the second most important factor. The effect of R37 habitat size differed among species and an increasing edge had a negative effect on the R38 R39 184 | Abstract

R1 occurrence of Carex lasiocarpa and Pedicularis palustris. My results clearly show that even if R2 abiotic conditions are suitable for certain species, isolation of habitat patches and an increased R3 habitat edge caused by habitat fragmentation affect negatively the viability of characteristic R4 fen plant species. Therefore, it is important not only to improve habitat quality but also R5 to consider spatial characteristics of the habitat of target species when deciding on plant R6 conservation strategies in intensively used landscapes, such as fen areas in Western Europe and North America. R7 Now that it became clear that habitat fragmentation, and specifically habitat isolation, does R8 negatively influence occurrence of fen plant species, I focussed on seed dispersal of this type R9 of species. Fen species occur in Holocene parts of the Netherlands, for example in remnant R10 fen nature reserves or along the banks of drainage ditches in landscapes dominated by R11 agriculture. In Pleistocene parts of the Netherlands, similar fen vegetation can be found in R12 river floodplains. R13 In chapter 3, I aimed at determining how species traits and abiotic factors influence the extent of R14 hydrochorous dispersal into and out of river floodplains. For this purpose, hydrochorous seeds R15 were captured before and after a small and isolated floodplain of a Dutch river throughout R16 a year. Composition and numbers of trapped seeds before and after the floodplain were R17 related to species traits, location of the species in the floodplain, elevation and river water R18 level. The results showed that the floodplain functioned both as a seed source and sink. High R19 levels of river water seemed to promote seed transport into or out of the floodplain. Seeds of river bank species occurred significantly more often in the river water than expected. Species R20 that were more often found in the nets downstream of the floodplain than in the upstream R21 nets, indicating that they were dispersing from the floodplain effectively, had significantly R22 higher seed production, taller stature and higher seed buoyancy, but lower site elevation than R23 species with a net inflow of seeds into the floodplain. R24 These results show that inundation, and therefore more natural river water management, R25 is a prerequisite for seed transport to and from a floodplain. The restoration of target R26 floodplain vegetation may be successful for common species that produce many seeds and R27 grow in proximity to the river. Consequently, it is expected that the probability of restoring R28 vegetation types that occur further away from the river, such as wet grasslands, is lower than R29 the restoration of taller growing tall sedge or reedland floodplain communities. R30 To gain more insight into the mechanisms by which hydrochorous seeds are transported in R31 drainage ditches in landscapes dominated by agriculture, the effects of the velocity of wind R32 and water on the rate of transport of seeds of three wetland species (Carex pseudocyperus, Iris pseudacorus and Sparganium erectum) were investigated using an experimental set-up R33 (chapter 4). I hypothesised that in addition to an indirect effect of wind shear stress on the R34 movement of floating seeds -via surface flow-, wind would drive hydrochorous seed movement R35 directly as well. If so, the effect of wind should have a greater effect on the movement of R36 seeds which protrude from the water than for those which float lower in the water. The R37 results showed that, unlike in rivers, seed transport in ditches was indeed determined by R38 wind speed and direction, enabling multidirectional seed dispersal. In accordance with my R39 Abstract | 185

hypothesis, the effect of wind speed on the rate of transport of floating seeds was greater R1 for Sparganium erectum seeds, for which a greater ratio of their volume protrudes from the R2 water, than for Carex pseudocyperus and Iris pseudacorus seeds. No relation between water R3 flow at mid-depth in the ditches and seed transport was found. R4 Furthermore, in release and retrace experiments with painted C. pseudocyperus seeds, a R5 number of factors potentially determining the probability of seed deposition were investigated R6 (chapter 4). The principal factors that determined seed deposition were aquatic plant cover, R7 ditch slope and indentations in the ditch bank. Seeds changed direction if the wind direction R8 changed, or if there was a bend in the ditch. The final deposition distances were positively R9 related to mean net wind speed. Mean transport distance after 2 days varied between 34 and 451 m. R10 I concluded that in slow flowing waters wind is a more important driver for hydrochorous seed R11 transport than the flow of water. This sheds a new light on hydrochory, and has important R12 consequences for the management of otherwise fragmented wetland remnants. R13 Chapters 3 and 4 showed that seeds of (semi-)terrestrial species disperse via surface water and R14 that hydrochorous dispersal distances may exceed several hundreds of meters. Considering R15 these results, and the importance of plant seed dispersal in fragmented freshwater wetlands R16 for species distribution (chapter 2), it might be valuable to add a hydrochorous (and an R17 anemochorous) seed dispersal component to habitat suitability models. In chapter 5, a simple R18 linked habitat suitability-seed dispersal model (hydrochory and anemochory) was developed R19 to investigate the importance of adding seed dispersal models to habitat suitability models R20 for optimising spatial planning of restoration strategies. This model predicts potential species R21 distribution as a function of current species distribution, species-specific dispersal traits, R22 dispersal infrastructure and habitat configuration. Using the linked model, we could compare the effectiveness of different hydrological fen R23 restoration strategies. Moreover, we showed that stand-alone habitat suitability models, R24 which assume unlimited dispersal, may considerably overestimate restoration prospects. R25 We conclude that linked habitat suitability-dispersal models can provide useful insights into R26 spatially differentiated potentials and constraints of nature restoration measures aiming at R27 the sustainable conservation of endangered plant populations whose habitats have been R28 deteriorated due to detrimental effects of land and water management on abiotic conditions. R29 In chapter 5, only a simplified anemochory and hydrochory module was used. However, to gain R30 more detailed insight into dispersal mechanisms and connectivity, a more complex, process R31 based dispersal model would be needed. In chapter 6, a more elaborated, innovative, coupled R32 anemochory-hydrochory model was developed. The aim of this chapter was to investigate the R33 relative contribution of wind dispersal and dispersal via surface water to total seed dispersal R34 distances in agricultural landscapes with slow flowing ditches. Furthermore, the model was R35 used to investigate to what extent drainage ditches are effective corridors for dispersal of riparian species and how the dispersal success is related to properties of the landscape and R36 ditches and to species traits. Different model scenarios, representing different landscape R37 configurations and ditch states were simulated for a riparian species typically adapted to R38 R39 186 | Abstract

