Long-Distance Seed Dispersal

46 downloads 0 Views 442KB Size Report
Darwin (1859) and has recently received renewed attention (Munoz et al. 2004, de Queiroz. 55 .... parameterized models for range expansion by ant dispersal suggest that since the last glacial. 131 ...... CAB International, Wallingford, UK. 697.
Chapter 6: Long-Distance Seed Dispersal

1 2

Frank M. Schurr1,2*, Orr Spiegel3, Ofer Steinitz3, Ana Trakhtenbrot3, Asaf Tsoar3 and Ran

3

Nathan3

4 5

1

6

Institute of Biochemistry and Biology

7

University of Potsdam

8

Maulbeerallee 2

9

14469 Potsdam, Germany

Plant Ecology and Conservation Biology

10 11

2

12

Institut des Sciences de l’Evolution de Montpellier, UMR-CNRS 5554

13

Université Montpellier II

14

34095 Montpellier cedex 5, France

Equipe Génétique et Environnement

15 16

3

17

Department of Evolution, Systematics and Ecology

18

Alexander Silberman Institute for Life Sciences

19

The Hebrew University of Jerusalem

20

Edmond J. Safra Campus at Givat Ram

21

Jerusalem 91904, Israel

The Movement Ecology Laboratory

22 23

*Corresponding author: Frank Schurr ([email protected])

1

24

Introduction

25

In most plant species, the vast majority of seeds is dispersed over short distances that rarely

26

exceed a few dozen metres, and only very few seeds travel over long distances (Willson

27

1993). While the long-distance dispersal of seeds (LDD) is thus typically rare, it has

28

disproportionately large effects on the long-term and large-scale dynamics of plants.

29

In this chapter, we first highlight the importance of LDD for various aspects of plant biology,

30

discuss problems with quantifying LDD and advocate a new framework for LDD research

31

that may help to overcome some of these problems. We then present six generalizations about

32

LDD mechanisms that can be derived using this framework (Nathan et al. submitted), and

33

address two fundamental questions about LDD: can we identify all relevant LDD vectors and

34

can plant traits influence LDD? We end by discussing the implications of the framework and

35

the generalizations for our understanding of LDD evolution and our ability to forecast the

36

large-scale dynamics of plants. An overview of the key terms relevant for the study of LDD is

37

given in Table 6.1.

38 39

The importance of long-distance seed dispersal

40

Because most (terrestrial) plants are mobile only as seeds, the spatial dynamics of plants is

41

largely driven by seed dispersal (Nathan & Muller-Landau 2000). Short-distance seed

42

dispersal shapes the local dynamics of plant populations and communities and provides a

43

template for subsequent local processes such as predation, competition and mating (Nathan &

44

Muller-Landau 2000). The long-distance dispersal of seeds (LDD), on the contrary, is of

45

primary importance for the large-scale dynamics of plant species and communities. For a

46

plant species to persist in a fragmented landscape, local extinction from occupied habitat

47

fragments has to be balanced by the colonization of unoccupied fragments, which requires

2

48

LDD to these fragments (e.g. Higgins & Cain 2002, Levin et al. 2003). LDD is also critical

49

for the rates at which transgenes spread in plants that are perennial and/or form feral

50

populations (Williams & Davis 2005, Kuparinen & Schurr 2007), for the velocity at which

51

plant species can migrate in response to climate change (Clark et al. 1998, Higgins et al.

52

2003a), and for the speed at which invasive plants expand their range (Hastings et al. 2005).

53

Due to its importance for range expansion, LDD is an important determinant of

54

biogeographical and macroevolutionary dynamics, a fact that was already recognized by

55

Darwin (1859) and has recently received renewed attention (Munoz et al. 2004, de Queiroz

56

2005, Alsos et al. 2007). Moreover, LDD is not only important for the dynamics of single

57

species, it also shapes community structure by determining migration rates between local

58

communities and the regional species pool (or metacommunity) (Hubbell 2001, Levine &

59

Murrell 2003). LDD is thus crucial both for fundamental ecological, biogeographical and

60

evolutionary processes, as well as for applied questions of biodiversity conservation under

61

environmental change (Levin et al. 2003, Levine & Murell 2003, Trakhtenbrot et al. 2005,

62

Nathan 2006).

63 64

Defining long-distance seed dispersal

65

Two approaches are commonly used to define LDD (Cain et al. 2000, Nathan 2005, Figure

66

6.1). The absolute definition considers LDD events as dispersal events that are longer than a

67

specified threshold distance (e.g., 1 km). According to the alternative proportional definition,

68

LDD events are those that exceed a certain high quantile of dispersal distance (e.g. the 99%

69

quantile as the distance exceeded by only 1% of all seeds). It is important to notice that the

70

proportional definition actually identifies extreme dispersal events rather than LDD events: in

71

a species with highly restricted seed dispersal, the 99% quantile of dispersal distance might

72

only be located at a few metres, a distance that typically will not be considered as "long"

3

73

(Figure 6.1a). On the other hand, in a species whose seeds are frequently dispersed over long

74

distances, only the most extreme dispersal events will meet the proportional definition, even

75

though somewhat less extreme dispersal distances might also be long enough to affect the

76

large-scale dynamics of this species (Figure 6.1b). For studies interested in LDD rather than

77

extreme dispersal, we thus recommend applying the absolute definition. The threshold

78

distance for the absolute definition should be chosen based on the question studied: when

79

studying the dynamics of a plant species in a fragmented landscape such a LDD threshold

80

could be the typical distance between neighbouring habitat fragments. For certain questions

81

(e.g. to predict the spread of populations, Clark et al. 2001), it can be important to quantify

82

not only the proportion of dispersal events exceeding this threshold, but to also report the

83

frequency distribution of dispersal distances above the threshold. Because this review

84

encompasses LDD processes operating at different scales, ranging from landscape (102-104

85

m), to regional (103-105 m) and biogeographical (105-107 m) scales, a single all-encompassing

86

absolute threshold for LDD cannot be defined. Instead, we identify the range of spatial scales

87

over which each LDD mechanism highlighted in this review typically operates (see below).

88 89

The challenge of studying long-distance seed dispersal

90

The importance of LDD for the spatial dynamics of plants has long been recognized (e.g.

91

Darwin 1859). Nevertheless, LDD research has long been based on the compilation of

92

anecdotal observations and has lacked a rigorous theoretical background. LDD has also rarely

93

been examined experimentally, and there are very few attempts at quantitatively synthesizing

94

observational data on LDD. This paucity in quantitative LDD research can be explained

95

through the concurrence of several circumstances: First, direct empirical observations of LDD

96

are difficult if not impossible due to the severe methodological problems encountered both in

97

tracking seeds during LDD and in finding those seeds that have been long-distance dispersed

4

98

(Cain et al 2000, Nathan et al. 2003). Second, even if LDD events can be observed, the fact

99

that these events are typically rare, severely limits our ability to empirically estimate their

100

probability of occurrence (and other properties). Third, two different approaches are

101

commonly used to define LDD (see section "Defining long-distance seed dispersal" above),

102

meaning that there is no general currency to compare different studies. Fourth, the way in

103

which seed dispersal has been studied traditionally makes it very difficult to detect and

104

quantify LDD mechanisms. In our opinion, a detailed examination of this fourth point is the

105

basis for the formulation of a new quantitative framework for LDD research. In the following

106

section we thus describe the traditional approaches to dispersal research and examine why

107

these approaches fail for LDD.

108

The traditional framework of seed dispersal research

109

The traditional framework of seed dispersal research can be described as "seed-centred". It

110

focuses on a set of seeds (e.g. all seeds produced by a mother plant) and asks by which

111

mechanisms and over which distances these seeds are dispersed. This seed-centred framework

112

typically assumes that a plant species has a "standard" dispersal vector which disperses most

113

seeds and whose identity can be inferred from specific morphological attributes of the

114

dispersal unit (e.g. the fact that seeds are equipped with an elaiosome suggests that ants are

115

their standard dispersal vector) (van der Pijl 1982, Higgins et al. 2003c). Because seed

116

morphology often seems to suggest a particular dispersal vector, the concept of haplochory,

117

dispersal mediated by a single (standard) dispersal vector, has long dominated plant dispersal

118

research (van der Pijl 1982). To describe this notion of a strong association between a plant

119

species and its standard dispersal vector, Berg (1983) coined the term 'specialized dispersal'.

120

If haplochory and specialized dispersal prevail, a plant species' potential for LDD should be

121

predictable from the identity of its standard dispersal vector. For example, it has commonly

5

122

been assumed that LDD requires morphological adaptations for seed dispersal by birds, water

123

or wind, but is extremely unlikely for species exhibiting morphological adaptations for seed

124

dispersal by ants, ballistic dispersal (explosive discharge of seeds) or for species lacking any

125

obvious dispersal adaptation (Ridley 1930, van der Pijl 1982, Berg 1983).

126

While the seed-centred framework with its focus on standard dispersal vectors might produce

127

reliable results for short-distance dispersal (which depends on the majority of dispersal

128

events), it frequently fails for LDD (Box 6.1, Figure 6.2). The case of the woodland herb

129

Asarum canadense provides a good example for such failure (Cain et al. 1998). A. canadense

130

has been classified as ant-dispersed because its seeds bear elaiosomes. However, empirically

131

parameterized models for range expansion by ant dispersal suggest that since the last glacial

132

maximum A. canadense should only have spread by 10-11 km, whereas the species actually

133

covered hundreds of kilometres during this time (ca. 16000 years). This suggests that

134

occasional LDD by vectors other than ants dominated the postglacial expansion of this

135

species.

136

The application of the seed-centred framework to LDD is thus problematic because the

137

vectors that transport seeds over long distances are typically not those which disperse the

138

majority of seeds (Higgins et al. 2003c, Box 6.1). Consequently, a research approach focusing

139

on the majority of seeds is likely to miss those seeds that are long-distance dispersed.

140

[Insert Box 6.1 here]

141 142

A vector-centred framework for research on long-distance seed dispersal

143

Recent years have seen a shift from the traditional seed-centred framework to a vector-centred

144

framework for LDD research (e.g. Nathan et al. 2002, Munoz et al. 2004, Levey et al. 2005,

145

Jordano et al. 2007). This vector-centred framework focuses on a dispersal vector capable of

6

146

generating LDD and asks how many seeds this vector disperses over which distance. This

147

permits to concentrate research efforts on those processes that cause LDD rather than having

148

to study a large number of seeds of which at best a few will be long-distance dispersed.

