1

1

2

3

4

Currents connecting communities: nearshore community similarity 5

and ocean circulation. 6

7

8

J.R. Watson 1, *, C.G. Hays2, P.T. Raimondi3, S. Mitarai4, C. Dong5, J.C. McWilliams5, C.A. 9

Blanchette6, J. E. Caselle6, D.A. Siegel1,7 10

11

12 1 Institute for Computational Earth System Science, University of California Santa Barbara, Santa Barbara, CA, 13

93106, USA. 14 2 School of Natural Sciences, University of California Merced, Merced, CA, USA. 15 3Department of Ecology and Evolutionary Biology, University of California Santa Cruz, Santa Cruz, CA, USA. 16 4 Okinawa Institute of Science and Technology, 7542 Onna-son Onna, Okinawa 904-0411, Japan. 17 5 Institute of Geophysics and Planetary Physics, University of California, Los Angeles, CA 90095, USA. 18 6 Marine Sciences Institute, University of California Santa Barbara, Santa Barbara, CA, 93106, USA. 19 7Department of Geography, University of California Santa Barbara, Santa Barbara, CA, 93106, USA. 20

21 *Author for correspondence: watson@icess.ucsb.edu 22

23

Keywords: Marine biodiversity, marine community structure, larval dispersal, ocean circulation. 24

25

2

Abstract 25

Understanding the mechanisms that create spatial heterogeneity in species distributions is 26

fundamental to ecology. For nearshore marine systems, most species have a pelagic larval stage 27

where dispersal is strongly influenced by patterns of ocean circulation. Concomitantly, nearshore 28

habitats and the local environment are also influenced by ocean circulation. Because of the 29

shared dependence on the seascape, distinguishing the relative importance of the local 30

environment from regional patterns of dispersal for community structure remains a challenge. 31

Here, we quantify the oceanographic ‘distance’ and ‘asymmetry’ between nearshore sites using 32

ocean circulation modeling results. These novel metrics quantify spatial separation based on 33

realistic patterns of ocean circulation and we explore their explanatory power for intertidal and 34

subtidal community similarity in the Southern California Bight. We find that these metrics show 35

significant correspondence with patterns of community similarity and that their combined 36

explanatory power exceeds that of the thermal structure of the domain. Our approach identifies 37

the unique influence of ocean circulation on community structure and provides evidence for 38

oceanographically mediated dispersal limitation in nearshore marine communities. 39

40

Introduction 41

The decay of similarity with geographic distance has long been identified (Tobler, 1970), yet 42

reasons why this relationship exists in ecological communities are still debated (Menge et al. 43

1997, Nekola and White 1999, Cottenie 2005). Spatial patterns in biodiversity and community 44

similarity reflect the influence of multiple processes broadly categorized as either niche-based, 45

through species sorting and the effect of environmental gradients on individual species 46

demography (Armstrong and McGehee 1980, Leibold 1995) or so-called neutral processes, 47

3

namely dispersal limitation (Hubbell 2001). In most ecological systems, because there is often 48

strong correspondence between patterns of dispersal and environmental variables, quantifying the 49

relative contribution of niche and neutral processes in determining community composition 50

remains a challenge (Gilbert et al. 2004). Our aim is to identify the unique influence of dispersal 51

on spatial patterns of community similarity for nearshore marine species in the Southern 52

California Bight (SCB). In order to address this challenge, we have developed novel metrics of 53

dispersal using ocean circulation models. 54

55

In benthic marine systems the role of dispersal in determining community dynamics, population 56

demography and genetic structure has had a long and rich history (Hjort 1914, Gaines and 57

Roughgarden 1985, Palumbi 2003, Alberto et al. 2010, White et al. 2010). As adults, many 58

nearshore marine species are sedentary, with limited home ranges (e.g. < 10 km). It is primarily 59

during their larval stage that they disperse further, potentially hundreds of kilometers (Kinlan and 60

