Received: 26 November 2022 | Revised: 7 April 2023 | Accepted: 17 April 2023

DOI: 10.1111/1755-0998.13807

RESOURCE ARTICLE

The role of oceanic currents in the dispersal and connectivity

of the mangrove Rhizophora mangle on the Southwest Atlantic
region

André Guilherme Madeira1 | Yoshiaki Tsuda2 | Yukio Nagano3,4 | Takaya Iwasaki5 |

Maria Imaculada Zucchi6 | Tadashi Kajita4,7 | Gustavo Maruyama Mori1

1
São Paulo State University (Unesp),
Institute of Biosciences, São Vicente, Abstract
Brazil
Dispersal is a crucial mechanism to living beings, allowing them to reach new resources
2
Sugadaira Research Station, Mountain
Science Center, University of Tsukuba,
such that populations and species can occupy new environments. However, directly
Nagano, Japan observing the dispersal mechanisms of widespread species can be costly or even im-
3
Analytical Research Center for practical, which is the case for mangrove trees. The influence of ocean currents on
Experimental Sciences, Saga University,
Saga, Japan mangrove dispersal is increasingly evident; however, few studies mechanistically re-
4
The United Graduate School of late the patterns of population distribution with the dispersal by oceanic currents
Agricultural Sciences, Kagoshima
under an integrated framework. Here, we evaluate the role of oceanic currents on
University, Kagoshima, Japan
5
Faculty of Science, Ochanomizu connectivity of Rhizophora mangle along the Southwest Atlantic. We inferred popula-
University, Tokyo, Japan tion genetic structure and migration rates, simulated the displacement of propagules
6
Agência Paulista de Tecnologia dos
and tested our hypotheses with Mantel tests and redundancy analysis. We observed
Agronegócios, Piracicaba, Brazil
7
Iriomote Station, Tropical Biosphere
populations structured in two major groups, north and south, which is corroborated
Research Center, University of the by other studies with Rhizophora and other coastal plants. Inferred recent migration
Ryukyus, Nishihara, Okinawa, Japan
rates do not indicate ongoing gene flow between sites. Conversely, long-­term migra-
Correspondence tion rates were low across groups and contrasting dispersal patterns within each one,
Gustavo Maruyama Mori, São Paulo
State University (Unesp), Institute of
which is consistent with long-­distance dispersal events. Our hypothesis tests sug-
Biosciences, São Vicente, Brazil. gest that both isolation by distance and isolation by oceanography (derived from the
Email: [email protected]
oceanic currents) can explain the neutral genetic variation of R. mangle in the region.
Tadashi Kajita, The United Graduate
Our findings expand current knowledge of mangrove connectivity and highlight how
School of Agricultural Sciences,
Kagoshima University, Kagoshima, Japan. the association of molecular methods with oceanographic simulations improve the
Email: [email protected]
interpretation of the dispersal process. This integrative approach is a cost-­and time-­
Funding information efficient strategy to include dispersal and connectivity data into marine protected
Conselho Nacional de Desenvolvimento
areas planning and management.
Científico e Tecnológico, Grant/Award
Number: 448286/2014-­9; Coordenação
de Aperfeiçoamento de Pessoal de KEYWORDS
Nível Superior, Grant/Award Number: LDD, migration rates, RAD-­seq, Rhizophoraceae, seascape genetics
Finance code 001; Fundação de Amparo à
Pesquisa do Estado de São Paulo, Grant/
Award Number: 13/08086-­1, 14/22821-­9,
18/02655-­8 and 2020/07967-­8; Japan
Society for the Promotion of Science,
Grant/Award Number: 17H01414 and
25290080; Kagoshima University

Handling Editor: Sophie von der Heyden

Mol Ecol Resour. 2023;00:1–16. wileyonlinelibrary.com/journal/men © 2023 John Wiley & Sons Ltd. | 1
2 | MADEIRA et al.

1 | I NTRO D U C TI O N of Africa, northern South America and a small region of the Pacific
coast of Central America, occurring discontinuously (Lo et al., 2014;
Dispersal is the movement of organisms or units—­like seeds and de Menezes et al., 2008). In regions where R. mangle and R. race-
fruits—­from their place of origin to a new home range. It is a funda- mosa occur in sympatry, there is the occurrence of a putative hybrid,
mental process for a species, as it influences its distribution and its R. × harrisonii (Cerón-­Souza et al., 2010, 2014; Cornejo, 2013; Mori
resilience to environmental changes (Travis et al., 2013) and estab- et al., 2021; Tomlinson, 2016). Furthermore, populations from the
lishes connectivity between spatially structured populations (Lowe Atlantic and Pacific coasts of the Americas seem to represent distinct
et al., 2017). The dispersal process can be divided into three stages: lineages, formed after the emergence of the Isthmus of Panama, so
(a) emigration, which is the departure of the dispersing individual or that the populations of R. mangle from the Atlantic coast are more
unit, (b) transit through the environmental matrix and (c) immigra- closely related to the populations of R. racemosa from the Atlantic
tion, which is the arrival and establishment in the new area of life than with populations of R. mangle found on the Pacific coast (Mori
(Cayuela et al., 2018). Given its biological importance and implica- et al., 2021; Takayama et al., 2013, 2021). In the IWP, hybrid species
tions, the study of dispersal mechanisms and their consequences is of Rhizophora are also observed; R. × lamarckii, a hybrid of R. apiculata
of great scientific importance and has practical applications, as in and R. stylosa (Chen, 1996), and R. × sellala, a hybrid of R. samoensis
the development of management plans and conservation strategies and R. stylosa (Takayama et al., 2021; Tomlinson, 2016). R. × sellala
(Driscoll et al., 2014). represents, therefore, a secondary contact between the IWP and
To assess effective dispersal, it is necessary to combine strate- AEP lineages, since R. samoensis must have reached the South Pacific
gies to evaluate establishment, assessing the individuals that con- after one or more LDD events from the west coast of the American
tribute to the gene pool, and displacement, monitoring or simulating continent, uniting the two lineages after millions of years of sep-
the process of immigration. Establishment can be inferred at multi- aration (Takayama et al., 2021). The complex evolutionary history
ple time scales, from recent events through kinship tests and par- of the genus, with transoceanic dispersal events and cosmopolitan
entage assignment, to longer-­term evaluations, using, for example, presence, makes it a good model to study the history of mangrove
the fixation index (FST ) or demographic inferences based on genetic distribution and connectivity as a whole.
approaches (Cayuela et al., 2018; Nathan et al., 2003). Displacement Recently, mangrove dispersal, distribution and population genetic
can be evaluated by telemetry and capture and recapture methods, structure have received substantial attention. Studies based on ge-
for example. Both have been used to study seasonal migration and netic variation highlighted that multiple factors play important roles
foraging but are less suitable to deal with less frequent and more shaping the distribution and connectivity of mangrove species and
unpredictable movements such as long-­distance dispersal (LDD; their populations, like the influence of landscape (Canty et al., 2022;
Cayuela et al., 2018; Nathan et al., 2003; Shafer et al., 2016). LDD Wee et al., 2014, 2020), human activity (Geng et al., 2021; Hasan
events, on the order of thousands of kilometres, are rare events for et al., 2018), land barriers (Mori et al., 2021; Takayama et al., 2021)
most species, but they can have major demographic consequences and oceanic currents (Da Silva et al., 2021; Hodel et al., 2018; Ngeve
and therefore should not be underestimated (Nathan, 2006). et al., 2021; Sereneski-­Lima et al., 2021; Van der Stocken, Carroll,
The study of LDD events is costly, and it is often impractical to et al., 2019). Despite the relatively extensive literature on mangrove
directly monitor dispersal mechanisms, and it is usually impossible dispersal using molecular markers, studies using more than one
to extrapolate regional ecological data for scales of this magnitude source of information are rare, and existing research deals with scales
(Bullock & Nathan, 2008). LDD is especially important for marine of a few 100 km. Using the genetic information obtained by microsat-
species (Villarino et al., 2018), as oceanic currents play a major ellite markers coupled with oceanographic physical simulation, it was
role in their displacements (Bertola et al., 2020; Hays, 2017; Hays shown that the distribution of Rhizophora racemosa along the coast
et al., 2016; Lalire & Gaspar, 2019). The same is true for coastal corresponds better to characteristics of local currents than to ances-
plants that disperse via seawater. Recent studies have used particle tral geological events (Ngeve et al., 2016). At a finer scale, capture–­
modelling to simulate the dispersal of, among others, coastal grasses recapture methods coupled with the same molecular tools showed
(Smith et al., 2018), nettles (Wu et al., 2018) and mangroves (Van der that transport of propagules in the Wouri River is bidirectional, sug-
Stocken, Carroll, et al., 2019). gesting that tide and wind play a major role in the dispersal (Ngeve
Mangroves present adaptations to live between land and sea. et al., 2017). Similar studies that integrate different methodologies
Among these adaptations are propagules, which are fruits or seeds such as those carried out for the Cameroonian coast are rare, but in-
that can stay afloat for weeks or months at sea without losing via- creasingly common in other groups of organisms (Bertola et al., 2020;
bility (Tomlinson, 2016), thus allowing LDD. The genus Rhizophora Jahnke & Jonsson, 2022; Liu et al., 2021; Nikolic et al., 2020).
is one of the main mangrove genera, dominating landscapes in the Integrative approaches bridging biophysical models and genetic/
biogeographic regions of Indo West-­Pacific (IWP) and Atlantic East-­ genomic data in a comprehensive framework may allow the identi-
Pacific (AEP; Tomlinson, 2016). In the AEP, the genus is represented fication of differences in potential and realized dispersal, unveiling
by two described species: R. mangle, with wide distribution along the the factors that control processes like population persistence, demo-
East and West coasts of the American continent to the islands of graphic fluctuations and local adaptation (Jahnke & Jonsson, 2022). At
the South Pacific, where it is traditionally identified as R. samoen- larger geographic scales, these ecological and evolutionary processes
sis (Tomlinson, 2016), and R. racemosa, restricted to the west coast are usually driven by climatic factors like changes in precipitation,
MADEIRA et al. | 3