R1 wind dispersal (Pragmithes australis) and a riparian species adapted to hydrochorous dispersal R2 (Carex pseudocyperus). The simulations show that within local scale (2 x 2 km) water dispersal R3 distances for typical wind dispersers are similar to those of typical water dispersers. 90th R4 percentile water dispersal distances were between 100 m and >1000 m, depending on ditch Samenvatting R5 roughness and obstructions, after wind and water dispersal. Wind dispersal distance was nil th R6 for the typical water disperser, whereas 90 percentile wind dispersal distance was only 28 m for the typical wind disperser. Density or direction of the ditch network did not seem to R7 influence simulated water dispersal distances substantially, whereas roughness of the ditch R8 and obstructions in the ditch strongly limited seed dispersal distances. Our study suggests that R9 in a modern agricultural landscape with a network of ditches, water is an important dispersal R10 vector for both typical wind and typical water dispersers. From a biodiversity restoration R11 perspective, it is important that riparian populations and suitable sites are connected by R12 a network of surface water and that obstructions in this network are avoided as much as R13 possible. R14 Summarising, my results suggest that habitat fragmentation per se indeed has negative R15 effects on species occurrence in freshwater wetlands, and that - even for wetland species R16 specifically adapted to wind dispersal - water dispersal is an effective dispersal mode. The R17 importance of hydrochorous dispersal for connectivity within metapopulations of riparian R18 species in a fragmented landscape may be greater than the importance of wind dispersal, R19 which is almost exclusively local. Therefore, improving the infrastructure of linear aquatic systems for hydrochrorous seeds could increase metapopulation viability of riparian species, R20 wetland biodiversity, and the success of wetland restoration projects. R21 R22 The results of this thesis provide several implications for ecosystem conservation and R23 restoration. It is advisable to create new nature reserves in the proximity of existing ones, R24 to enable gene flow and colonisation between and from occupied patches. The optimal R25 spatial location of new patches or the viability of metapopulations could be assessed using R26 habitat suitability models incorporating realistic dispersal models of both anemochorous and R27 hydrochorous (and ideally also zoochorous) dispersal (chapter 5 and 6). Instead of considering R28 Euclidian distance only, the focus should be on how to increase connectivity by diminishing R29 roughness and adapt the configuration of dispersal corridors, on dispersal characteristics of R30 target species, and on location of existing populations, apart from local abiotic conditions. R31 When aiming at restoring rich-fen vegetation, it could be recommended to eliminate enough R32 ditches to restore groundwater flux towards the area and thereby restore abiotic conditions (chapter 5), but keep few shallow ditches in tact along the area, to enable seeds to enter the R33 water to disperse to other habitat patches. When aiming for optimal connectivity between R34 nature areas, or between sub populations of riparian plant species at ditch banks, it will be R35 advisable to replace culverts in ditches by, for instance, bridges where possible, so that seeds R36 can disperse unhindered through the dispersal corridors. R37 The results in this thesis that show the importance of ditches and rivers as dispersal corridors R38 for characteristic riparian wetland species underline the importance of networks, such as the R39 national ecological network (EHS) and green-blue veining, for biodiversity. Samenvatting 188 | Samenvatting

R1 R2 R3 R4 R5 R6 R7 R8 R9 R10 R11 R12 R13 R14 R15 R16 R17 R18 R19 R20 R21 R22 R23 R24 R25 R26 R27 R28 R29 R30 R31 R32 R33 R34 R35 R36 R37 R38 R39 Samenvatting | 189