Gaines 2003, Watson et al. 2010). During this stage ocean currents play a vital role creating, at 61

the same time, barriers to dispersal and highly connected regions (Baums et al. 2006; Cowen et al. 62

2006). It is unknown whether this seascape structure exerts control on community similarity 63

through dispersal limitation. Recently, Lagrangian particle simulations have been used to capture 64

the spatial complexity of oceanographically mediated dispersal and have provided insight into the 65

potential scales of population connectivity (e.g. Cowen et al. 2006, Mitarai et al. 2009) and in 66

understanding the drivers of genetic structure (e.g. Galindo et al. 2006, Kool et al. 2009, White et 67

al. 2010, Selkoe et al. 2010). Here, we use Lagrangian particle simulations to investigate the 68

influence of ocean circulation on the community similarity of nearshore marine species in the 69

SCB. 70

4

71

The SCB is a region on the west coast of North America that stretches from Point Conception to 72

the U.S. – Mexican border (Fig. 1a). It is characterized by the meeting of two major water masses 73

at Point Conception – the relatively cool, offshore, equatorward flowing California Current and 74

the warmer, near coastal, poleward flowing Southern California Counter Current (Fig. 1a). Spatial 75

patterns in community similarity have been shown to correspond strongly with the thermal 76

structure of the SCB (Blanchette et al. 2008), suggesting that species are adapted to this thermal 77

gradient. However, because multiple mechanisms driving community structure are functions of 78

ocean circulation, environmental variables such as temperature will correspond with patterns of 79

oceanographically mediated dispersal. Hence, it remains an open question whether the thermal 80

structure of the SCB is the primary force shaping community structure in the SCB. The goal of 81

this work is to partition the statistical contribution of dispersal by ocean currents from that of the 82

thermal structure of the SCB. In order to do so, we use simulations of Lagrangian water mass 83

trajectories and develop novel metrics of dispersal based on the complex ocean circulation of the 84

SCB. 85

86

Methods 87

We analyzed spatial patterns in marine community composition (i.e. species identity and 88

abundance) from two empirical sources: in 2004, subtidal rocky reef communities were surveyed 89

using underwater visual census as part of the Cooperative Research Assessment of Nearshore 90

Ecosystems program (CRANE 2004), and the intertidal communities were surveyed as part of the 91

Community Biodiversity Survey conducted by the Partnership for the Interdisciplinary Study of 92

Coastal Oceans, an ongoing project which started in 2001 (PISCO, Blanchette et al., 2008). Fifty-93

5

nine subtidal rocky reef sites and thirty-nine intertidal sites in the SCB were sampled, spanning 94

the mainland and the Channel Islands (Fig. 1a). Here, we analyzed a total of 214 species (105 95

subtidal and 109 intertidal species). The data were square root transformed to reduce the influence 96

of highly abundant species and a community similarity matrix was calculated using the Bray-97

Curtis similarity coefficient (Bray and Curtis 1957). 98

99

Lagrangian particle simulations were used to quantify the movement of water parcels, between 100

nearshore regions. Approximately 400,000 passive Lagrangian particles were released in 135 101

circular patches (5 km in radius), distributed uniformly throughout the nearshore of the SCB, over 102

the period January 1, 1996 - December 31, 2002 (see Mitarai et al. 2009 for more information on 103

the Lagrangian particle simulations). The size of these patches reflects limitations in the 104

numerical simulations of Lagrangian particles. Lagrangian particles advect passively with 105

velocity fields produced from a Regional Oceanic Modeling System (ROMS) solution to the SCB 106