temperature or storminess and may be linked to LDD events and col- sites and that simulated propagules do not disperse far from their
onization (Van der Stocken et al., 2019). Thus, this framework to infer places of origin. In a third hypothesis (H2), responses to different
connectivity and dispersal patterns has the potential to add consider- multiple factors would be key to the connectivity of R. mangle pop-
able value to conservation and management planning and monitoring ulations. The corollary of this hypothesis is that one would observe
as it allows the identification of priority areas that are likely to be less discrepancies between the pattern of inferred migration and the
resilient because of lower connectivity or areas that are central in the pattern of displacement of the simulated propagules. Our last hy-
metapopulation (Jahnke et al., 2020; Jahnke & Jonsson, 2022; Leiva pothesis (H3) assumes that ocean currents are the sole major factor
et al., 2022; Wee et al., 2018). Its application is especially relevant in in the dispersal and establishment of R. mangle. Here, we predict that
the Southwest Atlantic basin, where the use of connectivity has been the pattern of inferred migration and genetic structure mirrors the
limited in marine protected area design (Balbar & Metaxas, 2019). pattern of transport of simulated propagules.
Here, we evaluate the role of ocean currents in the dispersal
and connectivity of Rhizophora mangle along the Southwest Atlantic
basin. We used single-­nucleotide polymorphisms (SNPs) to describe 2 | M ATE R I A L S A N D M E TH O DS
population genetic structure and estimate recent and ancient mi-
gration rates as proxies of effective dispersal. We combined these 2.1 | Sampling and RAD-­seq library preparation
findings based on molecular variation with estimates of particle and sequencing
transport via oceanic currents using two biophysical models of par-
ticle displacement, one mechanistic and one probabilistic. Finally, we We sampled 78 individuals of R. mangle from 11 sites across the
tested our hypothesis for the distribution of R. mangle populations Southwest Atlantic coast, ranging from 1°23′ S to 28°25′ S (Table 1,
using Mantel tests and redundancy analysis, which is expected to Figure 1a). Due to resources and personnel limitations, and DNA ex-
provide a mechanistic explanation of the dispersal patterns we ob- traction of challenging samples, our study presents a sampling gap
served (Jahnke & Jonsson, 2022). This study provides a framework between the localities of Tamandaré (TMD) and Guapimirim (GPM),
for the integrated use of population genetics and oceanic currents and relatively small number of individuals sampled for some sam-
for future studies with mangroves and other water-­dispersed or- ples. Thus, an approach with thousands of single-­nucleotide poly-
ganisms in a cost-­and time-­effective approach, setting the basis for morphisms is recommended, as thousands of molecular markers may
studies to support marine protected areas planning or management increase the robustness of population structure inferences despite
in the Southwest Atlantic coast. small sample sizes, especially for the estimation of fixation indexes
We tested the null hypothesis (H 0) that assumes that R. mangle and migration rates (Aguirre-­Liguori et al., 2020; McLaughlin &
trees from the Southwest Atlantic basin constitute a single homog- Winker, 2020; Nazareno et al., 2017).
enous population. H 0 predicts that the genetic data will not indicate The individuals were identified in the field based on leaf and
any population structure with substantial migration among the sam- inflorescence morphology (Cerón-­Souza et al., 2010). Despite
pling sites and that the biophysical simulations will show an intense taxonomic issues regarding the genus Rhizophora in the Western
exchange of propagules between them. Alternatively (H1), dispersal hemisphere (Cerón-­Souza et al., 2010;Francisco et al., 2018; Mori
would be limited mostly in the immigration phase. According to this et al., 2021), in the studied area, plants morphologically identified as
hypothesis, one may expect to observe strong, small-­scale popula- R. mangle comprise a rather homogenous genetic group (Francisco
tion genetic structure with little to no migration among sampling et al., 2018; Mori et al., 2021). We were granted licences to

TA B L E 1 Rhizophora mangle sampling

Sampling site
sites along the Southwest Atlantic basin.
Acronym (N) (municipality-­state) Latitude Longitude

SAL (5) Salinópolis-­PA 1°23′44.88″ S 47°22′5.16″ W

PAR (4) Bragança-­PA 1°10′34.43″ S 46°36′59.4″ W
ALC (9) Alcântara-­MA 3°35′25.03″ S 44°24′20.62″ W
PNB (7) Parnaíba-­PI 3°13′10.16″ S 41°49′24.89″ W
PRC (8) Paracuru-­CE 4°35′14.32″ S 39°3′25.49″ W
TMD (8) Tamandaré-­PE 9°24′36.94″ S 35°3′52.02″ W
GPM (2) Guapimirim-­R J 23°18′3.96″ S 43°0′5.47″ W
UBA (10) Ubatuba-­SP 24°30′36″ S 45°9′46.8″ W
CNN (7) Cananéia-­SP 25°6′10.44″ S 47°50′49.92″ W
PPR (10) Pontal do Paraná-­PR 26°22′37.56″ S 48°21′19.44″ W
FLN (8) Florianópolis-­SC 28°25′55.92″ S 48°31′8.04″ W

Abbreviation: N, number of sampled individuals in each site.

4 | MADEIRA et al.

(a) (b)

(c) (d)

F I G U R E 1 Sampling and population genetic structure of Rhizophora mangle. (a) geographic distribution of the sampling sites along the
Southwest Atlantic basin. Sample sizes are represented by the size of the circles and their colours denote sampling sites. (b) population
genetic structure identified with admixture, with K = 2–­4. Each line represents one individual, and each colour represents an inferred
ancestry. (c) pairwise FST between all samples, excluding GPM. Warmer colours indicate higher FST values. Black lines split the sampling sites
between the north and south groups identified in the population structure analyses. (d) effective migration surface was used to identify
regions of sustained gene flow and barrier. Colours denote log10-­transformed effective migration rates (m) where m = 1 means a migration
rate 10-­fold faster than average. Warmer colours indicate areas with relatively lower migration, whereas colder colours denote areas with
relatively higher migration. Acronyms for the sampling sites follow Table 1.

sample Rhizophora plants (Instituto Chico Mendes de Conservação (G^AATTC; Takara, Japan) and fragments were ligated with the
da Biodiversidade—­ICMBio—­17,130 and 17,159) and assess their ge- adapters P1 (Illumina) that contain an amplification site, the Illumina
netic variation (Sistema Nacional de Gestão do Patrimônio Genético sequencing site and a specific barcode for posterior reads demulti-
e do Conhecimento Tradicional Associado—­SisGen—­A885104). The plexing. The fragments were then sheared, size-­selected and ligated
samples we used in this study were previously analysed on three with the adapters P2 (Illumina) that ensure that only P1-­ligated frag-
independent works using different molecular markers (Francisco ments are enriched in the subsequent PCR amplification. We then
et al., 2018; Mori et al., 2021; Takayama et al., 2013). sequenced the libraries with the HiSeq2000 sequencer (Illumina) at
We isolated DNA samples from each individual using a regular Macrogen (Seoul, South Korea).
cetyltrimethylammonium bromide (CTAB) method (Doyle and Doyle,
1990). We evaluated DNA integrity with 1% agarose gel electropho-
resis and quantity in Quantus™ Fluorometer with QuantiFluor® 2.2 | SNP calling, filtering and neutral
dsDNA System (Promega, Madison, WI, EUA). Only samples with SNP selection
high quantities (>250 ng) of high-­quality DNA were used to build
three RAD-­seq libraries following the original protocol (Baird For SNP calling, we used STACKS 2.4 (Rochette & Catchen, 2017)
et al., 2008) with small changes to allow the use of 250 ng of DNA. to demultiplex reads, cutadapt 2.2 (Martin, 2011) to remove adapt-
Briefly, the samples were digested with the restriction enzyme EcoRI ers and TRIMMOMATIC 0.39 (Bolger et al., 2014) to filter reads by
MADEIRA et al. | 5

quality (Phred score > 33). Following, with bowtie 2.3.5.1 (Langmead presumes correlated allelic frequencies and ancestral mixture. We
& Salzberg, 2012), we aligned reads to the reference genome of R. api- ran admixture for values of K (number of genetic groups) varying
culata (NCBI Project PRJEB8423; Xu et al., 2017), used SAMtools from 1 to 11 and compared the cross-­validation error of different
1.9 (Li, 2011) to index, list and organize reads and bcftools 1.9 runs to determine the best values for K. Complementarily, we used
(Narasimhan et al., 2016) to identify SNPs. Finally, VCFtools v0.1.16 adegenet 2.1 (Jombart & Ahmed, 2011) to perform a discriminant
(Danecek et al., 2011) was used to filter out individuals and loci based analysis of principal components (DAPC) to identify populations re-
on their missing data. Different filters with different missing data gardless of genetic models, focussing on describing the differences
permissibility were tested to maximize loci and individual retained between the inferred groups and minimizing variation within clus-
and minimize the amount of missing data. We filtered out individuals ters (Jombart et al., 2010). K, in this case, was chosen by the small-
with more than 20% missing data and loci missing from more than est Bayesian information criterion (BIC) score after an initial PCA.
20% individuals and with less than three minor allele counts. We calculated the Slatkin-­linearized pairwise FST (Slatkin, 1995;
As an effort to only analyse neutral markers, we used several ap- Weir & Cockerham, 1984) between sampling sites and between
proaches to identify signatures of natural selection because methods inferred groups (see Results) using arlequin v3.5.2.2 (Excoffier &
to test marker neutrality may generate large numbers of false posi- Lischer, 2010).
tives (François et al., 2016; Grünwald et al., 2017). We used LOSITAN We also described population structure of R. mangle across the
(Antao et al., 2008), which detects markers at the edge of the FST dis- Southwest Atlantic basin by inferring effective migration surfaces
tribution curve with the FDIST2 method (Beaumont & Nichols, 1996). with EEMS v1.0 (Petkova et al., 2015). This method utilizes the
This method evaluates the expected neutral relationship between FST isolation-­by-­distance model as null hypothesis and returns the de-
and the expected heterozygosity across all loci to identify outlier loci, viations of this model in the observed gene flow for georeferenced
that is, those that have excessively high or low FST. We considered a genetic samples. Using this model-­based method, one may estimate
false discovery rate (FDR) lower than 0.1 as threshold. Second, we a migration parameter m, which allows the characterization of ge-
used pcadapt 4.1.0 (Luu et al., 2017), which implements a principal netic divergence between demes. EEMS enables the visualization of
component analysis (PCA) to identify whether a marker has a signifi- population structure in a surface that highlights regions with high
cantly different FST, and considered as outliers SNPs that presented gene flow and migration barriers. We assembled the grid of in-­water
a q-­value lower than 0.1. By using a PCA to determine the groupings demes by excluding on-­land demes within the Southwest Atlantic
of individuals, pcadapt does not require prior population structure region (considering that the propagules can only migrate through
information to identify outliers and can handle admixed individu- the water and not by land), and ran five independent runs, with 107
als better than other strategies (Luu et al., 2017). Lastly, we used a Markov chain Monte Carlo steps (MCMC) and 106 burn-­in, with
Bayesian approach to identify FST outliers with Bayescan 2.1 (Foll & deme size of 103.
Gaggiotti, 2008). Bayescan is especially effective at handling small
samples, since its Bayesian method incorporates the uncertainty of
allele frequencies due to small sample sizes, reducing bias. In this 2.4 | Migration inference
case, we also chose a q-­value lower than 0.1 as threshold.
In addition to the FST methods, we used latent factors mixed mod- To quantify effective connectivity among populations, we used
els (LFMM) implemented in the packages LEA 3.6.0 and LFMM 1.0 two approaches. The first estimates recent (i.e. contemporary) di-
(Frichot et al., 2013), which allow us to associate allelic frequencies rectional migration rates, over few generations, implemented in
and environmental factors to identify putative loci under selection. BA3-­SNPs 1.1 (Mussmann et al., 2019), which is a modification of
For this analysis, we also considered an FDR of 0.1. We obtained BayessAss 3 (Wilson & Rannala, 2003) to work with large SNPs
data regarding temperature, precipitation, water vapour pressure, datasets. BayessAss implements a Bayesian MCMC method to es-
solar radiation (collected from WordClim2; Fick & Hijmans, 2017), timate asymmetric recent migration rates, which are the proportion
sea surface salinity and temperature (collected from MARSPEC; of each population with migrant ancestry (Wilson & Rannala, 2003).
Sbrocco & Barber, 2013), tide amplitudes and number and dura- We ran 10 8 iterations, with 106 burn-­in and sampling interval of
tion of tidal cycles (collected from the Environmental Climate Data 100. The mixing parameters for allelic frequency, inbreeding coef-
Sweden, ECDS; Obst, 2017). For downstream analyses, we utilized ficient and migration rate were set as 0.9, 0.05 and 0.1, respectively,
only the SNPs with no evidence of natural selection in at least three adjusted for an acceptance rate between 20% and 40% (Wilson &
of these methods. Rannala, 2003).
To complement BA3-­SNPs analyses, we estimated ancestral (i.e.
long-­term) migration rates based on site frequency spectrum (SFS,
2.3 | Population structure and effective obtained with arlequin v3.5.2.2; Excoffier & Lischer, 2010) with
migration surfaces fastsimcoal26 (Excoffier et al., 2021). Given a complex demographic
model, fastsimcoal26 simulates the neutral genetic diversity of sam-
To assign individuals to populations, we used admixture 1.3.0 ples and their ancestors trough a coalescent method. Thus, it can be
(Alexander et al., 2009), which implements a genetic model that used to test different complex evolutionary scenarios, accounting
6 | MADEIRA et al.