De mens heeft door zijn handelen een grote impact op de natuur. De biodiversiteit is dramatisch R1 afgenomen sinds het begin van de industriële revolutie. Verandering in landgebruik wordt R2 gezien als de belangrijkste oorzaak van dit biodiversiteitverlies. Voor veel plantensoorten R3 heeft de omvorming van natuurlijke ecosystemen tot landbouwgebied of stedelijk gebied R4 geleid tot verlies van hun habitat, of tot de achteruitgang van de kwaliteit daarvan. Daarnaast R5 resulteert verandering in landgebruik over het algemeen in een versnipperde verdeling R6 van het resterende natuurlijke habitat; habitatfragmentatie. Habitatfragmentatie wordt R7 gedefinieerd als een proces gedurende welke een groot oppervlak aan natuurlijk habitat R8 wordt getransformeerd tot een aantal kleinere habitatgebieden, die geïsoleerd van elkaar R9 zijn door gebied dat ongelijk is aan het oorspronkelijke habitat (de zogenaamde ‘matrix’). Habitatfragmentatie leidt tot een afname van het totale oppervlak aan natuurlijk habitat, en R10 over het algemeen tot een afname in de grootte van de resterende lokale plantenpopulaties. R11 Tevens leidt habitatfragmentatie meestal tot een relatieve vergroting van de lengte van de R12 ‘rand’ van het habitat ten opzichte van de oppervlakte (het zogenaamde ‘edge-effect’), en tot R13 een toename van isolatie van de lokale populaties. Als gevolg hiervan kunnen gefragmenteerde R14 populaties onder andere te lijden hebben onder een achteruitgang van genetische diversiteit, R15 verhoging van de negatieve invloed van demografische- of milieustochasticiteit en een R16 verhoging van de (negatieve) invloed van de omgeving op de lokale populatie. Deze processen R17 kunnen de uitsterfkans van lokale plantenpopulaties vergroten (zie hoofdstuk 1). R18 Zoals hierboven gezegd is de toename van isolatie van lokale populaties een van de oorzaken R19 van de achteruitgang van genetische diversiteit. Connectiviteit tussen lokale populaties, en R20 daarmee dus ook een effectieve zaadverspreiding (dispersie), is van vitaal belang om dit tegen R21 te gaan. Derhalve is het verwerven van kennis over de dispersiecapaciteit van plantensoorten R22 in een reeks van gefragmenteerde habitats, die variëren in grootte en ruimtelijke configuratie van de patches, noodzakelijk voor het begrijpen van metapopulatiedynamica van R23 gefragmenteerde plantenpopulaties en voor ruimtelijke optimalisatie van herstelmaatregelen. R24 Tot nu hebben echter weinig studies zich gericht op de dispersie van plantenzaden in relatie R25 tot ruimtelijke patronen in het landschap. R26 Plantenzaden kunnen verspreid worden via verschillende vectoren; wind (anemochory), R27 oppervlakte water (hydrochory), dieren (zoochory), de mens (anthropochory), en door de R28 plant zelf (autochory). Sommige soorten zijn specifiek aangepast aan een bepaald type R29 dispersie, terwijl anderen kunnen verspreiden via verscheidene vectoren. R30 Ondanks het feit dat veel studies reeds het belang van hydrochory voor plantenpopulaties R31 en voor de samenstelling van plantengemeenschappen, en dus voor biodiversiteit, hebben R32 aangetoond zijn de precieze mechanismen die de dispersiecapaciteit van plantenzaden via R33 hydrochory in verschillende landschappen bepalen onderbelicht in de wetenschappelijke R34 literatuur. R35 Het doel van dit proefschrift is om duidelijkheid te scheppen over het aandeel van hydrochory in de dispersie van plantenzaden in zoetwaterwetlands beïnvloed door menselijke activiteit R36 en over het effect van landschapsconfiguratie en karakteristieken van waterlichamen op R37 gerealiseerde dispersie afstanden. R38 R39 190 | Samenvatting

R1 Om te controleren of habitatfragmentatie ook in zoetwaterwetlands inderdaad een negatief R2 effect heeft op het voorkomen van plantensoorten, onderzocht ik in hoofdstuk 2 voor zes R3 karakteristieke laagveenplantensoorten het relatieve effect van isolatie, patch grootte, en R4 ‘habitat-edge’, ten opzichte van het effect van habitatkwaliteit op het voorkomen van deze R5 plantensoorten. Voor dit doel is er gebruik gemaakt van grote datasets met informatie over R6 habitatkwaliteit en soortenverspreiding in een semi-natuurlijk laagveengebied in Nederland (De Vechtstreek). R7 Voor op een na alle soorten waren naast de abiotische factoren een of meer variabelen R8 gerelateerd aan habitatfragmentatie van significant belang in het verklaren van aan- of R9 afwezigheid van de soort. Voor Carex lasiocarpa (Draadzegge) was isolatie de belangrijkste R10 en voor Juncus subnodulosus (Paddenrus) en Menyanthes trifoliata (Waterdrieblad) de op een R11 na belangrijkste beperkende variabele voor de verspreiding van de soort. Een relatief grotere R12 rand van de patch ten opzichte van de oppervlakte had een negatief effect op het voorkomen R13 van Carex lasiocarpa en Pedicularis palustris (Moeraskartelblad). Het effect van patchgrootte R14 varieerde per soort. De resultaten tonen duidelijk aan dat isolatie van de habitatpatch en een R15 relatief grote rand van de patch, veroorzaakt door habitatfragmentatie, een negatief effect R16 hebben op het voorkomen van karakteristieke laagveensoorten, ook al zijn de abiotische R17 condities ter plaatse gunstig voor de betreffende soorten. Derhalve is het van groot belang R18 om niet alleen de habitatkwaliteit te verbeteren maar ook de ruimtelijke configuratie van R19 het habitat van doelsoorten in ogenschouw te nemen wanneer men maatregelen neemt met betrekking tot natuurbehoud in sterk door de mens beïnvloede landschappen, zoals R20 laagveengebieden in West-Europa en Noord-Amerika. R21 Nu het duidelijk is dat habitatfragmentatie, en in het bijzonder isolatie, inderdaad een R22 negatief effect heeft op het voorkomen van laagveenplantensoorten, focus ik me in de R23 vervolghoofdstukken van dit proefschrift op dispersie van plantenzaden van soorten R24 voorkomend in zoetwaterwetlands. Laagveensoorten komen voor in de Holocene delen van R25 Nederland, zoals bijvoorbeeld in overblijfselen van laagveennatuurgebieden of oevers van R26 sloten in landschappen die worden gedomineerd door landbouw. In de Pleistocene delen R27 van Nederland kan soortgelijke vegetatie worden gevonden in uiterwaarden van rivieren of R28 beken. R29 In hoofdstuk 3 stelde ik mezelf tot doel om te bepalen hoe soorteigenschappen en abiotische R30 factoren de mate van dispersie via rivierwater naar en vanuit een uiterwaard beïnvloeden. R31 Voor dit doel werden gedurende een jaar periodiek zaden bemonsterd die in de rivier R32 dreven direct stroomopwaarts en -afwaarts van een klein geïsoleerd uiterwaardengebied (De Kappersbult) van een Nederlandse kleine rivier (De Drentsche Aa). De soortensamenstelling R33 en aantallen van de gevangen zaden voor en achter het uiterwaardengebied werd R34 gerelateerd aan soorteigenschappen, groeilocatie van de soort in de uiterwaard, hoogte R35 van de groeiplaats en rivierwaterstand. De resultaten van dit onderzoek tonen aan dat de R36 uiterwaard kan functioneren als ‘source’ en als ‘sink’ van zaden. Hoge rivierwaterstanden R37 leken de uitwisseling van zaden tussen uiterwaard en rivier te bevorderen. Zaden van soorten R38 die op de rivieroever groeien hadden een significant grotere kans om in het rivierwater R39 Samenvatting | 191