(Dong et al. 2009; Mitarai et al. 2009; see Fig. 1b). This solution is 3D with a horizontal spatial 107

resolution of 1 km and 40 vertical levels. All eight islands in the Southern California Bight are 108

resolved and the complex circulation of the SCB is well captured (see Dong et al. 2009 for further 109

information on the hydrodynamic model). 110

111

Lagrangian particle transit times, between all pairs of nearshore sites, provide a quantification of 112

spatial separation based on ocean currents (Mitarai et al, 2009, Fig. 1c) and describe patterns of 113

larval dispersal. An important property of Lagrangian particle transit times is their inherent 114

asymmetry, that is, Lagrangian particles traveling from patch A to B may have a different mean 115

transit time than those particles traveling from B to A. Thus the first metric used here, termed the 116

6

oceanographic distance, is the average of the two-way Lagrangian transit times. The second 117

metric is termed the oceanographic asymmetry and is defined as the absolute difference in the 118

two-way Lagrangian particle transit times. Both oceanographic distances and asymmetries are 119

measured in units of time (days). 120

121

The thermal environment of the empirical sampling sites was assessed using monthly sea surface 122

temperatures from the Pathfinder 5 Advanced Very-High Resolution Radiometer (Kahru and 123

Mitchell, 2002). Data at a spatial resolution of 4 km was averaged over the period 1997 – 2008. 124

Spatially, values were then smoothed using a 2-dimensional, 5 km radius, Gaussian filter. These 125

values were then linearly interpolated to the subtidal and intertidal sampling sites. Smoothing and 126

interpolation created a better representation of the local sea surface temperatures than the nearest 127

data point. The thermal structure between the empirical (community) sampling sites was then 128

defined as the difference in SST (ΔSST). Unlike Lagrangian transit times, ΔSST is a symmetric 129

quantity. 130

131

Univariate and multivariate linear models of community similarity were made using 132

oceanographic distance, asymmetry and ΔSST as independent variables. The Mantel correlation 133

statistic (r) and its permutation test of significance (p-value) were used to describe univariate 134

relationships. In the multivariate case, multiple linear regression on distance matrices was used to 135

explore the influence and interaction of independent variables (Legendre and Legendre 1998, 136

Lichstein 2007). Further, multivariate models were created for different regions of the SCB: a full 137

model (using all sampling sites) and a regional model (comparing island with mainland sites 138

only). Standard regression coefficients and semi-partial correlation coefficients (Legendre and 139

7

Legendre 1998) were used to describe the multivariate models. Standard regression coefficients 140

are produced using the z-score of the independent variables, which controls for differing units 141

among terms and allows for direct comparisons of their influence; larger standard regression 142

coefficients identify those independent variables with more influence. Confidence intervals on 143

these values were estimated with a bootstrap procedure. Semi-partial correlation coefficients were 144

used to quantify the relative contribution of each independent variable in a given model. These 145

are determined as the reduction in R2 when a given independent variable is left out of a 146

multivariate model. Values are reported as fractions of the original R2. For all models, the key 147

assumptions of linear regression were met: (1) The distribution of community similarities was 148

approximately normal (assessed with quantile-quantile plots), (2) community similarity and all 149

independent variables correspond linearly and (3) collinearity among independent variables, 150

quantified using the variance inflation factor, was negligible (all variance inflation factors were 151

found to be less than five). 152

153

Results 154

The univariate analysis revealed a significant, negative, linear relationship between community 155

similarity and ΔSST for both the subtidal and intertidal communities (r = -0.59 and -0.25, p < 156

0.01, Fig. 2a and b respectively). For both, variability in community similarity is largest at small 157

mean differences in SST. This could simply be a result of fewer data points at greater SST 158

differences or it may indicate the spatial scales over which ΔSST has explanatory power. For 159

instance, ΔSST can explain differences in species composition at the extremes of the SCB, the 160

very warm (~17oC) compared to the very cold (~12oC, Fig. 1a). However, in the center of the 161

SCB, where the temperature gradient is less steep, there is less significance to ΔSST's explanatory 162

