ECOGRAPHY

Research
Spatial graphs highlight how multi-generational dispersal shapes
landscape genetic patterns

Emilie Boulanger, Alicia Dalongeville, Marco Andrello, David Mouillot and Stéphanie Manel

E. Boulanger (https://orcid.org/0000-0002-6446-7342) and S. Manel (https://orcid.org/0000-0001-8902-6052) ✉ ([email protected]),

CEFE, Univ. Montpellier, CNRS, EPHE-PSL, IRD, Univ. Paul Valéry Montpellier, Montpellier, France. – A. Dalongeville (https://orcid.org/0000-0002-
8969-8243), People and The Sea, Logon, Daanbantayan, Cebu, Philippines. – M. Andrello (https://orcid.org/0000-0001-7590-2736), D. Mouillot
(https://orcid.org/0000-0003-0402-2605) and EB, MARBEC, Univ. Montpellier, CNRS, Ifremer, IRD, Montpellier, France.

Ecography Current approaches that compare spatial genetic structure of a given species and the
43: 1–13, 2020 dispersal of its mobile phase can detect a mismatch between both patterns mainly
doi: 10.1111/ecog.05024 due to processes acting at different temporal scales. Genetic structure result from
gene flow and other evolutionary and demographic processes over many generations,
Subject Editor: Pedro Peres-Neto while dispersal predicted from the mobile phase often represents solely one genera-
Editor-in-Chief: Miguel Araújo tion on a single time-step. In this study, we present a spatial graph approach to land-
Accepted 17 April 2020 scape genetics that extends connectivity networks with a stepping-stone model to
represent dispersal between suitable habitat patches over multiple generations. We
illustrate the approach with the case of the striped red mullet Mullus surmuletus in
the Mediterranean Sea. The genetic connectivity of M. surmuletus was not correlate
with the estimated dispersal probability over one generation, but with the stepping-
stone estimate of larval dispersal, revealing the temporal scale of connectivity across
the Mediterranean Sea. Our results highlight the importance of considering multiple
generations and different time scales when relating demographic and genetic con-
nectivity. The spatial graph of genetic distances further untangles intra-population
genetic structure revealing the Siculo-Tunisian Strait as an important corridor rather
than a barrier for gene flow between the Western- and Eastern Mediterranean basins,
and identifying Mediterranean islands as important stepping-stones for gene flow
between continental populations. Our approach can be easily extended to other sys-
tems and environments.

Keywords: genetic connectivity, Mediterranean Sea, Mullus surmuletus, seascape

genetics, spatial graphs, stepping-stone dispersal

Introduction
Connectivity and gene flow influence the evolutionary dynamics of spatially
structured populations, maintain genetic diversity and promote population adap-
tive potential and resilience after disturbance (Hughes and Stachowicz 2004,
Vandergast et al. 2008, Baguette et al. 2013, Donati et al. 2019). Connectivity refers
to the movement of individuals in a heterogenous landscape (Taylor et al. 1993).
––––––––––––––––––––––––––––––––––––––––
© 2020 The Authors. Ecography published by John Wiley & Sons Ltd on behalf of Nordic Society Oikos
www.ecography.org This is an open access article under the terms of the Creative Commons
Attribution License, which permits use, distribution and reproduction in any
medium, provided the original work is properly cited. 1
When individual dispersal is followed by successful repro- reef fishes (e.g. Plectropomus maculatus in the southern Great
duction, demographic connectivity results in genetic con- Barrier Reef, Bode et al. 2019) and lobsters (e.g. Panulirus
nectivity, a measure of gene flow and other evolutionary argus in the Caribbean sea, Truelove et al. 2016), and seden-
processes (Lowe and Allendorf 2010). Understanding how tary species such as sea cucumbers (e.g. Parastichopus californi-
populations are connected across space and time is essential cus in the North-Eastern Pacific, Xuereb et al. 2018), molluscs
to assess the impact that habitat change and fragmentation (e.g. Kelletia kelletii in the Santa Barbara Channel, White et al.
can have on population persistence, and to develop sustain- 2010) and seagrasses (e.g. Zostera marina in the North Sea,
able resource management and appropriate conservation Jahnke et al. 2018). The two latter studies, White et al. (2010)
strategies (Fischer and Lindenmayer 2007, Andrello et al. and Jahnke et al. (2018), integrated dispersal over multiple
2015, Magris et al. 2018). It can also help to quantify the generations by applying Markov Chain matrix multiplication
potential spreading of adaptive alleles in areas threatened but missed herein a spatially explicit consideration.
by climate change (Razgour et al. 2019). Landscape genetic Spatial graphs (Box 1) can help to better understand the
studies aim to characterize how habitat and environmental dynamics of complex systems composed of many interact-
features promote or impede the movement of individu- ing units (Rozenfeld et al. 2008, Dale and Fortin 2010,
als through landscapes, riverscapes and seascapes and thus Peterson et al. 2019). They offer a promising tool to study
influence microevolutionary processes, including gene multi-generational dispersal by integrating stepping-stone
flow and genetic drift (Manel et al. 2003, Anderson et al. dispersal through multiple sites and multiple pathways, and
2010). For species whose connectivity is mainly realised to study the spatial structure of gene flow between popula-
by a dispersing propagule stage (e.g. eggs, larvae, seeds, tions (Murphy et al. 2015). Applications of spatial graphs to
spores), much interest has been given to understanding the model multi-generational dispersal are scarce and restricted
drivers of population connectivity by estimating the dis- to the marine environment (e.g. modelling the dispersal of
persal of these propagules through water, wind or animal rafting brown algae at small spatial scale; Buonomo et al.
vectors and by assessing the extent to which propagule dis- 2017, and the dispersal of coral larvae at the scale of the Great
persal explains the observed genetic structure across space Barrier Reef; Riginos et al. 2019). To our knowledge, only
(Nathan and Muller-Landau 2000, Castorani et al. 2017, one recent study (Jahnke et al. 2018) applied spatial graphs
Escalante et al. 2018). to explicitly compare patterns of demographic and genetic
Comparing genetic connectivity and propagule dispersal connectivity, but this study did not calculate the multi-
requires matching propagule dispersal to the temporal scales generational connectivity.
of the processes shaping genetic connectivity. While propagule Here, we present a spatial graph approach that extends
dispersal occurs within one generation, genetic structure results propagule dispersal networks with a stepping-stone model
from gene flow and demographic processes over many genera- to represent dispersal between suitable habitat patches over
tions which can, in the absence of strong barriers to dispersal, multiple generations. This approach allows to estimate the
lead to connectivity over large spatial scales (Hedgecock et al. multi-generation long-distance connectivity that is poten-
2007). Considering only single-generation dispersal events tially detected using genetic data, and to compare explicitly
ignores the effect of stepping-stone dispersal through habitat the spatial structure of multi-generational genetic connec-
patches leading to genetic connectivity over multiple genera- tivity, single-generation propagule dispersal and multi-gen-
tions (Saura et al. 2014). We thus need to integrate propagule eration dispersal networks. Our approach also accounts for
dispersal across time to better understand genetic connectivity unsampled stepping-stone sites that contribute to the con-
patterns and understand how multi-generational stepping- nectivity patterns between the sampled sites (the so-called
stone dispersal shapes genetic population structure. ‘ghost population’ effect; Beerli 2004, Slatkin 2005). We
In the terrestrial environment, most landscape genetic then illustrate our approach by analyzing connectivity of a
studies analyse genetic connectivity using least-cost path esti- high gene-flow species, the red mullet Mullus surmuletus, in
mates of habitat connectivity (Epps et al. 2007, Castillo et al. the Mediterranean Sea. Finally, we discuss how to extend our
2014, Row et al. 2018, Schoville et al. 2018). Some recent approach to other systems.
advances have been made to combine spatial and tempo-
ral scales in landscape connectivity models. Saura et al.
(2014) developed a generalized habitat network connec- Methods
tivity model which accounts for the potential role of step-
ping-stones enhancing species dispersal across generations. General approach: multi-generational spatial graphs for
Martensen et al. (2017) modelled spatio-temporal networks genetic connectivity
of connectivity in dynamic landscapes where patch availabil-
ity changes through time. However, neither of these studies The first step of our approach to study multi-generational genetic
tested how such improved estimates of landscape connectiv- connectivity is to model propagule dispersal in a given environ-
ity correlate with genetic connectivity. ment or habitat for a single generation. This first step produces
In the marine environment, important advances combined dispersal probabilities between each pair of sites used to build a
biophysical models and genetic analyses to describe popula- spatial graph (Box 1) of single-generation dispersal (Treml et al.
tion connectivity. Examples include mobile species such as 2008, Kininmonth et al. 2010, Andrello et al. 2013).