terecht te komen dan verwacht. Plantensoorten waarvan de zaden vaker benedenstrooms R1 dan bovenstrooms aangetroffen werden – en dus effectief vanuit de uiterwaard verspreiden R2 - hadden een significant hogere zaadproductie in het gebied, een groter formaat, een hoger R3 zaaddrijfvermogen en lager gesitueerde groeilocatie dan soorten waarvan de zaden juist R4 netto de uiterwaard binnenstroomden. R5 Deze resultaten tonen aan dat overstroming van uiterwaarden, en derhalve een meer R6 natuurlijk rivierbeheer, een voorwaarde is voor transport van zaden van en naar een R7 uiterwaard via de rivier. Na herstel van de abiotiek, kan spontaan herstel van doelvegetatie R8 van een uiterwaardengebied optreden voor soorten die veel zaden produceren en in de R9 directe nabijheid van de rivier groeien. Het is te verwachten dat het herstel van vegetatietypen die verder van de rivier gelegen zijn, zoals natte of vochtige (schraal)graslanden, aanzienlijk R10 minder kansrijk is dan het herstel van langs de oever groeiende vegetatietypen van R11 bijvoorbeeld de Riet-klasse, met hoge grassen en grote zeggen. R12 Om meer inzicht te verkrijgen in de mechanismen die bepalen hoe hydrochore plantenzaden R13 worden getransporteerd via sloten in door landbouw gedomineerde landschappen zijn de R14 effecten van wind en watersnelheid op de transportsnelheid van zaden van drie wetland R15 soorten (Carex pseudocyperus; Hoge cyperzegge, Iris pseudacorus; Gele lis, en Sparganium R16 erectum; Grote egelskop) onderzocht met behulp van een veldexperiment (hoofdstuk 4). De R17 hypothese was dat naast een indirect effect van ‘wind shear stress’ (de beweging van het R18 wateroppervlak door de wind) op de beweging van drijvende zaden, wind ook een direct R19 effect heeft op het transport van drijvende zaden. Indien dit inderdaad het geval is, dan zou R20 wind een groter effect moeten hebben op zaden die met een groter deel van hun volume R21 boven het water uitsteken, dan op zaden die relatief dieper in het water liggen. R22 De resultaten van de veldexperimenten toonden aan dat, in tegenstelling tot in rivieren, zaadtransport in sloten inderdaad bepaald wordt door windrichting en -snelheid. Dit heeft R23 tot gevolg dat zaden die via sloten verplaatsen in staat zijn om in meerdere richtingen te R24 dispergeren en niet alleen in de stroomrichting. In overeenstemming met de hypothese was R25 het effect van windsnelheid op de transportsnelheid van de drijvende zaden van S. erectum R26 groter dan op die van C. pseudocyperus en I. pseudacorus. Voor zaden van S. erectum steekt R27 een groter deel van hun volume boven het wateroppervlak uit dan bij de andere twee R28 soorten. Er werd geen relatie tussen waterstroomsnelheid halverwege de diepte van de sloot R29 en zaadtransportsnelheid gevonden. R30 Verder werden er experimenten uitgevoerd waarbij geverfde C. pseudocyperus-zaden R31 werden in de sloot losgelaten en na drie uur, een dag en twee dagen teruggezocht. Het R32 doel van deze experimenten was om te bepalen welke factoren de depositie (vastlegging) R33 van zaden beïnvloeden (hoofdstuk 4). De belangrijkste factoren die de depositie bepaalden R34 waren de mate van plantenbedekking in het water langs de oever van de sloot, de helling R35 van de oever, en aanwezigheid van inkepingen in de oever. De via water transporterende zaden veranderden van richting indien de windrichting veranderde of bij een bocht in de R36 sloot. De dispersieafstand was positief gerelateerd aan gemiddelde netto windsnelheid. De R37 gemiddelde afstand waarop zaden werden vastgelegd varieerde tussen 34 en 451 meter na R38 2 dagen. R39 192 | Samenvatting