8

power. 163

164

Oceanographic distance shows a stronger, more significant, negative relationship with subtidal 165

and intertidal community similarity (r = -0.73 and -0.38, p-values < 0.01, Fig. 2c and d 166

respectively). This indicates that those local communities that are far apart, based on ocean 167

circulation, are less similar than those closer together. Further, in contrast to ΔSST, the variability 168

in the relationship between community structure and oceanographic distance is homogenous 169

throughout the domain. The significance and strength of the relationship indicate that 170

oceanographic distance is a better predictor of community similarity than ΔSST. Oceanographic 171

asymmetry was found to have a negative relationship with subtidal community similarity (r = -172

0.19, p < 0.01, Fig. 2e). Asymmetry did not produce a significant relationship with intertidal 173

community similarity (r = 0.03, p > 0.01, Fig. 2f) and hence, from a univariate analysis, it is 174

difficult to draw any conclusions as to the influence of asymmetry on community similarity. 175

176

Between all sampling sites (the full model), multivariate linear models using oceanographic 177

distance, asymmetry and ΔSST as independent variables explained 58% (p < 0.01) and 18% (p < 178

0.01) of the variance in subtidal and intertidal community structure (Figs. 3a and b) respectively. 179

For both communities, oceanographic distance was most influential with the largest standard 180

regression coefficient and semi-partial correlation coefficient (81.0% and 76.9%). ΔSST and 181

asymmetry swap importance between the subtidal and intertidal communities with ΔSST being 182

more influential on subtidal community similarity and less so for intertidal community similarity. 183

For the subtidal community ΔSST then has the second largest semi-partial coefficient (17.5%) 184

and oceanographic asymmetry has the smallest (1.5%), suggesting that asymmetry has little 185

9

influence in this model. In contrast, asymmetry has more unique explanatory power than ΔSST 186

for the intertidal community (17.0% compared to 6.1% respectively). 187

188

We found notable differences between the subtidal and intertidal groups when comparing only 189

pairs of sites between the mainland and the islands (the regional model, Figs. 3c and d 190

respectively). In both cases the explanatory power (R2) of the multivariate models decrease, from 191

0.58 to 0.41 in the subtidal community model and from 0.19 to 0.05 in the intertidal community. 192

In particular, the intertidal model is not as significant (p < 0.1). For the subtidal community, 193

ΔSST now has the largest influence with the largest standard regression coefficient and semi-194

partial correlation coefficient (62.0%, Fig 3c). Relative to full model, the influence of 195

oceanographic asymmetry increased (from 1.5% to 6.0%) and the oceanographic distance 196

dropped (from 81.0% to 32.1%). For the intertidal community, asymmetry has the largest 197

standard regression coefficient (68.8%). The influence of oceanographic distance decreased from 198

76.9% in the full model to 31.3% in the regional model. Interestingly, ΔSST’s contribution to the 199

regional model is zero. 200

201

Discussion 202

Here, we present evidence for the influence of oceanographically mediated dispersal on spatial 203

patterns of nearshore marine species community similarity. We developed two metrics of 204

dispersal, oceanographic distance and asymmetry, using high-resolution ocean circulation 205

simulation results. Both metrics describe spatial separation between nearshore locations based on 206

the complex patterns of ocean circulation of the SCB. Oceanographic distance was defined as the 207

mean of the two-way Lagrangian particle transit times and asymmetry as their absolute 208

10

difference. Oceanographic distances, relative to the thermal structure of the SCB, better explained 209

community similarity. Further in the multivariate models, oceanographic asymmetry was found to 210

be a statistically significant predictor variable, extending the explanatory power of oceanographic 211

distances. Absolute differences in any environmental variable will be symmetric. In contrast, 212

asymmetry is a quantity unique to oceanographically mediated dispersal and hence, provides 213

direct evidence for its influence on patterns of community similarity in the SCB. 214

215

The explanatory power of the multivariate models is a result of the interaction between 216

oceanographic distances and asymmetries; respectively, oceanographic distances and asymmetries 217

were negatively and positively associated with community similarity. For example, take two pairs 218

of sites [A, B] and [C, D] and assume that the oceanographic distance, the average of the two-way 219