2
 Box 1. How spatial graphs can illuminate connectivity

Spatial graphs stem from graph theory and help to better understand the dynamics of complex systems composed of many
interacting units (Rozenfeld et al. 2008, Dale and Fortin 2010, Peterson et al. 2019). They explicitly consider the spatial
context of data (Urban et al. 2009) and can be used to study dispersal and connectivity between multiple sites simultane-
ously (Dale and Fortin 2010, Eros et al. 2011). In graph theory, a graph G consists of a set of nodes or vertices that are
pairwise linked by a set of edges E(G) (Urban and Keitt 2001). A graph can be built to study connectivity in a network by
using populations or sampling localities as georeferenced nodes, rendering a spatial graph, and weighing the edges between
them by their modelled dispersal to represent connection probability, or by their pairwise genetic distance to study gene
flow (under the hypothesis that genetic differentiation is mainly the outcome of gene flow). In doing so, graph theory
allows to visually represent the network of genetic connectivity and analyse topological features to test hypotheses of popu-
lation structure (Dyer and Nason 2004, Dale and Fortin 2010, Xuereb et al. 2018). At the same time, graph theory allows
the quantitative estimation of network- and node-level properties (Rozenfeld et al. 2008, Albert et al. 2013).

Interpreting graph metrics

Node metrics allow the identification of nodes, e.g. populations, that are of increased importance to the whole network
by simultaneously considering all nodes and edges of the graph. High node degrees indicate populations that exchange
propagules with many others (in a dispersal graph) or that have already exchanged genes with many others (in a genetic
connectivity graph). Such populations have potentially sent or received many migrants over evolutionary time scales,
acting as sources or sinks in the network (Rozenfeld et al. 2008). Node betweenness centrality quantifies the importance
of single nodes in passing information through the network by identifying nodes that represent important stepping-
stones for dispersers (in propagule dispersal graphs) or genes (in genetic connectivity graphs) (Garroway et al. 2008,
Rozenfeld et al. 2008, Andrello et al. 2013). Network-level metrics allow us to quantitatively characterize the topology of
a graph. They furthermore allow us to test landscape-genetic hypotheses by comparing the topology and associated char-
acteristics of a genetic connectivity graph to graphs representing explanatory variables relating to structural or functional
connectivity (Dale and Fortin 2010).

The second step examines nodes that are not directly con- allows to calculate multi-generational connectivity through
nected in the single-generation dispersal graph and calculate stepping-stone nodes (Fig. 1, multi-generational propagule
the number of stepping-stones to connect the shortest paths dispersal). To calculate the shortest paths, dispersal prob-
between all pairs of nodes (Table 1). For example, if no direct abilities between nodes are transformed into distance mea-
connection exists between nodes a and c (Fig. 1, single-gener- sures by taking the natural logarithm of their inverse; this
ation propagule dispersal), the shortest path algorithm iden- transformation yields a distance matrix where pairs of nodes
tifies node b as the most suitable stepping-stone and indicates with the highest dispersal probabilities have the smallest dis-
that one stepping-stone is needed to connect nodes a and c tance, thereby ensuring that the shortest path between two
through b. In this way, the single-generation dispersal graph nodes is effectively the most probable one (Costa et al. 2017).

Table 1. Spatial graph metrics used in this study (Urban and Keitt 2001, Newman 2003, Urban et al. 2009).
Network assortativity Measure of how well nodes in a network show similar characteristics. Typically calculated as degree assortativity,
which quantifies whether high-degree nodes tend to attach to other high-degree nodes (assortative mixing) or
whether high-degree nodes rather attach to low-degree ones (disassortative mixing). Measured as the Pearson
correlation coefficient r of the degrees at either ends of an edge, averaged over edges.
Network clustering Probability that two nodes connected to a given third node are themselves connected, averaged for all the nodes in
a network.
Node betweenness Fraction of all shortest paths between nodes that passes through a given node. Quantifies the relative importance of
a single node in passing information through the network. The betweenness centrality (CB) of node i is calculated
by taking the proportion of shortest paths connecting nodes s and t (σst) that passes through node i (σst(i)), and
summing them over all possible node pairs:
s st (i )
C B (i ) = å
s ¹t ¹i
s st
Node degree Number of edges connected to a given node. Quantifies the local importance of a node.
Shortest path Minimum-weight path connecting two nodes. In an unweighted graph, the shortest path represents the path with
the smallest number of edges. In a weighted graph, the shortest path represents the path in which the total weight
of the sequence of edges is the smallest.

3
Figure 1. Spatial graphs to analyse single-generation propagule dispersal (top), multi-generational propagule dispersal (middle) and multi-
generational genetic connectivity networks (bottom). Dots and arrows represent the nodes and the directional edges of the graph. Black dots
are the sampled populations and grey dots are unsampled populations. Black and grey arrows are modelled propagule dispersal probabilities
between sampled and unsampled populations, respectively. The width of the arrows is proportional to propagule dispersal probability (in
the single-generation propagule dispersal graph) and inversely proportional to genetic distance (in the multi-generational genetic connectiv-
ity graph). In the multi-generational propagule dispersal graph, dispersal probabilities are used to estimate the number of stepping-stones
needed to connect nodes that are not connected in the single-generation graph. Dashed lines represent multi-generational connectivity
through stepping-stones and are highlighted for nodes a and c (brown) and nodes b and f (green). Adapted from Urban et al. (2009).

The shortest paths between all possible node pairs can then to as multi-generational genetic connectivity (Fig. 1)
be estimated using the Floyd–Warshall algorithm for directed (Rozenfeld et al. 2008, Fortuna et al. 2009, Albert et al.
graphs (Floyd 1962, Warshall 1962). Unsampled sites poten- 2013). The apparent discrepancies between genetic con-
tially acting as stepping-stones are included in the multi-gen- nectivity and propagule dispersal can in part be explained
erational propagule dispersal graph. by the process of multi-generational dispersal connect-
The third step of our approach estimates a spatial ing nodes over larger spatial scales than single-generation
graph of genetic distances among sites, hereafter referred dispersal.