R1 Uit de hierboven beschreven experimenten kan geconcludeerd worden dat in (zeer) R2 langzaam stromende wateren, zoals sloten, wind een belangrijkere bepalende factor voor R3 verspreiding van zaden via water is dan waterstroom. Dit werpt nieuw licht op het beheer van R4 gefragmenteerde wetlandrestanten, en heeft hier belangrijke consequenties voor. R5 Hoofdstuk 3 en 4 lieten zien dat zaden van (semi-)terrestrische plantensoorten via R6 oppervlaktewater kunnen dispergeren, en dat er op deze manier afstanden van enkele honderden meters of meer overbrugd kunnen worden. Gezien deze resultaten, en het belang R7 van zaadverspreiding voor gefragmenteerde zoetwaterwetlands (hoofdstuk 2), zou het R8 waardevol kunnen zijn om een hydrochory- (en anemochory-)component toe te voegen aan R9 habitatmodellen. R10 In hoofdstuk 5 werd een simpel gekoppeld habitat-dispersie model gepresenteerd dat R11 ontwikkeld werd om het belang van de toevoeging van een dispersie-component aan een R12 habitatmodel voor de optimalisatie van ruimtelijke planning van herstelmaatregelen te R13 onderzoeken. Dit model voorspelt potentiële soortenverspreiding als een functie van de R14 huidige soortenverspreiding, soortspecifieke dispersie-eigenschappen, dispersie-infrastructuur R15 en habitatconfiguratie. Gebruikmakende van dit gekoppelde model kon de effectiviteit R16 van verschillende hydrologische herstelmaatregelen vergeleken worden. Tevens laten de R17 resultaten zien dat habitatmodellen die een dispersie module ontberen, en derhalve uitgaan R18 van een ongelimiteerde dispersie, de kansen voor natuurherstel aanzienlijk overschatten. Er R19 wordt geconcludeerd dat gekoppelde habitat-dispersie modellen bruikbaar inzicht verlenen in mogelijkheden en beperkingen van natuurherstelmaatregelen, wanneer men tot doel R20 heeft om het behoud te waarborgen van bedreigde plantensoorten, voor welke het habitat R21 achteruit is gegaan als gevolg van schadelijke effecten van land- en waterbeheer op de R22 abiotische condities ter plaatse. R23 In hoofdstuk 5 werd gebruik gemaakt van simpele hydrochory- en anemochorymodules. Om R24 echter meer inzicht te verkrijgen in dispersiemechanismen en in connectiviteit, is er behoefte R25 aan een complexer, procesgestuurd, dispersiemodel. R26 In hoofdstuk 6 werd een complexer, innovatief, procesgestuurd gekoppeld anemochory- R27 hydrochorymodel gepresenteerd. Het doel van dit hoofdstuk was om de relatieve bijdrage R28 van anemochory en hydrochory aan dispersie in door landbouw gedomineerde gebieden met R29 langzaam stromende sloten te bepalen. Het model werd gebruikt om te bepalen in welke R30 mate sloten effectieve corridors zijn voor dispersie van oeversoorten, en hoe dispersiesucces is R31 gerelateerd aan de eigenschappen van het landschap en de sloten en aan soorteigenschappen. R32 Voor verschillende modelscenario’s, die elk verschillende landschapsconfiguraties en sloottoestanden representeerden, werd de dispersie gesimuleerd voor een soort die specifiek R33 is aangepast aan verspreiding via wind (Phramites australis; Riet), en een soort die specifiek is R34 aangepast aan zaadverspreiding via water (C. pseudocyperus; Hoge cyperzegge). R35 De modelsimulaties lieten zien dat op lokale schaal (2 x 2 km) dispersieafstanden die R36 zaden afleggen via water voor de typische windverspreider (Riet) gelijkwaardig zijn aan R37 die van de typische waterverspreider (Hoge cyperzegge). De gesimuleerde 90-percentiel R38 dispersieafstanden na wind- en waterverspreiding waren tussen 100 en meer dan 1000 m, R39 Samenvatting | 193