Lagrangian transit times, is 20 days for both. Now consider that the Lagrangian transit times for 220

the [A, B] pair are 20 days each way, but between [C and D] the two-way times are different, say 221

30 and 10 days. In the case of the [C, D] pair the minimum oceanographic distance is 10 days. 222

This lower bound to the oceanographic distance, captured by our metric of oceanographic 223

asymmetry, results in greater ocean connectivity and hence community similarity. Essentially, 224

larger values of asymmetry result in a smaller minimum oceanographic distance. Note that using 225

the minimum and maximum oceanographic distance in place of the mean created essentially the 226

same result. Oceanographically, the weak explanatory power of oceanographic asymmetry, 227

relative to oceanographic distance, indicates that dispersal between regions is dominated by 228

bidirectional mesoscale eddy dynamics rather than unidirectional persistent currents (e.g., Siegel 229

et al. 2008). 230

231

11

Our analysis assumes an implicit relationship between ocean circulation and patterns of larval 232

dispersal. This assumption could be inaccurate as many marine species’ larvae are thought to be 233

able to influence their dispersal, for example through vertical behavior or late-larval period active 234

swimming (e.g. Leis 2007, Shanks 2009). This larval behavior could have a large effect on 235

dispersal patterns. However, although our metrics of oceanographic distance and asymmetry were 236

derived from passive Lagrangian particles, it is likely that they describe the long-term patterns of 237

dispersal in the SCB (Mitarai et al. 2009). Further, other than patterns of ocean circulation and 238

dispersal, there are numerous physical and demographic factors that alter community similarity, 239

for example wave exposure, air temperature (especially for intertidal species), larval mortality in 240

the water column, settlement behavior and post-settlement processes. Thus, we may not expect 241

these two quantities to be closely linked. Indeed, the low explanatory power of the intertidal 242

multivariate models reveals that intertidal community similarity will be predominantly shaped by 243

factors not included in this analysis. However, our key finding is not the absolute explanatory 244

power of the multivariate models, rather it is the significance of the independent variables; the 245

existence of a statistically significant relationship between these our metrics of dispersal, 246

oceanographic distance and asymmetry, and community similarity suggests that 247

oceanographically mediated dispersal is important to spatial patterns of community similarity. 248

249

We found that subtidal community similarity was better explained by oceanographic distance and 250

asymmetry, relative to intertidal community similarity. There are several potential reasons for 251

this. First, although our Lagrangian particle simulations accurately capture regional patterns of 252

ocean circulation (Mitarai et al. 2009), surf-zone currents are not well represented. Hence, 253

patterns of intertidal larval dispersal may not be well described by our calculations of 254

12

oceanographic distance and asymmetry. Indeed, surf-zone hydrodynamics and their effect on 255

dispersal is an area of active debate; surf-zone currents could act to retain dispersing larvae and 256

limit long distance dispersal (Shanks 2009). Second, in our data sets there are many more subtidal 257

species, for example fish and invertebrates, which have long pelagic larval durations and the 258

potential for long distance dispersal. Thus oceanographic distance and asymmetry, which likely 259

describe patterns of long distance dispersal (i.e. > 1 km) rather than dispersal over short distances 260

(i.e. < 1 km), are expected to show a stronger relationship with the subtidal species we examined. 261

262

Neutral theory suggests that spatial autocorrelation in dispersal will create a decline in 263

compositional similarity between communities as distance between sites increases (Hubbell 2001, 264

Gilbert et al. 2004). It is important then that the appropriate spatial basis, by which ‘distance’ is 265

measured, be identified. For example, patterns of larval dispersal, at the temporal and spatial 266

scales of this investigation, do not necessarily correspond with geographic distance (Appendix 1). 267

