Vol. 554: 1–19, 2016 MARINE ECOLOGY PROGRESS SERIES Published July 28 doi: 10.3354/meps11792 Mar Ecol Prog Ser

OPENPEN ACCESSCCESS FEATURE ARTICLE: REVIEW

A decade of seascape genetics: contributions to basic and applied marine connectivity

Kimberly A. Selkoe1,2,3,*,**, Cassidy C. D’Aloia4,*,**, Eric D. Crandall5, Matthew Iacchei6, Libby Liggins7, Jonathan B. Puritz8, Sophie von der Heyden9, Robert J. Toonen1

1Hawai‘i Institute of Marine Biology, University of Hawai‘i, Kane‘ohe,- HI 97644, USA 2National Center for Ecological Analysis and Synthesis, University of California, Santa Barbara, CA 93101, USA 3Bren School of Environmental Science and Management, University of California, Santa Barbara, CA 93101, USA 4Department of Ecology & Evolutionary Biology, University of Toronto, Toronto, ON M5S 3G5, Canada 5School of Natural Sciences, California State University, Monterey Bay, 100 Campus Center, Seaside, CA 93955, USA 6Department of Oceanography, University of Hawai‘i at Manoa,- Honolulu, HI 96822, USA 7Institute of Natural and Mathematical Sciences, Massey University, Auckland 0745, New Zealand 8Marine Science Center, Northeastern University, Nahant, MA 01945, USA 9Evolutionary Genomics Group, Department of Botany and Zoology, University of Stellenbosch, Private Bag X1, Matieland 7602, South Africa

ABSTRACT: Seascape genetics, a term coined in 2006, is a fast growing area of population genetics that draws on ecology, oceanography and geography to address challenges in basic understanding of mar- ine connectivity and applications to management. We provide an accessible overview of the latest develop- ments in seascape genetics that merge exciting new ideas from the field of marine population connectivity with statistical and technical advances in population genetics. After summarizing the historical context leading to the emergence of seascape genetics, we detail questions and methodological approaches that are evolving the discipline, highlight applications to conservation and management, and conclude with An underwater seascape from the top of Steve’s Bommie in a summary of the field’s transition to seascape ge- the Great Barrier Reef, Australia. nomics. From 100 seascape genetic studies, we assess Photo by Jonathan B. Puritz trends in taxonomic and geographic coverage, sam- pling and statistical design, and dominant seascape drivers. Notably, temperature, oceanography and geo - connectivity and adaptation. The latest studies are graphy show equal prevalence of influence on spatial illuminating differences between demographic, func- genetic patterns, and tests of over 20 other seascape tional and neutral genetic connectivity, and informing factors suggest that a variety of forces impact connec- applications to marine reserve design, fisheries sci- tivity at distinct spatio-temporal scales. A new level of ence and strategies to assess resilience to climate rigor in statistical analysis is critical for disentangling change and other anthropogenic impacts. multiple drivers and spurious effects. Coupled with GIS data and genomic scale sequencing methods, this KEY WORDS: Seascape genetics · Genomics · rigor is taking seascape genetics beyond an initial Connectivity · Marine population genetics · Gene flow · focus on identifying correlations to hypothesis-driven Dispersal · Landscape genetics insights into patterns and processes of population

© The authors 2016. Open Access under Creative Commons by *Corresponding author: [email protected] Attribution Licence. Use, distribution and reproduction are un - **These authors contributed equally to this work restricted. Authors and original publication must be credited. Publisher: Inter-Research · www.int-res.com 2 Mar Ecol Prog Ser 554: 1–19, 2016

INTRODUCTION many concepts and analyses are shared between seascape and landscape versions of this discipline, In 2013, the field of landscape genetics turned seascape genetics has unique challenges that stem 10 years old (Manel & Holderegger 2013). Seascape from the biology of marine taxa and the fluid medium genetics could be viewed as a younger sibling, in which they disperse (Riginos & Liggins 2013, recently passing its own 10 years mark and showing Selkoe et al. 2015). For a majority of marine species, an accelerating rate of publication, enabled in part dispersal is undertaken during a larval phase, and by rapid growth in GIS data and advances in statisti- trajectories of individual larvae, as well as long-term cal methods (Fig. 1) (Galindo et al. 2006, Fontaine et patterns of dispersal, are difficult to predict (Siegel al. 2007, Hansen & Hemmer-Hansen 2007). A major et al. 2008). The seascape’s heterogeneity (e.g. in emphasis for landscape and seascape genetics has bathymetry, current speeds and water chemistry) is been taking an organismal perspective of the dis- largely hidden from our view; little is known about persal process, creating a need for new models and marine organisms’ behavior and ecological interac- methods (Manel et al. 2003, Storfer et al. 2007, tions during the dispersive phase; and even the most Holderegger & Wagner 2008, Balkenhol et al. 2015). complex oceanographic models are limited to coarse Many foundational population genetic and ecologi- scales relative to a larva’s perspective (Levin 2006, cal models assume that spatial heterogeneity is either Woodson & McManus 2007, Cowen & Sponaugle lacking or highly ordered and predictable. The rise of 2009, Morgan et al. 2014, Mazzuco et al. 2015, Nick- landscape and seascape genetics as named disci- ols et al. 2015). Consequently, the organism’s view of plines reflects interest in overcoming these simplify- the dispersal process is easier to understand and ing assumptions to understand how movement of an quantify in terrestrial settings compared to marine organism through the landscape or seascape impacts systems (Marshall & Morgan 2011, Burgess et al. realized dispersal and gene flow. These subdisci- 2015). In some ways, these challenges and unknowns plines distinguish themselves within the field of of the marine environment increase the potential population genetics by their focus on spatially contribution of seascape genetics to our understand- explicit processes and statistics (Dyer 2015), and a ing of marine population connectivity. focus on the linkages between environment, ecology and genetics. The topic resonates widely, in part because identifying landscape features critical to the THE RISE OF A SEASCAPE PERSPECTIVE persistence and adaptation of populations is central to effective strategies for conservation and resource Genetic tools have long played an instrumental management (Wagner & Fortin 2013). role in the study of marine connectivity (e.g. Berger Our focus here is on detailing the approaches and 1973, Waples 1987, Hellberg 1994, Palumbi 1996) findings of seascape genetics for a wide audience because they promise insights into the scale of mar- interested in the field of marine connectivity. While ine larval dispersal that are simply not possible with other natural or artificial tags (Berumen et al. 2010).

25 While this promise has not always been realized (Waples 1998), population genetic approaches have enabled tests of specific hypotheses about spatial 20 patterns of exchange (e.g. hierarchical or continuous) and drivers of exchange (e.g. whether a biogeo- 15 graphic boundary serves as a boundary to gene flow). Trends in hypotheses and study designs have 10 been shaped by recent changes in our understanding of marine connectivity, as well as an increasing abun-

# papers published dance and variety of genetic and environmental data. 5 At the turn of the century, the paradigm of open populations and uniform larval pools was beginning 0 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 to erode (Jones et al. 1999, Swearer et al. 1999, Kin- lan & Gaines 2003, Palumbi 2003). Diverse evidence Fig. 1. Seascape genetic papers published between 2006 and 2015. See the Supplement at www.int-res.com/articles/ solidified a new focus on complex dispersal kernels suppl/m554p001_supp.pdf for details on paper collection (Siegel et al. 2003), local retention (Swearer et al. and the full reference list 2002) and quantifying genetic variation across larval Selkoe et al.: A decade of seascape genetics 3

cohorts (Hedgecock 1994, Selkoe et al. 2006). These example, some studies were misleading because they seminal works supported the emergence of seascape interpreted genetic metrics to solely indicate differ- genetics by motivating the exploration of the patterns ences in dispersal when other processes may con- and drivers of nuanced spatial genetic variation tribute to spatial patterns of allele frequencies (e.g. across populations, recruits and larval samples. range expansions, see Hart & Marko 2010, Marko & Following the first published use of the term in Hart 2011). Others mistakenly assumed very high 2006, an early priority of seascape genetics was to test rates of larval exchange and long-distance dispersal the role of ocean currents in generating spatial ge- because of insufficient statistical power to detect netic patterns (e.g. Baums et al. 2006, Galindo et al. spatial genetic structure, or methods that are inap- 2006, reviewed in Liggins et al. 2013). More recently, propriate for the data (as discussed in Meirmans seascape genetics has expanded to embrace a rich 2014, Richardson et al. 2016). Although genome se- suite of potential ecological and genetic forces dis- quencing advances are helping to address low power cernable by spatial analysis of genetic data when by increasing locus sample size (Gagnaire et al. used in analytical frameworks borrowed from land- 2015), simply adding more genetic loci will not neces- scape genetics (Fig. 2). This more holistic ap proach sarily improve inference of connectivity without a will aid the field of marine connectivity in continued change in analytical philosophy (Ryman et al. 2006, efforts to disentangle the relative effects of ocean cur- Larsson et al. 2009, Bowen et al. 2014). rents, larval behavior and processes occurring at set- Landscape and seascape genetics have catalyzed a tlement and recruitment in driving patterns of popu- shift from describing patterns of genetic structure in a lation exchange. Current genetic technology enables null-hypothesis testing framework to distinguishing a range of spatio-temporal scales of inference on the many forces contributing to observed spatial pat- marine connectivity, including estimates of single terns, including interactions between ecology and ge- dispersal events, recent and historical migration rates, netics, in a formal model selection framework (Manel hybridization and speciation (see Box 1). The multi- & Holderegger 2013, Wagner & Fortin 2013). While faceted perspective on spatial genetic structure many studies related seascape features to population summarized in Fig. 2 has helped the field of marine genetic data before the term ‘seascape genetics’ was population genetics overcome a few key pitfalls. For coined (e.g. Johnson & Black 1991, Riginos & Nach- man 2001, Gilg & Hilbish 2003), there are • Reproductive strategy • OnOntogenictogenic a variety of new and newly popularized hhabitatabitat shishiftf GGeographiceo distance multivariate statistical ap proaches to OOceanographicce distance • LarvalLarval beh HabitatHab configuration integrating diverse data types that are • DispersalDispersal scale BBarriersar to dispersal rapidly bringing the field of empirical • Larval productionction rrateate EnvEnvironmental gradients population genetics beyond an explor- • Recruitment ratratee atory era (reviewed by Balkenhol et al. • Population sizeze • Age structure Disturbance events 2015, Rellstab et al. 2015). At the same • Metapopula Colonization history time, evolutionary geneticists have un- • Historical gene flow • Genome covered strong linkages between rates of • Mutation rat • Climatic changes selection, drift and gene flow that likely • Recent human • Sampling iinfluencenf impact connectivity inference, and dem - • Spatial scale • Measurementt eerrorrror onstrated how selection can strongly af- fect neutral genomic regions (reviewed • SSamplingampl by Nosil et al. 2009, Orsini et al. 2013). • Evolutionary stochasticity • Species interactions The rising number of seascape genetic • Homoplasic & null mutations • Habitat structure studies using single nucleotide polymor- • Physical context (e.g. depth and slope) • Site productivity and diversity phisms (SNPs) sampled across the • Local selective pressures genome is improving our ability to ex- plore these linkages (Fig. 3). Compared Fig. 2. Spatial genetic structure (SGS) of marine populations is the com- bined result of a suite of interacting forces and traits that can be categorized to seascape genetics, seascape genomics into topics (listed on sea star). These forces include demography, species’ creates even greater imperative for life history traits, rates of migration influenced by spatial factors, lingering grounding analyses in model selection to signals of history, influences of local ecology and/or local adaptation, some minimize spurious effects and ensure in- degree of noise and study design factors. Seascape genetics focuses on un- covering support for effects of these forces on SGS. Examples of possible terpretability of findings (Meirmans 2015, predictors of these varied effects are listed adjacent to each ray Riginos et al. 2016). 4 Mar Ecol Prog Ser 554: 1–19, 2016