4
Example: multi-generational genetic connectivity in In brief, three-dimensional sea current velocities were
the striped red mullet across the Mediterranean Sea obtained from the hydrodynamics model NEMOMED12
(Beuvier et al. 2012) and used to simulate passive larval dis-
To illustrate our approach, we re-analysed a published genetic persion using the software Ichthyop 3.1 (Lett et al. 2008).
dataset of an economically important fish species with a Larvae were released in 1/10th degree cells covering the
pelagic larval stage, the striped red mullet Mullus surmule- Mediterranean continental shelf (7703 cells; Andrello et al.
tus, in the Mediterranean Sea (Dalongeville et al. 2018b). We 2015) every three days during the species’ spawning season
hypothesize that building a spatial graph of potential larval (1–28 May) and allowed to passively disperse during 30 d.
dispersal over multiple generations, can improve our under- Pairwise probabilities of larval dispersal c(i,j) between cells i
standing of the genetic structure across the Mediterranean and j were calculated from the numbers of simulated larvae
Sea and of the temporal scale at which connectivity operates. released in cell j that arrived in cell i after 30 d. Probabilities
We build the spatial graph of genetic connectivity to charac- of dispersal were averaged over a grid of 100 cells covering
terize the spatial pattern of genetic structure and to compare the full coastal range of the Mediterranean Sea, resulting in
it to the graph of multi-generational larval dispersal. a 100 × 100 asymmetric connectivity matrix C100. Not every
grid cell could be sampled for the genetic analysis described
Species description, study area, sampling and genetic data above. To allow comparison between the larval dispersal
The striped red mullet Mullus surmuletus is a demersal fish probabilities and the genetic connectivity estimates between
species that is commonly found in the Mediterranean Sea the 47 sampled sites, a 47 × 47 connectivity matrix C47 was
particularly in areas characterised by a narrow continental extracted from the C100 matrix with the cell centroids cor-
shelf with rough substrate and at depths ranging from 10 to responding to the 47 sites sampled for the genetic analysis
100 m (Lombarte et al. 2000). Adults move to deeper sites (Supplementary material Appendix 1 Fig. A1).
for spawning (70–150 m, Machias et al. 1998), which occurs We constructed directed spatial graphs of single-generation
from April to May and produces larvae with a pelagic larval larval dispersal both for the 47 × 47 and 100 × 100 larval con-
duration (PLD) of approximately 30 d (Reñones et al. 1995, nectivity matrices. Edges of the spatial graphs were weighted
Macpherson and Raventos 2006, Arslan and İş 2015). Mullus by the modelled larval dispersal probabilities and nodes rep-
surmuletus is one of the most economically valuable species in resented the cell-centroids of the 47 populations and the 100
commercial landings of coastal Mediterranean demersal fish- modelled sites, respectively. The 47 populations are connected
eries (Reñones et al. 1995, Félix-Hackradt et al. 2013). through single-generation larval dispersal by 108 edges.
The Mediterranean Sea harbours a complex oceanic circu- Marine geographic distances (minimum distances con-
lation that shapes larval dispersal patterns and creates puta- strained by water) were computed as least-cost path distances
tive barriers for gene flow (Pascual et al. 2017). To maximise between all pairs of populations by assigning infinite resis-
the detection of these various processes influencing popula- tance to land areas and constant zero resistance to water using
tion structure, M. surmuletus samples were collected along the functions ‘transition’, ‘geoCorrection’ and ‘costDistance’
the entire Mediterranean coastal range including islands of the R package ‘gdistance’ v. 1.1-1 (van Etten 2017). The
(47 sites, Fig. 2a). Fin clips were taken from specimens nearest geographical neighbour was additionally determined
obtained at local artisanal fisheries landing sites. DNA was for each population.
extracted using the DNeasy Blood & Tissue Kit (Qiagen)
according to the manufacturer’s protocol, and individuals Step 2. Multi-generational propagule dispersal
were genotyped using a Genotyping by Sequencing method Both larval dispersal matrices (C100 and C47) reflect dispersal
(Elshire et al. 2011). Individual genotypes were then pooled only from the current generation. We calculated larval con-
by their sampling locality to maximize sequencing coverage nections for the matrix C100 between non-connected nodes
(see Dalongeville et al. 2018b for a complete description of based on a stepping-stone approach. We hypothesize that lar-
sampling, molecular and bioinformatics procedure). This vae that dispersed from site j to i successfully settle in their
resulted in a dataset containing the allele frequencies of 1153 arrival site and, upon reaching maturity, produce larvae that
Single Nucleotide Polymorphism (SNP) loci for 47 pools of potentially disperse to other sites. As M. surmuletus popula-
nine to 18 individuals, hereafter referred to as populations. tions occur across the full Mediterranean coastal range, we
Genetic differentiation among these populations was cal- hypothesize that all the 100 grid cells represent suitable habi-
culated in the ‘ade4’ package in R v3.2.3 (Dray and Dufour tat and all sites contribute equally to larval production.
2007, R Core Team) using Cavalli-Sforza and Edwards’ After transforming the single-generation larval dispersal
chord distance Dc (Cavalli-Sforza and Edwards 1967), which probabilities of C100 into distance measures, we calculated
is well-suited to distinguish genetically similar populations in the shortest paths between all possible node pairs using the
high gene flow species (Libiger et al. 2009). Floyd–Warshall algorithm embedded in the R package ‘Rfast’
(Papadakis et al. 2018) (Supplementary material Appendix
Step 1. Single-generation propagule dispersal 1 Fig. A2). We then extracted the multi-generational dis-
Larval dispersal was simulated using the biophysical model of persal measures between the 47 sampled populations from
Andrello et al. (2013), which was adapted to fit the life his- the C100 multi-generational dispersal graph (Supplementary
tory parameters of M. surmuletus (Dalongeville et al. 2018b). material Appendix 1 Fig. A3). The final spatial graph of

5
Figure 2. Spatial graph of genetic connectivity. (a) Map of the Mediterranean Sea with the location of the 47 sampling sites, the Siculo-
Tunisian Strait (STS), Mediterranean regions and ecoregions (ALB, Alboran Sea; ADR, Adriatic Sea; ION, Ionian Sea; AEG, Aegean Sea;
LEV, Levantine Sea). (b) Spatial graph representing the genetic connectivity network of Mullus surmuletus in the Mediterranean Sea. Edges
between nodes (populations) are weighted by their pairwise genetic distance (Dc), represented by a colour gradient (dark red for the smallest
distances and thus the highest genetic similarity). The size of the nodes is proportional to their betweenness centrality.