Indeed, the predictive power of our multivariate models decreases when oceanographic distance 268

in exchanged with geographic distance (Appendix 2). Further, asymmetry is not captured by 269

geographic distance and hence it cannot be used to quantify a critical aspect of dispersal in the 270

ocean. 271

272

Spatial patterns in biodiversity and community similarity are inherently dynamic. There is 273

significant evidence suggesting that key oceanographic phenomena, such as El Niño events, are 274

vital to larval recruitment and demographic processes of nearshore marine species in the SCB 275

(Cowen et al. 1985, Selkoe et al. 2007). However, most analyses of community similarity are 276

confined to static descriptors of spatial separation (i.e. geographic distance) and hence do not 277

13

account for such dynamism. In contrast, oceanographic distances and asymmetries are derived 278

from multi-year simulations of ocean circulation patterns and hence quantify spatial separation as 279

a dynamic property. In the SCB there is significant interannual and seasonal variation in patterns 280

of ocean circulation (Dong et al. 2009; Mitarai et al. 2009, Watson et al. 2010) and these are 281

reflected in our estimates of oceanographic distances and asymmetries. For example, some 282

nearshore sites are effectively 'closer' together during El Niño conditions than during La Niña 283

conditions (Appendix 3). Given the results of this paper, it is expected that such changes in 284

'distance' will result in corresponding changes in community similarity for these years. 285

286

Understanding the processes that drive spatial patterns in community similarity is crucial to 287

spatial fisheries management and conservation. The knowledge generated from quantitative 288

investigations secure adequate representation of biodiversity in, for instance, marine protected 289

areas (Lubchenco et al. 2003). For example in the SCB, knowledge of spatial patterns of 290

community structure has already proven informative to spatial fisheries management (MLPA 291

2009). Further, making use of real-time simulation data products (e.g. Chao et al. 2009) will 292

greatly improve our ability to account for changes in patterns of community similarity. 293

Ultimately, incorporating modeled oceanographic products, like those used here, will improve 294

decision-making processes and our governance of nearshore living resources. 295

296

Acknowledgements. The authors would like to thank the SWAT team, J. Elliott and E. Fields, 297

NASA for graduate student support for JRW and NSF for support of the Flow, Fish and Fishing 298

project. The authors also thank PISCO and the CRANE Program (specifically, the principal 299

investigators: T. Anderson, M. Edwards, J. Car- roll, D. Schroeder, M. Carr, C. Syms, and D. 300

14

Pondella) for the use of their data. This is PISCO contribution number XXXX. 301

302

References 303

1. Alberto, F., P. T. Raimondi, D. C. Reed, N. C. Coelho, R. Leblois, A. Whitmer, and E. A. Serrão. 304

2010. Habitat continuity and geographic distance predict population genetic differentiation in 305

giant kelp. Ecology 91: 49-56. 306

2. Armstrong, R. A., and R. McGehee. 1980. Competitive exclusion. The American Naturalist 115: 307

151-170. 308

3. Baums, I. B., C. B. Paris, and L. M. Cherubin. 2006. A bio-oceanographic filter to larval dispersal 309

in a reef-building coral. Limnol. Oceanogr. 51: 1969-1981. 310

4. Blanchette, C. A., C. M. Miner, P. T. Raimondi, D. Lohse, K. E. K. Heady, and B. R. Broitman. 311

2008. Biogeographical patterns of rocky intertidal communities along the Pacific coast of North 312

America. Journal of Biogeography 35: 1593 – 1607. 313

5. Bray, J. R., and J. T. Curtis. 1957. An ordination of the upland forest communities of southern 314

Wisconsin. Ecol Monogr 27: 325 – 349 315

6. California Marine Life Protection Act, California Department of Fish and Game, CA, USA. 2009. 316

http://www.dfg.ca.gov/mlpa/. 317

7. Chao, Yi., Z. Li, J. Farrara, J. C. McWilliams. J. Bellingham, X. Capet, F. Chavex, J-K. Choi, R. 318

Davis, J. Doyle, D. M. Fratantoni, P. Li. P. Marchesiello. M. A. Moline, J. Paduan, and S. Ramp. 319