149

The vector-centred approach is by no means new: as many other branches of ecology, this

150

field was pioneered by Charles Darwin who conducted quantitative experiments on seed

151

dispersal by ocean currents (Darwin 1859). More recently, Berg (1983) compellingly

152

suggested to shift the focus of dispersal research to polychory (dispersal by multiple vectors)

153

and generalized dispersal by nonstandard vectors. Berg's principles, which correspond well

154

with recent findings, emphasize that seeds of a given species are usually dispersed by multiple

155

vectors (Ozinga et al. 2004, Nathan 2007), including LDD vectors that tend to be different

156

from the standard vector which disperses the vast majority of seeds at short distances (Higgins

157

et al. 2003c, Box 6.1, Figure 6.2). For example, seeds with morphological adaptations for

158

dispersal by ants can be long-distance dispersed by deer (Odocoileus virginianus) (Vellend et

159

al. 2003) and emus (Dromaius novaehollandiae) (Calviño-Cancela et al. 2006), and seeds

160

with adaptations for bird dispersal can be transported by floods (Hampe et al. 2003). In the

161

dandelion (Taraxacum officinale), as the archetype of a wind-dispersed species with a pappus

162

promoting aerial transport, the vast majority (99%) of wind-dispersed seeds are predicted to

163

travel over less than 2.15 m (Soons & Ozinga 2005). Yet, the dandelion hairy seed has a half-

164

time buoyancy of 2.5 days in water (Boedeltje et al. 2003) and a relatively high potential for

165

long-term retention in furs (Tackenberg et al. 2006). This implies that hydrochory (dispersal

166

by water) and epizoochory (externally by animals) have much greater potential for LDD of

167

dandelion seeds than anemochory (dispersal by wind). However, LDD need not be caused by

168

non-standard vectors - it can also arise from irregular behaviour of the standard vector. For

169

example, strong turbulent vertical updrafts induced by specific atmospheric conditions can

170

transport seeds over much greater distances than typical winds (Nathan et al. 2002, Kuparinen

7

171

et al. 2007).

172

To identify mechanisms that are capable of generating LDD, we use a general model for

173

passive dispersal (Figure 6.3). This model (adapted from Isard & Gage 2001) describes three

174

main phases of vector-mediated dispersal (Picture 1): the initiation phase, in which the

175

dispersal vector removes seeds from the mother plant; the transport phase, in which the

176

vector transports seeds away from the source; and the termination phase, in which seeds are

177

deposited. This chain can be repeated, possibly involving different vectors in the transport of

178

a single seed (vander Wall & Longland 2004). The key parameter of the initiation phase is the

179

vector seed load (Q), the number of seeds taken by a dispersal vector in a certain time period.

180

Q depends on plant fecundity, fruiting schedule in relation to the behaviour of the vector, and

181

the loading ability of the vector. The mean vector displacement velocity (V) after seed uptake

182

is the key parameter of the transport phase. It depends primarily on the movement properties

183

of the vector, such as its travel velocity, directionality and intermittence. The key parameter of

184

the termination phase is the seed passage time (P), the duration of seed transport by the

185

vector, which depends on both seed and vector traits, and their interactions. All three key

186

parameters also depend on external factors such as landscape structure and climatic

187

conditions.

188

Displacement velocity V and seed passage time P together define the shape of the dispersal

189

distance kernel, whereas seed load Q defines how many seeds the vector disperses. General

190

considerations (Box 6.2, Figure 6.5) show that important LDD vectors are those which

191

generate a relatively high per-seed probability of LDD, because retention time (P) and/or

192

displacement velocity (V) are at least occasionally high. Such vectors can be important for

193

LDD even if they have a small seed load (Q).

194

[Insert Box 6.2 here]

8

195

Six generalizations on long-distance dispersal mechanisms

196

Using the vector-centred framework outlined above, we reviewed the seed dispersal literature

197

in search of mechanisms promoting LDD. This led us to formulate six generalizations on

198

LDD mechanisms (Nathan et al. submitted). Arranged by the spatial scale for which they are

199

most relevant (Figure 6.6) these generalizations state that LDD is more frequent in open

200

terrestrial landscapes (G1), and is likely to be caused by large animals (G2), migratory

201

animals (G3), extreme meteorological events (G4), ocean currents (G5) and human

202

transportation (G6). The first generalization (G1) thus concerns the environmental conditions

203

affecting LDD whereas the remaining generalizations (G2-G6) identify five major LDD

204

vectors. In addition to the information given in Nathan et al. submitted, we provide here

205

additional evidence, in particular for the importance of large animals (G2, Box 6.3).

206

For each generalization, we will in the following summarize the underlying mechanisms,

207

review the available evidence, and discuss the scope of the generalization in terms of the plant

208

species for which it is most relevant.

209 210

G1) Open landscapes

211

Mechanisms: Open landscapes are defined here as regions of no, sparse or very short

212

vegetation, such as grassland meadows, arid steppes and arctic tundra, as opposed to more

213

closed landscapes such as forests and woodlands. In open landscapes, LDD is facilitated by

214

the low density of obstacles to the movement of seeds and their vectors. All else being equal,

215

vectors moving through more open landscapes should thus have higher displacement velocity

216

V and longer retention time P, but not necessarily larger seed load Q.

217

Evidence: Mechanistic models strongly suggest higher LDD in more open landscapes both for

218

primary (Nathan & Katul 2005) and secondary (Schurr et al. 2005) seed dispersal by wind.

9

219

Over smooth playa surfaces (Fort & Richards 1998), on snow (Greene & Johnson 1997), and

220

in post-fire environments (Bond 1988), dispersal distances of wind-dispersed seeds are much

221

longer than in closed landscapes. A lack of vegetation also seems to promote LDD by water

222

courses: cleared channels and relatively wide rivers provide highways for the LDD of seeds

223

and vegetative fragments of many plant species (Boedeltje et al. 2003, Truscott et al. 2006).

224

Furthermore, an analysis of potential LDD vectors for 123 plant communities (mostly

225

grasslands) in the Netherlands showed a clear positive correlation between the potential for

226

LDD and light availability as an index for landscape openness (Ozinga et al. 2004).

227

Clearly, different dispersal vectors may require different degrees of landscape openness: for

228

instance, a grassland may lack obstacles for the movement of large mammalian seed

229

dispersers, but because wind speed within the grass canopy is low, it holds many obstacles to

230

the wind dispersal of herbaceous plants.

231

Scope: All plants. It should, however, be noted that LDD is not restricted to open landscapes.

232

In closed forests, highly mobile animals such as cassowaries (Casuarius casuarius) and spider

233

monkeys (Ateles paniscus) can transport seeds over long distances (Westcott et al. 2005,

234

Russo et al. 2006). Explicit analyses of how landscape openness affects the V and P of seed

235

dispersing animals are still missing.

236 237

G2) Large animals

238

Mechanisms: Animals with larger body mass tend to have larger home range, higher travel

239

velocity, larger gut capacity and longer seed retention time, compared to smaller animals

240

within the same taxonomic group (Calder 1996, Box 6.3). Larger home ranges and higher

241

travel velocities lead to higher V, and longer gut retention times imply longer P. The

242

relationship between animal body mass and Q is less clear cut: while the larger intake rate of

10

243

large animals means that they are likely to have higher Q per individual animal, smaller

244

animals may compensate by visiting fruiting plants more frequently or by having denser local

245

populations (e.g. Spiegel & Nathan 2007).

246

Evidence: The recent literature abounds with quantitative examples of large animals

247

dispersing seeds over relatively long distances (e.g. Vellend et al. 2003, Westcott et al. 2005,

248

Russo et al. 2006, Calviño-Cancela et al. 2006, Spiegel & Nathan 2007). Moreover, large

249

animals have been shown to take up seeds of a large variety of plant species, irrespective of

250

their dispersal morphology (e.g. Myers et al. 2004, Westcott et al. 2005). The few available

251

comparisons of seed dispersal by animals of different body mass suggest that larger animals

252

cause more LDD (Dennis & Westcott 2007, Jordano et al. 2007, Spiegel & Nathan 2007).

253

Furthermore, we present here a meta-analysis of data on endozoochorous dispersal by birds

254

which shows that seed dispersal distance increases with the body mass of the seed-dispersing

255

animal as predicted from allometric relationships for P and V (Figure 6.7, Box 6.3).

256

Scope: Plant species dispersed by animals, especially such dispersed endozoochorously

257

(Jordano et al. 2007). The relationship does not necessarily hold for inhomogeneous

258

taxonomic and/or dietary groups of animals. Although larger body mass is associated with

259

higher V in both epi- and endozoochorous dispersal, it is currently unclear whether the

260

allometric relationship between body mass and P found for endozoochorous dispersal (Figure

261

6.7, Box 6.3), also holds for epizoochory.

262 263

[Insert Box 6.3 here]

264 265

G3) Migratory animals

11

266

Mechanisms: Migratory animals, specifically birds and mammals, can transport seeds both

267

externally and internally. During migration, animals move relatively fast and in a directional

268

manner. They therefore have considerably higher V than equivalent non-migratory animals.

269

Furthermore, migratory animals are more likely to transport seeds across dispersal barriers.

270

Also, Q could be very high given the large number of migratory animals. However, general

271

conclusions about the magnitude of Q and P for migratory animals still seem premature.

272

Evidence: A large set of anecdotal evidence (e.g. Carlquist 1981) has been reported since

273

Darwin (1859) highlighted LDD by migratory animals. Experimental approaches have mostly

274

focused on seed viability and passage time (P) (e.g. Charalambidou & Santamaria 2002), and

275

the expected dispersal distances were then calculated assuming that V can be estimated from

276

observed migratory flight speeds (Charalambidou et al. 2003). This assumption is likely to

277

lead to gross overestimation of LDD, because it ignores delays between seed uptake and the

278

start of migratory flights as well as pauses in stopover sites (Sánchez et al. 2006). Laboratory

279

estimates of P for captive animals may fail to represent effects of pre-migratory fasting and

280

reduction of the digestive system (Clausen et al. 2002). The only available experiment

281

incorporating such effects suggests that seeds dispersed by migrating ducks have relatively

282

long P (Figuerola & Green 2005). However, despite the recent upsurge in research on this

283

topic (Clausen et al. 2002, Charalambidou & Santamaria 2002, Figuerola & Green 2002,

284

Charalambidou et al. 2003, Figuerola & Green 2005, Sánchez et al. 2006), quantitative

285

estimates of V, P and Q are still largely missing for migratory animals.

286

Scope: Plants fruiting during migration periods (Hanya 2005). Fleshy-fruited plants are likely

287

to be dispersed by migrating passerines (Jordano 1982), while waterfowl mostly disperse

288

small-seeded plants (Figuerola & Green 2002). Since studies have largely focused on aquatic

289

plants dispersed internally by waterbirds; we have almost no information on other plants, on

12

290

other groups of migrating animals, and on the importance of epizoochorous dispersal by

291

migrating animals.

292 293

G4) Extreme meteorological events

294

Mechanisms: As extreme meteorological events we denote phenomena that cause highly

295

energetic wind or water flow. Extreme meteorological events have exceptionally high V for

296

plant seeds and are also likely to exhibit relatively long P (storms/floods carry seeds for

297

longer periods than normal conditions) and high Q (storms/floods induce mass release of

298

seeds). Extreme events can also damage plant seeds, thereby lowering effective seed load (Q)

299

and reducing the establishment probability of transported seeds.