6
multi-generational dispersal is constructed using the 47 pop- subsequent graph–theoretic analyses were conducted with the
ulations as nodes and weighting the edges by the number of R package ‘igraph’ (functions ‘degree’, ‘betweenness’, ‘assor-
stepping stones connecting them. tativity_degree’ and ‘transitivity’) (Csardi and Nepusz 2006,
Jombart 2008). Random networks were generated using the
Step 3. Spatial graph of multi-generational genetic ‘sample_gnm’ function.
connectivity Correlations between genetic distances and marine
The network of genetic connectivity was constructed by des- geographic distances, single-generation larval dispersal
ignating the 47 sampled populations as nodes, and their pair- probabilities as well as multi-generational larval dispersal
wise genetic distance (Dc) as weighted edges. The construction distances were estimated using the Mantel coefficient along
of a spatial graph from all Dc values results in a fully saturated with 9999 permutations (R package ‘vegan’, Oksanen et al.
network, where every node is connected to every other node. 2016). As the Mantel test is known to have an inflated type
The calculation of most graph metrics, such as node degree, I error rate and low statistical power (Legendre and Fortin
network assortativity and network clustering, depends on the 2010, Guillot and Rousset 2013), we also computed maxi-
number of edges connected to the graph’s nodes and can- mum-likelihood population-effects (MLPE) mixed models
not be calculated in a saturated network (Table 1). As the (Clarke et al. 2002) which account for the non-indepen-
weights of the genetic graph have a small range (see further), dence of pairwise data and provide the greatest probability
the shortest-path algorithm always identified the direct path of identifying the true model from competing alternatives
between two nodes as the most efficient one and impeded the (Shirk et al. 2018). We used the ‘lmer’ function of the
calculation of informative node betweenness. We thus pro- R package ‘lme4’ to fit the MLPE mixed models with a ran-
ceeded to prune the graph edges to get a more informative dom-effects term accounting for population-level influence
graph topology (Garroway et al. 2008). Edges were pruned (Bates et al. 2015, Row et al. 2017). Model fit was evaluated
by retaining only the 108 edges with the smallest genetic using the coefficient of determination R2 calculated with
distance (approximately 10% of all edges), thus obtaining a the ‘MuMIn’ package (Bartoń 2019).
multi-generational genetic connectivity graph with the same
number of edges as the single-generation larval dispersal
graph. The retained connections are considered as the rel- Results
evant genetic relationships for further network analyses.
As this pruning parameter is arbitrary, we also imple- Single-generation larval dispersal
mented an edge removal scenario where edges of decreasing
genetic distances were discarded one by one. The resulting The larval dispersal matrix C47 is highly sparse, with only 108
genetic connectivity spatial graphs were compared after each realised connections out of 1081 (Supplementary material
removal to test the influence of the pruning parameter choice Appendix 1 Fig. A4). The constructed larval network con-
on the graph topology. Additionally, we ranked each node nects all nodes to at least one other node with connections
according to their betweenness values for each node removal primarily made to the nearest neighbours. It is fragmented
scenario and compared the variation in node ranks. in three main clusters roughly representing the western
Mediterranean, the Aegean sea and the Levantine sea (Fig. 3a,
Analysis of the spatial graphs Supplementary material Appendix 1 Fig. A5). The graph has
Network topology was analysed by estimating several graph a high level of clustering and a positive degree assortativity
metrics (Table 1). For the multi-generational genetic connec- (Supplementary material Appendix 1 Table A1).
tivity graph, we calculated node degree and node between- The C100 network of modelled larval dispersal however
ness centrality to identify important stepping-stone nodes in fully connects the Mediterranean Sea, i.e. the network con-
the network (Box 1) (Rozenfeld et al. 2008). Network assor- sists of one large cluster connecting all nodes to at least three
tativity and network clustering were calculated to evaluate other nodes (Supplementary material Appendix 1 Fig. A6).
the structure of the genetic network. To determine whether
assortativity and clustering resulted from biological processes Multi-generational larval dispersal
(such as habitat patch location and quality and between-hab-
itat migration) versus random configuration, we generated The stepping-stone approach to larval connectivity converted
10 000 random networks with the same number of nodes our sparse connectivity matrix C47 into a fully saturated con-
and edges as our genetic connectivity network following nectivity matrix, although a substructuring separating the
the Erdös–Rényi model (Erdös and Rényi 1959). The mean western and eastern basin was still visible (Fig. 3a). The spa-
graph metrics (assortativity and clustering) and their stan- tial graph of the multi-generational larval dispersal network
dard deviations were calculated and compared to the values shows that the Aegean and Levantine Sea are connected when
of the genetic graph. Network assortativity and clustering including one stepping-stone (Fig. 3b), and that the west–
were also calculated for the single-generation propagule dis- east barrier is breached when including two stepping-stones
persal graph. (Fig. 3c). Allowing for more stepping-stones increased graph
The pruned genetic graph was constructed with the connectivity (Fig. 3d–e), saturating the network with 24
‘gengraph’ function in the R package ‘adegenet’, and all stepping-stones (suggesting 25 generations).

7
Figure 3. Temporal perspective of marine connectivity: spatial graphs representing the amount of generations needed to connect our 47
populations through stepping-stone larval dispersal. (a) Graph with all pairwise edges, the colours represent the number of edges in the
shortest path (= number of generations) connecting each node pair. The darkest green colour (1 edge) represents the direct larval dispersal
over one generation. The Mediterranean Sea is fully connected through larval dispersal when allowing up to 25 edges (or 24 stepping stones)
in the shortest paths (a). (b–e) Graph built using edges representing two, three, four and five generations respectively.

Multi-generational genetic connectivity analysis of the graph reveals no substructuring into clus-
ters but instead more and stronger connections between
Pairwise genetic distance Dc varied between 0.10 and 0.20. the western and eastern Mediterranean basin than within
The pruned spatial graph built with these genetic dis- each basin. Network metrics showed a high level of disas-
tances connected 29 out of 47 nodes (Fig. 2b). Topological sortative mixing with r = −0.57, which was considerably

