1 2 Received Date: 09-Jul-2016 3 Revised Date: 19-Dec-2016 4 Accepted Date: 16-Feb-2017 5 Article Type: Articles 6 Short-range dispersal maintains a volatile marine metapopulation: the brown alga 7 Postelsia palmaeformis 8 9 Robert T. Paine 10 Department of Biology 11 University of Washington 12 Seattle, WA 98195-1800 13 14 Eric R. Buhle 15 Northwest Fisheries Science Center 16 2725 Montlake Blvd. East 17 Seattle, WA 98112-2097 18 19 Simon A. Levin 20 Department of Ecology and Evolutionary Biology 21 Princeton University 22 Princeton, NJ 08544Author Manuscript This is the author manuscript accepted for publication and has undergone full peer review but has not been through the copyediting, typesetting, pagination and proofreading process, which may lead to differences between this version and the Version of Record. Please cite this article as doi: 10.1002/ecy.1798 This article is protected by copyright. All rights reserved 23 24 Peter Kareiva 25 Institute of the Environment and Sustainability 26 UCLA 27 Los Angeles, CA 90095 28 [email protected] 29 (corresponding author) 30 31 32 33 Running Head: Local dispersal allows persistence 34 Abstract. The annual brown alga Postelsia palmaeformis is dependent for its survival on 35 short-distance dispersal (SDD) where it is already established, as well as occasional long- 36 distance colonization of novel sites. To quantify SDD, we transplanted Postelsia to sites 37 lacking established plants within ≥10 m. The spatial distribution of the first naturally 38 produced sporophyte generation was used to fit dispersal kernels in a hierarchical Bayesian 39 framework. Mean dispersal distance within a year ranged from 0.16 to 0.50 m across sites; 40 95% of the recruits were within 0.38 to 1.32 m of the transplant. The fat-tailed exponential 41 square root kernel was the best among the candidate models at describing offspring density 42 and dispersal. Independent measurements of patch size over 2-5 generations permitted an 43 evaluation of whether models parameterized by individual-level data could adequately 44 predict longer-termAuthor Manuscript persistence and spread at the patch scale. The observed spread rates 45 generally fell within the 95% predictive intervals. Finally, Postelsia was eliminated from 14 46 occupied sites that were then followed for ≥27 years. The probability of invasion when This article is protected by copyright. All rights reserved 47 unoccupied declined and the probability of extinction when occupied increased with distance 48 from the nearest propagule source. Sites >10 m from a source were rarely invaded, and one 49 initially densely populated site isolated by 39 m has remained Postelsia-free since 1981. In 50 spite of dispersal that is almost entirely within 2 m of the parent, the ability of our models to 51 capture the observed dynamics of Postelsia indicates that short-range dispersal adequately 52 explains the persistent and thriving Postelsia metapopulation on Tatoosh Island. However, 53 the presence of Postelsia over a 2000-km coastal range with many gaps >1 km makes it clear 54 that rare long-distance dispersal must be required to explain the geographic range of the 55 species. 56 57 Key words: Postelsia palmaeformis, dispersal, metapopulation, colonization, ephemeral 58 populations, persistence 59 60 Running head: Local dispersal allows persistence 61 INTRODUCTION 62 63 Postelsia palmaeformis Ruprecht, or sea palm, is a widely distributed and locally 64 abundant brown alga found in the upper rocky intertidal zone from southern California to 65 Vancouver Island, and is a characteristic species of wave-exposed, highly disturbed sites 66 (Dayton 1973, Paine 1988). It is an annual that must recolonize each year without any sort of 67 seed bank as well as a fugitive species that depends upon wave-mediated gap formation to 68 remove the competitivelyAuthor Manuscript dominant mussel Mytilus californianus and release space, the 69 limiting resource (Paine 1979). The persistence of vigorous Postelsia populations in a highly 70 capricious and severe environment is due to dispersal and recolonization, but unlike annual This article is protected by copyright. All rights reserved 71 plants with seed banks or perennials with overlapping generations, Postelsia must rely 72 entirely on spatial dispersal for persistence. The absence of resting stages or seed banks and 73 the inevitability of local competitive exclusion by Mytilus makes Postelsia the quintessential 74 dispersal-dependent metapopulation. Yet unlike many “classic” fugitive species, Postelsia’s 75 dispersal ability appears to be quite limited (Dayton 1973, Paine and Levin 1981). The key 76 question is whether or not local, short-distance dispersal is adequate for this species’ 77 persistence, or whether long-distance rescue is necessary. 