Box 1. Defining marine connectivity across spatial and temporal scales. A variety of terms distinguish key agents, time-scales and viewpoints of marine connectivity, and collectively contribute to a holistic understanding of connectivity

Alternative definitions of connectivity population-level metric that reflects the cumulative impact of functional connectivity on allele frequencies in a popu- In its broadest sense, ‘connectivity’ refers to any rela- lation over one or more generations. tionship between spatially or temporally distinct entities Adaptive genetic connectivity measures movement of (Kool et al. 2013). The field of marine connectivity largely alleles with consequence to individual fitness. It can focuses on the materials and processes that connect eco- diverge greatly from neutral genetic connectivity and logical communities across space and those that disrupt or relates to the adaptive potential of a population. bound connectivity. Neutral genetic connectivity measures movement of Structural connectivity quantifies the physical relation- alleles that have no consequence to fitness. Neutral con- ships of landscape elements, including spatial positioning nectivity estimates can be related to the long-term out- of habitat (e.g. Kool et al. 2013), geo-morphological fea- come of demographic connectivity (Funk et al. 2012). tures of the seafloor and/or hydrodynamic flow features impacting movements, pathways and boundaries. Struc- tural connectivity creates potential for biological connec- Relating genetics to connectivity tivity. Functional connectivity is a product of the organism Typically, seascape genetic studies estimate metrics of interacting with the landscape. It concerns the patterns genetic connectivity; however, some methods may enable and rates of dispersal that result from the response of indi- the measurement of demographic connectivity (see Box 2) viduals to the structural matrix, mediated by behavioral (Lowe & Allendorf 2010). ‘Genetic tagging’ studies that traits and dispersal success. Questions of functional con- distinguish individuals and kin using genetic markers are nectivity focus on whether landscape elements impede or used to discover dispersal events. Most other uses of facilitate dispersal, given the species’ traits. Biophysical genetic markers use indirect inference to estimate genetic oceanographic modeling of larval dispersal is commonly connectivity. Indirect inference depends on theoretical used to predict functional connectivity (e.g. Treml et al. population genetic models with simplifying assumptions 2008, 2015, Kool et al. 2011). about possible effects of mutation, selection and drift on Demographic connectivity relates the dispersal of indi- genetic patterns to be able to infer migration (Waples & viduals to their cumulative effects on population-level pro- Gaggiotti 2006). Estimators based on Wright's fixation cesses such as population growth, extinction or recoloniza- index (FST) integrate over longer time-scales, possibly tion. Thus, measures require estimates of demographic representing historical connectivity more than present day rates and the relative contributions of self-recruitment and because of response lags dependent on rates of marker immigration to these rates (see Lowe & Allendorf 2010). mutation and drift (Epps & Keyghobadi 2015). Landscape Genetic connectivity tracks the dispersal of genes (and and seascape genetic studies commonly aim to estimate the genomes), which only accounts for those individuals that correlations among variables representing different types successfully reproduce after dispersing. Genetic connec- of connectivity. This strategy can identify important physi- tivity is the opposite of reproductive isolation. Like demo- cal features that influence multiple types of connectivity, or graphic connectivity, genetic connectivity is inherently a important ways that the types of connectivity diverge.

APPROACHES AND STUDY DESIGNS 2014). Choice of genetic metrics should be guided by hypotheses about mechanisms generating the correl- There are almost always multiple potential causes ative patterns tested, including effects on genetic of genetic discontinuities (i.e. spatial shifts in allele links, nodes, boundaries and neighborhoods, as de - frequencies or diversity). These alternative factors scribed in Wagner & Fortin (2013). These 4 terms are have often been treated in a post hoc or qualitative defined below in sections that expand on their dis- way. Seascape genetics provides the tools and per- tinct value and applications in marine settings. spective needed to design a study a priori to identify the drivers of genetic discontinuities, with a suite of complementary methodologies (see Box 2). All sea- Links: parsing drivers of genetic exchange scape genetic study designs should assess spatial co- variation of seascape factors of interest, so that sites Link-based designs, which make up the majority of for genetic sampling are chosen to maximally repre- empirical studies, focus on pairwise genetic distance sent the full range of these gradients with minimal and address how seascape features either create confounding (Wang & Bradburd 2014, Riginos et al. discontinuities or promote gene flow between sites. 2016). Individual-based sampling designs may better Within a rigorous alternative hypothesis testing frame- capture these gradients compared to highly clustered work, multiple coincident or competing factors, e.g. ‘population-based’ sampling, especially when many habitat patchiness, environmental gradients and/or loci are used (Prunier et al. 2013, Wang & Bradburd oceanographic flow fields, can be disentangled. Mar- Selkoe et al.: A decade of seascape genetics 5

cent model revealed the effects of stepping-stone 14 versus long-distance dispersal in generating high gene flow. Liggins et al. (2016) used a multiple- 12 regression model-testing approach to show that spatial genetic patterns of different species could be 10 attributed to the same seascape features despite little congruence in spatial genetic patterns across spe- 8 cies. These and other studies provide important tests of existing theories, such as the role of upwelling, 6 headlands and embayments in local retention of larvae (e.g. Mace & Morgan 2006, Banks et al. 2007,

# studies per year 4 Lotterhos & Markel 2012, Iacchei et al. 2013, Pfaff et al. 2015). In sum, seascape studies are increasingly able to overcome the obstacle of weak genetic signal 2 to help push forward concepts about the possible mechanisms influencing larval production, emigra- 0 tion, dispersal and settlement. 2006 2008 2010 2012 2014 Exploring the relationship between ocean currents Marker Type: and link-based genetic data has been a cornerstone Microsatellites SNPs Other of seascape genetics since its inception (see Liggins Multiple Types mtDNA et al. 2013 for an overview). Estimates of larval Fig. 3. Trends in marker usage over time. Microsatellites exchange from biophysical transport models can be are the most common marker type used, but single nucleo- directly compared to observed genetic patterns, or tide polymorphism (SNP) studies are rising. Multiple first incorporated into population genetic simulations Types: studies that use more than one marker type, with that are then compared to empirical genetic data. the most common combination being microsatellites + mtDNA. Other: marker types used infrequently in sea- The latter is especially valuable for illuminating scape genetic studies (e.g. inter simple sequence repeats, how multi-generational, often asymmetrical stepping- allozymes, single nuclear sequences) stone dispersal can lead to high genetic connectivity despite a lack of direct larval exchange (Treml et al. 2008, Kool et al. 2010, White et al. 2010, Crandall et ine populations often have large effective population al. 2012, Foster et al. 2012, Munguia-Vega et al. sizes and relatively high rates of gene flow, resulting 2014). These simulations can also explore the effects in greater difficulty to assess population structure of spawning, life history and behavioral traits on than in terrestrial systems due to low genetic differen- genetic connectivity and source–sink metapopula- tiation between populations (Mills & Allendorf 1996, tion dynamics (Galindo et al. 2006, Lee et al. 2013, Waples 1998). Sometimes, low differentiation is actu- Young et al. 2015, Selkoe et al. 2016). Continued ally caused or enhanced by high mutation rates and expansion of oceanographically informed genetic extreme marker heterozygosity that mask genetic simulations will help elucidate why and when gene distinction among populations, so interpreting levels flow patterns are expected to diverge from transport of differentiation requires careful contextualization patterns. They can also uncover the limitations of (Hellberg 2007, Bird et al. 2011, Gagnaire et al. available oceanographic data (e.g. coarse spatial res- 2015). Importantly, recent seascape genetics studies olution, short time series, missing nearshore features) demonstrate that despite genetic estimates and ‘iso- that interfere with comparisons between empirical lation by distance’ (IBD) tests that are not statistically genetic data and biophysical dispersal estimates distinguishable from zero, significant spatial pattern- (Pineda et al. 2007, Pringle & Wares 2007). ing in differentiation is uncovered when seascape features are used (1) to inform appropriate null hypotheses, (2) as parameters in a Bayesian frame- Nodes: local effects on spatial genetic patterns work, or (3) as predictors in model testing. For exam- ple, Crandall et al. (2012) showed that despite very Node-based studies use metrics that characterize low genetic structure for 4 neritid snails sampled a location, rather than pairwise differences between throughout the Pacific, incorporating gene flow pre- locations. Node-based genetic metrics typically de- dictions from oceanography into a Bayesian coales- scribe how allele frequencies or diversity of a site 6 Mar Ecol Prog Ser 554: 1–19, 2016

Box 2. Advances in methodology. We highlight select techniques that are bringing rigor and creativity to hypothesis formation and data analysis

Kinship analyses: relating dispersal to gene flow summary statistics of the species' valaues (Hickerson et al. 2006, Hickerson & Meyer 2008). Coalescent approaches Kinship analyses generate direct estimates of dispersal offer new opportunities for community-level seascape distances, in contrast to indirect inference based on popu- genetics, but their application re quires considerable com- lation-level genetic data (see Box 1). Dispersal distances puting resources, informed model design and careful can be reconstructed by linking individuals’ multi-locus investigation of outputs. With these caveats, coalescent genotypes to natal populations (Hogan et al. 2012), parents analyses provide the opportunity to wring far more infor- (Christie 2010) or sibling groups (Schunter et al. 2014). The mation from hard won data than traditional population most popular kinship approach is parentage analysis in which pools of potential parents and offspring are geno- genetic statistics. typed to assign offspring to parents (Christie 2013, D’Aloia et al. 2015). Assuming minimal post-settlement movement, the distance between parent and offspring is the net dis- Analysis of pairwise data: moving beyond Mantel persal distance. Parentage studies can describe the magni- tude and direction of dispersal between sites (Jones et al. Given the prevalence of link-based studies, a key statis- 2005, Planes et al. 2009, Harrison et al. 2012), and with tical issue is the treatment of non-independence for pair- intense sampling can estimate dispersal kernels (Almany wise data. After decades of widespread use, the Mantel et al. 2013, D’Aloia et al. 2015). The emerging theme is that test was recently shown to be an inappropriate way to most marine larvae disperse shorter distances than address non-independence for most population genetic inferred from estimates of genetic structure, supporting a data (Legendre et al. 2015). We estimated that 63% of sea- conclusion from the larval biology literature that long- scape genetic studies have used Mantel tests, begging for distance marine dispersal may be maladaptive (e.g. Strath- an assessment of how our understanding of isolation by mann et al. 1981, Burgess et al. 2012). Importantly, the dis- distance in the sea may be biased. Importantly, extensions crepancy between direct and indirect dispersal estimates of the Mantel, such as multiple regression on distance may result from the fact that genetic structure is affected by multiple generations of gene flow, and is influenced by matrices, may also be prone to biased model selection rare long-distance gene flow events that parentage meth- (Goldberg & Waits 2010, van Strien et al. 2012, Guillot & ods do not capture. Comparison of kinship-derived disper- Rousset 2013). In a recent review, Rellstab et al. (2015) sal kernels to genetic structure estimates will hone our suggested that valid inference can still be obtained within conceptual understanding of how to relate dispersal to the Mantel framework if effect sizes are used to interpret various connectivity metrics, given their distinct temporal results, or if a rank-based partial Mantel test is utilized, as and spatial scales. in Bayenv2 (Günther & Coop 2013). Alternatively, instead