8
lower than that of the randomized networks (r0 = −0.05) both p < 0.001) (Table 2) indicating a pattern of isolation by
(Supplementary material Appendix 1 Table A1), which distance (IBD). The Mantel test and MLPE model between
indicates a centralized network. This means that the genetic genetic distances and single-generation larval dispersal were
network has a central core of a few highly connected popu- not significant (Table 2). Conversely, the Mantel test and
lations (i.e. with a high degree), that are in turn linked to MLPE model between genetic distances and multi-genera-
lesser connected populations (i.e. with a low degree) in the tional larval dispersal revealed a significant positive correla-
periphery (Supplementary material Appendix 1 Fig. A7). tion (Mantel r = 0.18, MLPE R2m = 0.12, both p < 0.001)
The level of clustering c = 0.58 was considerably higher than (Table 2), indicating an isolation by larval dispersal when
the mean of the randomized graphs (c0 = 0.10). These stron- considering multiple generations. The multi-generational
ger levels of disassortativity and clustering demonstrate that estimates of larval dispersal were also significantly correlated
the genetic connectivity network is more structured than to the geographic distances (Mantel r = 0.70, p = 0.001).
would be expected from a random process alone. The cen-
tralised structure was robust under varying levels of pruning
(Supplementary material Appendix 1 Fig. A8). The sequen- Discussion
tial removal of edges also shows that the unconnected nodes
in the pruned graph are not completely isolated but rather Here we demonstrate how stepping-stone dispersal graphs
increasingly differentiated from the central structure. can be used as a proxies of multi-generational connectivity
The values of betweenness centrality also suggest a central- and increase our understanding of genetic connectivity. We
ized structure, which attribute an importance to only six out present a spatial graph approach applied to red mullet in the
of the 29 connected nodes. In particular, populations 11_86 Mediterranean Sea that calculates shortest paths between
in Sicily (present in 50.0% of all shortest paths) and 55_56 populations unconnected by direct modelled larval dispersal
in Turkey (23.3%) appear to be the main stepping stones for but showing strong genetic connectivity. The increased cor-
the gene flow through the network (Fig. 1b). The remain- relation between the resulting multi-generational connectiv-
ing populations with a nonzero betweenness are 19_27 in ity and genetic connectivity, compared to single-generation
Corsica (9.9%), 13_15 in Sardinia (8.1%), 29_96 in Crete dispersal, highlights the underlying discrepancy in spatio-
(7.8%) and 93_94 in Cyprus (0.9%). These six nodes con- temporal scales in our landscape genetic study.
sistently ranked highest out of all nodes by their betweenness The spatial graph of single-generation larval dispersal
values under varying pruning criteria (Supplementary mate- was mostly realized between nearest neighbours, resulting
rial Appendix 1 Fig. A9). in a highly clustered structure. The spatial graph of genetic
connectivity revealed an absence of subdivision between the
Comparison between single-generation larval western and eastern Mediterranean sub basins and a stronger
dispersal, multi-generational larval dispersal and between-basin than within-basin connectivity. No significant
multi-generational genetic connectivity correlation could be found between the genetic distances
and the direct larval dispersal probabilities at the scale of the
The mean larval dispersal was estimated to 266 km (SD = 137) Mediterranean Sea.
for the single-generation dispersal graph and to 1139 km Visualisation of the larval dispersal and gene flow net-
(SD = 694) for the genetic connectivity graph, revealing the work showed that genetic connectivity is realised at a much
different spatial scales of both components of connectivity. larger spatial scale than larval dispersal, as already found by
Sixty-six of the 72 nearest neighbour pairs were connected in Dalongeville et al. (2018b), reflecting different temporal
the single-generation larval dispersal graph, whereas only 16 scales. Biophysical connectivity is usually modelled for a single
of the nearest neighbour pairs were connected in the genetic generation of larvae (White et al. 2010), whereas genetic con-
connectivity graph. nectivity integrates gene flows over multi-generational time
The Mantel test and maximum-likelihood population- scales (Hedgecock et al. 2007). Intermediate stepping-stones
effects (MLPE) mixed model between the genetic distances for dispersal are crucial in maintaining genetic connectivity
(pairwise Dc) and the geographic distances for the 47 popu- across large spatial scales (Crandall et al. 2012). The short-
lations were significant (Mantel r = 0.31, MLPE R2m = 0.17, est-path approach to larval dispersal explores the temporal

Table 2. Mantel test and maximum-likelihood population-effects (MLPE) mixed model results comparing the multi-generational genetic
connectivity (Cavalli–Sforza and Edwards’ chord distance Dc) to the geographic distance, single-generation propagule dispersal (larval dis-
persal probability cij) and the multi-generational propagule dispersal (stepping-stone larval distance). Only the geographic distance and
stepping-stone larval distance were significantly correlated to the genetic chord distance (p < 0.001).
Genetic chord distance Dc
Predictor variable Mantel correlation p value MLPE R2m p value
Geographic distance r = 0.31 < 0.001 0.17 < 2.2e-16
Larval dispersal probability (cij) r = 0.02 0.24 0.0007 0.37
Stepping-stone larval distance (Floyd–Warshall shortest r = 0.18 < 0.001 0.12 < 2.2e-16
path on ln(1/cij))

9
aspect of marine connectivity by suggesting how sites would importance of a single node in relaying information through
be connected if M. surmuletus larvae were successfully spread the network (Rozenfeld et al. 2008). The populations with
through the Mediterranean Sea over multiple generations fol- high genetic betweenness could thus be interpreted as
lowing a stepping-stone model. Multigenerational dispersal prominent stepping-stones for gene flow. These are popu-
can thus increase the explanatory power of dispersal models lations where individuals from multiple locations seem to
for population genetic analyses, and our understanding of the have aggregated and reproduced, to subsequently further
drivers of population genetic structure (White et al. 2010, spread their genes to multiple other locations. Six popula-
Buonomo et al. 2017). tions, located in Sicily (11_86), Turkey (55_56), Corsica
Connectivity can promote the spread of adaptive alleles (19_27), Sardinia (13_15), Crete (29_96) and Cyprus
from populations locally adapted to presently extreme envi- (93_94) were identified as important stepping-stones in our
ronmental conditions that could become more common in network. These populations lay along the longitudinal axis
the future. Such spreading of adaptive alleles can strengthen and seem to form a path connecting the western and eastern
the adaptation of populations to future environmental con- Mediterranean. Remarkably, five out of these six populations
ditions, and in some cases even result in genetic rescue of are located along islands and coincide with the location of the
declining populations (Whiteley et al. 2015, Xuereb et al. high degree nodes. These results detect islands as important
2019). In the marine environment, populations can be stepping-stones in the network-wide genetic connectivity of
locally adapted to climate-related variables such as salin- M. surmuletus, and suggest that islands can be important sites
ity (Dalongeville et al. 2018a). Climate projections under for gene flow between continental populations by provid-
a business-as-usual scenario (RCP8.5) predict that sea sur- ing intermediate suitable habitats for larval settlement in the
face salinity (SSS) in the Mediterranean Sea will increase, deeper parts of the Mediterranean basin.
with marked regional differences, by 0.13 practical salinity The construction of a genetic spatial graph requires defin-
unit (± 0.13 PSU) for the 2021–2050 period and to come ing a criterion to prune the initially saturated graph. The
back to its current global climate value (± 0.01 PSU) for the appropriate strategy depends largely on the questions asked.
2071–2100 period (Moullec et al. 2019). In some areas, the For example Dyer and Nason (2004) developed the popula-
increase in SSS might result in extreme environmental condi- tion graph approach where graph edges are trimmed based
tions that cannot be tolerated by populations of M. surmule- on the genetic covariance among populations. This approach
tus, because the future conditions will be too different from requires individual genotypes and is not applicable to our
the present conditions to which the populations are adapted population-level data. Because one of our hypotheses was
(Rellstab et al. 2016). In order to persist, these populations that genetic connectivity is in part the result of larval dis-
will have to move to more suitable areas or adapt to the new persal, edges were pruned by retaining only the 108 edges
local conditions from standing genetic variation and/or immi- with the smallest genetic distance (approximately 10% of all
gration of adaptive genotypes. Analysis of the association edges), thus obtaining a multi-generational genetic connec-
between SNP and salinity have suggested a signal of adap- tivity graph with the same number of edges as the single-
tation to local water salinity in Mediterranean populations generation larval dispersal graph. The disassortative structure
of M. surmuletus (Dalongeville et al. 2018a). The results of with a strong cross-basin connectivity in the gene flow graph
our spatial graph analysis indicate that about five generations remained robust under the edge removal scenario. The nodes
will be necessary to spread beneficial alleles from populations with the higher betweenness values in the pruned gene flow
adapted to current high salinity in the east, to populations graph also consistently ranked higher under varying prun-
with similar projected future physiochemical conditions in ing values. The consistency of these results validates that our
the central Mediterranean (Fig. 3e). Given an average age at pruned graph correctly represents the gene flow patterns of
first reproduction of 1.5 yr for M. surmuletus (Reñones et al. M. surmuletus in the Mediterranean Sea.
1995), this corresponds to one decade which has been judged A large portion of the genetic differences still remains
as a small enough timescale to keep pace with the temporal unexplained. Bottom-trawl surveys suggest that M. sur-
scale of climatic changes (Jönsson and Watson 2016). muletus migrates to deeper water during maturation but
The strong basin-wide genetic connectivity highlighted realises rather short-distance than long-distance movement
by the genetic spatial graph identified the Siculo-Tunisian (Machias et al. 1998). Together with the small home ranges
Strait (STS) as an important corridor for gene flow. Beyond of sympatric demersal species (reviewed by Calò et al. 2013)
analysing the topology of a spatial network, graph theory this suggests that adult dispersal might be rather limited.
allows to quantitatively analyse the properties of the nodes Other processes that could explain the remaining genetic
and edges in the spatial graph (Rozenfeld et al. 2008). The structure are larval behaviour and higher-resolution ocean-
calculation of network- and node-metrics revealed a disas- ographic circulation patterns not resolved by the hydrody-
sortative, centralised network. This means that the network namic model used to simulate larval dispersal (Briton et al.
consists of a central core with a few high-degree hubs, which 2018, Faillettaz et al. 2018). Larvae at the settlement stage
are in turn linked to lesser connected nodes in the periph- can use oriented swimming driven by e.g. odour or sound
ery (Rozenfeld et al. 2008). These highly connected hubs to detect favourable settlement habitats (Simpson 2005,
could act as sources and/or sinks in the gene flow network Paris et al. 2013). Shoreward swimming at the end of the
(Peery et al. 2008). Betweenness centrality measures the pelagic stage can strongly influence the rate of recruitment