for population resizes and hybridization, and to infer parameters We simulated the release of a total of 106 propagules distributed in
like divergence time, effective population sizes and migration rates equal time intervals from 2010 to 2020 from each sampling site and
(Excoffier et al., 2021). To reduce computational demand and allow tracked each particle for at least 1 year. To avoid having all particles
time-­consuming simulations, we inferred the ancestral migration trapped inside a bay, release coordinates for PRC, TMD, GPM, CNN
rates based on the population genetic structure we observed (see and FLN were slightly shifted to open areas to identify the fate of
Results): only among the north group sampling sites (SAL, PAR, ALC, propagules that reach oceanic currents. We disregarded data from
PNB and PRC) and among the south group sampling sites (TMD, particles that drifted for more than 1 year and registered the num-
GPM, UBA, CNN, PPR and FLN). We evaluated three different sce- ber of particles that reached land in a 1° radius around each other
narios for each setting: panmixia, steppingstone and directional sampling site.
migration (Supplementary Information S1, S2 and S3, respectively).
In the panmixia scenario, we inferred the ancestral migration rates
between all sites in their own region. In the steppingstone scenario, 2.6 | Drivers of isolation
we inferred only the ancestral migration rates between neighbour-
ing sites, considering the remaining migration rates as zero. Finally, We tested the relationship between the ancestral migration rates
in the directional migration scenario, we assumed east-­to-­west mi- and the simulated propagule displacement through oceanic currents
gration rates to be 10 times higher than in the opposite direction, as using Mantel tests (Mantel, 1967) modified to consider asymmetri-
the oceanic currents would suggest. In addition to these scenarios, cal matrices (Bergmann & McElroy, 2014), with 99,999 iterations.
we also estimated the ancestral migration rates between the two We also compared the migration rates with the Euclidian distance
groups, north and south. We ran 50 independent runs, with 106 it- among sites and with the minimum in-­water distance, computed
erations each for each model, and selected the best run based on with package gdistance v1.3-­6 for R (van Etten, 2017) with the least-­
the smallest difference between maximum likelihood and observed cost distance between sites using land mass as an infinite resistance
likelihood (deltaL). We then compared the scenarios based on the surface (Assis et al., 2013).
Akaike information criterium (AIC; Aho et al., 2014) of the best run. In addition to the Mantel tests, we used redundancy analysis
(RDA) to model linear relationships between spatial eigenvector
maps describing the symmetric (i.e. Euclidian and minimum in-­water
2.5 | Simulation of propagule displacement distances) and asymmetric (simulated propagule displacement)
variables with genetic variation. This approach is more efficient at
We used two methods to estimate the propagule displacement among describing fine-­scale spatial genetic structure than Mantel tests,
sites and, thus, provide an oceanographic context to be tested with as it partitions the contribution of many evolutionary processes
the genetic-­based results. The first is a probabilistic oceanic transport across loci (Diniz-­Filho et al., 2013). Briefly, the symmetric variables
estimate with Adrift (van Sebille, 2014), which combines observa- were transformed into distance-­based Moran's eigenvector maps
tional data from seventeen thousand floaters released between 1979 (dbMEM; Dray et al., 2006) while the simulated probabilities of prop-
and 2013 to track the movement of floating objects with two-­month agule displacement were transformed into an asymmetric eigenvec-
intervals and 1° resolution. This approach disregards the effects of tor map (AEM; Blanchet et al., 2011) with the R package adespatial
seasonality or year-­specific oceanographic conditions, which makes v0.3-­16 (Dray et al., 2022), as described in Xuereb et al. (2018). We
Adrift an adequate tool for historical biogeography studies (Bertola ran a PCA on the Hellinger-­transformed allele frequencies (Legendre
et al., 2020). We simulated the release of mangrove propagules de- & Gallagher, 2001) and used the most explanatory components as
parting from each of the 11 sampling sites to obtain the cumulative response variables of the predictors, AEM and dbMEM, for the RDA.
probability of the presence of particles in each other localities in a We identified the most explanatory variables with the R package
2-­year period. This value was then divided by the sum of probabil- vegan v2.5-­7 (Oksanen et al., 2020) and used ANOVA with 1000
ities in every site to build a relative index of directional transport. permutations to access the significance of the model and the con-
This 24-­month period is a conservative measure of floating period of tribution of each variable. These analyses were not performed for
Rhizophora species (Van der Stocken, Wee, et al., 2019). recent migration rates, as they were negligible (see Results).
To complement the oceanic transport estimate analysis, we
used a mechanistic propagule displacement approach based on the
biophysical model of particle dispersal OceanDrift implemented in 3 | R E S U LT S
OpenDrift v1.9.0 (Dagestad et al., 2018). OpenDrift is a framework
for trajectory models, which tracks the fate of objects drifting in the 3.1 | SNP calling, filtering and neutral SNP
ocean based on given current velocity data. We used a 1/3° grided selection
horizontal sea surface current velocity data from OSCAR (Ocean
Surface Current Analysis Real-­time), which are based on sea surface We sequenced 222,782,299 reads for all samples obtained in the
height, surface vector wind and sea surface temperature data from Southwest Atlantic region, with 70%–­82% alignment rate with the
various satellites and in situ instruments (Bonjean & Lagerloef, 2002). reference genome. We retained 263,279 SNPs. After the filtering
MADEIRA et al. | 7

steps, we kept 6568 SNPs. We found 112, 1, and 241 outlier SNPs (6.286 × 10−3 and 1.162 × 10−3, respectively; Figure 2). The ancestral
with pcadapt, Bayescan and LOSITAN, respectively, and 23 loci as- migration rates between both groups were two to three orders of mag-
sociated with each environmental variable on the LFMM analyses, nitude smaller than the inferred migration rates within groups. Also,
on average. The most relevant environmental variable, associated long-­term migration estimates between groups revealed a 50% higher
with 35 SNPs, was the mean annual sea surface temperature. We migration from south to north than the opposite direction (S ➔ N:
identified 58 SNPs as putatively under selection in at least three of 1.525 × 10−5; N ➔ S: 9.687 × 10−6). The highest ancestral migration
these methods and therefore kept 6510 uncoupled neutral SNPs to rates inferred within each group were from ALC to SAL in the north
the downstream analyses. group (7.275 × 10−3) and from TMD to GPM in the south group (0.052).

3.2 | Population genetic structure 3.4 | Oceanic transport estimates

Both DAPC and admixture analyses showed two major populations Adrift and OpenDrift analyses provided similar results (Mantel
(K = 2) of R. mangle on the Southwest Atlantic coast: one popula- r = 0.7, p < 0.0001; Figure S2). Both showed a clear separation
tion in the north and another south of the northeastern extremity between the north and the south groups (Figure 3), without sub-
of South America (NEESA), henceforth regarded as ‘north group’ stantial exchange of particles (i.e. propagules) between groups.
and ‘south group’, respectively. We found more admixed individu- In Adrift analysis (Figure 3a), populations from the north group
als in the north group on both independent analyses (Figure 1b and and TMD, which is located south to the NEESA, showed high rates
Figure S1). Considering finer-­scale structure (K = 3 and K = 4) in the of transport of particles from east to west. Probabilities of prop-
admixture analysis, we also could observe the two major groups, agules drifting in the other direction were null, with the excep-
north and south, with more admixed individuals in the north (i.e. more tion of the neighbouring sites PAR and SAL. TMD site was the
unique genetic groups were rarely found in the south group). The FST only one showing non-­null probabilities of transport to all other
between the north and south groups was significant (FST = 0.032, localities, with a higher probability of transport to sites of the
p < 0.00001), which indicates population structure between both north group (min = 6.6%; max = 23.2%; mean = 12%) than to those
groups. Only three pairwise FST values regarding the GPM sampling within south group (min = 0.2%; max = 1.2%; mean = 1%). TMD is
site were significant. We recognize that this result is likely an artefact also the only site that did not receive propagules from any of the
due to the small sample size for this locality. With GPM excluded, other sites, acting solely as a propagule donor. In the mechanistic
91% of the pairwise FST values were significant, ranging from 0.056 simulation of propagule displacement with OpenDrift (Figure 3b),
to 0.149 (Figure 1c). This north–­south pattern was also recovered by TMD is further isolated and has only contributed to GPM site,
effective migration surfaces found with EEMS, which also revealed with only 10 propagules out of 100,000. Unlike the outcomes of
a barrier to migration within the north group and greater gene flow Adrift, the east-­to-­west directionality of propagules trajectories
within the south group (Figure 1d). within the north group was not observed. The overall propagule
exchange between sites within the south group was 1.4× higher
on average than the exchange between localities within the north
3.3 | Migration rates group. Both mechanistic and probabilistic analyses highlighted
the intense propagule exchange between SAL and PAR within the
The recent migration rates estimated between samples were neg- north group and between PPR and CNN within the south group.
ligible, as none presented a confidence interval that did not include Additionally, both methods showed higher probability of particle
zero. The ancestral migration models that best fitted our data were transport within the site of release, 56.5% for Adrift and 82.8%
the steppingstone model for the north group and the panmixia model for OpenDrift. This difference is likely due to the assumptions of
for the south group (Table 2). Migration rates within the south group each method, like stopping the simulation of particles that reach
were, on average, six times higher than those within the north group land only for OpenDrift.