78 To answer this question we combine experiments and models to ask whether local 79 dispersal processes can account for the year-to-year pattern of persistence and spread of 80 Postelsia on Tatoosh Island, WA. While several researchers have asked whether local 81 observations of dispersal can explain broader patterns of spread (Andow et al 1990, 82 Shigesada and Kawasaki 1997, Clark 1998), none have done so with direct annual 83 observations spanning >20 years and frequent extinction and colonization events. Similarly, 84 models of patch occupancy and recolonization in metapopulations (Hanski 1994) have not 85 been based on detailed dispersal characterizations of individual organisms. Thus this study is 86 unique in building an empirically based microscale (i.e., at the scale of individual dispersal 87 events) model of population turnover and colonization in a highly volatile population 88 experiencing frequent disturbance. 89 Our approach builds upon a long history of theoretical and empirical examinations of 90 the dynamical consequences of space and movement in population theory, tracing back to 91 Haldane, Fisher and Wright, and later Skellam (1951). Levins (1969) coined the term 92 “metapopulation”Author Manuscript to describe populations hierarchically structured in space, and the 93 discipline of spatial ecology has developed rapidly since (Levin 1976, Tilman and Kareiva 94 1997, Turchin 1998, Hanski and Gaggiotti 2004). The theme uniting these approaches is that This article is protected by copyright. All rights reserved 95 dispersal permits populations to recover from local extinction and thus allows persistence 96 even though local populations are ephemeral. 97 A number of modeling frameworks have been used to relate the dispersal kernel (i.e., 98 the distribution of individual dispersal endpoints about a source) to various aspects of spatial 99 population dynamics. Classical reaction-diffusion (RD) models (Skellam 1951, Okubo and 100 Levin 2001) implicitly assume that dispersal distances are normally distributed and show 101 population spread converging to a constant-speed traveling wave, provided Allee effects are 102 absent. Integrodifference equations (IDEs; Kot et al. 1996) are the discrete-time analog of 103 RD models, but more importantly they allow the shape of the dispersal kernel to be specified 104 explicitly. For exponentially bounded kernels such as the normal distribution, IDEs also 105 predict asymptotically constant spread rates; however, leptokurtic or “fat-tailed” kernels 106 (which contain a greater frequency of short distances near the source and long distances in 107 the tails compared to a normal distribution) can produce accelerating, asymptotically 108 unbounded population spread (Kot et al. 1996, 2004). IDEs have seen increasing use among 109 ecologists in response to growing recognition that many organisms have leptokurtic 110 distributions of dispersal distance (Clark et al. 1999) and that rare long-distance dispersal 111 events, often mediated by different mechanisms than local dispersal, can drive patterns of 112 population spread (Kot et al. 1996, Clark 1998, Nathan and Muller-Landau 2000). However, 113 by their very nature such events are difficult to observe and quantify, and thus pose a 114 challenge for parameterizing fat-tailed kernels empirically (Nathan and Muller-Landau 115 2000). Moving from the spread of a single continuous population to the dynamics of patchy 116 population aggregatesAuthor Manuscript can involve stochastic patch occupancy models (SPOMs; Ovaskainen 117 and Hanski 2004), which describe metapopulation dynamics driven by local extinction and 118 migration between patches in a network. Both the colonization and extinction rates in these This article is protected by copyright. All rights reserved 119 models may depend on the probability of immigration from neighboring patches, which is 120 directly related to the dispersal kernel. 121 Here we describe a series of field experiments used to parameterize empirical models 122 spanning three scales in a hierarchy of spatial population dynamics. In the first section 123 (Quantifying local dispersal at the individual scale), we examine local, short distance 124 dispersal over a single generation in which the distribution of offspring can be attributed to 125 known source populations within a few meters. These data allow us to parameterize simple 126 models which are then employed in the second section (Predicting multi-generation spread 127 at the patch scale) to model the growth of discrete patches over multiple generations, and to 128 compare these predictions against independent observations. In the third section (Dispersal 129 limitation and patch turnover