300

Evidence: Dispersal biogeographers have long considered extreme meteorological events as

301

potential LDD vectors, even at intercontinental scales (e.g. Visher 1925). Indirect evidence

302

from studies of plant recruitment or genetic analyses suggests causal relationship between

303

extreme events and LDD. For example, a high frequency of kelp (Pterygophora californica)

304

recruitment events at relatively long (up to 4 km) distances from the nearest zoospore source

305

was documented after severe winter storms (Reed et al. 1988). At far larger scales, a

306

biogeographical analysis (Munoz et al. 2004) found that the floristic similarities of moss,

307

liverwort, lichen, and pteridophyte floras for 27 locations across the Southern Hemisphere

308

conform better with how well these locations are connected by 'wind highways' than with

309

their geographic proximity. This study thus suggests that wind can be an important LDD

310

vector at the global scale. However, it does not permit to assess the relative contribution of

311

standard and extreme conditions to LDD by wind. While quantitative analyses that directly

312

link extreme events to plant LDD are critically missing, some insights can be drawn from a

313

long-term data set on various objects that have been found as fallout debris from tornadic

314

thunderstorms and could be traced back to their source location (Snow et al. 1995). Most

13

315

plant seeds will fall into Snow et al.'s (1995) "paper" category, for which the mean transport

316

distance was 111.2 km and the maximum (of 48 records) was 338 km. Even "heavy" objects

317

(>450 g) were transported over a mean distance of 31.6 km and a maximum distance of 177

318

km (n= 30 records). For obvious reasons, Snow et al. (1995) could not trace back seeds, but

319

their data set nevertheless demonstrates the ability of tornadoes to cause LDD (Nathan et al.

320

submitted).

321 322

Scope: Extreme meteorological events probably disperse plant seeds (and larger diaspores)

323

irrespective of their taxonomy and morphological classification (Nathan et al. submitted), but

324

species inhabiting regions where extreme events are relatively frequent (e.g. ocean coasts and

325

islands) are more likely to be transported by this LDD mechanism.

326 327

G5) Ocean currents

328

Mechanisms: Surface ocean currents serve as efficient LDD vectors for floating seeds and for

329

rafts transporting seeds or entire plants. Oceanic currents are relatively slow (mean V = 0.1 -

330

0.3 m s-1), but their Q is essentially limitless. Moreover, seeds with good floating capacity

331

typically have a high P of several tens of days (exceptionally extending to 19 years), and for

332

rafting, P is typically one order of magnitude longer (Thiel & Haye 2006).

333

Evidence: A plethora of floating and rafting records suggests that at macro-evolutionary

334

timescales, transoceanic LDD of terrestrial plants is a recurrent phenomenon (Darwin 1859,

335

Ridley 1930, Carlquist 1981, de Queiroz 2005, Thiel & Gutow 2005, Thiel & Haye 2006).

336

But even at ecological timescales, ocean currents can transport plant propagules of various

337

non-specialized dispersal morphologies, over tens and even hundreds of kilometres. For

338

example, of all plant species colonizing the volcanic island Surtsey (35 km from Iceland) in

14

339

the first decade after its emergence, 78% were recorded to arrive by oceanic currents,

340

although only 25% had apparent morphological adaptations for dispersal by water (Higgins et

341

al. 2003c).

342

Scope: The Surtsey example shows that seeds do not have to have obvious morphological

343

adaptations for floating to be long-distance dispersed by ocean currents (Higgins et al. 2003).

344

Moreover, a wide range of plant species has been recorded as rafting. These species mostly

345

originate from islands and coastal habitats (e.g. estuaries and saltmarshes) (Thiel & Gutow

346

2005, Thiel & Haye 2006).

347 348

G6) Human transportation

349

Mechanisms: Compared to nonhuman dispersal vectors, global trade and transportation by

350

humans have much higher V and longer P as well as higher Q. Human transportation also

351

increases the probability of seeds to dispersed over natural barriers (e.g. from one continent to

352

another).

353

Evidence: The rapid spread of agricultural crops across Eurasia after 8500 BC is evidence of

354

early human-mediated LDD (Diamond 2002, possible reference to Chapter 7). Over the last

355

centuries, massive human-mediated plant LDD is evident in the large number of plant species

356

that have become naturalized in regions outside their native range (Hodkinson & Thompson

357

1997, Mach & Lonsdale 2001). In most cases, human-mediated transport is given as the most

358

plausible explanation for the unprecedented scale and frequency of dispersal over very broad

359

barriers, leading to conclusions such as "Only humans could have transported scores of

360

western European species to Australia, Argentina, and western North America, and vice

361

versa, in less than 500 years" (Novak & Mack 2001). In certain cases, the available evidence

362

points to particular modes of human-mediated dispersal. For instance, of the 67 species of

15

363

Southern African Iridaceae that have become naturalized outside their native range, 93% are

364

known to be used in horticulture, whereas this percentage is only 30% for the entire set of

365

1036 species (van Kleunen et al. 2007). At smaller scales, the identification of human-

366

mediated dispersal mechanisms is even clearer: it has been shown that LDD of seeds is caused

367

by cars (von der Lippe & Kowarik 2007), agricultural machinery (Bullock et al. 2003), and

368

livestock herds (Manzano & Malo 2006).

369

Scope: Probably most plants, especially those closely associated with human activities (e.g.

370

crop and ornamental plants, weeds and ruderals).

371 372

A vector-based perspective on the evolution and predictability of long-

373

distance seed dispersal

374

The importance of the above generalizations for LDD evolution and our ability to understand

375

and predict the large-scale dynamics of plants will depend on the answers to two fundamental

376

questions:

377

(1) Is it possible to identify all important LDD vectors? and

378

(2) How important are plant and seed traits for LDD?

379

In the following we first address these two questions from a vector-based perspective and then

380

discuss their implications for the evolution and predictability of LDD.

381 382

Is it possible to identify all important long-distance dispersal vectors?

383

The above generalizations certainly do not provide an exhaustive list of LDD mechanisms.

384

But the potential of the vector-based framework to advance our understanding of LDD will

385

depend on whether it is possible to identify all important LDD vectors for a given plant

386

species. It seems obvious that we cannot identify all vectors that might potentially disperse

16

387

seeds over long distances. This is because there seems to be an essentially limitless number of

388

vectors that could potentially cause LDD. For instance, seeds of temperate plant species could

389

be long-distance dispersed by an escaped circus elephant, and the shock wave following a

390

major asteroid impact could cause LDD of those seeds that are not destroyed by the impact.

391

Such 'freak events' occur in such an irregular manner that it seems impossible to assess

392

whether or not they will be involved in the LDD of a given plant species. The notion that

393

LDD is driven by "chance dispersal" (Carlquist 1981) seems to suggest that these freak events

394

dominate LDD. However, as pointed out by Berg (1983), we need to distinguish two types of

395

"chance dispersal": LDD can either arise from "an unusually favourable combination of the

396

regular dispersal factors" or from "an unusual coincidence involving a dispersal factor not

397

normally operating together with the taxon in question". The key distinction between these

398

two forms is whether LDD results from the rare but essentially predictable behaviour of a

399

regular dispersal vector (such as extremely turbulent wind conditions), or whether it results

400

from the rare and unpredictable involvement of an irregular dispersal vector (such as an

401

escaped elephant).

402

The distinction between regular and irregular LDD vectors is important, because it marks the

403

boundary between what can and what cannot be predicted by a vector-based approach to LDD

404

research. We can hope to obtain a quantitative understanding of regular LDD vectors (such as

405

the ones in G2-G6), but research into the mechanisms of LDD by irregular LDD vectors

406

seems futile because of the sheer number of irregular vectors. Hence, the overall contribution

407

of regular vectors to LDD defines an upper limit of what can be explained by a vector-based

408

approach (Figure 6.8).

409

The success of the vector-based approach will thus depend on the composition of a plant's

410

LDD vector spectrum: LDD can - at least in principle - be explained if it is dominated by a

411

limited number of regular vectors (Figure 6.8 a and b). On the contrary, the vector-based

17

412

approach will fail if LDD results from the joint action of an essentially unlimited number of

413

irregular vectors (Figure 6.8 c and d). The vector-based approach thus directs our attention to

414

the shape of LDD vector spectra.

415

But what do these LDD vector spectra look like for particular plant species and what is the

416

relative contribution of regular and irregular vectors to the LDD of these species? While these

417

are clearly very difficult questions, we are starting to gather information that helps to answer

418

them. Such information comes from two studies comparing mechanistic predictions for

419

regular LDD vectors with data on large-scale plant distributions. The first study is the

420

biogeographical analysis of Munoz et al. (2004) (see "G4) Extreme meteorological events"

421

above) which suggests that wind as a regular LDD vector dominates the floristic similarity

422

between cryptogam communities across the Southern Hemisphere. A second study (Schurr et

423

al. 2007) found that the combination of mechanistic models for LDD by primary and

424

secondary wind dispersal (Tackenberg 2003, Schurr et al. 2005) explains a substantial

425

proportion of the amount to which 37 species of South African Proteaceae fill their potential

426

range. In these two cases, LDD thus seems to be dominated by one regular dispersal vector. It

427

is not surprising that the dominant vector in both studies is wind: the quantitative

428

understanding of LDD mechanisms - as a prerequisite for the comparison of mechanistic

429

predictions to large-scale patterns - is most advanced for seed dispersal by wind.

430

Further quantitative analyses that also include mechanistic predictions for other regular

431

dispersal vectors are needed to assess how frequently LDD is dominated by one or a few

432

regular vectors. The limited evidence available to date does, however, suggest that LDD is not

433

always dominated by irregular vectors and that we do not always have to consider many

434

regular vectors to adequately describe LDD. If LDD is dominated by a limited number of

435

regular vectors, the vector-based framework will help to identify these vectors, to measure

436

their Q, V and P and hence to quantify their contribution to LDD.

18

437 438

How important are plant and seed traits for long-distance dispersal?

439

Another critical question for our understanding of LDD evolution and the large-scale

440

dynamics of plants is to which extent plant and seed traits control LDD. For instance, LDD

441

can only evolve adaptively if it is to some extent controlled by phenotypic traits, and if these

442

traits have in part a genetic basis. A similar question can be asked within an ecological

443

context: do interspecific differences in LDD depend in part on species traits or is LDD

444

essentially identical for all species inhabiting a given environment with a given set of LDD

445

vectors? If LDD depends on species traits, an understanding of interspecific variation in

446

biogeographical distributions and dynamics requires the quantification of these traits. If, on

447

the other hand, species traits are unimportant for LDD, all species co-occurring in a given

448

environment with a given set of LDD vectors should have the same rates of LDD. In this case,

449

the assumption of the neutral theory of biogeography (Hubbell 2001) that all species in a

450

metacommunity have the same migration rate might be a good description of reality.