TA B L E 2 Within groups long-­

North group South group
term migration models tested with
fastsimcoal26. deltaL AIC deltaL AIC

Steppingstone 1030.526 15029.562 4997.803 97369.966

Directional 1294.901 16271.054 4869.849 96820.716
Panmixia 1043.597 15113.756 4727.517 96165.253

Note: Best AIC values for each region are highlighted in bold.
Abbreviations: AIC, Akaike information criterion; deltaL, difference between the obtained
likelihood in the best run and the maximum possible likelihood for the model.
8 | MADEIRA et al.

F I G U R E 2 Within groups ancestral

migration rates. Long-­term migration rates
inferred with fastsimcoal26 within each
group are shown. Arrows point from one
sampling site in direction of the inferred
migration destination, and its width
represents the intensity of migration.
Between groups long-­term migration rates
are not shown. Acronyms for sampling
sites follow Table 1.

(a) (b)

F I G U R E 3 Oceanic transport estimates. (a) Probabilistic oceanic transport estimates with Adrift. (b) Mechanistic simulation of propagule
displacement with OpenDrift. The arrows point in the direction of the propagule transit, and its width represents the intensity of the particle
exchange. Self-­pointing arrows were excluded from the representation. Acronyms for the sampling sites follow the notation on Table 1.

3.5 | Drivers of isolation r = 0.273, p = 0.048), suggesting a stronger effect of oceanography

on migration.
We used the results obtained only with OpenDrift to test the Regarding the RDA analyses, we retained four axes of the PCA
effects of oceanic currents, because we observed a significant with the Hellinger-­transformed allele frequencies, which cumulatively
correlation between Adrift and OpenDrift outcomes. We found explained 56.28% of the total neutral genetic variance. The selection
evidence for isolation by distance based on the correlation be- of asymmetrical explanatory variables (i.e. the AEM obtained with the
tween the Euclidian (Mantel r = −0.189, p = 0.029) and minimum propagule displacement simulation results) retained one AEM variable
in-­water distances (Mantel r = −0.207, p = 0.008) with the inferred (AEM1; Figure 4a). The RDA considering AEM1 and the four PCs re-
ancestral migration rates. The correlation between the propagule tained had a coefficient of determination (adjusted R2) of 8.1% (p = 0.03).
displacement simulation with these rates was higher (Mantel We identified one significant Euclidian dbMEM (dbMEM1; Figure 4b)
MADEIRA et al. | 9

and two in-­water dbMEMs (dbMEM1 and dbMEM10; Figure 4c,d). observed the subdivision between north and south groups with
The RDA regarding the selected dbMEM variables showed signifi- more admixed individuals in the former with multiple analyses.
cant coefficients of determination (for Euclidian distances: R2 = 6.7%, This pattern was recovered with the model-­based inference of in-
p = 0.045; for in-­water distances: R2 = 20.3%, p = 0.013). The retention dividual ancestry with admixture, a multivariate clustering of related
of the first eigenvector map for AEM, Euclidian and in-­water dbMEM individuals (DAPC), the genetic distance among samples (pairwise
indicates a broad-­scale spatial structure, because eigenvector maps FST ) and the effective migration surfaces (EEMS). Additionally, this
are increasingly specific in the spatial scale they explain. structure is recovered at smaller scales (K = 3 and K = 4 in the ad-
mixture analysis), indicating that this is a major pattern of R. mangle
population genetic structure in the Southwest Atlantic basin. The
4 | DISCUSSION geographic positioning of the barrier to gene flow within the region
of the NEESA was also compatible across methods. Resembling pat-
We recovered the north–­south population genetic structure terns of organization of the genetic variation were previously re-
found for R. mangle in previous studies (Francisco et al., 2018; Pil ported for R. mangle (Francisco et al., 2018; Pil et al., 2011) and other
et al., 2011; Takayama et al., 2013) with substantial population coastal plants (Cruz et al., 2019, 2020; Da Silva et al., 2021; Mori
admixture (Francisco et al., 2018; Takayama et al., 2013) in the et al., 2015). Additionally, the EEMS results revealed a smooth west-
Southwest Atlantic coast using genomic tools. Recovering a similar ward population differentiation within the north group, mirroring a
pattern using thousands of single-­base markers, which may be under previous study that used microsatellites (Francisco et al., 2018). The
different evolutionary regimes than microsatellites used in previ- observation of similar patterns with a considerably larger number
ous works, strengthens the evidence for such structure. Our find- of molecular markers (~6500 SNPs) compared with previous studies
ings based on multiple lines of evidence were consistent with this conducted with less than 10 microsatellites (Francisco et al., 2018;
pattern. We observed low long-­term migration rates between north Pil et al., 2011) may be interpreted as a consistent evidence of limited
and south groups, low probability of oceanic transport between gene flow at large geographic scales.
groups, and different migration patterns within each group. Our re- Our analyses showed that contemporary migration is rare or
sults show that both geographic and oceanographic distances are absent, whereas rates of long-­term migration were not only non-­
crucial factors to explain the neutral genetic variation of mangrove negligible but largely asymmetrical. We observed low levels of his-
populations. torical migration between north and south groups when compared
to the historical migration rates within groups, which supports the
population genetic structure found. Our findings also indicated a
4.1 | Large-­scale connectivity 1.5× higher ancestral migration rate from south-­to-­north relative to
north-­to-­south, which is consistent with the higher number of ad-
At large geographic scale, our results based on 6510 unlinked neu- mixed individuals in the north group. This pattern could be explained
tral SNPs revealed a north–­south population genetic structure. We by the prevalent oceanic currents in the studied region; the split of

(a) (b) (c) (d)

F I G U R E 4 Visual representation of the significative eigenvector maps retained for the redundancy analysis. (a) Asymmetric Eigenvector
Map 1, derived from the propagule displacement simulation (i.e. biophysical simulation); (b) Euclidean distance-­based Moran's Eigenvector
Map 1, derived from the geographic distance between sample locations; (c and d) in-­water distance-­based Moran's Eigenvector Maps 1
and 10, respectively, derived from the in-­water minimal distance between sampling site locations. For each individual map, similar colours
represent similar eigenvector map values.
10 | MADEIRA et al.

the South Equatorial Current into the North Brazilian Current and to 2020. These methodological differences allow us to interpret the
the Brazilian Current would separate the north and south groups. outcomes of each method as complementary timeframes and time
Also, the weaker Brazilian Current would allow the occasional north- scales. Our analyses with OpenDrift considered a narrower and more
ward dispersal more often than the stronger North Brazilian Current recent timeframe and generated particle trajectory patterns that
would allow southward dispersal (Lumpkin & Johnson, 2013). These indicated an absence of dispersal between groups. Contrastingly,
results coupled with the population genetic structure results allow Adrift outcomes revealed drift patterns that suggest that long-­term
us to reject the null hypothesis of panmixia (H 0) and H1, which as- propagule dispersal from south group to north group is not only a
sumes small-­scale population structure leading to isolated popula- possible, but a plausible phenomenon. These findings are in line
tions, since we observed a predominant large-­scale structure, with with the population genetic structure and asymmetric long-­term mi-
non-­negligible connectivity between sampling sites within each gration patterns and the negligible contemporary migration rates.
group. Interpreted together, they are consistent with LDD in R. mangle.
The overall incongruence of the results obtained with BA3-­SNPs Long-­distance dispersal events are mostly infrequent but have
and fastsimcoal26 we observed may be interpreted as is likely an important consequences for population dynamics and evolution
outcome of the rarity of LDD events (Nathan, 2006), because es- (Cain et al., 2000; Jordano, 2017; Kremer et al., 2012; Nathan
timates of gene flow based on allele frequencies are expected to et al., 2008). A combined interpretation of lines of evidence regard-
represent shorter time scales as they approach equilibrium more ing oceanographic physical modelling and molecular data allows us
rapidly than coalescent approaches (Bohonak & Vandergast, 2011). to bridge ecological processes of dispersal, from emigration and
It means that LDD events are likely rare even for organisms such transfer to immigration (Van der Stocken, Wee, et al., 2019), with its
as mangroves, which are capable of transoceanic dispersal (e.g. genetic outcomes at genomic scale. This combination also allowed us
Mori et al., 2015, 2021; Nettel & Dodd, 2007; Ngeve et al., 2021; to unveil connectivity patterns at finer spatial scales.
Takayama et al., 2021). R. mangle populations would remain majorly
isolated, and rare LDD events would contribute to the mixing of
populations. The ancestral migration rates inferred were low (mean 4.2 | Small-­scale connectivity
4.02 × 10−3), indicating that these admixture events are rare even
within the inferred genetic groups (south and north). Over time, this Our results suggest contrasting modes of dispersal within each group
admixture would become evident, but in a few-­generations scale, it based on multiple sources of information. A higher gene flow within
would not be possible to identify this phenomenon. the south group is evident from the effective migration surface.
Likewise, the mechanistic and probabilistic propagule displace- Within this group, we also observed 6× higher ancestral migration
ment simulations revealed a clear distinction between north and rates and 1.4x higher displacement rates of propagules estimated
south groups. Both showed that propagules most likely remain within with OpenDrift, compared with those estimated for the north group.
its release group. One key difference between the findings obtained Complementarily, model testing with fastsimcoal26 indicated that
with Adrift and OpenDrift was the probability of transit away from a steppingstone model better fitted our genetic data for the north
the release site, which was higher for the former. For this reason, group whereas panmixia model best explained the genetic varia-
neighbouring sites presented a much higher exchange of propagules tion of the south group. This was corroborated by the OpenDrift
with OpenDrift, which can be readily observed, for instance, in the analyses. It would be reasonable to expect neighbouring sites to
intensity of the particle exchange between SAL and PAR in the north extensively exchange propagules, but, within the north group, the
group and between CNN and PPR in the south group (Figure 3). rates of exchange between neighbours were higher in comparison
Another difference we observed between the results of these bio- with the south group. The ratio of average number of propagules
physical oceanographic methods regards the TMD site. This south transported to neighbouring sites over the average number of prop-
group site is geographically close to the NEESA and to the split of agules transported to non-­neighbouring sites was 4.7 for the north
the South Equatorial Current (Lumpkin & Johnson, 2013). The anal- group and 3.4 for the south group. These results together empha-
yses with Adrift showed that particles released from TMD can reach size the higher connectivity within the south group compared with
all remaining sites in both groups. Contrastingly, OpenDrift analy- the north group and highlight that the connectivity patterns within
ses showed a sharper distinction, in which propagules released from each region are structurally different. Therefore, these groups are
TMD stranded only at the GPM site. not only highly differentiated based on different molecular markers
Such contrasts may be understood by the underlying differ- (Francisco et al., 2018; Pil et al., 2011; Takayama et al., 2013); they
ences between methodologies. Adrift, which relies on surface also disperse differently.
drifter trajectories spanning from 1979 to 2013, calculates probabil-
ities of transit of particles in a transition matrix (van Sebille, 2014).
Conversely, OpenDrift is deterministic; a particle released in a given 4.3 | Drivers of isolation
location and time will always have the same trajectory (Dagestad
et al., 2018). Furthermore, for the OpenDrift analyses, we simulated Oceanic currents have been used to explain the genetic divergence
the release of 106 particles, evenly released for 10 years, from 2010 between north and south groups in mangrove trees (Cruz et al., 2019,
MADEIRA et al. | 11