of using summary statistics like FST, populations’ allele fre- quencies can be modeled as a linear function of environ- Coalescent techniques: explicit treatment of history mental factors in a multiple linear regression framework, or with multivariate frameworks like canonical correspon- Coalescent approaches consider demography and dence analysis (CCA) or redundancy analysis (RDA) meta population history separately (Rosenberg & Nord- borg 2002), and thus can overcome the simplifying (Legendre & Legendre 2012). assumptions necessary for estimating gene flow from F- statistics (Whitlock & McCauley 1999). Coalescent models can specify directional and asymmetric gene flow, distinct Future directions for increasingly complex datasets population sizes and the timing of population divergence (Hey & Machado 2003). By estimating effective popula- Mixed models, such as maximum-likelihood population- tion size and proportion of migrants per generation sepa- effects (MLPE) models, may offer a robust way to evaluate rately, coalescent models are able to estimate rates of predictors of pairwise population genetic data (van Strien gene flow at much higher levels than F-statistics, and et al. 2012, Row et al. 2015, reviewed in Rellstab et al. 2015). with appropriate estimates of uncertainty. Coalescent Mixed models treat sampling location as a random effect, samplers such as Migrate or BEAST calculate a marginal and they can isolate neutral genetic structure when testing likelihood for any specified coalescent model, which may for the influence of environmental factors on allele frequen- then be compared to other modeled hypotheses using cies (e.g. latent factor mixed models, Frichot et al. 2013) standard information theoretic methods (Beerli & Pal- New techniques based on Bayesian modeling of allele co- czewski 2010, Baele et al. 2012). These methods have variance structures, eigenvector mapping and others are been used to test hypotheses generated from biogeogra- phy (Teske et al. 2011), biophysical dispersal models also gaining traction (reviewed in Wang & Bradburd 2014, (Crandall et al. 2012) and mark−recapture data (Jue et al. Gautier 2015). Whether you use well-known or cutting- 2015). Approximate Bayesian computation (ABC) meth- edge statistics, we urge researchers to carefully consider ods applied to coalescent modeling can handle hierarchi- model assumptions before applying them to seascape ge- cal models, wherein parameter estimates for individual netic data sets, to test multiple approaches for concordant species are nested within hyperparameters based on results and to clearly describe the methods they adopt. Selkoe et al.: A decade of seascape genetics 7

differs from the mean of all sites, and node-based defining fisheries management units (Roy et al. seascape metrics assign values to each sampling site. 2012). An important study design issue for boundary- When the seascape is hypothesized to have effects on level analyses is determining whether spatial gradi- current population size, rates of drift, local selection ents in genetic patterns occur in tandem with bound- and historical events, node-based studies may be aries in genetic patterns, as gradients can affect the most appropriate and effective for analysis. So far, output of barrier detection and clustering algorithms node-based studies are less common, but arguably (Manni et al. 2004, Meirmans 2012). they can be more powerful and reliable, as node- based genetic metrics often have less sampling error than most link-based metrics, and statistical testing is Neighborhoods: defining local scales comparatively unproblematic. A rigorous method for node-based analyses is a Bayesian approach to asso- Neighborhood-based studies focus on how the ciating local (i.e. site-specific) estimates of FST with local seascape and habitat array influence the scale node-based predictors in the software GESTE (Foll & of connectivity or, more generally, the scale of covari- Gaggiotti 2006). Arizmendi-Mejía et al. (2015) used ation in genetic metrics across samples or patches local FST and other node-based metrics to test the (Wagner & Fortin 2013). Investigations can take a interplay between ecological and evolutionary fac- variety of formats and use an assortment of metrics. tors in a Mediterranean gorgonian and define con- For example, D’Aloia et al. (2015) compared kinship- servation units. The results showed genetic drift derived dispersal kernels to microhabitat character- linked to partial mortality to be a primary driver of istics (e.g. depth) and local context (e.g. directional- genetic structuring. Describing seascape effects on ity, latitude) to determine whether local features patterns of local genetic diversity also fits a node- predicted the spatial scale of dispersal for a coral reef based design. For example, Liggins et al. (2015) goby within a 41 km neighborhood. Munguia-Vega assessed distributions of shared and private haplo- et al. (2014) explored a variety of network metrics types using measures of nestedness and partitioned derived from a biophysical dispersal model for leo- beta-diversity to support the hypothesis that the pard grouper in the Gulf of California and found that southern limit of a damselfish’s range is a demo- the topographic relationships of sites (i.e. the neigh- graphic ‘sink.’ borhood layout) is more critical than the rate of ex- change among sites for explaining genetic patterns. Employing biophysical models or simulated gene Boundaries: uncovering lines in the sea flow outputs is a critical means of understanding the influence of the spatial configuration of habitat Boundary-based studies define genetic groups patches, spawning aggregations or marine reserves across space and relate boundaries between groups on connectivity to inform marine management and to seascape features. Understanding the forces creat- parameterize metapopulation models (e.g. Crandall ing genetic boundaries informs fundamental ques- et al. 2012, Wood et al. 2014, Davies et al. 2015). tions in biogeography, evolution and population dynamics as well as marine management. The long history of statistical approaches to boundary detec- Multi-approach studies: relating pattern and process tion has resulted in a range of user-friendly software (e.g. ARLEQUIN, STRUCTURE and BARRIER), but Including more than one type of analysis is often relatively few studies have taken the next step in sta- required for robust interpretation of correlative stud- tistically testing for the relationship be tween bound- ies. Ecologically, feedbacks exist between processes aries and relevant seascape features. These few stud- that take place at nodes, neighborhoods, links and ies report temperature, salinity, irradiance, turbidity, boundaries. For example, Dawson et al. (2014) found depth and sediment type as significant drivers (e.g. that differences in population genetic structure González-Wangüemert et al. 2009, Roy et al. 2012, among intertidal species of the eastern North Pacific Viricel & Rosel 2014, Johansson et al. 2015). For were due to the combination of 2 node-based met- example, a study of white hake in the Northwest rics—census size and fecundity of populations—and Atlantic found that depth was the only significant species’ pelagic larval duration, a link-based metric environmental predictor of genetic group member- representing relative rates of migration. Importantly, ship in a highly fished population across time, so the pairwise genetic differentiation will be amplified if authors recommended a depth-based approach to there are large differences in the genetic diversities 8 Mar Ecol Prog Ser 554: 1–19, 2016

of the focal populations, thus intertwining node- Shafer & Wolf 2013, Wang & Bradburd 2014). Com- based effects into a link-based analysis. Selkoe et al. parative analyses with neutral and non-neutral pan- (2010) found that differences in local genetic diver- els of SNPs can elucidate the processes driving IBE sity primarily drove pairwise genetic structuring for patterns. For instance, Tepolt & Palumbi (2015) com- temperate reef species in Southern California. In pared SNPs from the neutral genome and the cardiac both these study systems, recolonization following transcriptome to address questions about the roles of disturbance likely impacts standing genetic diversity colonization bottlenecks and local temperature adap- (although with distinct mechanisms and time-scales). tation in both the native and expanding ranges of an Because recolonization is influenced by life history invasive green crab. Marine populations may com- adaptations to environmental regimes, we can posit monly harbor unique adaptations associated with that local selection may indirectly contribute to these oceanographic features, temperature, productivity or putatively neutral genetic patterns. more subtle features like sand inundation (Teske et Selection is often considered separately from genetic al. 2011, Hecht et al. 2015, Young et al. 2015). drift and gene flow. Importantly, however, the 3 Increasingly, evidence is accumulating for genetic forces are linked: strong selection can enhance rates impacts of anthropogenic threats to marine systems of drift by disrupting gene flow and increasing isola- (Puritz & Toonen 2011). Seascape genetic approaches tion (Nosil et al. 2008, Bird et al. 2012). The large can elucidate these spatial patterns and drivers of effective population sizes typical of many marine adaptation, which in turn can have important impli- species generally allow selection to have a large cations for marine conservation (see Box 3 for exam- influence on the genome relative to drift, yet the role ple applications to marine management efforts). of selection has historically been underappreciated in marine genetic studies (Allendorf et al. 2010). One example of how selection indirectly affects neutral SURVEY OF SEASCAPE GENETIC STUDIES genetics is termed monopolization. Namely, a posi- tive feedback between priority effects and local As the number of seascape genetic studies has adaptation during colonization events disrupts the grown, we are gaining insight into the most common influence of oceanographic connectivity on neutral drivers of genetic structuring. To foster a synoptic population genetic structure by reducing the success perspective on the direction that the field has taken of later immigrants (De Meester et al. 2002, Waters et so far and its potential insights, we conducted a liter- al. 2013, Fraser et al. 2015). Drawing from the previ- ature search and retained 100 empirical studies from ous example, Selkoe et al. (2010) found that genetic 2006 to 2015 that self-identified with ‘seascape’ or diversity of territorial predators like kelp bass and ‘marine landscape’ genetics (see the Supplement at lobster scaled inversely to kelp forest size. These spe- www.int-res.com/articles/suppl/m554p001_supp. pdf cies cannibalize and exclude new recruits, and their for details). From this survey, we assessed the taxo- territory size expands as they age and grow. With nomic and geographic coverage in the first genera- recolonization of a reef following storm disturbance, tion of seascape genetic studies. The vast majority of the expanding size of territories and rate of cannibal- studies focused on chordates (primarily fishes) (48%) ism may contribute to a winnowing of genetic diver- and invertebrates (38%) as study organisms (Fig. 4a). sity in the site’s population, even as the kelp bed Far less attention has been paid to the seascape itself expands and adds new habitat (Selkoe et al. genetics of algae (9%), marine angiosperms (5%) 2010). Thus, an ecological process connected to life and other marine phyla over the past 10 yr. A bias history and recolonization may indirectly influence towards temperate waters (68%) is partly driven by neutral genetic diversity. intense study of temperate diadromous fishery Many studies focus on detecting direct effects of species (despite likely underrepresentation of dia- selection by assessing ‘isolation by environment’ dromous studies with our search terms) (Fig. 4b). (IBE), a pattern of positive correlation between Tropical studies (26%) concentrated on coral reef genetic metrics and environmental factors, both abi- seascapes, and polar studies are still rare (6%). Most otic and biotic (reviewed by Wang & Bradburd 2014). studies were subtidal (65%) (Fig. 4c); fewer studies Related conceptualizations of IBE that focus on pro- focused on intertidal (15%) or estuarine (14%) envi- cess include ‘isolation by adaptation’ (i.e. due to ronments, and the fewest focused on pelagic ecosys- adaptive phenotypic divergence) and ‘isolation by tems (5%), perhaps due to the mobile lifestyle of ecology’ (i.e. due to selective reduction in gene flow many pelagic species. Although seascape genetic resulting from fitness differentials) (Nosil et al. 2008, studies have occurred in most of the world’s major Selkoe et al.: A decade of seascape genetics 9