451

Moreover, if species traits are unimportant for LDD, efforts to predict range dynamics under

452

environmental change can neglect variation in these traits. Instead these efforts should then

453

focus on interspecific differences in demographic processes (population growth rates,

454

probability of establishment) and on how environmental change will affect the joint spectrum

455

of LDD vectors shared by all species. To assess the consequences of the generalizations G1-

456

G6 for LDD evolution, for the large-scale dynamics of plants, and for our ability to predict

457

these dynamics under environmental change the critical question is thus to which extent plant

458

and seed traits control LDD.

459 460

As outlined above, the traits used in traditional morphological classifications have no or little

461

effect on LDD. However, this does not necessarily mean that traits in general are irrelevant -

19

462

traits other than those traditionally considered may well control LDD by particular vectors.

463

For instance, the LDD of Trillium grandiflorum seeds does not depend on the fact that they

464

are equipped with an elaiosome, but rather on their ability to survive ingestion by white-tailed

465

deer (Vellend et al. 2003). The high importance of human-mediated dispersal at very large

466

scales (G6) makes an entirely different suite of traits relevant for LDD: for example, the

467

attractiveness of Iridaceae species to gardeners seems to strongly determine their spread at

468

global scales (van Kleunen et al. 2007). These examples show how a vector-based approach

469

can identify LDD-relevant traits which escaped the traditional seed-centred approach.

470

Moreover, they show that plant and seed traits can exert some control on LDD by at least

471

certain dispersal vectors.

472

Therefore the question does not seem to be whether phenotypic traits can in principle affect

473

LDD, but rather how much of the variation in LDD they explain. It seems likely that different

474

LDD mechanisms differ in the amount of trait control (for instance phenotypic traits might

475

have a bigger effect on LDD by large animals than on LDD by tornadoes). It also seems likely

476

that, for a given LDD mechanism, the amount of trait control decreases with dispersal

477

distance. However, to give definite answers to these questions we need a quantitative

478

understanding of multiple LDD mechanisms which currently is still lacking.

479 480

Evolution of long-distance seed dispersal

481

A necessary condition for LDD to evolve in response to natural selection is that phenotypic

482

traits of plants and seeds must exert a certain degree of control over LDD. As discussed

483

above, this condition seems to be fulfilled at least for certain plant species and certain LDD

484

vectors. Yet, in comparison to environmental effects, phenotypic traits are likely to have only

485

a small effect on the LDD probability of an individual seed. Because phenotypic traits

486

themselves are also affected by environmental variation, this could be taken to mean that the

20

487

heritability of LDD (as the proportion of LDD variation explained by additive genetic

488

variation) is even smaller. However, when quantifying the heritability of LDD one has to

489

carefully distinguish between different levels at which environmental variation acts: if we

490

assume that the relevant phenotypic traits are mostly determined by the genotype of the

491

mother plant, the heritability of LDD should not be calculated from between-seed variation in

492

LDD, but rather from between-mother plant variation in the LDD of their offspring. At the

493

mother plant level, the environmental stochasticity involved in the LDD of individual seeds

494

will to some extent average out, with environmental stochasticity becoming less important the

495

more fecund individuals are. Hence, while the importance of environmental effects for LDD

496

means that the heritability of LDD is unlikely to be very high, it also seems premature to

497

conclude that LDD inevitably has a heritability close to 0. Due to the difficulties of directly

498

observing LDD (see "The challenge of studying long-distance seed dispersal" above), it seems

499

virtually impossible to quantify its heritability from direct empirical observations (as was

500

done for short-distance dispersal by Donohue et al. 2005). Instead, it seems more promising to

501

estimate the heritability of LDD by using mechanistic models of LDD vectors that propagate

502

observed phenotypic variation in dispersal traits and represent realistic levels of

503

environmental variation.

504

The fact that LDD can have strong fitness consequences means that it is likely to be exposed

505

to strong directional selection pressures. Evolutionary models show that LDD can be selected

506

for in species that expand their geographic range (Travis & Dytham 2002), are exposed to

507

specialized pests (Muller-Landau et al. 2003), face frequent catastrophic extinction from local

508

habitat fragments (e.g. van Valen 1971, Hamilton & May 1977, Comins et al. 1980; see also

509

Ronce et al. 2000, Poethke et al. 2003) or inhabit spatially autocorrelated landscapes

510

(Hovestadt et al. 2001). Selection against LDD may occur in island situations where LDD

511

carries high costs because long-distance dispersers are likely to end up in hostile

21

512

environments (Cody & Overton 1996). In metapopulations, the costs and benefits of LDD

513

may interact so that between-population selection for colonization favours increased LDD in

514

young populations whereas within-population selection favours reduced LDD as populations

515

age (Olivieri et al. 1995).

516

The short-term evolutionary response of LDD depends on the product of its heritability and

517

the strength of selection acting on it (Lynch and Walsh 1998). As discussed in the previous

518

paragraphs, LDD is likely to have rather low heritability, but it may also be exposed to strong

519

directional selection. A priori we can thus neither exclude the possibility that LDD might

520

respond quite quickly to natural selection nor the possibility that it might hardly react at all.

521

To assess the speed at which LDD can evolve adaptively, we instead have to quantify both its

522

heritability and the selective forces acting on it.

523

New avenues for studying the evolution of multivariate dispersal phenotypes are suggested by

524

the realization that most plant species have multiple LDD vectors. Seed movement over

525

different spatial scales is exposed to different selective pressures (Ronce et al. 2001), which

526

raises the question whether LDD and short-distance dispersal can evolve independently, or

527

whether they are linked through positive or negative genetic correlations (Ronce 2007). The

528

fact that LDD vectors are often different from the standard vectors driving short-distance

529

dispersal (Higgins et al. 2003c, Box 6.1) suggests that LDD and short-distance dispersal

530

might be able to evolve independently. However, seed dispersal by different vectors can still

531

be linked through phenotypic correlations between the relevant dispersal traits. For instance,

532

for seeds of a given mass, increased pappus size both reduces the seed's terminal falling

533

velocity and increases its attachment to animal furs (Couvreur et al. 2005, Tackenberg et al.

534

2006), which should lead to a positive correlation between wind dispersal and epizoochorous

535

dispersal. Such interactions between seed dispersal by different vectors are likely to be

536

mediated by quantitative, continuously varying traits (such as pappus size). However, they

22

537

can also arise from single mutations: the domestication of wild wheat and barley might have

538

been triggered by the emergence of single-gene mutants that did not shatter their seeds and

539

thus had reduced probabilities of abiotic seed dispersal but increased human-mediated LDD

540

(Diamond 2002, possible reference to Chapter 7). Changes at individual loci might thus have

541

profound consequences for the composition of Total Dispersal Kernels (TDKs) and can

542

substantially promote LDD.

543 544

Forecasting the large-scale dynamics of plants

545

The composition of TDKs may be important for the future dynamics of plants because

546

different LDD vectors are differentially suited for the spatial response of plants to

547

environmental change. For instance, many northern temperate plants produce ripe seeds in

548

summer and autumn when birds migrate southwards (Hampe et al. 2003). For poleward

549

expansions necessary under climate change, migrating birds may thus be less relevant than

550

other vectors (such as extreme storms).

551

Different LDD mechanisms are also likely to be differentially affected by environmental

552

change: climate change is predicted to increase the frequency and intensity of hurricanes,

553

severe floods and other extreme weather conditions (Easterling et al 2000, Knutson & Tuleya

554

2004, Webster et al 2005; but see Pielke et al. 2005), which should promote LDD (G4).

555

Rising levels of atmospheric CO2 raise the fecundity of certain plant species, thereby

556

increasing the seed load (Q) of their LDD vectors and hence the number of LDD events

557

(LaDeau & Clark 2001). On the other hand, the loss or decline of large and migratory animals

558

may substantially reduce LDD of many plant species (G2, G3). Certain drivers of

559

environmental change may have complex effects on LDD: for example, habitat destruction

560

and fragmentation may reduce LDD by reducing the movement of large animals (G3, Higgins

561

et al 2003b), but they may also promote LDD by creating more open landscapes (G1). To

23

562

quantitatively assess such interacting effects of environmental change on LDD, we need a

563

mechanistic understanding of the major LDD processes.

564

A quantitative description of LDD combined with knowledge on the probability of population

565

establishment after LDD (Nathan 2006) is the basis for forecasting the large-scale dynamics

566

of plants (e.g. Clark et al. 2001). However, LDD forecasts will inevitably involve uncertainty

567

that arises from three main sources, namely from model, parameter and inherent uncertainty

568

(Higgins et al. 2003a). Model uncertainty is caused by uncertainty about the identity of

569

dominant LDD vectors and by uncertainty about the exact mechanisms by which these

570

dominant vectors move seeds. This component of uncertainty can be greatly reduced by

571

research into LDD mechanisms and by the adequate modelling of these mechanisms (Higgins

572

et al. 2003a). However, the proportion of LDD accounted for by irregular LDD vectors

573

(Figure 6.8) will limit our ability to reduce model uncertainty (see " Is it possible to identify

574

all important long-distance dispersal vectors?" above). Parameter uncertainty is the result of

575

imprecise knowledge about the key parameters (Q,V and P) for each LDD vector. For most

576

LDD vectors, we seriously lack quantitative knowledge about these key parameters (see " Six

577

generalizations on long-distance dispersal mechanisms" above) meaning that parameter

578

uncertainty can be substantially reduced through the collection of empirical data. Finally,

579

inherent uncertainty is the consequence of stochasticity in LDD processes or in the models

580

used to describe them. For vectors with high seed load (Q) and fat-tailed dispersal kernels, the

581

inherent uncertainty about the location of the furthest dispersal distance is high. This leads to

582

high uncertainty about population spread rates that cannot be reduced by through an

583

improvement of models or parameter estimates (Clark et al. 2003). Despite inevitable

584

uncertainty in LDD predictions, informative quantitative forecasts of range dynamics seem

585

possible at least for certain species: the fact that LDD mechanisms can explain interspecific

586

variation in biogeographical distributions of Proteaceae (Schurr et al. 2007, see above),

24

587

suggests that mechanistic forecasts of range dynamics are within reach (Thuiller et al. 2007).

588 589

Future directions

590

In this review we have pointed out some key questions for future research on LDD of plants:

591

(1) Can we identify all important LDD vectors for a given plant species?

592

(2) What is the shape of LDD vector spectra (Figure 6.8) for different plant species?

593

(3) Which phenotypic traits of plants and seeds affect LDD?

594

(4) How important are these phenotypic traits for LDD?

595

(5) How large are the heritability of LDD and the strength of selection acting on it, and how

596

fast can LDD evolve?