2020; Da Silva et al., 2021; Mori et al., 2015; Pil et al., 2011) and Robledo-­Arnuncio, 2011). To describe the dynamics of connectivity
other coastal plants (Takayama et al., 2008). Because this pattern of patterns at fine scales and to translate them across scales, future
neutral population genetic structure is consistent across independ- studies with wind dispersal modelling and complementary sampling
ent lineages, it is reasonable to argue that a shared extrinsic factor, resolutions, from landscape to biogeographic scales, are needed.
or a set of common abiotic factors, independently shapes the con- The last eigenfunction retained, dbMEM10 for the minimum in-­
nectivity of these coastal species. Using a landscape genomics ap- water distance, showed a structure where TMD is isolated from the
proach, it has been recently shown that the neutral genetic variation other localities (Figure 4d). As previously discussed, TMD is at an
of the mangrove tree Avicennia schaueriana is largely driven by the interesting geographic spot, at the border of both groups. It is asso-
barrier created by the bifurcation of the South Equatorial Current ciated with the south group based on population genetic structure
and by geographic distances (Da Silva et al., 2021). In that study, this analyses (admixture, DAPC and EEMS) and on propagule displace-
oceanographic bifurcation was treated as an absolute and static bar- ment simulations. Conversely, it is associated with the north group in
rier and, therefore, it could not address how directionality, strength the symmetric in-­water (dbMEM1) and Euclidean (dbMEM1) eigen-
and variability of ocean currents influence mangroves' dispersal (Da functions. It also may be positioned at an intermediate or isolated
Silva et al., 2021). Here, the use of oceanographic modelling and status regarding the Adrift analyses and the in-­water dbMEM10 ei-
genomic data under an integrated statistical framework allowed us genfunction. Jointly, these findings indicate that TMD is genetically
to take a step further to provide more realistic insights into the ef- assigned to the south group but can possibly function as a bridge
fects of oceanic currents on the connectivity of R. mangle, and more for LDD events that maintain the admixture between both groups,
broadly, of mangrove and coastal plant species. mainly north to the NEESA.
Mantel tests showed that spatial distances play an important The discrepancies between the results of Mantel tests and
role shaping the long-­term migration of R. mangle (Euclidian distance: RDA, for instance, suggest that the role played by oceanic currents
Mantel r = −0.189; minimum distance in-­water: Mantel r = −0.207). is more relevant at larger scales compared with finer scales (i.e.
However, a stronger correlation was observed when propagule dis- within groups). Mantel tests provide a global summary of effective
placement rates were used as a predictor of ancestral migration in- connectivity (migration rates), revealing large-­scale patterns, while
ferred with fastsimcoal26 (Mantel r = 0.273). Accordingly, the RDA eigenvector maps enable the partition of genetic variation in mul-
results showed that both geography and the direction of ocean tiple loci, revealing fine-­scale processes (Diniz-­Filho et al., 2013;
currents influence the spatial patterns of long-­term migration. Both Dray et al., 2006). The combined interpretation of different methods
symmetrical (i.e. Euclidian and in-­water dbMEMs) and asymmetrical is necessary to identify processes that operate at different scales,
(i.e. AEM, derived from propagule displacement analyses) eigenfunc- as the stochasticity of dispersal dynamics at smaller scales hinders
tions retained describe a broad spatial scale structure, accounting the identification of general patterns and tendencies across scales
for the north–­south population genetic divergence (Figure 4a–­c). (Chave, 2013).
We did not recover fine-­scale structure with our eigenvector maps. Overall, our findings provide evidence to reject H0 and H1.
It indicates that spatial distances and oceanic currents play a greater Additionally, they allow one to reject H3, which assumes that oce-
role separating both groups than finer-­scale connectivity patterns anic currents are the only driver of dispersal of R. mangle propagules,
within each group. In fact, based on oceanic currents alone (Lumpkin because we did not observe a match between genetic and ocean-
& Johnson, 2013), one could expect greater east-­to-­west migration ographic patterns. Thus, based on our isolated and combined find-
rates and propagule displacement. We only found this pattern with ings, we could not reject our third hypothesis (H2), which stated that
Adrift analyses. We argue that other factors, such as nearshore the connectivity of R. mangle would be explained by multiple factors
and coastal currents regimes, pollen dispersal and shoreline geo- like nearshore and coastal currents and pollen-­mediated gene flow.
morphology are likely more relevant to short-­to-­medium scales (i.e. Although we observed that oceanic currents are important drivers
from tens to few hundreds of kilometres) connectivity of R. mangle. of isolation in a dominant mangrove tree, we could not rule out the
Such nearshore processes could influence the immigration stage of influence of geographic and in-­water distances.
propagule dispersal filtering the access, retention and recruitment Our study showcases how the combination of genetic-­b ased
of seedlings (Van der Stocken, Wee, et al., 2019). Rhizophora genus and biophysical oceanographic models improves the interpreta-
is ambophilous (pollinated by both wind and insects; Sánchez-­Núñez tion of connectivity patterns across temporal and spatial scales.
& Mancera-­Pineda, 2012), but anemophily is likely the main form The use of complementary statistical methods to assess to what
of pollination (de Menezes et al., 1997; Kondo et al., 1987; Nadia extent spatial and oceanographic features explain the genomic
& Machado, 2014). The role of insect pollination varies depend- variation allowed a broader understanding of how ecological
ing on the density of R. mangle in the forest; more sparse R. mangle and evolutionary processes shape mangrove connectivity in the
forests have a greater contribution of insect pollination (Nadia & Southwest Atlantic basin. Given the broad scale of distribution of
Machado, 2014). Wind-­carried pollen can, in rare conditions, travel mangrove species, their conservation is a challenging endeavour.
thousands of kilometres (Campbell et al., 1999), and for exclusively The understanding of connectivity and dispersal is crucial to the
wind-­pollinated species, this long-­range pollination can contribute effective design of a network of marine protected areas (Leiva
significantly to the gene flow at larger scales (~hundreds of km; et al., 2022), and a combined approach with independent lines of
12 | MADEIRA et al.

evidence can contribute for the solid identification of key areas for future conditions, plants from both groups will likely face novel in-­
conservation (Jahnke et al., 2020; Jahnke & Jonsson, 2022; Wee transit challenges, with ecological and evolutionary consequences.
et al., 2018). We believe that research in organisms with water-­
based dispersal will benefit from an integrative approach involving
genomic and genetic data, biophysical models, open access data 4.5 | Conclusion
and a comprehensive statistical framework. This approach is par-
ticularly appealing for community-­, species-­ and population-­level The combination of genomic-­scale data with biophysical oceano-
conservation efforts as it provides a time-­and cost-­effective strat- graphic simulations is in line with the hypothesis that oceanic cur-
egy to unveil dispersal patterns of marine and coastal organisms rents, spatial and in-­water distances play a role as drivers of dispersal
and communities, like mangrove forests. More broadly, we expect of Rhizophora mangle. Although these distances are important fac-
this work to contribute disproportionally to the seascape genetics tors that explain the species' connectivity, considering the nonlin-
and genomics literature because it focusses on coastal plants, a ear, dynamic and asymmetric nature of oceanic currents provided
functional group with few studies, and it was carried out in one a broader perspective on the connectivity of coastal plants across
of the ‘very underrepresented’ areas of the globe, the Southwest scales. As sea surface properties change, dispersal will likely be af-
Atlantic basin (Jahnke & Jonsson, 2022). fected, with potentially major ecological and evolutionary conse-
quences. The realism provided by the mechanistic explanation of
biophysical models to the population genetic structure and migra-
4.4 | Perspectives tion patterns we observed adds to the increasing literature regarding
the seascape genetics and genomics of coastal plants.
Our genetic-­b ased findings largely support the genetic structure
patterns described in previous studies on R. mangle (Francisco AC K N O​W L E​D G E​M E N T S
et al., 2018; Pil et al., 2011; Takayama et al., 2013) and other The authors are grateful to PR Laborda, SCS Andrade, AG Nazareno,
coastal plants (Cruz et al., 2019, 2020; Da Silva et al., 2021; Mori MPP Romeiro, T Tjui-­Yeuw, EM Moraes and MK Visentini for the
et al., 2015; Takayama et al., 2008). However, there are differ- constructive critiques and suggestions on earlier versions of this
ences across taxa. For instance, Avicennia, also a mangrove genus, manuscript, to MC Almeida for assistance with preliminary analysis,
and H. pernambucensis, a coastal plant that grows inland of man- and to Ms. A Sakai for technical assistance. This study was supported
grove forests and along brackish rivers, present sharper genetic by research awards granted by the Brazilian National Council for
breaks between groups found north and south of the NEESA (Cruz Scientific and Technological Development (CNPq) to GMM (CNPq
et al., 2019, 2020; Da Silva et al., 2021; Mori et al., 2015; Takayama 448286/2014-­9), by the Japan Society for the Promotion of Science
et al., 2008) compared with R. mangle. Remarkably, Laguncularia (JSPS) to TK (JSPS KAKENHI 25290080 and 17H01414) and by the
racemosa, a mangrove species whose geographic range largely United Graduate School of Agricultural Sciences from Kagoshima
overlaps with the one R. mangle presents, showed a homogenous University to YN. Also, we thank São Paulo Research Foundation
distribution of the genetic variation across the Southwest Atlantic (FAPESP) for the research fellowships to AGM (FAPESP 18/02655-­8
coast (Sereneski-­L ima et al., 2021), unlike its counterparts. A. ger- and 2020/07967-­8) and GMM (FAPESP 13/08086-­1 and 14/22821-­
minans and L. racemosa are entomophilous and have a higher fruit-­ 9). This research was supported by resources supplied by the UNESP
set than R. mangle (Sánchez-­N úñez & Mancera-­Pineda, 2012), Center for Scientific Computing (NCC/GridUNESP). Additionally, this
while having propagules with much lower viability period, of only study was financed in part by the Coordenação de Aperfeiçoamento
45 and 90 days, respectively, compared with more than 1 year for de Pessoal de Nível Superior—­Brasil (CAPES)—­Finance Code 001.
R. mangle (Van der Stocken, Wee, et al., 2019). Likewise, A. schaue-
riana is likely dependent on a broad range of insect pollinators as C O N F L I C T O F I N T E R E S T S TAT E M E N T
no fruit is spontaneously formed (Nadia et al., 2012). It reinforces The authors declare no conflict of interest.
that species-­specific traits, like pollination syndrome and prop-
agules' flotation and longevity, also contribute to the connectivity DATA AVA I L A B I L I T Y S TAT E M E N T
of mangrove plants (Van der Stocken, Wee, et al., 2019), which Individual genetic data are available as a VCF file as Supplementary
remains an open avenue for future investigations. Information S4. Also, Adrift and OpenDrift data are available in the
As climate rapidly changes, so do oceans. Recent evidence in- authors' GitHub repository https://github.com/gumor​i/Mangr​oveDi​
dicates that future environmental conditions under representative spers​al/.
concentration pathway 8.5 by 2100 will likely change dispersal po-
tential of mangroves. Changes in sea surface salinity, density and B E N E F I T-­S H A R I N G S TAT E M E N T
temperature of waters bordering mangrove forests are expected to This study is a consequence of a collaborative work developed by
make LDD events even rarer and promote short-­distances dispersal scientists from the countries; all collaborators are included as co-­
(Van der Stocken et al., 2022). Because north and south groups are authors. Our research group is committed to international scien-
genetically divergent and possess contrasting dispersal patterns, in tific partnerships, mainly via human resources capacity building.
MADEIRA et al. | 13