Box 3. Applied seascape genetics: building roadmaps for conservation and management. We highlight 3 arenas where seascape genetics brings unique value to conservation and management of marine populations and communities

Species-based management Hawai‘i, an analysis of mitochondrial genetic diversity in approximately 40 reef species identified large reef size Demographic parameter estimates such as lifetime and high coral cover as key predictors of high diversity, fecundity, self-recruitment and age-specific survival rates and these features could be used to select areas for long- can all be generated from neutral genetic data, and are term protection (Selkoe et al. 2016). Genetic diversity and critical to marine reserve success (Mace & Morgan 2006, species diversity often co-vary, indicating that ecological Burgess et al. 2014, Bonin et al. 2016). Seascape genetics resilience and genetic resilience are linked (Messmer et al. can specifically identify ecological and biophysical attrib- 2012, Wright et al. 2015, Selkoe et al. 2016). Syntheses of utes associated with high spill-over from marine reserves, genomic data across multiple species have the potential to asymmetries in source−sink dynamics within reserve net- identify suites of species with similar local adaptations that works (Harrison et al. 2012, Young et al. 2015), and the could lead to more nuanced delineation of conservation potential positive and negative effects of altered gene flow units and habitat restoration plans (Funk et al. 2012). In the (Richardson et al. 2016). Locations can be characterized by future, using genetic tools to monitor biodiversity and key attributes such as high centrality or high outgoing con- measure resilience holds promise to rapidly advance mar- nectivity in prioritization algorithms for network design ine conservation and restoration (e.g. Port et al. 2016). (Beger et al. 2014, Mendez et al. 2014). Understanding the spatial distribution of genomic variation and drivers of local adaptation is critical to the protection of genetic Climate adaptation resources and population resilience of fisheries (e.g. Limborg et al. 2012). Pinsky & Palumbi (2014) showed a Climate-induced range shifts of individual species and decline in genetic diversity for numerous exploited marine communities are becoming well documented (Beaugrand fishes, attributed to over-harvesting. These trends imply 2009). Studies are starting to investigate how these shifts widespread fisheries-induced evolution, and underscore impact connectivity between populations (Andrello et al. the idea that spatial patterns of human impacts on sea- 2015, Young et al. 2015) and reshape the overall genetic scapes can strongly influence genetic diversity and long- structuring of species (Cossu et al. 2015), thereby impact- term resilience. ing conservation features (McLeod et al. 2009). Seascape genetics can identify climate variables with the greatest influence on both neutral and functional regions of the Biodiversity conservation genome. For example, an apparent negative correlation between temperature stress and mean genetic diversity of When the target of management is protecting intact Hawaiian reef species highlights the potential for global habitats and biodiverse areas, a community genetics warming to compromise the genetic resilience of entire approach to integrate genetic data across species is key. communities (Selkoe et al. 2016). A seascape genetics per- Bayesian modeling can leverage multiple studies to assess spective can inform strategies for genotype-based coral regional patterns in genetic diversity across many taxa by restoration using trans-generation acclimatization, selec- accounting for spatial effects, species traits and disparate tion of either host or symbiont genotypes (van Oppen et al. sampling designs (Pope et al. 2015). Over the long term, 2015), and targeted gene flow that aims to increase adap- marine managers must consider how to maintain ecologi- tive variation and fitness in populations (Kelly & Phillips cal resilience in an uncertain future. Identifying seascape 2016). As genomic data become more widely available, features that promote persistence and diversity can inform incorporating ecological niche models and projections of this goal. Neutral genetic diversity indicates a population’s climate-induced environmental extremes into genetic potential to resist inbreeding depression, recover from dis- studies will help identify marine populations most likely turbances and preserve adaptive potential (Young et al. to persist under future conditions (Sgrò et al. 2011, Fitz- 1996, Hughes & Stachowicz 2004, Reusch et al. 2005). In patrick & Keller 2015, Razgour 2015). oceans and seas (Fig. 4d), a bias exists towards the taxon, life history, geography and scale. Neverthe- Northern Hemisphere, and underrepresented regions less, these studies provide a striking suggestion that remain (e.g. the Southern Atlantic). seascape effects on genetics are varied and ubiqui- tous. Link-based study designs were most prevalent. A total of 43 studies with link-based designs tested Emerging patterns in seascape effects the influence of 2 to 9 seascape factors against pair- wise genetic distance metrics. Considering each Using a sub-sample of 53 studies that tested for unique combination of species, marker type, spatial multiple seascape predictors of spatial genetic pat- scale, statistical test and genetic response variable, terns (see the Supplement), we examined which study these 43 studies included 66 tests. Nearly two-thirds designs and seascape factors have received the most of these tests were partial Mantel or multiple regres- attention by the field. This sample size was insuffi- sion on distance matrices, which recent work has cient to comment on how seascape drivers differ by shown to lead to inflated false positives (Legendre et 10 Mar Ecol Prog Ser 554: 1–19, 2016

a) Taxon 50

20

# of studies 10

0 Chordata Mollusca Cnidaria Echino- Arthropoda Porifera Annelida Hetero- Archae- Angio- dermata kontophyta plastida spermae

b) Global position c) Ecosystem

60 60

40 40 # of studies # of studies 20 20 Fig. 4. Taxonomic and geographic breadth of seascape genetic studies. (a) Taxonomy reveals a focus on 0 0 chordates (mainly fish) and inverte- Temperate Tropical Polar/ Subtidal Intertidal Diadra- Pelagic or brate phyla. Note the break in the Subpolar mous Estuarine y-axis. (b) Global position in terms d) Oceans and major seas of latitude reveals a bias towards temperate regions and underrepre- sentation of polar and subpolar re- gions. (c) Ecosystem type reveals a 20 focus on subtidal ecosystems (but note our search terms may under- represent diadromous studies; see 10 the Supplement at www.int-res.com/ articles/ suppl/m554 p001_ supp.pdf). 5 (d) Major oceans and seas, with cir- 1 cle size proportional to number of studies in each body of water Marine

Diadromous al. 2015). With this caveat in mind, we report that for Nearly all link-based studies included geographic 25 types of seascape predictors, 20 were found to be distance as a predictor and nearly one-half of these a significant driver in at least one test. This sample is reported significance. For comparison, a previous re- far from random—bias towards positive results is view of IBD in marine studies found that one-third of expected from informed choice of important sea- sampled studies showed significant IBD; however, scape drivers, but biased reporting and false posi- study selection criteria differed (Selkoe & Toonen tives are also possible. We note that most studies 2011). Several software programs such as Bayenv focused on abiotic predictors; only a handful of biotic (Coop et al. 2010) assume and adjust for IBD before factors were apparent in the literature survey (e.g. testing effects of other factors. Given that only half of ‘chlorophyll a’ and ‘presence of vegetation’). A vast seascape studies find significant spatial autocorrela- array of ecological interactions likely also play a role tion in the genetic response variable, careful choice in shaping marine species distributions and deserve and appropriate use of software is critical (i.e. only ad- greater focus. just for IBD if it is present). For the 32 link-based stud- Selkoe et al.: A decade of seascape genetics 11

ies in our survey that lacked evidence of geographical A study’s selection of seascape predictors and influence, half reported a significant influence of at interpretation of findings depends on consideration least one other seascape factor. Arguably, this indi- of how life history, ecology and environment influ- cates that seascape effects may account for a substan- ence gene flow. Our sample of multi-factor studies tial number of cases of what is commonly called showed that the most common seascape factors ‘chaotic genetic heterogeneity’—a lack of genetic tested were metrics of temperature and ocean trans- spatial autocorrelation despite significant genetic port; 43% of tests of temperature were reported to be structuring (Johnson & Black 1982, Toonen & Gros- significant, while only 31% of ocean transport met- berg 2011). The flipside of this view is that the cause rics were significant (Table 1). These results suggest of spatial genetic structure in roughly one-quarter of that temperature may be as influential as geography marine studies was unresolved, suggesting that there on regional scale population genetics of marine spe- is still much to be learned about the drivers of genetic cies, and more so than ocean currents. However, it structuring in the sea. Addressing the role of time may also be either easier or more common to design alongside space may be key to gains in this arena. For effective sampling arrays across temperature gradi- instance, by analyzing 3 successive years and over ents than across ocean transport gradients. A half- 2000 fish, R. Henriques et al. (unpubl.) show that dozen other categories of seascape factors were anomalous low oxygen water events contribute to tested 10 to 25 times across the sample of studies, and chaotic genetic patchiness of hake in the southern all showed significant results in 40 to 80% of the tests Benguela system. Most seascape studies encompass (Table 1). In light of these trends, the idea of using only 1 sampling event and so cannot illuminate how patterns of potential connectivity based on ocean natural fluctuations in seascape drivers and genetic flows as an informed null model may be no more patterns may result in unexplained spatial genetic defensible than using geography, depth or tempera- variance. ture. Some of the driver metrics, such as habitat

Table 1. Number and outcome of tests that evaluated each type of driver studied, taken from 53 studies that statistically tested correlation of genetic metrics with at least 2 seascape drivers

Predictor All studiesLink-based Node-based Tests Significant (%) Tests Significant (%) Tests Significant (%)