597

(6) To what extent are LDD forecasts limited by inherent uncertainty that cannot be reduced

598

through improvement of model structure and parameter estimation?

599

We believe that our ability to answer these questions can be greatly advanced through the

600

wide adoption of a vector-based framework for LDD research that defines the spatial scales of

601

interest for a particular problem, identifies the dispersal vectors operating at these scales, and

602

quantifies the key parameters Q, V and P (Figure 6.3) for these vectors.

603

To quantify how Q, V and P are determined by the interaction of environmental conditions

604

(e.g. landscape openness, G1) and phenotypic traits we need to develop and/or refine

605

mechanistic models for specific LDD vectors, notably for those mentioned in G2-G6.

606

Mechanistic models for seed dispersal by animals (G2, G3) are so far restricted mostly to

607

specific animal species. However, the allometric model presented in Box 6.3 demonstrates

608

that it is possible to formulate more generic models for seed dispersal by animals. Models for

609

LDD by abiotic vectors such as extreme meteorological conditions (G4) and ocean currents

610

(G5) can build on existing models for fluid dynamics in meteorology and oceanography.

25

611

Finally, it seems possible to construct mechanistic models for LDD by human transportation

612

(G6) that use data on traffic and commodity flows.

613

The development of mechanistic models for LDD vectors will have to go hand in hand with

614

the collection of data on relevant environmental variables, vector properties and phenotypic

615

traits. For instance, the development of large databases on functional species traits (e.g.

616

Knevel et al. 2003, Poschlod et al. 2003) should interact closely with the development of

617

LDD models to ensure that the trait databases provide parameter estimates necessary to

618

forecast large-scale dynamics under environmental change. To maximize the value of trait

619

databases for conservation, it will furthermore be important to ensure that they cover the

620

geographical regions and the species that are most affected by environmental changes.

621

The development of mechanistic models for LDD vectors will also have to be integrated with

622

the collection of relevant data at large spatial scales. Large scale data can serve both to

623

validate mechanistic LDD models and to assess whether models for regular LDD vectors

624

provide an adequate description of overall LDD (see section "Is it possible to identify all

625

important long-distance dispersal vectors?" above). Large-scale information on LDD can be

626

collected by regional-scale sampling of long-distance dispersed seeds (Shields et al. 2006),

627

through technological advances in the tracking of seeds and their LDD vectors (Wikelski et al.

628

2007) and through the combination of large-scale genetic (Jordano et al. 2007) and

629

biogeographical (Munoz et al. 2004) data with information on LDD mechanisms.

630

In conclusion, we believe that the combination of new quantitative tools in a vector-centred

631

approach holds great promise for LDD research. The pace at which this field has developed in

632

recent years lets us hope that further advances in our understanding of long-distance seed

633

dispersal are not far away.

634

26

635 636

Acknowledgements We thank Jordi Figuerola, Daniel García, Arndt Hampe, Anna Kuparinen, Wim Ozinga,

637

Ophélie Ronce, Sabrina Russo, Louis Santamaría and Steve Wagstaff for helpful comments

638

and suggestions. This study has been supported through the Israeli Science Foundation (ISF

639

474/02 and ISF-FIRST 1316/05), the International Arid Land Consortium (IALC 03R/25), the

640

Israeli Nature and Parks Protection Authority, the US National Science Foundation (IBN-

641

9981620 and DEB-0453665), the German Ministry of Education and Research (BMBF) in the

642

framework of Biota Southern Africa (FKZ 54419938), the European Union through Marie

643

Curie Transfer of Knowledge Project FEMMES (MTKD-CT 2006-042261), the Minerva

644

short-term fellowship program, the Simon and Ethel Flegg Fellowship, and the Friedrich

645

Wilhelm Bessel Research Award of the Humboldt Foundation.

27

646

References

647

Alsos, I.G., Eidesen, P.B., Ehrich, D., et al. (2007) Frequent long-distance plant colonization

648

in the changing Arctic. Science, 316, 1606-1609.

649

Berg, R.Y. (1983). Plant distribution as seen from plant dispersal: general principles and basic

650

modes of plant dispersal. In: Dispersal and Distribution Vol.,(ed K. Kubitzki)edn. pp.

651

13-36. Paul Parey, Hamburg.

652

Boedeltje, G., Bakker, J.P., Bekker, R.M., et al. (2003) Plant dispersal in a lowland stream in

653

relation to occurrence and three specific life-history traits of the species in the species

654

pool. Journal of Ecology, 91, 855-866.

655 656 657

Bond, W.J. (1988) Proteas as "tumbleseeds": wind dispersal through the air and over soil. South African Journal of Botany, 54, 455-460. Bullock, J.M., Moy, I.L., Coulson, S.J., et al. (2003) Habitat-specific dispersal: environmental

658

effects on the mechanisms and patterns of seed movement in a grassland herb

659

Rhinanthus minor. Ecography, 26, 692-704.

660 661 662 663

Cain, M.L., Damman, H., & Muir, A. (1998) Seed dispersal and the Holocene migration of woodland herbs. Ecological Monographs, 68, 325-347. Cain, M.L., Milligan, B.G., & Strand, A.E. (2000) Long-distance seed dispersal in plant populations. American Journal of Botany, 87, 1217-1227.

664

Calder, W.A. (1996) Size, function and life history, 2nd edn. Dover Publications, New York.

665

Calviño-Cancela, M., Dunn, R.R., van Etten, E.J.B., et al. (2006) Emus as non-standard seed

666

dispersers and their potential for long-distance dispersal. Ecography, 29, 632-640.

667

Carlquist, S. (1981) Chance dispersal. American Scientist, 69, 509-516.

668

Charalambidou, I. & Santamaria, L. (2002) Waterbirds as endozoochorous dispersers of

669

aquatic organisms: a review of experimental evidence. Acta Oecologica-International

670

Journal of Ecology, 23, 165-176.

28

671

Charalambidou, I., Santamaria, L., & Langevoord, O. (2003) Effect of ingestion by five avian

672

dispersers on the retention time, retrieval and germination of Ruppia maritima seeds.

673

Functional Ecology, 17, 747-753.

674 675

Clark, J.S., Fastie, C., Hurtt, G., et al. (1998) Reid's paradox of rapid plant migration: dispersal theory and interpretation of paleoecological records. BioScience, 48, 13-24.

676

Clark, J.S., Lewis, M., & Horvath, L. (2001) Invasion by extremes: population spread with

677

variation in dispersal and reproduction. American Naturalist, 157, 537-554.

678 679 680

Clark, J.S., Lewis, M., McLachlan, J.S., et al. (2003) Estimating population spread: what can we forecast and how well? Ecology, 84, 1979-1988. Clausen, P., Nolet, B.A., Fox, A.D., et al. (2002) Long-distance endozoochorous dispersal of

681

submerged macrophyte seeds by migratory waterbirds in northern Europe: a critical

682

review of possibilities and limitations. Acta Oecologica-International Journal of

683

Ecology, 23, 191-203.

684 685 686 687 688 689 690 691 692 693 694 695

Cody, M.L. & Overton, J.M. (1996) Short-term evolution of reduced dispersal in island plant populations. Journal of Ecology, 84, 53-61. Comins, H.N., Hamilton, W.D., & May, R.M. (1980) Evolutionarily stable dispersal strategies. Journal of Theoretical Biology, 82, 205-230. Couvreur, M., Verheyen, K., & Hermy, M. (2005) Experimental assessment of plant seed retention times in fur of cattle and horse. Flora, 200, 136-147. Darwin, C. (1859) The Origin of Species by Means of Natural Selection John Murray, London. de Queiroz, A. (2005) The resurrection of oceanic dispersal in historical biogeography. Trends in Ecology & Evolution, 20, 68-73. Dennis, A.J. & Westcott, D.A. (2007). Estimating dispersal kernels produced by a diverse community of vertebrates. In: Seed Dispersal: Theory and its Application in a

29

696

Changing World Vol. (eds A.J. Dennis, R.J. Green, E.W. Schupp, et al.)edn. pp. 201-

697

228. CAB International, Wallingford, UK.

698 699

Diamond, J. (2002) Evolution, consequences and future of plant and animal domestication. Nature, 418, 700-707.

700

Donohue, K., Polisetty, C. R., & Wender, N. J. (2005) Genetic basis and consequences of

701

niche construction: Plasticity-induced genetic constraints on the evolution of seed

702

dispersal in Arabidopsis thaliana. American Naturalist, 165, 537-550.

703 704 705 706 707

Easterling, D.R., Meehl, G.A., Parmesan, C., et al. (2000) Climate extremes: Observations, modeling, and impacts. Science, 289, 2068-2074. Figuerola, J. & Green, A.J. (2002) Dispersal of aquatic organisms by waterbirds: a review of past research and priorities for future studies. Freshwater Biology, 47, 483-494. Figuerola, J. & Green, A.J. (2005) Effects of premigratory fasting on the potential for long

708

distance dispersal of seeds by waterfowl: An experiment with marbled teal. Revue D

709

Ecologie-La Terre Et La Vie, 60, 283-287.

710 711 712 713

Fort, K.P. & Richards, J.H. (1998) Does seed dispersal limit initiation of primary succession in desert playas? American Journal of Botany, 85, 1722-1731. Greene, D.F. & Johnson, E.A. (1997) Secondary dispersal of tree seeds on snow. Journal of Ecology, 85, 329-340.

714

Hamilton, W.D. & May, R.M. (1977) Dispersal in stable habitats. Nature, 269, 578-581.

715

Hampe, A., Arroyo, J., Jordano, P., et al. (2003) Rangewide phylogeography of a bird-

716

dispersed Eurasian shrub: contrasting Mediterranean and temperate glacial refugia.

717

Molecular Ecology, 12, 3415-3426.

718

Hanya, G. (2005) Comparisons of dispersal success between the species fruiting prior to and

719

those at the peak of migrant frugivore abundance. Plant Ecology, 181, 167-177.

720

Hastings, A., Cuddington, K., Davies, K.F., et al. (2005) The spatial spread of invasions: new

30

721 722

developments in theory and evidence. Ecology Letters, 8, 91-101. Higgins, S.I. & Cain, M.L. (2002) Spatially realistic plant metapopulation models and the

723

colonization-competition trade-off. Journal of Ecology, 90, 616-626.

724

Higgins, S.I., Clark, J.S., Nathan, R., et al. (2003a) Forecasting plant migration rates:

725 726 727

managing uncertainty for risk assessment. Journal of Ecology, 91, 341-347. Higgins, S.I., Lavorel, S., & Revilla, E. (2003b) Estimating plant migration rates under habitat loss and fragmentation. Oikos, 101, 354-366.

728

Higgins, S.I., Nathan, R., & Cain, M.L. (2003c) Are long-distance dispersal events in plants

729

usually caused by nonstandard means of dispersal? Ecology, 84, 1945-1956.