afhankelijk van de ruwheid van de sloot, obstructies in de sloot. Windverspreidingsafstand R1 was nihil voor de typische waterverspreider, terwijl windverspreiding voor de typische R2 windverspreider eveneens niet boven de 28 meter uitkwam (90-percentiel afstand). De R3 dichtheid van het slotennetwerk of de hoofdrichting van het slotennetwerk hadden R4 geen duidelijk effect op de gesimuleerde dispersieafstanden, terwijl een vermeerderde R5 weerstand van de sloten en obstructies in de sloten (duikers) een sterk limiterend effect op R6 de gesimuleerde dispersie afstand had. Deze studie suggereert dat oppervlaktewater een R7 belangrijke dispersievector is voor zowel typische wind als typische waterverspreiders die R8 langs slootkanten groeien in landschappen waarin moderne landbouw met een netwerk van R9 sloten dominant aanwezig is. Vanuit het perspectief van biodiversiteitsherstel is het derhalve van belang dat bestaande oeverpopulaties en geschikte gebieden waar nog geen populatie R10 aanwezig is verbonden zijn via een netwerk van oppervlaktewater, en dat er zo min mogelijk R11 obstructies in dit netwerk aanwezig zijn. R12 Samenvattend suggereren mijn resultaten dat habitatfragmentatie inderdaad een negatief R13 effect heeft op het voorkomen van soorten in zoetwaterwetlands, en dat waterverspreiding R14 – zelfs voor wetlandsoorten die specifiek aangepast zijn aan de verspreiding via wind – R15 een effectief verspreidingsmechanisme is. De resultaten suggereren dat het belang van R16 hydrochore verspreiding voor de connectiviteit binnen metapopulaties van oeversoorten in R17 gefragmenteerde landschappen groter is dan het belang van windverspreiding, wat vrijwel R18 uitsluitend op een lokale schaal plaatsvindt. Het verbeteren van de infrastructuur van lineaire R19 aquatische ecosystemen kan derhalve bijdragen aan een verhoogde levensvatbaarheid R20 van metapopulaties van oeverplantensoorten, aan een verhoging van de biodiversiteit in R21 zoetwaterwetlands en aan het succes van natuurherstelprojecten voor zoetwaterwetlands. R22

De resultaten van dit proefschrift geven implicaties voor het behoud en herstel van het R23 onderzochte ecosysteem. Zo is het raadzaam om nieuwe natuurgebieden te creëren in de R24 nabijheid van bestaande populaties van doelsoorten, om genenuitwisseling tussen populaties R25 en kolonisatie van nieuwe habitatpatches mogelijk te maken. De optimale locatie van nieuwe R26 natuurgebieden, evenals de levensvatbaarheid van metapopulaties, kan worden bepaald R27 met behulp van habitatmodellen waarin realistische dispersiemodellen geïncorporeerd zijn R28 (hoofdstuk 5 en 6). In plaats van het enkel beschouwen van de afstand tussen patches in R29 vogelvlucht of via corridors, zou men met name moeten focussen op de manier waarop R30 connectiviteit verbeterd kan worden door het terugdringen van de ruwheid van dispersie R31 corridors en op soorteigenschappen, en daarnaast eveneens op de abiotische condities. R32 Wanneer het doel is om soortenrijke laagproductieve laagveenvegetatie te herstellen, R33 dan is het aan te bevelen om genoeg sloten te elimineren om de grondwaterflux naar het R34 doelgebied te herstellen, en daarmee de geschikte abiotische condities te herstellen. Men R35 dient echter een aantal ondiepe sloten langs het doelgebied intact te laten om zo dispersie van zaden van doelsoorten via water van en naar het te herstellen gebied mogelijk te maken R36 (hoofdstuk 5). Wanneer het doel is om een optimale connectiviteit te bewerkstelligen tussen R37 natuurgebieden of tussen lokale populaties van oeverplantensoorten op slootkanten, dan is R38 R39 194 | Samenvatting

R1 het aan te bevelen om daar waar mogelijk duikers te vervangen door bijvoorbeeld bruggen, R2 zodat zaden ongehinderd kunnen dispergeren door de sloten. R3 R4 De resultaten van dit proefschrift tonen het belang van sloten en rivieren aan als Dankwoord R5 dispersiecorridors voor karakteristieke oeverplantensoorten van wetlands, en onderschrijven R6 daarmee het belang van netwerken als de Ecologische Hoofdstructuur (EHS) en het groen-blauwe netwerk, voor het behoud en herstel van biodiversiteit in waardevolle R7 zoetwaterwetlands. R8 R9 R10 R11 R12 R13 R14 R15 R16 R17 R18 R19 R20 R21 R22 R23 R24 R25 R26 R27 R28 R29 R30 R31 R32 R33 R34 R35 R36 R37 R38 R39 Dankwoord 196 | Dankwoord

R1 R2 R3 R4 R5 R6 R7 R8 R9 R10 R11 R12 R13 R14 R15 R16 R17 R18 R19 R20 R21 R22 R23 R24 R25 R26 R27 R28 R29 R30 R31 R32 R33 R34 R35 R36 R37 R38 R39 Dankwoord | 197