Geography 86 40 62 48 22 18 Temperature 47 43 26 54 21 29 Ocean transport 35 31 26 27 9 44 Habitat patch size 25 48 10 60 15 40 Depth 23 65 15 73 8 50 Other climatic variables 18 50 10 70 8 25 Salinity 16 69 13 69 3 67 Precipitation 15 80 7 100 8 63 Biogeographic breaks/ecoregions 14 43 12 50 0 0 Chlorophyll a 11 45 10 50 1 0 Habitat continuity 9 89 9 89 0 0 Turbidity 8 75 6 67 2 100 Nutrients 7 25 6 33 2 0 Sediment type 7 86 4 100 3 67 Hydrodynamic regime 6 33 3 33 3 33 Habitat exploitation 5 20 1 0 4 25 Dissolved matter 4 50 3 67 1 0 Other hydrodynamic variables 4 100 3 100 1 100 Geological history (e.g. glaciation) 3 33 1 100 2 0 Oxygen 3 67 3 67 0 0 Solar irradiance 3 67 2 50 1 100 Tidal flux 3 67 3 67 0 0 pH 2 50 1 0 1 100 Coastal pollution 1 50 1 0 1 100 Presence of vegetation 1 0 0 0 1 0 Wave energy 1 0 1 0 0 0 12 Mar Ecol Prog Ser 554: 1–19, 2016

patch size, wave energy and sediment type, may be data from marine organisms are also improving more obviously related to mechanisms affecting pop- (Puritz et al. 2014b). Consequently, SNP-based sea- ulations within sites than dispersal and gene flow scape genetic analyses are growing rapidly (Fig. 3). between sites, meaning that study designs focused Below we highlight examples demonstrating how on node-based analyses may be more appropriate genomic sampling has 2 distinct benefits for sea- than link-based analyses (Wagner & Fortin 2013). scape analyses: (1) increased power and likelihood to Fewer significant drivers were identified in node- detect gene−environment interactions, and (2) the based studies than in link-based studies, consistent ability to simultaneously investigate the roles of drift, with the finding that the non-independence of link- migration and selection in shaping genetic structure. based data can lead to biased model selection and spurious predictor variables (Wagner & Fortin 2015). Alternatively, the seascape may influence migration Power through replication more strongly than drift and local adaptation, although this is unlikely. Node-based studies showed In the majority of seascape genetic studies, the that geography was a significant predictor only 18% amount of sampled genetic variation is relatively low of the time, and temperature 29% (Table 1). Ocean compared to the potential number of seascape driv- transport metrics were more commonly significant in ers. Consequently, investigators are required to test node-based studies, at a 44% rate of significance. only a small number of seascape factors or to collapse them down into coordinate axes to balance Type I and II error. Scaling up to genomic data adds Study design effects response variation that can be parsed among addi- tional factors and lead to more detailed interpreta- Considering all multi-factor studies, an average of tion. For example, 2 studies recently examined popu- 3.6 drivers were tested, with statistical support for 1 lation discontinuities in clownfish Amphiprion or 2 of them. Interestingly, we found that when a bicinctus over a similar region in the Red Sea. Nan- larger number of potential drivers was tested against ninga et al. (2014) used 38 microsatellite loci and a single response variable, a smaller fraction of driv- Saenz-Agudelo et al. (2015) used 4559 SNPs with ers was found to be significant. This trend suggests partially overlapping population samples, and both that only 1 or a few drivers tend to dominate genetic detected a pattern of IBE between clownfish popula- structuring, or that variables are sometimes added tions above and below an oligotrophic–eutrophic tran- without strong rationale. Studies using nuclear mark- sition zone, but no additional drivers of genetic dif- ers found larger fractions of tested drivers to be sig- ferentiation among the populations within each nificant than those using mtDNA, with no difference sample regime. Nanninga et al. (2014) chose chloro- in apparent power for SNP versus microsatellite stud- phyll a concentrations as the best representative for ies. Of course, these trends depend on choices of environmental distance and were able to evaluate a study design, statistical testing and which drivers hypothesis of IBE versus IBD using partial Mantel and taxa are easily sampled in the marine environ- tests and multiple matrix regressions with random- ment. Nevertheless, they lend insight into the range ization. With an order of magnitude increase in of seascape predictors reported in recent studies genetic variance, Saenz-Agudelo et al. (2015) used a and the effects of study design (see Fig. 2), and high- different powerful statistical approach including an light underrepresented predictors that deserve future information theoretic model selection framework and study. the evaluation of 49 different linear models of sea- scape drivers including the presence of genetic breaks. This approach greatly improved the overall THE ADVENT OF SEASCAPE GENOMICS model fit, and revealed a new dimension (presence of genetic discontinuities) that a simple IBD/IBE analy- Genomic studies of non-model organisms are sis would have overlooked. becoming more accessible thanks to techniques that cheaply and randomly sub-sample the genome, such as restriction site associated DNA sequencing (RAD- Parsing neutral and adaptive genetics seq) (Miller et al. 2007, Baird et al. 2008, Toonen et al. 2013, Puritz et al. 2014a, Andrews et al. 2016). Bioin- One of the key benefits of genome-level analyses formatic pipelines specifically designed to handle on populations is sampling more unlinked (i.e. inde- Selkoe et al.: A decade of seascape genetics 13

pendently sorting) portions of the genome. This Future directions allows more robust characterization of a species’ demography that overcomes idiosyncratic gene his- The next challenge for seascape genomics is to tories (i.e. differentiation between responses to envi- progress from pattern to process, i.e. from seascape ronmental selection, migration, drift, etc.). For exam- correlates of spatial genetic patterns to specific ple, in a recent study on bonnethead sharks, Portnoy mechanisms of population differentiation (reviewed et al. (2015) isolated the entire mtDNA control in Riginos et al. 2016). New analytical frameworks region, and used RADseq to produce 5914 SNPs; 49 will play a key role in uncovering selective forces. SNPs were outliers in their level of spatial genetic For instance, linking shifts in allele frequencies to structure compared to the other loci. Outlier SNPs adaptive phenotypes will require creative approaches exhibited signatures of latitude-driven selection that draw on gene ontology databases and experi- while the mtDNA data, in conjunction with the rest of mental tests of fitness advantage (re viewed in de the SNPs, showed a signature of female philopatry Villemereuil et al. 2015, Pardo-Diaz et al. 2015, Rell- and male-mediated gene flow among the sampled stab et al. 2015). Research on responses of a sea populations. The non-outlier SNPs also supported a urchin model system to ocean acidification (e.g. Kelly classic biogeographic separation between the Gulf of et al. 2013, Pespeni & Palumbi 2013, Pespeni et al. Mexico and the western Atlantic Ocean populations. 2013, Evans et al. 2015) provides a glimpse of how Importantly, the combination of multiple individual ecological experiments and genomic and transcrip- unlinked gene genealogies enabled the characteri- tomic data together illuminate adaptive evolutionary zation of population dynamics and historical bio- response across a seascape. geography of bonnethead sharks. Outlier loci, especially those that are associated with the environment, have special utility for sea- CONCLUSIONS scape analyses, but theory predicts that extreme outlier loci should be rare. Importantly, complex Fueled by GIS data, genomics and statistical landscapes can lead to high rates of false positives advances, the first wave of seascape genetic studies when studies are not properly designed (Lotterhos has changed our conceptual understanding of the & Whitlock 2014, 2015, Forester et al. 2016). In multiple historical and contemporary processes shap- general, false positives are a major challenge for ing dispersal and gene flow in the sea. Importantly, it seascape genomics, given the deluge of genetic and can be tempting to overlook potential false positives, environmental data (De Mita et al. 2013, de Ville- and over-interpret nuanced spatial patterns and mereuil et al. 2014, Lotterhos & Whitlock 2014). slight but significant explanatory power (Meirmans New analyses coupling outlier detection with envi- 2015, Richardson et al. 2016), but not all spatial ronmental gradients reduce false positives (de Ville- genetic structure has ecological meaning (Hedrick mereuil & Gaggiotti 2015). Furthermore, a focus 1999). At present, we still largely lack the tools and on extreme outliers misses more common selective perspective needed to gauge where to limit inference actions that should be distributed as small selective on the meaning of subtle effects. Complementing effects at multiple genomic regions responding to empirical studies with simulations holds promise to similar environmental forcing in concert (Schwartz illuminate these boundaries. et al. 2010). These polygenic effects would not In many ways, the rise of landscape and seascape likely be detected via outlier analyses. For example, genetics represents the maturation of population Laporte et al. (2016) surveyed 23 659 and 14 755 genetics, with a stronger emphasis on spatial ecolog- SNPs in North Atlantic eels across ‘polluted’ and ical predictors (Dyer 2015). Perhaps unsurprisingly, control environments. Outlier detection revealed we found that many seascape genetic studies have only 2 outlier loci; however, a machine learning been oriented towards the population genetic litera- technique (random forest algorithm) coupled with a ture, as evidenced by the strong bias in journal pub- distance-based redundancy analysis detected over lication (Fig. S1 in the Supplement at www.int- 140 loci under polygenic selection. Whether disen- res.com/articles/suppl/m554p001_supp.pdf). We pro- tangling the effects of demography, life history and pose that broadening the audience of seascape genetic migration, or detecting the subtle genetic effects of studies is a critical step in improving their utility and environmental polygenic selection, genomic-scale value across disciplinary lines. data sets will enhance the scope and power of sea- As the field of seascape genetics progresses, we scape analysis. anticipate that the most exciting advances will come 14 Mar Ecol Prog Ser 554: 1–19, 2016