Canty, S. W. J., Kennedy, J. P., Fox, G., Matterson, K., González, V. L.,
Additionally, we are openly sharing our data and results as described Núñez-­Vallecillo, M. L., Preziosi, R. F., & Rowntree, J. K. (2022).
above. The scientific information generated by the authors may be Mangrove diversity is more than fringe deep. Scientific Reports, 12,
relevant to the conservation and sustainable use of the biological 1695.
Cayuela, H., Rougemont, Q., Prunier, J. G., Moore, J. S., Clobert, J.,
diversity, which include taxonomic studies and biological collections.
Besnard, A., & Bernatchez, L. (2018). Demographic and genetic ap-
proaches to study dispersal in wild animal populations: A method-
ORCID ological review. Molecular Ecology, 27, 3976–­4 010.
Gustavo Maruyama Mori https://orcid. Cerón-­Souza, I., Rivera-­Ocasio, E., Medina, E., Jiménez, J. A., McMillan,
org/0000-0003-2308-2224 W. O., & Bermingham, E. (2010). Hybridization and introgression in
new world red mangroves, Rhizophora (Rhizophoraceae). American
Journal of Botany, 97, 945–­957.
REFERENCES Chave, J. (2013). The problem of pattern and scale in ecology: What have
Aguirre-­Liguori, J. A., Luna-­Sánchez, J. A., Gasca-­Pineda, J., & Eguiarte, we learned in 20 years? Ecology Letters, 16, 4–­16.
L. E. (2020). Evaluation of the minimum sampling design for popu- Cornejo, X. (2013). Lectotypification and a new status for Rhizophora X
lation genomic and microsatellite studies: An analysis based on wild Harrisonii (Rhizophoraceae), a natural hybrid between R. Mangle
maize. Frontiers in Genetics, 11, 1–­16. and R. Racemosa. Harvard Papers in Botany, 18(1), 37.
Aho, K., Derryberry, D., & Peterson, T. (2014). Model selection for ecolo- Cruz, M. V., Mori, G. M., Oh, D. H., Dassanayake, M., Zucchi, M. I., Oliveira,
gists: The worldviews of AIC and BIC. Ecology, 95, 631–­636. R. S., & de Souza, A. P. (2020). Molecular responses to freshwater
Alexander, D. H., Novembre, J., & Lange, K. (2009). Fast model-­based limitation in the mangrove tree Avicennia germinans (Acanthaceae).
estimation of ancestry in unrelated individuals. Genome Research, Molecular Ecology, 29, 344–­362.
19, 1655–­1664. Cruz, M. V., Mori, G. M., Signori-­Müller, C., da Silva, C. C., Oh, D.-­H.,
Antao, T., Lopes, A., Lopes, R. J., Beja-­Pereira, A., & Luikart, G. (2008). Dassanayake, M., Zucchi, M. I., Oliveira, R. S., & de Souza, A. P.
LOSITAN: A workbench to detect molecular adaptation based on a (2019). Local adaptation of a dominant coastal tree to freshwater
Fst-­outlier method. BMC Bioinformatics, 9, 323. availability and solar radiation suggested by genomic and ecophys-
Assis, J., Castilho Coelho, N., Alberto, F., Valero, M., Raimondi, P., Reed, iological approaches. Scientific Reports, 9, 19936.
D., & Alvares Serrão, E. (2013). High and distinct range-­edge ge- Da Silva, M. F., Cruz, M. V., Vidal Júnior, J. D. D., Zucchi, M. I., Mori, G.
netic diversity despite local bottlenecks. PLoS One, 8, e68646. M., & De Souza, A. P. (2021). Geographical and environmental con-
Baird, N. A., Etter, P. D., Atwood, T. S., Currey, M. C., Shiver, A. L., Lewis, tributions to genomic divergence in mangrove forests. Biological
Z. A., Selker, E. U., Cresko, W. A., & Johnson, E. A. (2008). Rapid Journal of the Linnean Society, 132, 573–­589.
SNP discovery and genetic mapping using sequenced RAD markers. Dagestad, K.-­F., Röhrs, J., Breivik, Ø., & Ådlandsvik, B. (2018). OpenDrift
PLoS One, 3, 1–­7. v1.0: A generic framework for trajectory modelling. Geoscientific
Balbar, A. C., & Metaxas, A. (2019). The current application of ecological Model Development, 11, 1405–­1420.
connectivity in the design of marine protected areas. Global Ecology Danecek, P., Auton, A., Abecasis, G., Albers, C. A., Banks, E., DePristo, M.
and Conservation, 17, e00569. A., Handsaker, R. E., Lunter, G., Marth, G. T., Sherry, S. T., McVean,
Beaumont, M. A., & Nichols, R. A. (1996). Evaluating loci for use in the G., & Durbin, R. (2011). The variant call format and VCFtools.
genetic analysis of population structure. Proceedings of the Royal Bioinformatics, 27, 2156–­2158.
Society of London, Series B: Biological Sciences, 263, 1619–­1626. de Menezes, M. P. M., de Oliveira, D., & de Mello, C. F. (1997). Pollination
Bergmann, P. J., & McElroy, E. J. (2014). Many-­to-­many mapping of phe- of red mangrove, Rhizophora mangle, in northern Brazil. Acta
notype to performance: An extension of the F-­matrix for studying Horticulturae, 437, 431–­434.
functional complexity. Evolutionary Biology, 41, 546–­560. de Menezes, M. P. M., Berger, U., & Mehlig, U. (2008). Mangrove vege-
Bertola, L. D., Boehm, J. T., Putman, N. F., Xue, A. T., Robinson, J. D., tation in Amazonia: a review of studies from the coast of Pará and
Harris, S., Baldwin, C. C., Overcast, I., & Hickerson, M. J. (2020). Maranhão States, north Brazil. Acta Amazonica, 38(3), 403–­420.
Asymmetrical gene flow in five co-­distributed syngnathids ex- Diniz-­Filho, J. A. F., Diniz, J. V. B. P. L., Rangel, T. F., Soares, T. N., Telles,
plained by ocean currents and rafting propensity. Proceedings of the M. P. C., Collevatti, R. G., & Bini, L. M. (2013). A new eigenfunction
Royal Society B: Biological Sciences, 287, 1–­10. spatial analysis describing population genetic structure. Genetica,
Blanchet, F. G., Legendre, P., Maranger, R., Monti, D., & Pepin, P. (2011). 141, 479–­489.
Modelling the effect of directional spatial ecological processes at Doyle, J. J., & Doyle, J. L. (1990). Isolation of plant DNA from fresh tissue.
different scales. Oecologia, 166, 357–­368. Focus, 12, 13–­15.
Bohonak, A. J., & Vandergast, A. G. (2011). The value of DNA sequence Dray, S., Bauman, D., Blanchet, F. G., Borcard, D., Clappe, S., Guenard,
data for studying landscape genetics. Molecular Ecology, 20(12), G., Jombart, T., Larocque, G., Legendre, P., Madi, N., & Wagner, H.
2477–­2479. H. (2022). Adespatial: Multivariate multiscale spatial analysis. R
Bolger, A. M., Lohse, M., & Usadel, B. (2014). Trimmomatic: A flexible package.
trimmer for Illumina sequence data. Bioinformatics, 30, 2114–­2120. Dray, S., Legendre, P., & Peres-­Neto, P. R. (2006). Spatial modelling: A
Bonjean, F., & Lagerloef, G. S. E. (2002). Diagnostic model and analy- comprehensive framework for principal coordinate analysis of
sis of the surface currents in the tropical Pacific Ocean. Journal of neighbour matrices (PCNM). Ecological Modelling, 196, 483–­493.
Physical Oceanography, 32, 2938–­2954. Driscoll, D. A., Banks, S. C., Barton, P. S., Ikin, K., Lentini, P., Lindenmayer,
Bullock, J. M., & Nathan, R. (2008). Plant dispersal across multiple scales: D. B., Smith, A. L., Berry, L. E., Burns, E. L., Edworthy, A., Evans,
Linking models and reality. Journal of Ecology, 96, 567–­568. M. J., Gibson, R., Heinsohn, R., Howland, B., Kay, G., Munro, N.,
Cain, M. L. M. L., Milligan, B. G., & Strand, A. E. A. E. (2000). Long-­ Scheele, B. C., Stirnemann, I., Stojanovic, D., … Westgate, M. J.
distance seed dispersal in plant populations. American Journal of (2014). The trajectory of dispersal research in conservation biology.
Botany, 87, 1217–­1227. Systematic review. PLoS One, 9, e95053.
Campbell, I. D., McDonald, K., Flannigan, M. D., & Kringayark, J. (1999). Excoffier, L., & Lischer, H. E. L. (2010). Arlequin suite ver 3.5: A new se-
Long-­distance transport of pollen into the Arctic. Nature, 399(6731), ries of programs to perform population genetics analyses under
29–­3 0. Linux and windows. Molecular Ecology Resources, 10, 564–­567.
14 | MADEIRA et al.