10
of larvae, resulting in longer dispersal distances than pre- References
dicted with purely passive dispersal (Faillettaz et al. 2017).
In contrast, vertical migration of larvae at night as well as Albert, E. M. et al. 2013. Assessing the robustness of networks of
homing behaviour of larvae returning to their natal reef result spatial genetic variation. – Ecol. Lett. 16: 86–93.
in greater retention than predicted by advection models and Anderson, C. D. et al. 2010. Considering spatial and temporal scale
smaller dispersal distances (Gerlach et al. 2007, Andrello et al. in landscape-genetic studies of gene flow. – Mol. Ecol. 19:
3565–3575.
2013, Bottesch et al. 2016). This would decrease the amount Andrello, M. et al. 2013. Low connectivity between Mediterranean
of connectivity in the larval dispersal system and would high- marine protected areas: a biophysical modeling approach for the
light even more the importance of stepping-stone migration dusky grouper Epinephelus marginatus. – PLoS One 8: e68564.
over multiple generations to realize the observed long-dis- Andrello, M. et al. 2015. Extending networks of protected areas to
tance genetic connectivity. Predicting how larval behaviour optimize connectivity and population growth rate. – Ecography
would affect larval dispersal patterns thus requires including 38: 273–282.
larval swimming and orientation mechanisms in the larval Arslan, M. and İş, A. 2015. Age, growth and reproduction of Mul-
dispersal model, but the parametrization of these processes lus surmuletus (Linnaeus, 1758) in Saros Bay (northern Aegean
would be difficult because of the lack of biological knowledge Sea). – J. Black Sea/Mediterr. Environ. 19: 217–233.
on larval dispersal in M. surmuletus. Baguette, M. et al. 2013. Individual dispersal, landscape connectiv-
ity and ecological networks. – Biol. Rev. 88: 310–326.
Although applied here to the larval dispersal of a demer-
Bartoń, K. 2019. MuMIn: multi-model inference. – R package ver.
sal fish species, the stepping-stone approach to dispersal 1.43.6.
can be generalized to other environments or living systems. Bates, D. et al. 2015. Fitting linear mixed-effects models using
Connectivity estimates can be applied to any organisms lme4. – J. Stat. Softw. 67: 1–48.
with a dispersing propagule stage, by modelling wind dis- Beerli, P. 2004. Effect of unsampled populations on the estimation
persal (Nathan et al. 2011), animal-mediated seed dispersal of population sizes and migration rates between sampled popu-
(García-Fernández et al. 2019) or riverine resistance (Schick lations. – Mol. Ecol. 13: 827–836.
and Lindley 2007, Oliveira et al. 2019). The approach can Beuvier, J. et al. 2012. MED12, oceanic component for the mod-
be adapted by modelling dispersal between species-specific eling of the regional Mediterannean Earth system. – Mercat.
suitable habitats such as coral reefs (Williamson et al. 2016), Ocean Q. Newsl. 46: 60–66.
forests (Wang et al. 2008) or breeding ponds (Fortuna et al. Bode, M. et al. 2019. Successful validation of a larval dispersal
model using genetic parentage data. – PLoS Biol. 17: e3000380.
2006, Decout et al. 2012). Additionally, the model could Bottesch, M. et al. 2016. A magnetic compass that might help coral
be improved by taking into account species-specific settle- reef fish larvae return to their natal reef. – Curr. Biol. 26:
ment success rates, population-specific productivity rates and R1266–R1267.
behavioural responses. Briton, F. et al. 2018. High-resolution modelling of ocean circula-
tion can reveal retention spots important for biodiversity con-
Data availability statement servation. – Aquat. Conserv. Mar. Freshwater Ecosyst. 28:
882–893.
Data for this study are available from the Dryad Digital Buonomo, R. et al. 2017. Habitat continuity and stepping-stone
Repository: < https://doi.org/10.5061/dryad.31tk592 > oceanographic distances explain population genetic connectiv-
ity of the brown alga Cystoseira amentacea. – Mol. Ecol. 26:
(Dalongeville et al. 2018c). The R script for the shortest-path 766–780.
analysis and to produce Fig. 3 is available at < https://github. Calò, A. et al. 2013. A review of methods to assess connectivity and
com/eboulanger/seaConnect--spatialGraph >. dispersal between fish populations in the Mediterranean Sea.
– Adv. Oceanogr. Limnol. 4: 150–175.
Castillo, J. A. et al. 2014. Landscape effects on gene flow for a
Acknowledgements – We thank Raquel Torres, Laure Velez, Jennifer climate-sensitive montane species, the American pika. – Mol.
Tournois, Eva Maire, Sebastien Villeger, Johanna Brun-Vitelli, Ecol. 23: 843–856.
Manuel Muntoni, Claire Seceh, Sofiène Ben Abdelhamid, Achraf Castorani, M. C. N. et al. 2017. Fluctuations in population fecun-
Barbena, Itai Granot, Ori Frid and Shahar Malamud for collecting dity drive variation in demographic connectivity and metap-
M. surmuletus samples, as well as all the fishermen that provided opulation dynamics. – Proc. R. Soc. B 284: 20162086.
the fishes. We are grateful to the SMGE platform, and especially Cavalli-Sforza, L. L. and Edwards, A. W. 1967. Phylogenetic anal-
Veronique Arnal and Marie-Pierre Dubois for technical support ysis. Models and estimation procedures. – Evolution 21:
during molecular analyses. We also thank Stéphane Lobreaux for 550–570.
the bioinformatic analyses. We thank the anonymous referees as Clarke, R. T. et al. 2002. Confidence limits for regression relation-
well as the editor for their insightful comments. ships between distance matrices: estimating gene flow with dis-
Funding – This study was funded by the ‘ANR’ through the tance. – J. Agric. Biol. Environ. Stat. 7: 361–372.
RESERVEBENEFIT project, PSL Environment through the project Costa, A. et al. 2017. On the calculation of betweenness centrality
RESMOD, and EB was partially funded by the Region Occitanie. in marine connectivity studies using transfer probabilities.
Author contributions – All authors generated hypotheses and – PLoS One 12: e0189021.
designed research. AD conducted the sampling and molecular Crandall, E. D. et al. 2012. Coalescent and biophysical models of
analyses; EB analysed the data and wrote the first draft. All authors stepping-stone gene flow in neritid snails. – Mol. Ecol. 21:
helped to improve manuscript writing. 5579–5598.