730 731

Hodkinson, D.J. & Thompson, K. (1997) Plant dispersal: the role of man. Journal of Applied Ecology, 34, 1484-1496.

732

Holbrook, K.M. & Smith, T.B. (2000) Seed dispersal and movement patterns in two species

733

of Ceratogymna hornbills in a West African tropical lowland forest. Oecologia, 125,

734

249-257.

735

Hovestadt, T., Messner, S., & Poethke, H.J. (2001) Evolution of reduced dispersal mortality

736

and 'fat-tailed' dispersal kernels in autocorrelated landscapes. Proceedings of the Royal

737

Society of London Series B-Biological Sciences, 268, 385-391.

738 739 740 741 742 743 744 745

Hubbell, S.P. (2001) The Unified Neutral Theory of Biodiversity and Biogeography Princeton Univ. Press, Princeton, NJ. Isard, S.A. & Gage, S.H. (2001) Flow of Life in the Atmosphere Michigan State Univ Press, East Lancing. Jordano, P. (1982) Migrant birds are the main seed dispersers of blackberries Rubus ulmifolius. Oikos, 43, 149-153. Jordano, P., Garcia, C., Godoy, J.A., et al. (2007) Differential contribution of frugivores to complex seed dispersal patterns. Proceedings of the National Academy of Sciences of

31

746

the United States of America, 104, 3278-3282.

747

Knevel, I.C., Bekker, R.M., Bakker, J.P., et al. (2003) Life-history traits of the northwest

748

European flora: The LEDA database. Journal of Vegetation Science, 14, 611-614.

749

Knutson, T.R. & Tuleya, R.E. (2004) Impact of CO2-induced warming on simulated hurricane

750

intensity and precipitation: Sensitivity to the choice of climate model and convective

751

parameterization. Journal of Climate, 17, 3477-3495.

752

Kuparinen, A., Markkanen, T., Riikonen, H., et al. (2007) Modeling air-mediated dispersal of

753

spores, pollen and seeds in forested areas. Ecological Modelling, 208, 177-188.

754

Kuparinen, A. & Schurr, F.M. (2007) A flexible modelling framework linking the spatio-

755

temporal dynamics of plant genotypes and populations: Application to gene flow from

756

transgenic forests. Ecological Modelling, 202, 476-486.

757 758 759 760 761

LaDeau, S.L. & Clark, J.S. (2001) Rising CO2 levels and the fecundity of forest trees. Science, 292, 95-98. Levey, D.J., Bolker, B.M., Tewksbury, J.J., et al. (2005) Effects of landscape corridors on seed dispersal by birds. Science, 309, 146-148. Levin, S.A., Muller-Landau, H.C., Nathan, R., et al. (2003) The ecology and evolution of seed

762

dispersal: a theoretical perspective. Annual Review of Ecology Evolution and

763

Systematics, 34, 575-604.

764

Levine, J.M. & Murrell, D.J. (2003) The community-level consequences of seed dispersal

765

patterns. Annual Review of Ecology Evolution and Systematics, 34, 549-574.

766

Lynch, M. & Walsh, B. (1998) Genetics and analysis of quantitative traits. Sinauer

767 768 769 770

Associates, Inc, Sunderland, MA. Mack, R.N. & Lonsdale, W.M. (2001) Humans as global plant dispersers: getting more than we bargained for. Bioscience, 51, 95-102. Manzano, P. & Malo, J.E. (2006) Extreme long-distance seed dispersal via sheep. Frontiers in

32

771 772

Ecology and the Environment, 5, 244-248. Muller-Landau, H.C., Levin, S.A., & Keymer, J.E. (2003) Theoretical perspectives on

773

evolution of long-distance dispersal and the example of specialized pests. Ecology, 84,

774

1957-1967.

775 776 777

Muñoz, J., Felicísimo, Á.M., Cabezas, F., et al. (2004) Wind as a long-distance dispersal vehicle in the Southern Hemisphere. Science, 304, 1144-1147. Myers, J.A., Vellend, M., Gardescu, S., et al. (2004) Seed dispersal by white-tailed deer:

778

implications for long-distance dispersal, invasion, and migration of plants in eastern

779

North America. Oecologia, 139, 35-44.

780 781

Nathan, R. (2005) Long-distance dispersal research: building a network of yellow brick roads. Diversity and Distributions, 11, 125-130.

782

Nathan, R. (2006) Long-distance dispersal of plants. Science, 313, 786-788.

783

Nathan, R. (2007). Total dispersal kernels and the evaluation of diversity and similarity in

784

complex dispersal systems. In: Seed Dispersal: Theory and its Application in a

785

Changing World Vol. (eds A.J. Dennis, E.W. Schupp, R.J. Green, et al.)edn. pp. 252-

786

276. CAB International, Wallingford, UK.

787

Nathan, R. & Katul, G.G. (2005) Foliage shedding in deciduous forests lifts up long-distance

788

seed dispersal by wind. Proceedings of the National Academy of Sciences of the

789

United States of America, 102, 8251-8256.

790 791 792

Nathan, R., Katul, G.G., Horn, H.S., et al. (2002) Mechanisms of long-distance dispersal of seeds by wind. Nature, 418, 409-413. Nathan, R. & Muller-Landau, H.C. (2000) Spatial patterns of seed dispersal, their

793

determinants and consequences for recruitment. Trends in Ecology & Evolution, 15,

794

278-285.

795

Nathan, R., Perry, G., Cronin, J.T., et al. (2003) Methods for estimating long-distance

33

796 797 798 799 800 801 802 803 804 805 806 807

dispersal. Oikos, 103, 261-273. Nathan, R., Schurr, F.M., Spiegel, O., et al. (Submitted) Mechanisms of Long-Distance Seed Dispersal. Trends in Ecology and Evolution. Novak, S.J. & Mack, R.N. (2001) Tracing plant introduction and spread: genetic evidence from Bromus tectorum (Cheatgrass). Bioscience, 51, 114-122. Oliver, M. K. & Mckaye, K. R. (1982) Floating islands - a means of fish dispersal in Lake Malawi, Africa, Copeia, 4, 748-754. Olivieri, I., Michalakis, Y., & Gouyon, P.H. (1995) Metapopulation genetics and the evolution of dispersal. American Naturalist, 146, 202-228. Ozinga, W.A., Bekker, R.M., Schaminée, J.H.J., et al. (2004) Dispersal potential in plant communities depends on environmental conditions. Journal of Ecology, 92, 767-777. Pielke, R.A., Agrawala, S., Bouwer, L.M., et al. (2005) Clarifying the attribution of recent

808

disaster losses. A response to Epstein and McCarthy. Bulletin of the American

809

Meteorological Society, 86, 1481-1483.

810

Poethke, H.J., Hovestadt, T., & Mitesser, O. (2003) Local extinction and the evolution of

811

dispersal rates: Causes and correlations. American Naturalist, 161, 631-640.

812

Poschlod, P., Kleyer, M., Jackel, A.K., et al. (2003) BIOPOP - a database of plant traits and

813

Internet application for nature conservation. Folia Geobotanica, 38, 263-271.

814

Reed, D.C., Laur, D.R., & Ebeling, A.W. (1988) Variation in algal dispersal and recruitment:

815

the importance of episodic events. Ecological Monographs, 58, 321-335.

816

Ridley, H.N. (1930) The Dispersal of Plants throughout the World Reeve, Ashford.

817

Robbins, C.T. (1993) Wildlife feedign and nutrition, 2nd edn. Academic Press, San Diego.

818

Ronce, O. (2007) How does it feel to be like a rolling stone? Ten questions about dispersal

819

evolution. Annual Review of Ecology Evolution and Systematics, 38, 231-253.

820

Ronce, O., Olivieri, I., Clobert, J., et al. (2001). Perspectives on the study of dispersal

34

821

evolution. In: Dispersal Vol. (eds J. Clobert, É. Danchin, A.A. Dhondt, et al.)edn. pp.

822

341-357. Oxford University Press, Oxford, UK.

823 824 825 826 827

Ronce, O., Perret, F., & Olivieri, I. (2000) Evolutionarily stable dispersal rates do not always increase with local extinction rates. American Naturalist, 155, 485-496. Russo, S.E., Portnoy, S., & Augspurger, C.K. (2006) Incorporating animal behavior into seed dispersal models: Implications for seed shadows. Ecology, 87, 3160-3174. Sánchez, M.I., Green, A.J., & Castellanos, E.M. (2006) Internal transport of seeds by

828

migratory waders in the Odiel marshes, south-west Spain: consequences for long-

829

distance dispersal. Journal of Avian Biology, 37, 201-206.

830

Schurr, F.M., Bond, W.J., Midgley, G.F., et al. (2005) A mechanistic model for secondary

831

seed dispersal by wind and its experimental validation. Journal of Ecology, 93, 1017-

832

1028.

833

Schurr, F.M., Midgley, G.F., Rebelo, A.G., et al. (2007) Colonization and persistence ability

834

explain the extent to which plant species fill their potential range. Global Ecology and

835

Biogeography, 16, 449-459.

836

Shields, E.J., Dauer, J.T., VanGessel, M.J., et al. (2006) Horseweed (Conyza canadensis) seed

837

collected in the planetary boundary layer. Weed Science, 54, 1063-1067.

838

Snow, J.T., Wyatt, A.L., McCarthy, A.K., et al. (1995) Fallout of debris from tornadic

839

thunderstroms: a historical perspectives and two examples from VORTEX. Bulletin of

840

the American Meteorological Society, 76, 1777-1790.

841

Soons, M.B. & Ozinga, W.A. (2005) How important is long-distance seed dispersal for the

842

regional survival of plant species? Diversity and Distributions, 11, 165-172.

843

Spiegel, O. & Nathan, R. (2007) Incorporating dispersal distance into the disperser

844

effectiveness framework: frugivorous birds provide complementary dispersal to plants

845

in a patchy environment. Ecology Letters, 10, 718–728.

35

846 847 848 849 850

Sun, C., Ives, A.R., Kraeuter, H.J., et al. (1997) Effectiveness of three turacos as seed dispersers in a tropical montane forest. Oecologia, 112, 94-103. Tackenberg, O. (2003) Modeling long distance dispersal of plant diaspores by wind. Ecological Monographs, 73, 173-189. Tackenberg, O., Romermann, C., Thompson, K., et al. (2006) What does diaspore

851

morphology tell us about external animal dispersal? Evidence from standardized

852

experiments measuring seed retention on animal-coats. Basic and Applied Ecology, 7,

853

45-58.

854

Thiel, M. & Gutow, L. (2005) The ecology of rafting in the marine environment. I. The

855

floating substrata. Oceanography and Marine Biology: an Annual Review, 42, 181-

856

263.

857

Thiel, M. & Haye, P.A. (2006) The ecology of rafting in the marine environment. III.

858

Biogeographical and evolutionary consequences. Oceanography and Marine Biology -

859

an Annual Review, Vol 44, 44, 323-429.