Excoffier, L., Marchi, N., Marques, D. A., Matthey-­Doret, R., Gouy, A., flow and adaptation of forest trees to rapid climate change. Ecology
& Sousa, V. C. (2021). fastsimcoal2: Demographic inference under Letters, 15, 378–­392.
complex evolutionary scenarios. Bioinformatics, 37, 4882–­4885. Lalire, M., & Gaspar, P. (2019). Modeling the active dispersal of juvenile
Fick, S. E., & Hijmans, R. J. (2017). WorldClim 2: New 1-­km spatial reso- leatherback turtles in the North Atlantic Ocean. Movement Ecology,
lution climate surfaces for global land areas. International Journal of 7, 1–­17.
Climatology, 37, 4302–­4315. Langmead, B., & Salzberg, S. L. (2012). Fast gapped-­read alignment with
Foll, M., & Gaggiotti, O. (2008). A genome-­scan method to identify se- bowtie 2. Nature Methods, 9, 357–­359.
lected loci appropriate for both dominant and codominant markers: Legendre, P., & Gallagher, E. D. (2001). Ecologically meaningful
A bayesian perspective. Genetics, 180, 977–­993. transformations for ordination of species data. Oecologia, 129,
Francisco, P. M., Mori, G. M., Alves, F. M., Tambarussi, E. V., & de Souza, 271–­280.
A. P. (2018). Population genetic structure, introgression, and hy- Leiva, C., Riesgo, A., Combosch, D., Arias, M. B., Giribet, G., Downey, R.,
bridization in the genus Rhizophora along the Brazilian coast. Kenny, N. J., & Taboada, S. (2022). Guiding marine protected area
Ecology and Evolution, 8, 3491–­3504. network design with comparative phylogeography and population
François, O., Martins, H., Caye, K., & Schoville, S. D. (2016). Controlling genomics: An exemplary case from the Southern Ocean. Diversity
false discoveries in genome scans for selection. Molecular Ecology, and Distributions, 28(9), 1891–­1907.
25, 454–­469. Li, H. (2011). A statistical framework for SNP calling, mutation discovery,
Frichot, E., Schoville, S. D., Bouchard, G., & Franc, O. (2013). Testing association mapping and population genetical parameter estima-
for associations between loci and environmental gradients using tion from sequencing data. Bioinformatics, 27, 2987–­2993.
latent factor mixed models. Molecular Biology and Evolution, 30, Liu, J., Lindstrom, A. J., Chen, Y., Nathan, R., & Gong, X. (2021).
1687–­1699. Congruence between ocean-­dispersal modelling and phylogeog-
Geng, Q., Wang, Z., Tao, J., Kimura, M. K., Liu, H., Hogetsu, T., & Lian, C. raphy explains recent evolutionary history of Cycas species with
(2021). Ocean currents drove genetic structure of seven dominant buoyant seeds. The New Phytologist, 232, 1863–­1875.
mangrove species along the coastlines of southern China. Frontiers Lo, E. Y., Duke, N. C., & Sun, M. (2014). Phylogeographic pattern of
in Genetics, 12, 1–­20. Rhizophora (Rhizophoraceae) reveals the importance of both vi-
Grünwald, N. J., Everhart, S. E., Knaus, B. J., & Kamvar, Z. N. (2017). Best cariance and long-distance oceanic dispersal to modern mangrove
practices for population genetic analyses. Phytopathology, 107, distribution. BMC Evolutionary Biology, 14(1), 1–­15.
1000–­1010. Lowe, W. H., Kovach, R. P., & Allendorf, F. W. (2017). Population genetics
Hasan, S., Triest, L., Afrose, S., & De Ryck, D. J. R. (2018). Migrant pool and demography unite ecology and evolution. Trends in Ecology &
model of dispersal explains strong connectivity of Avicennia offi- Evolution, 32, 141–­152 Elsevier Ltd.
cinalis within Sundarban mangrove areas: Effect of fragmentation Lumpkin, R., & Johnson, G. C. (2013). Global Ocean surface velocities
and replantation. Estuarine, Coastal and Shelf Science, 214, 38–­47 from drifters: Mean, variance, El Niño-­southern oscillation re-
Elsevier. sponse, and seasonal cycle. Journal of Geophysical Research, Oceans,
Hays, G. C. (2017). Ocean currents and marine life. Current Biology, 27, 118, 2992–­3 006.
R470–­R473 Elsevier. Luu, K., Bazin, E., & Blum, M. G. B. (2017). Pcadapt: An R package to
Hays, G. C., Ferreira, L. C., Sequeira, A. M. M., Meekan, M. G., Duarte, perform genome scans for selection based on principal component
C. M., Bailey, H., Bailleul, F., Bowen, W. D., Caley, M. J., Costa, analysis. Molecular Ecology Resources, 17, 67–­77.
D. P., Eguíluz, V. M., Fossette, S., Friedlaender, A. S., Gales, N., Mantel, N. (1967). The detection of disease clustering and a generalized
Gleiss, A. C., Gunn, J., Harcourt, R., Hazen, E. L., Heithaus, M. R., regression approach. Cancer Research, 27, 209–­220.
… Thums, M. (2016). Key questions in marine megafauna move- Martin, M. (2011). Cutadapt removes adapter sequences from high-­
ment ecology. Trends in Ecology & Evolution, 31, 463–­475 Elsevier throughput sequencing reads. EMBnet.Journal, 17, 10.
Ltd. McLaughlin, J. F., & Winker, K. (2020). An empirical examination of sam-
Hodel, R. G. J., Knowles, L. L., McDaniel, S. F., Payton, A. C., Dunaway, J. ple size effects on population demographic estimates in birds using
F., Soltis, P. S., & Soltis, D. E. (2018). Terrestrial species adapted to single nucleotide polymorphism (SNP) data. PeerJ, 8, e9939.
sea dispersal: Differences in propagule dispersal of two Caribbean Mori, G. M., Madeira, A. G., Cruz, M. V., Tsuda, Y., Takayama, K., Matsuki,
mangroves. Molecular Ecology, 27, 4612–­4626. Y., Suyama, Y., Iwasaki, T., de Souza, A. P., Zucchi, M. I., & Kajita,
Jahnke, M., & Jonsson, P. R. (2022). Biophysical models of dispersal con- T. (2021). Testing species hypotheses in the mangrove genus
tribute to seascape genetic analyses. Philosophical Transactions of Rhizophora from the Western hemisphere and South Pacific islands.
the Royal Society, B: Biological Sciences, 377, 1–­10. Estuarine, Coastal and Shelf Science, 248, 1–­11.
Jahnke, M., Moksnes, P., Olsen, J. L., Serra Serra, N., Jacobi, M. N., Mori, G. M., Zucchi, M. I., & Souza, A. P. (2015). Multiple-­geographic-­
Kuusemäe, K., Corell, H., & Jonsson, P. R. (2020). Integrating genet- scale genetic structure of two mangrove tree species: The roles of
ics, biophysical, and demographic insights identifies critical sites for mating system, hybridization, limited dispersal and extrinsic fac-
seagrass conservation. Ecological Applications, 30(6), e02121. tors. PLoS One, 10, 1–­23.
Jombart, T., & Ahmed, I. (2011). Adegenet 1.3-­1: New tools for the analy- Mussmann, S. M., Douglas, M. R., Chafin, T. K., & Douglas, M. E. (2019).
sis of genome-­wide SNP data. Bioinformatics, 27, 3070–­3 071. BA3-­SNPs: Contemporary migration reconfigured in BayesAss for
Jombart, T., Devillard, S., & Balloux, F. (2010). Discriminant analysis of next-­generation sequence data. Methods in Ecology and Evolution,
principal components: A new method for the analysis of genetically 10, 1808–­1813.
structured populations. BMC Genetics, 11, 94. Nadia, T. L., & Machado, I. C. (2014). Wind pollination and propagule
Jordano, P. (2017). What is long-­distance dispersal? And a taxonomy of formation in Rhizophora mangle L. (Rhizophoraceae): Resource or
dispersal events. Journal of Ecology, 105, 75–­8 4. pollination limitation? Anais da Academia Brasileira de Ciências, 86(1),
Kondo, K., Nakamura, T., Tsuruda, K., Saito, N., & Yaguchi, Y. (1987). 229–­238.
Pollination in Bruguiera gymnorrhiza and Rhizophora mucronata Nadia, T. L., Morellato, L. P. C., & Machado, I. C. (2012). Reproductive
(Rhizophoraceae) in Ishigaki Island, the Ryukyu Islands, Japan. phenology of a northeast Brazilian mangrove community:
Biotropica, 19(4), 377. Environmental and biotic constraints. Flora –­ Morphology,
Kremer, A., Ronce, O., Robledo-­Arnuncio, J. J., Guillaume, F., Bohrer, G., Distribution, Functional Ecology of Plants, 207(9), 682–­692.
Nathan, R., Bridle, J. R., Gomulkiewicz, R., Klein, E. K., Ritland, K., Narasimhan, V., Danecek, P., Scally, A., Xue, Y., Tyler-­Smith, C., & Durbin,
Kuparinen, A., Gerber, S., & Schueler, S. (2012). Long-­distance gene R. (2016). BCFtools/RoH: A hidden Markov model approach for
MADEIRA et al. | 15