Dit proefschrift zou niet tot stand gekomen zijn zonder de hulp van een aantal mensen die R1 ik hieronder hartelijk wil bedanken voor hun inzet, prettige samenwerking, of gewoon hun R2 gezellige aanwezigheid. R3 R4 Ten eerste ben ik natuurlijk mijn promotoren en co-promotor veel dank verschuldigd. Om R5 te beginnen mijn eerste promotor, Martin. Martin, ik vond het geweldig je als promotor en R6 baas te hebben. Door je informele manier van leidinggeven en goede sociale vaardigheden R7 is het erg prettig om met je samen te werken. We leerden elkaar in de loop van de tijd goed R8 kennen, en konden alles tegen elkaar zeggen. Regelmatig kreeg ik van je te horen ‘dat is R9 nou weer typisch Hester’, als ik me bijvoorbeeld weer eens ergens (in jouw ogen onnodig) druk over maakte. Stiekem moest ik daar dan erg om lachen. Je bent oprecht begaan met R10 je medewerkers en PhD’s, en ik wil je enorm bedanken voor je steun en hulp gedurende het R11 hele traject. R12 R13 Jos, mijn tweede promotor, jij was wat meer op afstand betrokken bij het proces. Ik waardeer R14 het dat je destijds hebt duidelijk gemaakt dat je wat meer betrokken wilde worden bij het R15 proefschrift, en vond het fijn en nuttig om inhoudelijke input van je te krijgen. Pita, mijn co- R16 promotor, ook jij bent op sommige momenten wat meer op de achtergrond geraakt, maar R17 aan het einde van het traject weer actiever betrokken geweest bij het schrijfproces. Ik heb R18 het altijd gezellig gevonden om met je te kletsen, al dan niet over het onderwerp van mijn R19 proefschrift. Zo was het bijvoorbeeld ook erg leuk om samen ‘de Indonesiërs’ te begeleiden R20 op excursie. De afgelopen jaren zijn turbulent geweest voor jouw departement; ik hoop dat R21 jullie nu in rustiger vaarwater komen en wens je veel succes. R22

Niet alleen mijn promotoren en co-promotor waren inhoudelijk betrokken bij mijn R23 proefschrift, ook mensen van buiten mijn begeleidingsteam hebben er hun aandeel in gehad. R24 Derek, je enthousiasme om mee te denken en je hulpvaardigheid waren belachelijk groot; R25 enorm bedankt! Desondanks heb jij waarschijnlijk nog wel het meeste van alle collega’s mijn R26 proefschrift-frustraties over je heen gekregen, maar daar leek je je gelukkig weinig van aan te R27 trekken. Ik vond het leuk om samen te puzzelen aan mijn model en met je samen te werken. R28 Je bent erg precies, en net als ik geneigd om (soms te) veel aandacht aan details te besteden, R29 maar gelukkig bracht Martin evenwicht in het proces door zijn goede kijk op grote lijnen. R30 Merel, ik vond het leuk om met iemand van buiten mijn onderzoeksgroep te kunnen sparren R31 over dispersie. Bedankt voor je betrokkenheid en het voor mij runnen van jouw model! Verder R32 bedankt aan Rudy voor het ter beschikking stellen van gegevens en het meedenken over het R33 betreffende hoofdstuk, en ook aan alle andere medeauteurs. R34 R35 Naast mijn begeleiders en medeauteurs wil ik natuurlijk ook al mijn collega’s van Milieu- natuurwetenschappen bedanken. Het was gezellig om bij jullie op de gang te zitten, te R36 overleggen en discussiëren, kletsen, lunchen en borrelen. In het bijzonder mijn (voormalige) R37 mede-AIO’s Maarten, Brian, Remko, Frank, Yuki, Hugo, Arnoud [ik luister ;-)], Greetje, R38 R39 198 | Dankwoord

R1 Verena, Sonia, Marieke, Jasper en Roland bedankt voor de gezellige tijd! In de tijd die ik op R2 de elfde verdieping van het lelijke Van Unnik-gebouw heb doorgebracht heb ik aardig wat R3 kamergenoten versleten. De eerste, en tevens degene met wie ik het langst een kamer heb R4 gedeeld was Arnaut. Arnaut, wij waren niet alleen kamergenoten maar ook projectgenoten, R5 waardoor we gelukkig wel aardig van elkaar begrepen wat we aan het doen waren, ook al R6 ben jij hydroloog en ik ecoloog. Het was leuk om zowel over onze werkzaamheden als over niet-werkgerelateerde dingen te kunnen kletsen, en zelfs nog een gezamenlijk hoofdstuk te R7 schrijven; ik vond het erg gezellig een kamer met je te delen. R8 R9 Toen Arnaut promoveerde en de hoek van het gebouw waar mijn eerste kamer lag een beetje R10 leeg begon te lopen verhuisde ik naar een meer centraal gelegen kamer. Daarbij kreeg ik R11 ook een nieuwe kamergenote; mijn paranimf Roos. Roos, we hebben lief en leed met elkaar R12 gedeeld in die gezamenlijke tijd op kamer 11.09. Ik denk dat we recordhouder van het R13 gebouw zijn wat betreft liters thee die we er doorheen joegen per dag. Terugkijkend vraag R14 ik me af hoe het mogelijk is dat we nog aan werken toekwamen, maar dat kwam er toch ook R15 nog van. Je bent een dierbare vriendin geworden, en ik hoop je nog vaak te blijven zien! In R16 the mean time, the third desk in the room was alternately occupied by Anna, Angy and Mara. R17 Girls, I enjoyed your company very much. What you three have in common is a very cheerful R18 personality; is was very enjoyable having you around, especially in the last, difficult, phase of R19 writing a dissertation! I hope to accompany you to a lot more concerts and dinners! R20 Tijdens het werken en theedrinken door was er natuurlijk ook wel eens wat beweging en frisse R21 lucht nodig. Daarom liep ik, tot vermaak van sommige collega’s, regelmatig een vlinderroute. R22 Gelukkig begreep Jerry deze rare biologentic wel, en telde hij gezellig mee op deze route. R23 Jerry, het was leuk om met je te ‘bomen’ over vlinders en fotograferen. Ik wacht nog steeds in R24 spanning af op een foto van het Koevinkje ;-) (en dan natuurlijk wel gefotografeerd op onze R25 route!). Hopelijk vind je iemand, of heb je al iemand gevonden, om de vlindertellingen mee R26 voort te zetten. R27 R28 Mijn modelleerwerkzaamheden had ik nooit kunnen uitvoeren zonder de behulpzaamheid R29 van Rob en Arno van het FAD, die vakkundig en met veel geduld mijn pc omgebouwd hebben R30 tot hij aan mijn wensen voldeed. Bedankt mannen! Ook Maarten van de vierde verdieping R31 wil ik bedanken voor zijn behulpzaamheid en zijn enthousiasme voor GIS, dat hij op mij R32 overgebracht heeft. R33 Als er voor iemand geldt dat dit proefschrift er niet geweest zou zijn zonder hen, dan geldt R34 dat wel voor mijn studenten Deborah, Yun, Saskia en Florus, en student-assistent Dagmar. R35 Bedankt voor jullie volharding in het veld, in weer en wind; van brandende zon tot en met R36 sneeuwstorm. Ik vond het leuk jullie te begeleiden. ‘Boswachter Bert’, veel dank voor het ter R37 beschikking stellen van je gebieden en boot voor de veldexperimenten. R38 R39 Dankwoord | 199