860

Thuiller, W., Albert, C., Araujo, M.B., et al. (2007) Predicting global change impacts on plant

861

species' distributions: future challenges. Perspectives in Plant Ecology, Evolution and

862

Systematics, doi:10.1016/j.ppees.2007.09.004.

863

Trakhtenbrot, A., Nathan, R., Perry, G., et al. (2005) The importance of long-distance

864

dispersal in biodiversity conservation. Diversity and Distributions, 11, 173-181.

865

Travis, J.M.J. & Dytham, C. (2002) Dispersal evolution during invasions. Evolutionary

866

Ecology Research, 4, 1119-1129.

867

Truscott, A.M., Soulsby, C., Palmer, S.C.F., et al. (2006) The dispersal characteristics of the

868

invasive plant Mimulus guttatus and the ecological significance of increased

869

occurrence of high-flow events. Journal of Ecology, 94, 1080-1091.

870

Tucker, V.A. (1973) Bird Metabolism During Flight - Evaluation of a Theory. Journal of

36

871

Experimental Biology, 58, 689-709.

872

van der Pijl, L. (1982) Principles of Dispersal in Higher Plants, 3rd edn. Springer, Berlin.

873

Van Kleunen, M. & Johnson, S.D. (2007) South African Iridaceae with rapid and profuse

874

seedling emergence are more likely to become naturalized in other regions. Journal of

875

Ecology, 95, 674-681.

876

Van Valen, L. (1971) Group selection and the evolution of dispersal. Evolution, 25, 591-598.

877

Vander Wall, S.B. & Longland, W.S. (2004) Diplochory: are two seed dispersers better than

878 879

one? Trends in Ecology & Evolution, 19, 155-161. Vellend, M., Myers, J.A., Gardescu, S., et al. (2003) Dispersal of Trillium seeds by deer:

880

implications for long- distance migration of forest herbs. Ecology, 84, 1067-1072.

881

Visher, S.S. (1925) Tropical cyclones and the dispersal of life from island to island in the

882 883 884

Pacific. American Naturalist, 59, 70-78. Von der Lippe, M. & Kowarik, I. (2007) Long-distance dispersal of plants by vehicles as a driver of plant invasions. Conservation Biology, 21, 986-996.

885

Ward, M.J. & Paton, D.C. (2007) Predicting mistletoe seed shadow and patterns of seed rain

886

from movements of the mistletoebird, Dicaeum hirundinaceum. Austral Ecology, 32,

887

113-121.

888

Webster, P.J., Holland, G.J., Curry, J.A., et al. (2005) Changes in tropical cyclone number,

889

duration, and intensity in a warming environment. Science, 309, 1844-1846.

890

Weir, J.E.S. & Corlett, R.T. (2007) How far do birds disperse seeds in the degraded tropical

891

landscape of Hong Kong, China? Landscape Ecology, 22, 131-140.

892

Westcott, D.A., Bentrupperbäumer, J., Bradford, M.G., et al. (2005) Incorporating patterns of

893

disperser behaviour into models of seed dispersal and its effects on estimated dispersal

894

curves. Oecologia, 146, 57-67.

895

Westcott, D.A. & Graham, D.L. (2000) Patterns of movement and seed dispersal of a tropical

37

896

frugivore. Oecologia, 122, 249-257.

897

Wikelski, M., Kays, R.W., Kasdin, N.J., et al. (2007) Going wild: What a global small-animal

898

tracking system could do for experimental biologists. Journal of Experimental

899

Biology, 210, 181-186.

900 901 902 903

Williams, C.G. & Davis, B.H. (2005) Rate of transgene spread via long-distance seed dispersal in Pinus taeda. Forest Ecology and Management, 217, 95-102. Willson, M.F. (1993) Dispersal mode, seed shadows, and colonization patterns. Vegetatio, 107/108, 261-280.

38

904

Box 6.1: Polychory and the study of long-distance seed dispersal

905

The fact that most plant species are not dispersed by a single vector (haplochory) but by

906

multiple vectors (polychory) has profound consequences for the study of LDD. To illustrate

907

this, we consider the hypothetical case of a plant species producing a large number of seeds

908

(108) that are dispersed by four different dispersal vectors (Figure 6.2a). The four vectors are

909

arranged in order of decreasing seed load Q and increasing fat-tailedness of the dispersal

910

kernel they produce (Figure 6.2a). For a plant species with elaiosome-bearing seeds these four

911

vectors could for example be (1) ants (the standard dispersal vector), (2) large animals, (3)

912

strong floods and (4) transportation by humans.

913

Vector 1 transports the vast majority of seeds (95%), but its thin-tailed dispersal kernel causes

914

it to generate only short dispersal distances (Figure 6.2a). Vectors 2 through 4 generate

915

increasingly larger maximum dispersal distances although they disperse successively fewer

916

seeds (Figure 6.2a). In fact, vector 4 disperses only a tiny fraction (0.0125%) of all seeds, but

917

the maximum dispersal distance it generates is more than five orders of magnitude greater

918

than that generated by the standard vector (vector 1).

919

When plotting the relative contribution of each vector to the total dispersal kernel (TDK) it

920

becomes apparent that different vectors dominate seed dispersal over different scales (Figure

921

6.2b). In particular, the vectors dominating seed dispersal at the scales typically investigated

922

in empirical studies are very different from the vectors dominating LDD.

923

This hypothetical example illustrates the limited ability of small-scale studies to detect LDD

924

vectors, and the fallacy of extrapolating from the small-scale behaviour of the standard vector

925

to LDD. The traditional seed-centred approach will almost inevitably miss the rare seeds

926

dispersed by non-standard LDD vectors. We thus advocate a vector-centred approach that

927

specifically targets those vectors that can transport seeds at the scales of interest.

39

928

Box 6.2: How to identify long-distance dispersal vectors

929

The general model for passive dispersal (Figure 6.3) helps to identify vectors that are likely to

930

cause LDD of a given plant species. To illustrate this, we first consider the dispersal of a

931

single seed. A single seed's dispersal distance is the product of its retention time at the vector

932

(P) and the vector's displacement velocity during this time (V). For a set of seeds dispersed by

933

the same vector, V and P are two random variates that follow statistical distributions (Figure

934

6.5a). The dispersal distance as the product of these two random variates again follows a

935

statistical distribution which is called the vector's dispersal distance kernel (Figure 6.5b).

936

The area under the dispersal distance kernel that lies to the right of the LDD threshold gives

937

the per-seed probability of LDD. Formally, this per-seed LDD probability is

938

P (LDD ) =



∫ f (x )dx = 1 − F (xLDD ) ,

x LDD

939

where xLDD is the LDD threshold, f is the probability density function and F the cumulative

940

density function of dispersal distance. The per-seed LDD probability thus depends on the

941

dispersal distance kernel which in turn depends on the distributions of V and P.

942

The conditions under which a vector generates a high per-seed probability of LDD fall into

943

two broad categories: in the rarer case, the median dispersal distance is close to or above the

944

LDD threshold so that dispersal events frequently exceed the LDD threshold (hatched grey

945

line in Figure 6.5b). This arises if the product of typical values for V and P is high (hatched

946

grey line in Figure 6.5a). An example for this rarer case might be provided by large animals

947

for which both typical displacement velocities and typical times of endozoochorous seed

948

retention are high (see Box 6.3). In the commoner case, the LDD threshold is far greater than

949

the typical dispersal distances, and a high per-seed LDD probability can only arise if the

950

dispersal kernel has a fat-tail (Figure 6.5b). A dispersal kernel is called fat-tailed if its tail

40

951

drops off more slowly than the tail of a negative exponential kernel (Table 6.1). In a plot of

952

log probability density against dispersal distance (Figure 6.5b), a fat tail curves away from the

953

x-axis, whereas a thin (or exponentially bounded) tail bends towards the x-axis.

954

How does the fat-tailedness of the dispersal kernel depend on the distributions of V and P? To

955

examine this, we conduct a "crossing experiment" with the two typical distributions of V and

956

P plotted in Figure 6.5a: the solid black (lognormal) distribution has a lower mean than the

957

solid grey (truncated normal) distribution, but it is leptokurtic and therefore has a higher

958

probability of generating high values. The dispersal kernel as the product of V and P is fat-

959

tailed when one or both of V and P follow the leptokurtic distribution (Figure 6.5b). In

960

contrast, if both V and P conform to the mesokurtic solid grey distribution, the resulting

961

dispersal kernel has a thin tail (Figure 6.5b). Our example thus shows that a dispersal vector

962

causes a high per-seed probability of LDD if V and/or P typically or occasionally reach high

963

values (hatched and black solid lines in Figure 6.5a and b). Figure 6.5b also shows that - as

964

the LDD threshold increases - the per-seed LDD probability of thin-tailed kernels drops far

965

more quickly than that of fat-tailed kernels: if the LDD threshold was set to 1500 m, the two

966

fat-tailed kernels would generate a higher per-seed LDD probability than the hatched grey

967

kernel.

968

To calculate the expected number of LDD events for a given vector, the per-seed probability

969

of LDD has to be multiplied with the vector's seed load (Q). Figure 6.5c shows how Q and

970

LDD probability interact: for low per-seed probabilities of LDD (thin-tailed dispersal kernels

971

with typical distances below the LDD threshold), the expected number of LDD events is

972

essentially 0 even if Q is very high. However, for high per-seed probabilities of LDD

973

(dispersal kernels that are fat-tailed or have typical distances above the LDD threshold), the

974

number of LDD events increases steeply with Q.

975

Important LDD vectors are thus those which generate a high per-seed probability of LDD

41

976

because the product of retention time (P) and displacement velocity (V) is at least

977

occasionally high. Such vectors can have a considerable contribution to LDD even if their

978

seed load (Q) is small. On the contrary, vectors for which the product of V and P has a thin

979

tail and typically lies below the LDD threshold hardly contribute to LDD.

42

980

Box 6.3: The allometry of endozoochorous seed dispersal

981

Allometric relationships between the body mass and various other quantities of organisms are

982

ubiquitous (Calder 1996). Here we use such allometric relationships to build a simple general

983

model that relates the body mass of animals to the mean dispersal distance of the seeds they

984

disperse endozoochorously.

985

The dispersal distance of a seed transported internally by an animal is the product of the gut

986

retention time P (the passage time through the animal gut) and the animal displacement

987

velocity V (Figure 6.3, Box 6.2). Because the movement velocity U of animals, rather than

988

their displacement velocity V, is usually measured, we incorporated a straightness factor c

989

which is 1 if the animal moves in a straight line and decreases towards 0 as the movement

990

path becomes less positively autocorrelated. We also accounted for the fact that animals do

991

not always move, by assuming that the time allocated to movement is a constant fraction f of

992

the gut retention time P. Hence the displacement velocity is

993

V = f c U.