from a stronger integration with biological oceano - Balkenhol N, Cushman S, Storfer A, Waits L (2015) Land- graphy, metapopulation ecology, biogeography, larval scape genetics: concepts, methods, applications. John Wiley & Sons, Chichester ecology, physiology, evolution and other dis ciplines. Banks SC, Piggott LMP, Williamson JE, Bové U, Holbrook Finally, we emphasize that as marine eco systems NJ, Beheregaray LB (2007) Oceanic variability and around the globe face anthropogenic impacts on coastal topography shape genetic structure in a long- unprecedented levels, seascape genetics can play a dispersing sea urchin. Ecology 88: 3055−3064 Baums IB, Paris CB, Chérubin LM (2006) A bio-oceano- unique and powerful role in understanding effective graphic filter to larval dispersal in a reef-building coral. long-term conservation and management strategies. Limnol Oceanogr 51: 1969−1981 Beaugrand G (2009) Decadal changes in climate and eco - systems in the North Atlantic Ocean and adjacent seas. Acknowledgements. This contribution was initiated during Deep-Sea Res II 56:656−673 a symposium called ‘Waterscape Genetics: New Perspec- Beerli P, Palczewski M (2010) Unified framework to evaluate tives on Connectivity in Fluid Environments’ hosted by the panmixia and migration direction among multiple sam- 9th International Association of Landscape Ecology (IALE) pling locations. Genetics 185: 313−326 World Congress in Portland, Oregon, USA. The authors Beger M, Selkoe KA, Treml E, Barber PH and others (2014) thank contributors to initial discussions, including Aaron Evolving coral reef conservation with genetic informa- Adamack, Peter Buston, Annick Cros, Ingrid Knapp, Jesse tion. Bull Mar Sci 90: 159−185 Port and Eric Treml. The authors are also grateful for Berger EM (1973) Gene-enzyme variation in three sympatric methodological clarifications from Pablo Saenz-Agudelo, species of Littorina. Biol Bull (Woods Hole) 145:83−90 and the constructive feedback received from Philippe Borsa Berumen M, Walsh H, Raventos N, Planes S, Jones G, and 2 anonymous reviewers. K.A.S. thanks the National Starczak V, Thorrold S (2010) Otolith geochemistry does Center for Ecological Analysis and Synthesis, a center not reflect dispersal history of clownfish larvae. Coral funded by the University of California Santa Barbara and Reefs 29: 883−891 the State of California for support during the study. Funding Bird CE, Karl SA, Smouse PE, Toonen RJ (2011) Detecting was provided to K.A.S. and R.J.T. by the National Science and measuring genetic differentiation. In: Held C, Foundation (NSF; BioOCE Award Number 1260169); C.C.D. Koenemann S, Schubart CD (eds) Phylogeography and was supported by a NSERC Strategic Grant (430706-12); population genetics in Crustacea. CRC Press, Boca M.I. was supported by an NSF Postdoctoral Fellowship Raton, FLp 31−55 (BioOCE Award Number 1522572); L.L. was supported by a Bird C, Fernandez-Silva I, Skillings D, Toonen R (2012) Rutherford Foundation New Zealand Postdoctoral Fellow- Sympatric speciation in the post ‘modern synthesis’ era ship, an Allan Wilson Centre for Molecular Ecology and of evolutionary biology. Evol Biol 39: 158−180 Evolution Postdoctoral Fellowship, and conference travel Bonin MC, Harrison HB, Williamson DH, Frisch AJ, Saenz and attendance was funded by the Institute of Natural and Agudelo P, Berumen ML, Jones GP (2016) The role of Mathematical Sciences of Massey University. marine reserves in the replenishment of a locally impacted population of anemonefish on the Great Barrier Reef. Mol Ecol 25:487−499 LITERATURE CITED Bowen BW, Shanker K, Yasuda N, Malay CM and others (2014) Phylogeography unplugged: comparative geo- Allendorf FW, Hohenlohe PA, Luikart G (2010) Genomics graphic surveys in the genomic era. Bull Mar Sci 90: 13−46 and the future of conservation genetics. Nat Rev Genet Burgess SC, Treml EA, Marshall DJ (2012) How do dispersal 11: 697−709 costs and habitat selection influence realized population Almany GR, Hamilton RJ, Bode M, Matawai M and others connectivity? Ecology 93: 1378−1387 (2013) Dispersal of grouper larvae drives local resource Burgess SC, Nickols KJ, Griesemer CD, Barnett LAK and sharing in a coral reef fishery. Curr Biol 23: 626−630 others (2014) Beyond connectivity: how empirical meth- Andrello M, Mouillot D, Somot S, Thuiller W, Manel S (2015) ods can quantify population persistence to improve Additive effects of climate change on connectivity marine protected-area design. Ecol Appl 24:257−270 between marine protected areas and larval supply to Burgess SC, Baskett ML, Grosberg RK, Morgan SG, Strath- fished areas. Divers Distrib 21: 139−150 mann RR (2015) When is dispersal for dispersal? Unifying Andrews KR, Good JM, Miller MR, Luikart G, Hohenlohe marine and terrestrial perspectives. Biol Rev Camb PA (2016) Harnessing the power of RADseq for ecologi- Philos Soc (online early) doi: 10.1111/brv.12198 cal and evolutionary genomics. Nat Rev Genet 17: 81−92 Christie MR (2010) Parentage in natural populations: novel Arizmendi-Mejía R, Linares Prats C, Garrabou J, Antunes A methods to detect parent-offspring pairs in large data and others (2015) Combining genetic and demographic sets. Mol Ecol Resour 10: 115−128 data for the conservation of a Mediterranean habitat- Christie MR (2013) Bayesian parentage analysis reliably forming species. PLoS ONE 10: e0119585 controls the number of false assignments in natural Baele G, Lemey P, Bedford T, Rambaut A, Suchard MA, populations. Mol Ecol 22: 5731−5737 Alekseyenko AV (2012) Improving the accuracy of Coop G, Witonsky D, Di Rienzo A, Pritchard JK (2010) Using demographic and molecular clock model comparison environmental correlations to identify loci underlying while accommodating phylogenetic uncertainty. Mol local adaptation. Genetics 185:1411−1423 Biol Evol 29: 2157−2167 Cossu P, Dedola GL, Scarpa F, Sanna D and others (2015) Baird NA, Etter PD, Atwood TS, Currey MC and others Patterns of spatial genetic variation in Patella ulyssipo- (2008) Rapid SNP discovery and genetic mapping using nensis: insights from the western Mediterranean marine sequenced RAD markers. PLoS ONE 3: e3376 ecoregion. Hydrobiologia 755: 39−55 Selkoe et al.: A decade of seascape genetics 15

Cowen RK, Sponaugle S (2009) Larval dispersal and marine Fraser C, Banks S, Waters J (2015) Priority effects can lead to population connectivity. Annu Rev Mar Sci 1:443−466 underestimation of dispersal and invasion potential. Biol Crandall ED, Treml EA, Barber PH (2012) Coalescent and Invasions 17:1−8 biophysical models of stepping-stone gene flow in neritid Frichot E, Schoville SD, Bouchard G, François O (2013) snails. Mol Ecol 21:5579−5598 Testing for associations between loci and environmental D’Aloia CC, Bogdanowicz SM, Francis RK, Majoris JE, gradients using latent factor mixed models. Mol Biol Evol Harrison RG, Buston PM (2015) Patterns, causes, and 30: 1687−1699 consequences of marine larval dispersal. Proc Natl Acad Funk WC, McKay JK, Hohenlohe PA, Allendorf FW (2012) Sci USA 112: 13940−13945 Harnessing genomics for delineating conservation units. Davies SW, Treml EA, Kenkel CD, Matz MV (2015) Explor- Trends Ecol Evol 27: 489−496 ing the role of Micronesian islands in the maintenance of Gagnaire PA, Broquet T, Aurelle D, Viard F and others coral genetic diversity in the Pacific Ocean. Mol Ecol 24: (2015) Using neutral, selected, and hitchhiker loci to 70−82 assess connectivity of marine populations in the genomic Dawson MN, Hays CG, Grosberg RK, Raimondi PT (2014) era. Evol Appl 8: 769−786 Dispersal potential and population genetic structure in Galindo HM, Olson DB, Palumbi SR (2006) Seascape genet- the marine intertidal of the eastern North Pacific. Ecol ics: a coupled oceanographic-genetic model predicts Monogr 84: 435−456 population structure of Caribbean corals. Curr Biol 16: De Meester L, Gómez A, Okamura B, Schwenk K (2002) The 1622−1626 Monopolization Hypothesis and the dispersal−gene flow Gautier M (2015) Genome-wide scan for adaptive diver- paradox in aquatic organisms. Acta Oecol 23:121−135 gence and association with population-specific covari- De Mita S, Thuillet AC, Gay L, Ahmadi N, Manel S, Ronfort ates. Genetics 201:1555−1579 J, Vigouroux Y (2013) Detecting selection along environ- Gilg MR, Hilbish TJ (2003) The geography of marine larval mental gradients: analysis of eight methods and their dispersal: coupling genetics with fine-scale physical effectiveness for outbreeding and selfing populations. oceanography. Ecology 84: 2989−2998 Mol Ecol 22:1383−1399 Goldberg CS, Waits LP (2010) Comparative landscape genet- de Villemereuil P, Gaggiotti OE (2015) A new FST-based ics of two pond-breeding amphibian species in a highly method to uncover local adaptation using environmental modified agricultural landscape. Mol Ecol 19:3650−3663 variables. Methods Ecol Evol 6:1248−1258 González-Wangüemert M, Cánovas F, Marcos C, Pérez- de Villemereuil P, Frichot E, Bazin É, François O, Gaggiotti Ruzafa Á (2009) Phosphoglucose isomerase variability of OE (2014) Genome scan methods against more complex Cerastoderma glaucum as a model for testing the influ- models: when and how much should we trust them? Mol ence of environmental conditions and dispersal patterns Ecol 23: 2006−2019 through quantitative ecology approaches. Biochem Syst de Villemereuil P, Gaggiotti OE, Mouterde M, Till-Bottraud Ecol 37: 325−333 I (2015) Common garden experiments in the genomic Guillot G, Rousset F (2013) Dismantling the Mantel tests. era: new perspectives and opportunities. Heredity 116: Methods Ecol Evol 4: 336−344 249−254 Günther T, Coop G (2013) Robust identification of local Dyer RJ (2015) Is there such a thing as landscape genetics? adaptation from allele frequencies. Genetics 195: 205−220 Mol Ecol 24:3518−3528 Hansen MM, Hemmer-Hansen J (2007) Landscape genetics Epps CW, Keyghobadi N (2015) Landscape genetics in a goes to sea. J Biol 6: 6 changing world: disentangling historical and contempo- Harrison HB, Williamson DH, Evans RD, Almany GR and rary influences and inferring change. Mol Ecol 24: others (2012) Larval export from marine reserves and the 6021−6040 recruitment benefit for fish and fisheries. Curr Biol 22: Evans TG, Padilla-Gamiño JL, Kelly MW, Pespeni MH and 1023−1028 others (2015) Ocean acidification research in the ‘post- Hart MW, Marko PB (2010) It’s about time: divergence, genomic’ era: roadmaps from the purple sea urchin demography, and the evolution of developmental modes Stron gylo centrotus purpuratus. Comp Biochem Physiol in marine invertebrates. Integr Comp Biol 50: 643−661 A Mol Integr Physiol 185: 33−42 Hecht BC, Matala AP, Hess JE, Narum SR (2015) Environ- Fitzpatrick MC, Keller SR (2015) Ecological genomics meets mental adaptation in Chinook salmon (Oncorhynchus community-level modelling of biodiversity: mapping the tshawytscha) throughout their North American range. genomic landscape of current and future environmental Mol Ecol 24:5573−5595 adaptation. Ecol Lett 18:1−6 Hedgecock D (1994) Does variance in reproductive success Foll M, Gaggiotti O (2006) Identifying the environmental limit effective population size of marine organisms? In: factors that determine the genetic structure of popula- Beaumont AR (ed) Genetics and evolution of aquatic tions. Genetics 174:875−891 organisms. Chapman & Hall, London, p 122−134 Fontaine MC, Baird SJE, Piry S, Ray N and others (2007) Hedrick PW (1999) Perspective: highly variable loci and Rise of oceanographic barriers in continuous populations their interpretation in evolution and conservation. of a cetacean: the genetic structure of harbour porpoises Evolution 53: 313−318 in Old World waters. BMC Biol 5: 30 Hellberg ME (1994) Relationships between inferred levels of Forester BR, Jones MR, Joost S, Landguth EL, Lasky JR (2016) gene flow and geographic distance in a philopatric coral, Detecting spatial genetic signatures of local adaptation in Balanophyllia elegans. Evolution 48: 1829−1854 heterogeneous landscapes. Mol Ecol 25: 104−120 Hellberg ME (2007) Footprints on water: the genetic wake of Foster NL, Paris CB, Kool JT, Baums IB and others (2012) dispersal among reefs. Coral Reefs 26: 463−473 Connectivity of Caribbean coral populations: comple- Hey J, Machado CA (2003) The study of structured popula- mentary insights from empirical and modelled gene flow. tions—new hope for a difficult and divided science. Nat Mol Ecol 21:1143−1157 Rev Genet 4: 535−543 16 Mar Ecol Prog Ser 554: 1–19, 2016