Naast het werken door was er gelukkig ook af en toe nog tijd voor vrije tijd. Die bracht ik R1 het liefst door op de rug van een paard, met Mark of met vrienden. Ymke, bedankt dat ik R2 alweer een aantal jaar op die gekke pony van je mag rijden; zonder die afleiding had ik het R3 promoveren vast niet volgehouden. Ik zou die lieve Juul, en jou natuurlijk, niet meer willen R4 missen! Ook dank aan al mijn vrienden uit Wageningen en Lelystad, in het bijzonder paranimf R5 Maaike, voor hun belangstelling, steun en gezelligheid. Pap, mam, en Minko, ik heb het altijd R6 erg leuk gevonden dat jullie duidelijk geïnteresseerd en trots waren, bedankt daarvoor! R7 Ook bedankt aan mijn schoonouders Ton en Trix, en aan Ingrid, voor hun belangstelling en R8 gastvrijheid. R9

Mark, voor jou natuurlijk de laatste alinea van dit dankwoord. Je bent nu al bijna 15 (!) jaar R10 een constante factor in mijn leven, en kent me door en door. Bedankt voor de fijne tijd die we R11 samen hebben, je oneindige geduld met mijn proefschriftperikelen en je vertrouwen in me. R12 R13 R14 R15 R16 R17 R18 R19 R20 R21 R22 R23 R24 R25 R26 R27 R28 R29 R30 R31 R32 R33 R34 R35 R36 R37 R38 R39 Curriculum Vitae Curriculum Vitae 202 | Curriculum Vitae

R1 R2 R3 R4 R5 R6 R7 R8 R9 R10 R11 R12 R13 R14 R15 R16 R17 R18 R19 R20 R21 R22 R23 R24 R25 R26 R27 R28 R29 R30 R31 R32 R33 R34 R35 R36 R37 R38 R39 Curriculum Vitae | 203

Hester Soomers was born in Heerlen on the 5th of August 1977. After moving ‘north’, at the age R1 of 6, she went through primary school and high school (VWO) in Lelystad. In 1995 she moved R2 to Wageningen to study Biology at Wageningen University. During this study she specialized in R3 population biology, especially animal ecology and animal behaviour (ethology). Her first two R4 theses considered the social behaviour and time-budgets of (semi-)wild horses, and involved R5 in total approximately eight months of fieldwork in the Netherlands and Mongolia. Both R6 research projects were performed under supervision of dr. Paul Koene of the department of R7 Ethology. Afterwards, she was struck down by Pfeiffer’s disease for several years, but continued R8 studying in 2003. The third thesis was performed at the Dutch Butterfly Conservation, under R9 supervision of dr. Henk de Vries and dr. Frank van Langevelde (Wageningen University). This study involved the development of a spatial model which she used to simulate the viability R10 of a population of a rare butterfly species [the Large Copper Lycaena( dispar batava)] in R11 the Netherlands for several conservation scenario’s. In 2004, she graduated, and continued R12 working on the model as a volunteer for the Dutch Butterfly Conservation until she started her R13 PhD at Environmental Sciences of Utrecht University in 2005. The PhD research was supervised R14 by Prof. dr. Martin Wassen, Prof. dr. Jos Verhoeven, and Dr. Pita Verweij. The results of this R15 research are presented in this dissertation. From October 2011 to April 2012 Hester worked as R16 an ecologist at Taken Adviseurs and Ingenieurs in Bilthoven. In May 2012, she started working R17 as an ecologist at Natuurmonumenten in Rotterdam. R18 R19 R20 R21 R22 R23 R24 R25 R26 R27 R28 R29 R30 R31 R32 R33 R34 R35 R36 R37 R38 R39