994

Both the mean movement velocity U and the mean gut retention time P have strong

995

allometric relationships to animal body mass M (Robbins 1993, Calder 1996). These

996

relationships can be expressed as

997

U = U 0 M b1 and P = P0 M b 2 ,

998

where U0 and P0 are the allometric constants and b1 and b2 are the allometric exponents for U

999

and V, respectively.

1000

The expected mean dispersal distance x of seeds dispersed by an animal with body mass M is

1001

thus

1002

x = V P = fcU P = fcU 0 M b1P0 M b 2 = ( fcU 0 P0 )M b1+ b 2 .

43

(Eqn. 1)

1003

When parameterising this general mechanistic model with comparative data for the allometry

1004

of flight velocity and gut retention time in flying birds (Tucker 1973, Robbins 1993, Calder

1005

1996), we obtain

1006

x = 90432 fcM 0.5 in standard units of x [m] and M [kg],

1007

as a general mechanistic prediction for the mean distance of endozoochorous seed dispersal.

1008

To test this prediction, we extracted data from studies of seed dispersal by flying frugivorous

1009

birds (Figure 6.7). All these studies calculated seed dispersal kernels by combining empirical

1010

data on V and P, both estimated directly, and independently of M. The ten bird species

1011

included vary in body mass by two orders of magnitude and inhabit various ecosystems

1012

around the globe. A linear regression of log x against log M for these studies (Figure 6.7)

1013

estimated a scaling exponent of 0.502 that is very close to the independent expectation of 0.5

1014

(although the estimate has a broad 95% confidence interval ranging from 0.15 to 0.78). The

1015

allometric constant (f c U0 P0, see Eqn. 1) is estimated to be 1198.3 m, which together with

1016

Eqn. 2 implies that the product of the proportion of time for which birds fly (f) and the

1017

straightness factor of their movement (c) is on average 0.013.

1018

Our simple allometric model highlights the importance of large animals for LDD (G2). For

1019

large flying birds, even the mean distance of endozoochorous seed dispersal exceeds 1 km

1020

(Figure 6.7) showing that these species generate LDD frequently, as part of their ordinary

1021

behaviour. The wide confidence interval on the allometric exponent indicates that large

1022

deviations from the mean allometric trend are not uncommon. Positive deviations from the

1023

mean trend represent cases of particular importance for plant LDD.

44

(Eqn. 2)

1024

Table 6.1. An overview of the key terms used in this review (modified from Nathan et al.

1025

submitted). Term

Definition

Dispersal kernel

A probability density function characterizing the spatial distribution of seeds dispersed from a common source. The 'dispersal distance kernel' more precisely describes the distribution of seed dispersal distances (and hence the distribution of the product of displacement velocity, V, and retention time, P).

Dispersal vector

An agent transporting seeds or other dispersal units. Dispersal vectors can be biotic (e.g. birds) or abiotic (e.g. wind).

Vector displacement velocity

The speed at which a dispersal vector moves a seed away from the uptake location

(V)

(Figure 6.3).

Fat-tailed dispersal kernel

A dispersal kernel with a tail that drops off more slowly than that of any negative exponential kernel.

Haplochory

Seed dispersal by a single dispersal vector

Irregular dispersal vector

A dispersal vector acting in such a haphazard way that it cannot be predicted whether the vector will be involved in the seed dispersal of a given plant species.

Nonstandard dispersal vector

A dispersal vector different from the one inferred from traditional morphological. classifications (Higgins et al. 2003c)

Polychory

Seed dispersal by multiple dispersal vectors

Regular dispersal vector

A dispersal vector that is regularly (but potentially rarely) involved in the seed dispersal of a given plant species.

Seed passage time (P)

The time for which a seed is retained by a dispersal vector (Figure 6.3).

Seed

In a strict sense, the fertilized ovule of spermatophytes consisting of embryo, endosperm, and testa. We follow here the typical use of this term in the plant ecological literature as a synonym for a reproductive propagule.

Seed dispersal

The movement of seeds from the mother plant to a potential establishment site.

Vector seed load (Q)

The number of seeds dispersed by a particular dispersal vector.

Standard dispersal vector

A dispersal vector inferred from the phenotypic characters of the plant (e.g. the morphology of the dispersal unit) (Higgins et al. 2003c). Typically, this is the vector dispersing most seeds.

45

Total dispersal kernel (TDK)

The dispersal kernel generated by all vectors dispersing a certain plant species (Nathan 2007).

1026 1027

46

log Probability density

(a)

log Distance (x) log Probability density

(b)

1028

log Distance (x)

1029

Figure 6.1. Application of absolute (black solid arrows) and proportional (grey hatched

1030

arrows) definitions of long-distance seed dispersal (LDD) to two different dispersal distance

1031

kernels. Note that the proportional definition identifies extreme dispersal events rather than

1032

LDD events. Hence, in a species with highly restricted seed dispersal (a), the proportional

1033

definition will include dispersal distances of only a few metres, that typically are not

1034

considered as "long". On the other hand, in a species whose seeds are frequently dispersed

1035

over long distances (b), only the most extreme dispersal events meet the proportional

1036

definition, although somewhat less extreme distances might be long enough to influence

1037

large-scale dynamics.

1038

47

Q

Expected number of seeds

10

10

10

10

8

6

5

10

10

4

Vector1

Vector2

Vector3

Vector4

1

−5

10

−10

10

−2

10

2

1

4

10

6

10

10

Spatial scale (m)

1039

Contribution of vector to TDK (%)

Typical scale of empirical studies

1040

Scale of interest in landscape ecology/biogeography

100 80 60 40 20 0

−2

10

1

2

10

4

10

6

10

Spatial scale (m)

1041

Figure 6.2. (a) The total dispersal kernel (TDK, thick black line) for a hypothetical plant

1042

population dispersed by four vectors (different orange tints). The inset shows the unequal

1043

distribution of seed loads (Q) on a log scale. The arrows at the bottom indicate the range of

1044

distances of all seeds dispersed by each vector. (Figure modified from Nathan et al.

48

1045

submitted). (b) The relative contribution of each vector to the TDK as a function of spatial

1046

scale (x), showing that the vectors dominating LDD can be very different from those

1047

dominating seed dispersal at the typical scales of empirical studies.

49

1048

1049

Initiation

Transport

Vector seed load

Vector displacement velocity

Q

V

Termination

Seed passage time

P

1050

Figure 6.3. A general mechanistic model for passive dispersal (adapted from a general model

1051

for aerial transport, Isard & Gage 2001) that describes three phases of passive dispersal

1052

(boxes) through three key parameters (given below the boxes). The figure is modified from

1053

Nathan et al. (submitted).

50

(a)

1054 (b)

1055 (c)

1056 1057

Figure 6.4. Examples for the three general phases of passive dispersal (see Figure 6.3). (a)

1058

Initiation phase: Seeds of Sugarcane (Saccharum spp.) depart from their mother plants in a

1059

strong wind (© Ran Nathan). (b) Transport phase: A floating island composed of Papyrus

1060

(Cyperus papyurus L.) transports seeds, entire plants and other organisms across Lake Malawi

51

1061

(Oliver & McKaye 1982, © Kenneth McKaye). (c) Termination phase: After secondary wind

1062

dispersal over a burnt area in the South African Fynbos, a seed of the Common Sugarbush

1063

(Protea repens, L.) is trapped in burnt vegetation remains (© Frank Schurr).

52

Probability density

(a) 0.25 0.20 0.15 0.10 0.05 0.00 0

10

20

30

40

V (m/s) or P (s)

Probability density

(b) −2

10

−4

10

−6

10

0

500

1000

1500

Dispersal distance x (= V P) (m)

Expected number of LDD events

(c) 100 80 60 40 20 0 0

1064

50000

100000

Vector seed load Q

1065

Figure 6.5. Effects of vector seed load (Q), displacement velocity (V) and retention time (P)

1066

on the expected number of LDD events caused by different vectors dispersing a given plant

1067

species (different line types). (a) Distributions of V and P for different vectors. (b) Dispersal

1068

distance kernels that are the product of these distributions of V and P (note that the y-axis has

1069

log scaling). The solid black, solid grey and hatched grey kernels result from multiplying the

1070

respective distributions in (a) with themselves, and the black/grey line shows a kernel which

1071

is the product of the solid black and the solid grey distribution in (a). The black arrow

1072

indicates the LDD threshold chosen for this example. (c) The relationship between the

1073

expected number of LDD events and Q. Note that the hatched grey line increases so steeply

1074

that for Q = 100000 the expected number of LDD events exceeds 12000.

53

G1: Open landscapes (V,P) G2: Large animals (V,P) G3: Migratory animals (V) G4: Extreme meteorological events (V) G5: Ocean currents (Q,P) G6: Human transportation (Q,V,P)

10

2

10

3

4

5

10 10 Spatial scale (m)

10

6

10

7

1075 1076

Figure 6.6. A representation of six major generalizations about LDD mechanisms indicating

1077

the range of dispersal distances for which each generalization is most relevant. The brackets

1078

give the key parameters of the general dispersal model (Figure 6.3) that are predominantly

1079

affected by each mechanism.

54

1080 1081

Figure 6.7. Mean distance of endozoochorous seed dispersal versus disperser body mass in

1082

six field studies on ten flying frugivorous bird species. The line shows the fit of a regression

1083

of log mean dispersal distance ( x ) against log mean body mass (M). The estimated scaling

1084

exponent (the regression slope = 0.502) closely matches the value predicted by a generic

1085

allometric model (0.5). Numbers besides the points indicate the data source - 1: Westcott &

1086

Graham (2000), 2: Ward & Paton (2007), 3: Weir & Corlett (2007), 4: Spiegel & Nathan

1087

(2007), 5: Sun et al. (1997), 6: Holbrook & Smith (2000).

55

1088

Regular vectors dominate LDD (b) Cumulative contribution to LDD (%)

Contribution to LDD (%)

(a) 100

100

80 60 40 20 0

80 60 40 20 0

Vector

Vector

Irregular vectors dominate LDD (d) Cumulative contribution to LDD (%)

Contribution to LDD (%)

(c) 100

100

80 60 40 20 0

Vector

80 60 40 20 0

Vector

1089 1090

Figure 6.8. Two idealized types of dispersal vector spectra and their consequences for the

1091

predictability of LDD. Regular vectors are shown in grey and irregular vectors in white. (a)

1092

and (c) depict the two dispersal vector spectra by showing the relative contribution of the 20

1093

most important vectors to overall LDD. Note that the number of involved vectors is

1094

essentially unlimited. (b) and (d) show the respective cumulative contribution of vectors to

1095

overall LDD, with grey horizontal lines indicating the proportion of LDD explained by all

1096

regular vectors. If LDD is dominated by regular vectors, it can - at least in principle - be

56

1097

predicted through the quantification of these vectors. However, if LDD is dominated by

1098

irregular vectors, it is largely unpredictable.

57