11
Csardi, G. and Nepusz, T. 2006. The igraph software package for García-Fernández, A. et al. 2019. Herbivore corridors sustain
complex network research. – InterJounal Complex Syst. 1695: genetic footprint in plant populations: a case for Spanish drove
1–9. roads. – PeerJ 7: e7311.
Dale, M. R. T. and Fortin, M.-J. 2010. From graphs to spatial Garroway, C. J. et al. 2008. Applications of graph theory to land-
graphs. – Annu. Rev. Ecol. Evol. Syst. 41: 21–38. scape genetics. – Evol. Appl. 1: 620–630.
Dalongeville, A. et al. 2018a. Combining six genome scan methods Gerlach, G. et al. 2007. Smelling home can prevent dispersal of
to detect candidate genes to salinity in the Mediterranean reef fish larvae. – Proc. Natl Acad. Sci. USA 104: 858–863.
striped red mullet (Mullus surmuletus). – BMC Genomics 19: Guillot, G. and Rousset, F. 2013. Dismantling the Mantel tests.
1–13. – Methods Ecol. Evol. 4: 336–344.
Dalongeville, A. et al. 2018b. Geographic isolation and larval dis- Hedgecock, D. et al. 2007. Genetic approaches to measuring con-
persal shape seascape genetic patterns differently according to nectivity. – Oceanography 20: 70–79.
spatial scale. – Evol. Appl. 11: 1437–1447. Hughes, A. R. and Stachowicz, J. J. 2004. Genetic diversity
Dalongeville, A. et al. 2018c. Data from: Geographic isolation and enhances the resistance of a seagrass ecosystem to disturbance.
larval dispersal shape seascape genetic patterns differently – Proc. Natl Acad. Sci. USA 101: 8998–9002.
according to spatial scale. – Dryad Digital Repository, <https:// Jahnke, M. et al. 2018. Seascape genetics and biophysical connec-
doi.org/10.5061/dryad.31tk592>. tivity modelling support conservation of the seagrass Zostera
Decout, S. et al. 2012. Integrative approach for landscape-based marina in the Skagerrak-Kattegat region of the eastern North
graph connectivity analysis: a case study with the common frog Sea. – Evol. Appl. 11: 645–661.
(Rana temporaria) in human-dominated landscapes. – Land- Jombart, T. 2008. adegenet: a R package for the multivariate anal-
scape Ecol. 27: 267–279. ysis of genetic markers. – Bioinforamtics 24: 1403–1405.
Donati, G. F. A. et al. 2019. A process‐based model supports an Jönsson, B. F. and Watson, J. R. 2016. The timescales of global
association between dispersal and the prevalence of species traits surface-ocean connectivity. – Nat. Commun. 7: 1–6.
in tropical reef fish assemblages. – Ecography 42: 1–12. Kininmonth, S. J. et al. 2010. Graph theoretic topology of the
Dray, S. and Dufour, A. B. 2007. The ade4 package: implement- Great but small Barrier Reef world. – Theor. Ecol. 3: 75–88.
ing the duality diagram for ecologists. – J. Stat. Softw. 22: Legendre, P. and Fortin, M. J. 2010. Comparison of the Mantel
1–20. test and alternative approaches for detecting complex multi-
Dyer, R. J. and Nason, J. D. 2004. Population graphs: the graph variate relationships in the spatial analysis of genetic data.
theoretic shape of genetic structure. – Mol. Ecol. 13: 1713–1727. – Mol. Ecol. Resour. 10: 831–844.
Elshire, R. J. et al. 2011. A robust, simple genotyping-by-sequenc- Lett, C. et al. 2008. A Lagrangian tool for modelling ichthyoplank-
ing (GBS) approach for high diversity species. – PLoS One 6: ton dynamics. – Environ. Model. Softw. 23: 1210–1214.
e19379. Libiger, O. et al. 2009. Comparison of genetic distance measures
Epps, C. W. et al. 2007. Optimizing dispersal and corridor models using human SNP genotype data. – Hum. Biol. Int. Rec. Res.
using landscape genetics. – J. Appl. Ecol. 44: 714–724. 81: 389–406.
Erdös, P. and Rényi, A. 1959. On random graphs. – Publ. Math. Lombarte, A. et al. 2000. Spatial segregation of two species of
6: 290–297. Mullidae in relation to habitat. – Mar. Ecol. Prog. Ser. 206:
Eros, T. et al. 2011. Network thinking in riverscape conservation 239–249.
– a graph-based approach. – Biol. Conserv. 144: 184–192. Lowe, W. H. and Allendorf, F. W. 2010. What can genetics tell us
Escalante, M. A. et al. 2018. The interplay of riverscape features about population connectivity? – Mol. Ecol. 19: 3038–3051.
and exotic introgression on the genetic structure of the Mexican Machias, A. et al. 1998. Bathymetric distribution and movements
golden trout (Oncorhynchus chrysogaster), a simulation approach. of red mullet Mullus surmuletus. – Mar. Ecol. Prog. Ser. 166:
– J. Biogeogr. 45: 1500–1514. 247–257.
Faillettaz, R. et al. 2017. Swimming speeds of Mediterranean set- Macpherson, E. and Raventos, N. 2006. Relationship between
tlement-stage fish larvae nuance Hjort’s aberrant drift hypoth- pelagic larval duration and geographic distribution of Mediter-
esis. – Limnol. Oceanogr. 63: 509–523. ranean littoral fishes. – Mar. Ecol. Prog. Ser. 327: 257–265.
Faillettaz, R. et al. 2018. Larval fish swimming behavior alters dis- Magris, R. A. et al. 2018. Biologically representative and well-con-
persal patterns from marine protected areas in the north-west- nected marine reserves enhance biodiversity persistence in con-
ern Mediterranean Sea. – Front. Mar. Sci. 5: 1–12. servation planning. – Conserv. Lett. 11: e12439.
Félix-Hackradt, F. C. et al. 2013. Discordant patterns of genetic Manel, S. et al. 2003. Landscape genetics: combining landscape
connectivity between two sympatric species, Mullus barbatus ecology and population genetics. – Trends Ecol. Evol. 18:
(Linnaeus, 1758) and Mullus surmuletus (Linnaeus, 1758), in 189–197.
south-western Mediterranean Sea. – Mar. Environ. Res. 92: Martensen, A. C. et al. 2017. Spatio-temporal connectivity: assess-
23–34. ing the amount of reachable habitat in dynamic landscapes.
Fischer, J. and Lindenmayer, D. B. 2007. Landscape modification – Methods Ecol. Evol. 8: 1253–1264.
and habitat fragmentation: a synthesis. – Global Ecol. Biogeogr. Moullec, F. et al. 2019. An end-to-end model reveals losers and
16: 265–280. winners in a warming Mediterranean Sea. – Front. Mar. Sci. 6:
Floyd, R. W. 1962. Algorithms 97: shortest path. – Commun. 1–19.
ACM 5: 345. Murphy, M. et al. 2015. Graph theory and network models in land-
Fortuna, M. A. et al. 2006. Spatial network structure and amphib- scape genetics. – In: Balkenhol, N. et al. (eds), Landscape genet-
ian persistence in stochastic environments. – Proc. R. Soc. B ics: concepts, methods, applications. Wiley, pp. 165–180.
273: 1429–1434. Nathan, R. and Muller-Landau, H. C. 2000. Spatial patterns of
Fortuna, M. A. et al. 2009. Networks of spatial genetic variation seed dispersal, their determinants and consequences for recruit-
across species. – Proc. Natl Acad. Syst. USA 106: 19044–19049. ment. – Trends Ecol. Evol. 15: 278–285.