Hickerson MJ, Meyer CP (2008) Testing comparative phylo- Larsson LC, Charlier J, Laikre L, Ryman N (2009) Statistical geographic models of marine vicariance and dispersal power for detecting genetic divergence—organelle ver- using a hierarchical Bayesian approach. BMC Evol Biol sus nuclear markers. Conserv Genet 10: 1255−1264 8: 322 Lee PLM, Dawson MN, Neill SP, Robins PE, Houghton JDR, Hickerson MJ, Stahl EA, Lessios HA (2006) Test for simulta- Doyle TK, Hays GC (2013) Identification of genetically neous divergence using approximate Bayesian computa- and oceanographically distinct blooms of jellyfish. J R tion. Evolution 60: 2435−2453 Soc Interface 10: 20120920 Hogan JD, Thiessen RJ, Sale PF, Heath DD (2012) Local Legendre P, Legendre L (2012) Numerical ecology, 3rd edn. retention, dispersal and fluctuating connectivity among Elsevier, Oxford populations of a coral reef fish. Oecologia 168:61−71 Legendre P, Fortin M, Borcard D (2015) Should the Mantel Holderegger R, Wagner HH (2008) Landscape genetics. test be used in spatial analysis? Methods Ecol Evol 6: Bioscience 58: 199−207 1239−1247 Hughes AR, Stachowicz JJ (2004) Genetic diversity Levin LA (2006) Recent progress in understanding larval enhances the resistance of a seagrass ecosystem to dis- dispersal: new directions and digressions. Integr Comp turbance. Proc Natl Acad Sci USA 101: 8998−9002 Biol 46: 282−297 Iacchei M, Ben-Horin T, Selkoe KA, Bird CE, García- Liggins L, Treml EA, Riginos C (2013) Taking the plunge: an Rodríguez FJ, Toonen RJ (2013) Combined analyses of introduction to undertaking seascape genetic studies kinship and FST suggest potential drivers of chaotic and using biophysical models. Geogr Compass 7:173−196 genetic patchiness in high gene-flow populations. Mol Liggins L, Booth DJ, Figueira WF, Treml EA and others Ecol 22: 3476−3494 (2015) Latitude-wide genetic patterns reveal historical Johansson ML, Alberto F, Reed DC, Raimondi PT and others effects and contrasting patterns of turnover and nested- (2015) Seascape drivers of Macrocystis pyrifera popula- ness at the range peripheries of a tropical marine fish. tion genetic structure in the northeast Pacific. Mol Ecol Ecography 38: 1212−1224 24: 4866−4885 Liggins L, Treml EA, Possingham HP, Riginos C (2016) Sea- Johnson M, Black R (1982) Chaotic genetic patchiness in an scape features, rather than dispersal traits, predict spa- intertidal limpet, Siphonaria sp. Mar Biol 70:157−164 tial genetic patterns in co-distributed reef fishes. J Bio- Johnson MS, Black R (1991) Genetic subdivision of the geogr 43: 256−267 intertidal snail Bembicium vittatum (Gastropoda: Litto- Limborg MT, Helyar SJ, De Bruyn M, Taylor MI and others rinidae) varies with habitat in the Houtman Abrolhos (2012) Environmental selection on transcriptome-derived Islands, Western Australia. Heredity 67: 205−213 SNPs in a high gene flow marine fish, the Atlantic Jones GP, Milicich MJ, Emslie MJ, Lunow C (1999) Self- herring (Clupea harengus). Mol Ecol 21: 3686−3703 recruitment in a coral reef fish population. Nature 402: Lotterhos KE, Markel RW (2012) Oceanographic drivers of 803−804 offspring abundance may increase or decrease repro- Jones GP, Planes S, Thorrold SR (2005) Coral reef fish larvae ductive variance in a temperate marine fish. Mol Ecol 21: settle close to home. Curr Biol 15: 1314−1318 5009−5026 Jue NK, Brulé T, Coleman FC, Koenig CC (2015) From shelf Lotterhos KE, Whitlock MC (2014) Evaluation of demo- to shelf: assessing historical and contemporary genetic graphic history and neutral parameterization on the differentiation and connectivity across the Gulf of performance of FST outlier tests. Mol Ecol 23:2178−2192 Mexico in Gag, Mycteroperca microlepis. PLoS ONE 10: Lotterhos KE, Whitlock MC (2015) The relative power of e0120676 genome scans to detect local adaptation depends on Kelly E, Phillips BL (2016) Targeted gene flow for conserva- sampling design and statistical method. Mol Ecol 24: tion. Conserv Biol 30: 259−267 1031−1046 Kelly MW, Padilla-Gamiño JL, Hofmann GE (2013) Natural Lowe WH, Allendorf FW (2010) What can genetics tell us variation and the capacity to adapt to ocean acidification about population connectivity? Mol Ecol 19:3038−3051 in the keystone sea urchin Strongylocentrotus purpura- Mace AJ, Morgan SG (2006) Larval accumulation in the lee tus. Glob Change Biol 19:2536−2546 of a small headland: implications for the design of marine Kinlan BP, Gaines SD (2003) Propagule dispersal in marine reserves. Mar Ecol Prog Ser 318:19−29 and terrestrial environments: a community perspective. Manel S, Holderegger R (2013) Ten years of landscape Ecology 84: 2007−2020 genetics. Trends Ecol Evol 28:614−621 Kool JT, Paris CB, Andréfouët S, Cowen RK (2010) Complex Manel S, Schwartz MK, Luikart G, Taberlet P (2003) Land- migration and the development of genetic structure in scape genetics: combining landscape ecology and subdivided populations: an example from Caribbean population genetics. Trends Ecol Evol 18: 189−197 coral reef ecosystems. Ecography 33:597−606 Manni F, Guerard E, Heyer E (2004) Geographic patterns of Kool JT, Paris CB, Barber PH, Cowen RK (2011) Connectiv- (genetic, morphologic, linguistic) variation: how barriers ity and the development of population genetic structure can be detected by using Monmonier’s algorithm. Hum in Indo-West Pacific coral reef communities. Glob Ecol Biol 76: 173−190 Biogeogr 20:695−706 Marko PB, Hart MW (2011) The complex analytical land- Kool JT, Moilanen A, Treml EA (2013) Population connectiv- scape of gene flow inference. Trends Ecol Evol 26: ity: recent advances and new perspectives. Landscape 448−456 Ecol 28: 165−185 Marshall DJ, Morgan SG (2011) Ecological and evolutionary Laporte M, Pavey SA, Rougeux C, Pierron F and others consequences of linked life-history stages in the sea. (2016) RAD sequencing reveals within-generation poly- Curr Biol 21: R718−R725 genic selection in response to anthropogenic organic and Mazzuco A, Christofoletti R, Pineda J, Starczak V, Ciotti A metal contamination in North Atlantic eels. Mol Ecol 25: (2015) Temporal variation in intertidal community 219−237 recruitment and its relationships to physical forcings, Selkoe et al.: A decade of seascape genetics 17

chlorophyll-a concentration and sea surface tempera- mental mosaic. Mol Ecol 22:3580−3597 ture. Mar Biol 162: 1705−1725 Pespeni MH, Chan F, Menge BA, Palumbi SR (2013) Signs of McLeod E, Salm R, Green A, Almany J (2009) Designing adaptation to local pH conditions across an environmen- marine protected area networks to address the impacts tal mosaic in the California Current ecosystem. Integr of climate change. Front Ecol Environ 7: 362−370 Comp Biol 53: 857−870 Meirmans PG (2012) The trouble with isolation by distance. Pfaff MC, Branch GM, Fisher JL, Hoffmann V, Ellis AG, Mol Ecol 21:2839−2846 Largier JL (2015) Delivery of marine larvae to shore re - Meirmans PG (2014) Nonconvergence in Bayesian estima- quires multiple sequential transport mechanisms. Ecology tion of migration rates. Mol Ecol Resour 14: 726−733 96: 1399−1410 Meirmans PG (2015) Seven common mistakes in population Pineda J, Hare JA, Sponaungle S (2007) Larval transport genetics and how to avoid them. Mol Ecol 24: 3223−3231 and dispersal in the coastal ocean and consequences for Mendez M, Kershaw F, Palumbi S, Pinsky M, Ray C, Rosen- population connectivity. Oceanography (Wash DC) 20: baum H, Subramaniam A (2014) Marine Spatial Planning 22−39 2.0: genes and satellites to conserve seascape dynamics. Pinsky ML, Palumbi SR (2014) Meta-analysis reveals lower Aquat Conserv 24:742−744 genetic diversity in overfished populations. Mol Ecol 23: Messmer V, Jones GP, Munday PL, Planes S (2012) Concor- 29−39 dance between genetic and species diversity in coral reef Planes S, Jones GP, Thorrold SR (2009) Larval dispersal con- fishes across the Pacific Ocean biodiversity gradient. nects fish populations in a network of marine protected Evolution 66: 3902−3917 areas. Proc Natl Acad Sci USA 106:5693−5697 Miller MR, Dunham JP, Amores A, Cresko WA, Johnson EA Pope LC, Riginos C, Ovenden J, Keyse J, Blomberg SP (2007) Rapid and cost-effective polymorphism identifica- (2015) Population genetic diversity in the Australian ‘sea- tion and genotyping using restriction site associated scape’: a bioregion approach. PLoS ONE 10:e0136275 DNA (RAD) markers. Genome Res 17:240−248 Port JA, O’Donnell JL, Romero-Maraccini OC, Leary PR and Mills L, Allendorf FW (1996) The one-migrant-per-generation others (2016) Assessing vertebrate biodiversity in a kelp rule in conservation and management. Conserv Biol 10: forest ecosystem using environmental DNA. Mol Ecol 25: 1509−1518 527−541 Morgan SG, Fisher J, McAfee S, Largier J, Miller S, Sheri- Portnoy DS, Puritz JB, Hollenbeck CM, Gelsleichter J, dan M, Neigel J (2014) Transport of crustacean larvae Chapman D, Gold JR (2015) Selection and sex-biased between a low-inflow estuary and coastal waters. Estuar dispersal in a coastal shark: the influence of philopatry Coasts 37: 1269−1283 on adaptive variation. Mol Ecol 24:5877−5885 Munguia-Vega A, Jackson A, Marinone SG, Erisman B and Pringle JM, Wares JP (2007) Going against the flow: mainte- others (2014) Asymmetric connectivity of spawning nance of alongshore variation in allele frequency in a aggregations of a commercially important marine fish coastal ocean. Mar Ecol Prog Ser 335: 69−84 using a multidisciplinary approach. PeerJ 2: e511 Prunier JG, Kaufmann B, Fenet S, Picard D, Pompanon F, Nanninga GB, Saenz Agudelo P, Manica A, Berumen ML Joly P, Lena JP (2013) Optimizing the trade-off between (2014) Environmental gradients predict the genetic spatial and genetic sampling efforts in patchy popula- population structure of a coral reef fish in the Red Sea. tions: towards a better assessment of functional connec- Mol Ecol 23:591−602 tivity using an individual-based sampling scheme. Mol Nickols KJ, White JW, Largier JL, Gaylord B (2015) Marine Ecol 22: 5516−5530 population connectivity: reconciling large-scale disper- Puritz JB, Toonen RJ (2011) Coastal pollution limits pelagic sal and high self-retention. Am Nat 185:196−211 larval dispersal. Nat Commun 2:226 Nosil P, Egan SP, Funk DJ (2008) Heterogeneous genomic Puritz JB, Matz MV, Toonen RJ, Weber JN, Bolnick DI, Bird differentiation between walking-stick ecotypes: ‘isola- CE (2014a) Demystifying the RAD fad. Mol Ecol 23: tion by adaptation’ and multiple roles for divergent 5937−5942 selection. Evolution 62: 316−336 Puritz JB, Hollenbeck CM, Gold JR (2014b) dDocent: a RAD- Nosil P, Funk DJ, Ortiz-Barrientos D (2009) Divergent selec- seq, variant-calling pipeline designed for population tion and heterogeneous genomic divergence. Mol Ecol genomics of non-model organisms. PeerJ 2: e431 18: 375−402 Razgour O (2015) Beyond species distribution modeling: a Orsini L, Vanoverbeke J, Swillen I, Mergeay J, De Meester landscape genetics approach to investigating range shifts L (2013) Drivers of population genetic differentiation in under future climate change. Ecol Inform 30: 250−256 the wild: isolation by dispersal limitation, isolation by Rellstab C, Gugerli F, Eckert AJ, Hancock AM, Holderegger adaptation and isolation by colonization. Mol Ecol 22: R (2015) A practical guide to environmental association 5983−5999 analysis in landscape genomics. Mol Ecol 24:4348−4370 Palumbi SR (1996) Macrospatial genetic structure and speci- Reusch TBH, Ehlers A, Hämmerli A, Worm B (2005) Ecosys- ation in marine taxa with high dispersal abilities. In: Fer- tem recovery after climatic extremes enhanced by geno- raris JD, Palumbi SR (eds) Molecular zoology: advances, typic diversity. Proc Natl Acad Sci USA 102:2826−2831 strategies, and protocols. Wiley, New York, NY, p 101−117 Richardson JL, Brady SP, Wang IJ, Spear SF (2016) Navigat- Palumbi SR (2003) Population genetics, demographic con- ing the pitfalls and promise of landscape genetics. Mol nectivity, and the design of marine reserves. Ecol Appl Ecol 25: 849−863 13(S1): 146−158 Riginos C, Liggins L (2013) Seascape genetics: populations, Pardo-Diaz C, Salazar C, Jiggins CD (2015) Towards the individuals, and genes marooned and adrift. Geogr identification of the loci of adaptive evolution. Methods Compass 7: 197−216 Ecol Evol 6: 445−464 Riginos C, Nachman MW (2001) Population subdivision in Pespeni MH, Palumbi SR (2013) Signals of selection in out- marine environments: the contributions of biogeography, lier loci in a widely dispersing species across an environ- geographical distance and discontinuous habitat to 18 Mar Ecol Prog Ser 554: 1–19, 2016