detecting autozygosity from next-­generation sequencing data. instrumentation meets high-­throughput sequencing. PLoS Biology,
Bioinformatics, 32, 1749–­1751. 14, e1002350.
Nathan, R. (2006). Long-­distance dispersal of plants. Long Distance Slatkin, M. (1995). A measure of population subdivision based on micro-
Dispersal of Plants., 313, 786–­788. satellite allele frequencies. Genetics, 139, 457–­462.
Nathan, R., Perry, G., Cronin, J. T., Strand, A. E., Cain, M. L., & Methods, Smith, T. M., York, P. H., Broitman, B. R., Thiel, M., Hays, G. C., van Sebille,
M. L. (2003). Methods for estimating long-­distance dispersal. Oikos, E., Putman, N. F., Macreadie, P. I., & Sherman, C. D. H. (2018). Rare
103, 261–­273. long-­distance dispersal of a marine angiosperm across the Pacific
Nathan, R., Schurr, F. M., Spiegel, O., Steinitz, O., Trakhtenbrot, A., & Ocean. Global Ecology and Biogeography, 27, 487–­496.
Tsoar, A. (2008). Mechanisms of long-­distance seed dispersal. Takayama, K., Tamura, M., Tateishi, Y., Webb, E. L., & Kajita, T. (2013).
Trends in Ecology & Evolution, 23, 638–­6 47. Strong genetic structure over the American continents and
Nazareno, A. G., Bemmels, J. B., Dick, C. W., & Lohmann, L. G. (2017). transoceanic dispersal in the mangrove genus Rhizophora
Minimum sample sizes for population genomics: An empirical study (Rhizophoraceae) revealed by broad-­ scale nuclear and chloroplast
from an Amazonian plant species. Molecular Ecology Resources, DNA analysis. American Journal of Botany, 100, 1191–­1201.
17(6), 1136–­1147. Takayama, K., Tateishi, Y., & Kajita, T. (2021). Global phylogeography of
Nettel, A., & Dodd, R. S. (2007). Drifting propagules and receding a pantropical mangrove genus Rhizophora. Scientific Reports, 11,
swamps: Genetic footprints of mangrove recolonization and dis- 7228.
persal along tropical coasts. Evolution (N. Y), 61, 958–­971. Takayama, K., Tateishi, Y., Murata, J., & Kajita, T. (2008). Gene flow and
Ngeve, M. N., Koedam, N., & Triest, L. (2021). Genotypes of Rhizophora population subdivision in a pantropical plant with sea-­drifted seeds
propagules from a non-­mangrove beach provide evidence of recent Hibiscus tiliaceus and its allied species: Evidence from microsatellite
long-­distance dispersal. Frontiers in Conservation Science, 2, 1–­16. analyses. Molecular Ecology, 17, 2730–­2742.
Ngeve, M. N., Van der Stocken, T., Sierens, T., Koedam, N., & Triest, Tomlinson, P. B. (2016). The botany of mangroves. Cambridge University
L. (2017). Bidirectional gene flow on a mangrove river land- Press.
scape and between-­c atchment dispersal of Rhizophora racemosa Travis, J. M. J., Delgado, M., Bocedi, G., Baguette, M., Bartoń, K.,
(Rhizophoraceae). Hydrobiologia. Springer International Publishing, Bonte, D., Boulangeat, I., Hodgson, J. A., Kubisch, A., Penteriani,
790, 93–­108. V., Saastamoinen, M., Stevens, V. M., & Bullock, J. M. (2013).
Ngeve, M. N., Vanderstocken, T., Menemenlis, D., Koedam, N., & Triest, Dispersal and species' responses to climate change. Oikos, 122,
L. (2016). Contrasting effects of historical sea level rise and con- 1532–­1540.
temporary ocean currents on regional gene flow of Rhizophora rac- Van der Stocken, T., Carroll, D., Menemenlis, D., Simard, M., & Koedam,
emosa in eastern Atlantic mangroves. PLoS One, 11, 1–­24. N. (2019). Global-­scale dispersal and connectivity in mangroves.
Nikolic, N., Montes, I., Lalire, M., Puech, A., Bodin, N., Arnaud-­Haond, S., Proceedings of the National Academy of Sciences, 116, 915–­922.
Kerwath, S., Corse, E., Gaspar, P., Hollanda, S., Bourjea, J., West, W., Van der Stocken, T., Vanschoenwinkel, B., Carroll, D., Cavanaugh, K. C.,
& Bonhommeau, S. (2020). Connectivity and population structure & Koedam, N. (2022). Mangrove dispersal disrupted by projected
of albacore tuna across Southeast Atlantic and southwest Indian changes in global seawater density. Nature Climate Change, 12,
oceans inferred from multidisciplinary methodology. Scientific 685–­691 Springer US.
Reports, 10, 15657. Van der Stocken, T., Wee, A. K. S., De Ryck, D. J. R., Vanschoenwinkel,
Obst, M. (2017). Global tide variables. Swedish National Data Service. B., Friess, D. A., Dahdouh-­Guebas, F., Simard, M., Koedam, N., &
Oksanen, J., Blanchet, F. G., Friendly, M., Kindt, R., Legendre, P., McGlinn, Webb, E. L. (2019). A general framework for propagule dispersal in
D., Minchin, P. R., O'Hara, R. B., Simpson, G. L., Solymos, P., Stevens, mangroves. Biological Reviews, 94, 1547–­1575.
M. H. H., Szoecs, E., & Wagner, H. (2020). vegan: community ecol- van Etten, J. (2017). R package gdistance : Distances and routes on geo-
ogy package. R package. graphical grids. Journal of Statistical Software, 76(13), 1–­21.
Petkova, D., Novembre, J., & Stephens, M. (2015). Visualizing spatial van Sebille, E. (2014). Adrift.org.au —­A free, quick and easy tool to quan-
population structure with estimated effective migration surfaces. titatively study planktonic surface drift in the global ocean. Journal
Nature Genetics, 48, 94–­100. of Experimental Marine Biology and Ecology, 461, 317–­322 Elsevier
Pil, M. W. W., Boeger, M. R. T. T. R. T., Muschner, V. C. C., Pie, M. R. R., B.V.
Ostrensky, A., & Boeger, W. A. A. (2011). Postglacial north-­south Villarino, E., Watson, J. R., Jönsson, B., Gasol, J. M., Salazar, G., Acinas,
expansion of populations of Rhizophora mangle (Rhizophoraceae) S. G., Estrada, M., Massana, R., Logares, R., Giner, C. R., Pernice,
along the Brazilian coast revealed by microsatellite analysis. M. C., Olivar, M. P., Citores, L., Corell, J., Rodríguez-­Ezpeleta, N.,
American Journal of Botany, 98, 1031–­1039. Acuña, J. L., Molina-­Ramírez, A., González-­Gordillo, J. I., Cózar, A.,
Robledo-­Arnuncio, J. J. (2011). Wind pollination over mesoscale dis- … Chust, G. (2018). Large-­Scale Ocean connectivity and planktonic
tances: An investigation with scots pine. New Phytologist, 190(1), body size. Nature Communications, 9, 1–­13.
222–­233. Wee, A. K. S., Mori, G. M., Lira, C. F., Núñez-­Farfán, J., Takayama, K.,
Rochette, N. C., & Catchen, J. M. (2017). Deriving genotypes from RAD-­ Faulks, L., Shi, S., Tsuda, Y., Suyama, Y., Yamamoto, T., Iwasaki, T.,
seq short-­read data using Stacks. Nature Publishing Group, 12, Nagano, Y., Wang, Z., Watanabe, S., & Kajita, T. (2018). The integra-
2640–­2659. tion and application of genomic information in mangrove conserva-
Sánchez-­Núñez, D. A., & Mancera-­Pineda, J. E. (2012). Pollination and tion. Conservation Biology, 33(1), 206–­209.
fruit set in the main neotropical mangrove species from the south- Wee, A. K. S., Noreen, A. M. E., Ono, J., Takayama, K., Kumar, P. P., Tan,
western Caribbean. Aquatic Botany, 103, 60–­65. H. T. W., Saleh, M. N., Kajita, T., & Webb, E. L. (2020). Genetic
Sbrocco, E. J., & Barber, P. H. (2013). MARSPEC: Ocean climate layers for structures across a biogeographical barrier reflect dispersal po-
marine spatial ecology. Ecology, 94, 979. tential of four southeast Asian mangrove plant species. Journal of
Sereneski-­Lima, C., Baggio, R. A., Pil, M. W., Torres Boeger, M. R., & Biogeography, 47, 1258–­1271.
Boeger, W. A. (2021). Historical and contemporary factors affect Wee, A. K. S., Takayama, K., Asakawa, T., Thompson, B., Onrizal, S.,
the genetic diversity and structure of Laguncularia racemosa (L.) Sungkaew, N. X., Tung, M., Nazre, K. K., Soe, H. T. W., Tan, Y.,
Gaertn, along the western Atlantic coast. Estuarine, Coastal and Watano, S., Baba, T. K., & Webb, E. L. (2014). Oceanic currents,
Shelf Science, 249, 107055 Elsevier Ltd. not land masses, maintain the genetic structure of the mangrove
Shafer, A. B. A., Northrup, J. M., Wikelski, M., Wittemyer, G., & Wolf, Rhizophora mucronata lam. (Rhizophoraceae) in Southeast Asia.
J. B. W. (2016). Forecasting ecological genomics: High-­tech animal Journal of Biogeography, 41, 954–­964.
16 | MADEIRA et al.

Weir, B. S., & Cockerham, C. C. (1984). Estimating f -­statistics for the

analysis of population structure. Evolution (N. Y), 38, 1358–­1370. S U P P O R T I N G I N FO R M AT I O N
Wilson, G. A., & Rannala, B. (2003). Bayesian inference of recent migra- Additional supporting information can be found online in the
tion rates using multilocus genotypes. Genetics, 1191, 1177–­1191. Supporting Information section at the end of this article.
Wu, Z.-­Y., Liu, J., Provan, J., Wang, H., Chen, C.-­J., Cadotte, M. W., Luo,
Y.-­H., Amorim, B. S., Li, D.-­Z., & Milne, R. I. (2018). Testing Darwin's
transoceanic dispersal hypothesis for the inland nettle family
(Urticaceae). Ecology Letters, 21, 1515–­1529.
How to cite this article: Madeira, A. G., Tsuda, Y., Nagano, Y.,
Xu, S., He, Z., Zhang, Z., Guo, Z., Guo, W., Lyu, H., Li, J., Yang, M., Du, Z.,
Huang, Y., Zhou, R., Zhong, C., Boufford, D. E., Lerdau, M., Wu, C. Iwasaki, T., Zucchi, M. I., Kajita, T., & Mori, G. M. (2023). The
I., Duke, N. C., Shi, S., Lee, S. Y., Li, X., … Baereleo, R. T. (2017). The role of oceanic currents in the dispersal and connectivity of
origin, diversification and adaptation of a major mangrove clade the mangrove Rhizophora mangle on the Southwest Atlantic
(Rhizophoreae) revealed by whole-­genome sequencing. National
region. Molecular Ecology Resources, 00, 1–16. https://doi.
Science Review, 4, 721–­734.
Xuereb, A., Benestan, L., Normandeau, É., Daigle, R. M., Curtis, J. M. R., org/10.1111/1755-0998.13807
Bernatchez, L., & Fortin, M.-­J. (2018). Asymmetric oceanographic
processes mediate connectivity and population genetic structure,
as revealed by RADseq, in a highly dispersive marine invertebrate
(Parastichopus californicus). Molecular Ecology, 27, 2347–­2364.