2009. Development, implementation and evaluation of a data-assimilative ocean forecasting 320

system off the central California coast. Deep-Sea Research II 56: 100-126. 321

8. Cottenie, K. 2005. Integrating environmental and spatial processes in ecological community 322

dynamics. Ecology Letters 8: 1175-1182. 323

15

9. Cowen, R. K. 1985. Large scale pattern of recruitment by the labrid, Semicossyphus pulcher: 324

causes and implications. J Mar Res 43:719–742. 325

10. Cowen, R. K., C. Paris, and A. Srinivasan. 2006. Scaling of connectivity in marine populations. 326

Science 311: 522–53: 327

11. The Cooperative Research Assessment of Nearshore Ecosystems program (CRANE). 2004. 328

www.dfg.ca.gov/marine/fir/crane.asp. 329

12. Dong, C., E. Idica, and J. C. McWilliams. 2009. Circulation and multiple-scale variability in the 330

Southern California Bight. Progress in Oceanography 82: 168 - 190 331

13. Gaines, S. D., and J. Roughgarden. 1985. Larval settlement rate: A leading determinant of 332

structure in an ecological community of the marine intertidal zone. Proc. Natl. Acad. Sci. 82: 333

3707-3711. 334

14. Galindo, H. M., D. B. Olson, and S. R. Palumbi. 2006. Seascape genetics: a coupled 335

oceanographic-genetic model predicts population structure of Caribbean corals. Current Biology 336

16: 1622-1626. 337

15. Gilbert, B., and M. J. Lechowicz. 2004. Neutrality, niches, and dispersal in temperate forest 338

understory. PNAS. 101: 7651-7656. 339

16. Hjort, J. 1914. Fluctuations in the great fisheries of northern Europe. Cons. Permanent. Int. 340

Explor. Mer. Rapp. P-V Reun. 20:1–228. 341

17. Hubbell, S. P. 2001. The Unified Neutral Theory of Biodiversity and Biogeography. Princeton 342

Univ. Press, Princeton. 343

18. Kahru, M., and B. G. Mitchell. 2002. Influence of the El Niño – La Nina cycle on satellite derived 344

primary production in the California Current. Geophysical Research Letters 29: 1–4. 345

19. Kinlan, B. P., and S. D. Gaines. 2003. Propagule dispersal in marine and terrestrial environments: 346

16

a community perspective. Ecology 84: 2007-2020. 347

20. Kool, J. D., C. B. Paris, S. Andrefoet, and R. K. Cowen. 2010. Complex migration and the 348

development of genetic structure in subdivided populations: an example from Caribbean coral 349

reef ecosystems. Ecography (In press). 350

21. Legendre, P., and L. Legendre. 1998. Numerical Ecology. Developments in Environmental 351

Modelling, 2nd Edition. 352

22. Leibold, M. A. 1995. The niche concept revisited: mechanistic models and community context. 353

Ecology 76: 1371-1382. 354

23. Leis, J. M. 2007. Behavior as input for modelling dispersal of fish larvae: behavior, 355

biogeography, hydrodynamics, ontogeny, physiology and phylogeny meet hydrography. Mar. 356

Ecol. Prog. Ser. 347: 185 - 193. 357

24. Lichstein, J. W. 2007. Multiple regression on distance matrices: a multivariate spatial analysis 358

tool. Plant Ecology 188: 117 – 131. 359

25. Lubchenco, J., S. R. Palumbi, S. D. Gaines, and S. Andelman. 2003. Plugging the hole in the 360

ocean: the emerging science of marine reserves. Ecology Applications 13: S3-S7. 361

26. Menge, B. A., B. A. Daley, P. A. Wheeler, E. Dahlhoff, E. Sanford, and P. R. Strub. 1997. 362

Benthi-pelagic links and rocky intertidal communities: Bottom-up effects on top-down control? 363