genetic differentiation in a blennioid fish, Axoclinus Siegel DA, Mitarai S, Costello CJ, Gaines SD, Kendall BE, nigricaudus. Mol Ecol 10:1439−1453 Warner RR, Winters KB (2008) The stochastic nature of Riginos C, Crandall ED, Liggins L, Bongaerts P, Treml EA larval connectivity among nearshore marine populations. (2016) Navigating the currents of seascape genomics: Proc Natl Acad Sci USA 105: 8974−8979 how spatial analyses can augment population genomic Storfer A, Murphy MA, Evans JS, Goldberg CS and others studies. Curr Zool (online early) doi: 10.1093/cz/zow067 (2007) Putting the ‘landscape’ in landscape genetics. Rosenberg NA, Nordborg M (2002) Genealogical trees, Heredity 98: 128−142 coalescent theory and the analysis of genetic polymor- Strathmann RR, Branscomb ES, Vedder K (1981) Fatal errors phisms. Nat Rev Genet 3:380−390 in set as a cost of dispersal and the influence of intertidal Row JR, Oyler-McCance SJ, Fike JA, O’Donnell MS and flora on set of barnacles. Oecologia 48:13−18 others (2015) Landscape characteristics influencing the Swearer SE, Caselle JE, Lea DW, Warner RR (1999) Larval genetic structure of greater sage-grouse within the retention and recruitment in an island population of a stronghold of their range: a holistic modeling approach. coral-reef fish. Nature 402:799−802 Ecol Evol 5: 1955−1969 Swearer SE, Shima JS, Hellberg ME (2002) Evidence of self- Roy D, Hurlbut TR, Ruzzante DE (2012) Biocomplexity in a recruitment in demersal marine populations. Bull Mar demersal exploited fish, white hake (Urophycis tenuis): Sci 70(S1): 251−271 depth-related structure and inadequacy of current man- Tepolt CK, Palumbi SR (2015) Transcriptome sequencing agement approaches. Can J Fish Aquat Sci 69: 415-42 reveals both neutral and adaptive genome dynamics in a Ryman N, Palm S, André C, Carvalho GR and others (2006) marine invader. Mol Ecol 24: 4145−4158 Power for detecting genetic divergence: differences Teske PR, Papadopoulos I, Mmonwa KL, Matumba TG, between statistical methods and marker loci. Mol Ecol McQuaid CD, Barker NP, Beheregaray LB (2011) Cli- 15: 2031−2045 mate-driven genetic divergence of limpets with different Saenz-Agudelo P, Dibattista JD, Piatek MJ, Gaither MR, life histories across a southeast African marine biogeo- Harrison HB, Nanninga GB, Berumen ML (2015) Sea- graphic disjunction: different processes, same outcome. scape genetics along environmental gradients in the Mol Ecol 20:5025−5041 Arabian Peninsula: insights from ddRAD sequencing of Toonen RJ, Grosberg RK (2011) Causes of chaos: spatial and anemonefishes. Mol Ecol 24: 6241−6255 temporal genetic heterogeneity in the intertidal ano - Schunter C, Pascual M, Garza JC, Raventos N, Macpherson muran crab Petrolisthes cinctipes. In: Held C, Koene- E (2014) Kinship analyses identify fish dispersal events mann S, Schubart CD (eds) Phylogeography and popula- on a temperate coastline. Proc R Soc B 281: 20140556 tion genetics in crustacea. CRC Press, Boca Raton, FL, p Schwartz MK, McKelvey KS, Cushman SA, Luikart G (2010) 73−106 Landscape genomics: a brief perspective. In: Cushman Toonen RJ, Puritz JB, Forsman ZH, Whitney JL, Fernandez- SA, Huettmann F (eds) Spatial complexity, informatics, Silva I, Andrews KR, Bird CE (2013) ezRAD: a simplified and wildlife conservation. Springer Japan, Tokyo, method for genomic genotyping in non-model organ- p 165−174 isms. PeerJ 1: e203 Selkoe KA, Toonen RJ (2011) Marine connectivity: a new Treml EA, Halpin PN, Urban DL, Pratson LF (2008) Model- look at pelagic larval duration and genetic metrics of ing population connectivity by ocean currents, a graph- dispersal. Mar Ecol Prog Ser 436: 291−305 theoretic approach for marine conservation. Landscape Selkoe KA, Gaines SD, Caselle JE, Warner RR (2006) Cur- Ecol 23: 19−36 rent shifts and kin aggregation explain genetic patchi- Treml EA, Ford JR, Black KP, Swearer SE (2015) Identifying ness in fish recruits. Ecology 87: 3082−3094 the key biophysical drivers, connectivity outcomes, and Selkoe KA, Watson JR, White C, Horin TB and others (2010) metapopulation consequences of larval dispersal in the Taking the chaos out of genetic patchiness: seascape sea. Mov Ecol 3: 17 genetics reveals ecological and oceanographic drivers of van Oppen MJH, Oliver JK, Putnam HM, Gates RD (2015) genetic patterns in three temperate reef species. Mol Building coral reef resilience through assisted evolution. Ecol 19:3708−3726 Proc Natl Acad Sci USA 112: 2307−2313 Selkoe KA, Scribner KT, Galindo HM (2015) Waterscape van Strien MJ, Keller D, Holderegger R (2012) A new analyt- genetics—applications of landscape genetics to rivers, ical approach to landscape genetic modelling: least-cost lakes, and seas. In: Balkenhol N, Cushman S, Storfer A, transect analysis and linear mixed models. Mol Ecol 21: Waits L (eds) Landscape genetics: concepts, methods, 4010−4023 applications. John Wiley & Sons, Chichester, p 220−246 Viricel A, Rosel PE (2014) Hierarchical population structure Selkoe KA, Gaggiotti OE, Treml EA, Wren JKL, Donovan and habitat differences in a highly mobile marine spe- MK, Hawai’i Reef Connectivity Consortium, Toonen RJ cies: the Atlantic spotted dolphin. Mol Ecol 23:5018−5035 (2016) The DNA of coral reef biodiversity: predicting and Wagner HH, Fortin MJ (2013) A conceptual framework for protecting genetic diversity of reef assemblages. Proc R the spatial analysis of landscape genetic data. Conserv Soc B 283: 20160354 Genet 14: 253−261 Sgrò CM, Lowe AJ, Hoffmann AA (2011) Building evolu- Wagner HH, Fortin MJ (2015) Basics of spatial data analysis: tionary resilience for conserving biodiversity under cli- linking landscape and genetic data for landscape genetic mate change. Evol Appl 4: 326−337 studies. In: Balkenhol N, Cushman S, Storfer A, Waits L Shafer ABA, Wolf JBW (2013) Widespread evidence for (eds) Landscape genetics: concepts, methods, applica- incipient ecological speciation: a meta-analysis of isola- tions. John Wiley & Sons, Chichester, p 77−98 tion-by-ecology. Ecol Lett 16:940−950 Wang IJ, Bradburd GS (2014) Isolation by environment. Mol Siegel DA, Kinlan BP, Gaylord B, Gaines SD (2003) Lagran - Ecol 23: 5649−5662 gian descriptions of marine larval dispersion. Mar Ecol Waples RS (1987) A multispecies approach to the analysis of Prog Ser 260:83–96 gene flow in marine shore fishes. Evolution 41: 385−400 Selkoe et al.: A decade of seascape genetics 19

➤ Waples RS (1998) Separating the wheat from the chaff: pat- dispersal and connectivity of broadcast spawning corals terns of genetic differentiation in high gene flow species. at the global scale. Glob Ecol Biogeogr 23: 1−11 J Hered 89:438−450 ➤ Woodson CB, McManus MA (2007) Foraging behavior can ➤ Waples RS, Gaggiotti O (2006) What is a population? An influence dispersal of marine organisms. Limnol Oceanogr empirical evaluation of some genetic methods for identi- 52: 2701−2709 fying the number of gene pools and their degree of con- ➤ Wright D, Bishop JM, Matthee CA, Heyden S (2015) Genetic nectivity. Mol Ecol 15: 1419−1439 isolation by distance reveals restricted dispersal across a ➤ Waters JM, Fraser CI, Hewitt GM (2013) Founder takes range of life histories: implications for biodiversity con- all: density-dependent processes structure biodiversity. servation planning across highly variable marine envi- Trends Ecol Evol 28:78−85 ronments. Divers Distrib 21:698−710 ➤ White C, Selkoe KA, Watson J, Siegel DA, Zacherl DC, Too- ➤ Young A, Boyle T, Brown T (1996) The population genetic nen RJ (2010) Ocean currents help explain population consequences of habitat fragmentation for plants. Trends genetic structure. Proc R Soc B 277:1685−1694 Ecol Evol 11: 413−418 ➤ Whitlock MC, McCauley DE (1999) Indirect measures of ➤ Young EF, Belchier M, Hauser L, Horsburgh GJ and others gene flow and migration: FST ≠ 1/(4Nm+1). Heredity 82: (2015) Oceanography and life history predict contrasting 117−125 genetic population structure in two Antarctic fish spe- ➤ Wood S, Paris CB, Ridgwell A, Hendy EJ (2014) Modelling cies. Evol Appl 8: 486−509

Editorial responsibility: Philippe Borsa, Submitted: March 4, 2016; Accepted: June 6, 2016 Montpellier, France Proofs received from author(s): July 9, 2016