12
Nathan, R. et al. 2011. Mechanistic models of seed dispersal by Schick, R. S. and Lindley, S. T. 2007. Directed connectivity among
wind. – Theor. Ecol. 4: 113–132. fish populations in a riverine network. – J. Appl. Ecol. 44:
Newman, M. E. J. 2003. The structure and function of complex 1116–1126.
networks. – SIAM Rev. 45: 167–256. Schoville, S. D. et al. 2018. Preserving genetic connectivity in the
Oksanen, J. et al. 2016. vegan: community ecology package. – R European Alps protected area network. – Biol. Conserv. 218:
package ver. 2.4-1. 99–109.
Oliveira, J. dos A. et al. 2019. Model-based riverscape genetics: Shirk, A. J. et al. 2018. A comparison of regression methods for
disentangling the roles of local and connectivity factors in shap- model selection in individual-based landscape genetic analysis.
ing spatial genetic patterns of two Amazonian turtles with dif- – Mol. Ecol. Resour. 18: 55–67.
ferent dispersal abilities. – Evol. Ecol. 33: 273–298. Simpson, S. D. 2005. Homeward Sound. – Science 308: 221–221.
Papadakis, M. et al. 2018. Rfast: a collection of efficient and Slatkin, M. 2005. Seeing ghosts: the effect of unsampled popula-
extremely fast R functions. – R package ver. 1.99.1. tions on migration rates estimated for sampled populations. –
Paris, C. B. et al. 2013. Reef odor: a wake up call for navigation in Mol. Ecol. 14: 67–73.
reef fish larvae. – PLoS One 8: e72808. Taylor, P. D. et al. 1993. Connectivity is a vital element of landscape
Pascual, M. et al. 2017. Impact of life history traits on gene flow: structure. – Oikos 68: 571–573.
a multispecies systematic review across oceanographic barriers Treml, E. A. et al. 2008. Modeling population connectivity by
in the Mediterranean Sea. – PLoS One 12: e0176419. ocean currents, a graph–theoretic approach for marine conser-
Peery, M. Z. et al. 2008. Characterizing source–sink dynamics with vation. – Landscape Ecol. 23: 19–36.
genetic parentage assignments. – Ecology 89: 2746–59. Truelove, N. K. et al. 2016. Biophysical connectivity explains pop-
Peterson, E. E. et al. 2019. Spatially structured statistical network ulation genetic structure in a highly dispersive marine species.
models for landscape genetics. – Ecol. Monogr. 82: e01355. – Coral Reefs 36: 233–244.
Razgour, O. et al. 2019. Considering adaptive genetic variation Urban, D. and Keitt, T. 2001. Landscape connectivity: a graph–
in climate change vulnerability assessment reduces species theoretic perspective. – Ecology 82: 1205–1218.
range loss projections. – Proc. Natl Acad. Sci. USA 116: Urban, D. L. et al. 2009. Graph models of habitat mosaics. – Ecol.
10418–10423. Lett. 12: 260–273.
Rellstab, C. et al. 2016. Signatures of local adaptation in candidate van Etten, J. 2017. R Package gdistance: distances and routes on
genes of oaks (Quercus spp.) with respect to present and future geographical grids. – J. Stat. Softw. 76: 1–21.
climatic conditions. – Mol. Ecol. 25: 5907–5924. Vandergast, A. G. et al. 2008. Are hotspots of evolutionary poten-
Reñones, O. et al. 1995. Life history of the red mullet Mullus tial adequately protected in southern California? – Biol. Con-
surmuletus from the bottom-trawl fishery of the Island of serv. 141: 1648–1664.
Majorca (north-west Mediterranean). – Mar. Biol. 123: Wang, Y. H. et al. 2008. Habitat suitability modelling to correlate
411–419. gene flow with landscape connectivity. – Landscape Ecol. 23:
Riginos, C. et al. 2019. Asymmetric dispersal is a critical element 989–1000.
of concordance between biophysical dispersal models and spa- Warshall, S. 1962. A theorem on Boolean matrices. – J. ACM 9:
tial genetic structure in Great Barrier Reef corals. – Divers. 11–12.
Distrib. 25: 1684–1696. White, C. et al. 2010. Ocean currents help explain population
Row, J. R. et al. 2017. Developing approaches for linear mixed genetic structure. – Proc. R. Soc. B 277: 1685–1694.
modeling in landscape genetics through landscape-directed dis- Whiteley, A. R. et al. 2015. Genetic rescue to the rescue. – Trends
persal simulations. – Ecol. Evol. 7: 3751–3761. Ecol. Evol. 30: 42–49.
Row, J. R. et al. 2018. Quantifying functional connectivity: the role Williamson, D. H. et al. 2016. Large-scale, multidirectional larval
of breeding habitat, abundance and landscape features on connectivity among coral reef fish populations in the Great Bar-
range-wide gene flow in sage-grouse. – Evol. Appl. 11: rier Reef Marine Park. – Mol. Ecol. 25: 6039–6054.
1305–1321. Xuereb, A. et al. 2018. Asymmetric oceanographic processes medi-
Rozenfeld, A. F. et al. 2008. Network analysis identifies weak and ate connectivity and population genetic structure, as revealed
strong links in a metapopulation system. – Proc. Natl Acad. Sci. by RADseq, in a highly dispersive marine invertebrate (Paras-
USA 105: 18824–18829. tichopus californicus). – Mol. Ecol. 27: 2347–2364.
Saura, S. et al. 2014. Stepping stones are crucial for species’ long- Xuereb, A. et al. 2019. Marine conservation and marine protected
distance dispersal and range expansion through habitat net- areas. – In: Oleksiak, M. and Rajora, O. (eds), Population
works. – J. Appl. Ecol. 51: 171–182. genomics: marine organisms. Springer, pp. 423–446.

Supplementary material (available online as Appendix

ecog-05024 at <www.ecography.org/appendix/ecog-05024>).
Appendix 1.

13