Proc. Natl. Acad. Sci. 94: 14530-14535. 364

27. Mitarai, S., D. A. Siegel, J. R. Watson, C. Dong, and J. C. McWilliams. 2009. Quantifying 365

connectivity in the coastal ocean with application to the Southern California Bight. Journal of 366

Geophysical Research 114: C10026. 367

28. Nekola, J. C., and P. S. White. 1999. The distance decay of similarity in biogeography and 368

ecology. Journal of Biogeography 26: 867-878. 369

17

29. Palumbi, S. R. 2003. Population genetics, demographic connectivity and the design of marine 370

reservers. Ecological Applications 13: S146-S158. 371

30. Selkoe, K. A., A. Vogel, and S. D. Gaines. 2007. Effects of ephemeral circulation on recruitment 372

and connectivity of nearshore fish populations spanning Southern and Baja California. Mar Ecol 373

Prog Ser. 351: 209–220. 374

31. Selkoe, K. A., J. R. Watson, C. White, T. B. Horin, M. Iacchei, S. Mitarai, D. A. Siegel. S. D. 375

Gaines and R. J. Toonen. 2010. Taking the chaos out of genetic patchiness: seascape genetics 376

reveals ecological and oceanographic drivers of genetic patterns in three temperate reef species. 377

Molecular Ecology. In press. 378

32. Shanks, A. L. 2009. Pelagic larval duration and dispersal distance revisted. Biol. Bull. 216: 373-379

385. 380

33. Siegel, D.A., S. Mitarai, C.J. Costello, S.D. Gaines, B.E. Kendall, R.R. Warner and K.B. Winters, 381

2008. The stochastic nature of larval connectivity among nearshore marine populations. 382

Proceedings of the National Academy of Science, 105, 8974-8979. 383

34. Tobler, W. R. 1970. A computer movie simulating urban growth in the Detroit region. Econ. 384

Geogr. 46: 234-240. 385

35. Watson, J. R., S. Mitarai, D. A. Siegel, J. E. Caselle, C. Dong, and J. C. McWilliams. 2010. 386

Realized and potential larval connectivity in the Southern California Bight. Marine Ecology 387

Progress Series 401: 31-48. 388

36. White, C., K. A. Selkoe, J. Watson, D.A. Siegel, D.C. Zacherl, and R.J.Toonen, 2010. Ocean 389

currents help explain population genetic structure. Proceedings of the Royal Society, B: 390

Biological Sciences 277: 1685-1694. 391

392

18

392

Figure 1. (a) The Southern California Bight with average sea surface temperature (for the period 393

1998-2008). Empirical intertidal (white stars) and subtidal (black stars) sampling locations are 394

shown. (b) Example 30 day simulated Lagrangian particle trajectories. Particles were released 395

within the green circle on 1st of September 1997. Their locations 30 days after passive advection 396

by ocean currents are identified by blue dots. (c) Example oceanographic distances - the number 397

of days, on average, for a Lagrangian particle to travel from the pink site to all others.398

19

399

Figure 2. Univariate comparison of community similarity with oceanographic distance, 400

asymmetry and ΔSST for the (a, c, e) subtidal and (b, d, f) intertidal sites. Mantel correlation (r) 401

and p-values (p) are shown. 402

403

20

403

Figure 3. Standardized regression (diamonds) and semi-partial correlation coefficients 404

(percentages) for the multiple linear regression models. Independent variables are on the x-axis: 405

oceanographic distance (Oc-D), oceanographic asymmetry (Oc-Asym) and the difference in sea-406

surface temperature (ΔSST). Error-bars describe the 95% confidence interval for the regression 407

coefficients. Full models, using all sampling sites for the subtidal (a) and intertidal (b). Regional 408

models comparing community similarity between only islands and mainland sites for the subtidal 409

(c) and intertidal (d) communities. 410

411