Marine protected area restricts demographic connectivity: Dissimilarity in a marine environment can function as a biological barrier

Abstract The establishment of marine protected areas (MPAs) can often lead to environmental differences between MPAs and fishing zones. To determine the effects on marine dispersal of environmental dissimilarity between an MPA and fishing zone, we examined the abundance and recruitment patterns of two anemonefishes (Amphiprion frenatus and A. perideraion) that inhabit sea anemones in different management zones (i.e., an MPA and two fishing zones) by performing a field survey and a genetic parentage analysis. We found lower levels of abundance per anemone in the MPA compared to the fishing zones for both species (n = 1,525 anemones, p = .032). The parentage analysis also showed that lower numbers of fishes were recruited from the fishing zones and outside of the study area into each anemone in the MPA than into each anemone in the fishing zones (n = 1,525 anemones, p < .017). However, the number of self‐recruit production per female did not differ between the MPA and fishing zones (n = 384 females, p = .516). Because the ocean currents around the study site were unlikely to cause a lower settlement intensity of larvae in the MPA, the ocean circulation was not considered crucial to the observed abundance and recruitment patterns. Instead, stronger top‐down control and/or a lower density of host anemones in the MPA were potential factors for such patterns. Our results highlight the importance of dissimilarity in a marine environment as a factor that affects connectivity.


| INTRODUCTION
Many marine species have lifecycles with a pelagic larval phase, in which larvae disperse among habitat patches, and a benthic adult phase, in which relatively sedentary adults are found in habitat patches after settlement (Jones, Almany, et al., 2009). In the marine environment, larval dispersal plays an important role in demographic connections among patches that is fundamental for population persistence and resilience from disturbance Sale et al., 2005). Thus, quantifying the patterns of demographic connectivity via larval dispersal is essential for predicting population dynamics and for management of marine species.
The demographic connectivity of marine fishes has been directly estimated using a genetic parentage analysis and/or larval tagging (e.g., Almany et al. 2007;Jones, Planes, & Thorrold, 2005;Planes, Jones, & Thorrold, 2009). A comparison of marine dispersal estimates by these methods revealed a large variation in the dispersal distance for coral reef fishes. For example, estimates of the dispersal distance ranged from <50 m to 35 km for Amphiprion and from 1 to 33 km for Chaetodon vagabundus (Berumen et al., 2012;Jones et al., 2005; Saenz-Agudelo, . Marine dispersal studies have also shown a rapid decrease in the dispersal probability within the first few kilometers (Buston, Jones, Planes, & Thorrold 2012;D'Aloia, Bogdanowicz, Majoris, Harrison, & Buston, 2013;Saenz-Agudelo et al., 2012). Therefore, the spatial scale of dispersal distance is largely restricted within a certain radius, although some individuals disperse for long distances (i.e., more than 30 km). Demographic connectivity can be influenced by various factors, including physical and biological barriers. Physical barriers include geographic distance, topography, and oceanographic features, which are the most prominent factors limiting connectivity in marine environments (D'Aloia, Bogdanowicz, Harrison, & Buston, 2014;D'Aloia et al., 2013;Nakajima et al., 2014; Saenz-Agudelo, Jones, Thorrold, & Planes, 2011;Saenz-Agudelo et al., 2012;White et al. 2010). In contrast, biological barriers are less studied in the sea (but see Turgeon & Kramer, 2012) although they are expected to affect the emigration and immigration of marine species and eventually the demographic connectivity (Cowen & Sponaugle, 2009;Marshall, Monro, Bode, Keough, & Swearer, 2010). For example, predator abundance, habitat quality, and habitat heterogeneity have been shown to influence the connectivity of species in terrestrial systems (Fuller, Doyle, & Strayer, 2015;Wang, Glor, & Losos, 2013). Increased understanding of such biological barriers to marine connectivity will be helpful for conservation plans, such as the design of marine protected areas (MPAs).
The number of MPAs has been increasing rapidly around the world, and they are used as conservation and fisheries management tools (Edgar et al., 2014). Marine protected areas are generally designed to provide insurance against declines of species due to fishing and environmental disturbance and to enhance the production of species outside their boundaries through the spillover of adult individuals and larval subsidy (Sale et al., 2005). Such management practices often cause variation in the marine environment between MPAs and fishing zones (e.g., McCook et al., 2010;Mumby et al., 2006Mumby et al., , 2007White, 1986). Because the target species of MPAs are often large predatory fishes, the increase in predator abundance within MPAs has been often observed (reviewed in Babcock et al., 2010;Russ, 2002). The presence of predators can have negative effects on the reproduction, egg survival, and settlement of prey species through lethal and nonlethal modes (Nakaoka, 2000;Richardson, Hare, Fogarty, & Link, 2011;Stier, Hanson, Holbrook, Schmitt, & Brooks, 2014); therefore, a greater abundance of predators in an MPA may decrease the larval subsidy of prey species to the surrounding area and/or their recruitment into an MPA, resulting in reduced connectivity between them. In addition, healthy coral reefs can be maintained in MPAs because of restrictions on human activity and improvement of ecosystem functions (McCook et al., 2010;Mumby et al., 2006Mumby et al., , 2007White, 1986). However, other sessile organisms such as macroalgae, soft coral, and sea anemone, which compete with coral for space, may become more highly developed in a fishing zone. Greater coral cover in an MPA can provide more settlement sites for fish that depend on coral, but a fishing zone may attract more fish settlers that inhabit other substrates. Such habitat heterogeneity between MPAs and fishing zones can also be a biological barrier to demographic connectivity between them. Recent studies have empirically shown that an MPA provides a larval subsidy for fishing zones (Bonin et al., 2016;Harrison et al., 2012;Planes et al., 2009).
However, the effects of dissimilarity in marine environment between MPAs and fishing zones on demographic connectivity have not been examined.
In this study, we examined the differences in the abundance levels and numbers of recruitment per habitat of two anemonefish species (Amphiprion frenatus and A. perideraion) between an MPA and two fishing zones using a field survey and genetic parentage analysis. Anemonefish are low trophic level species in coral reefs, and their distribution patterns are easily monitored due to their strong habitat association (i.e., anemonefish inhabit specific anemone species). Genetic parentage analysis, using highly polymorphic markers (e.g., microsatellite markers), was recently applied to marine systems, which has allowed great progress in quantifying the dispersal patterns of coral reef fishes (e.g., Bonin et al., 2016;D'Aloia et al., 2013;Harrison et al., 2012;Jones et al., 2005;Planes et al., 2009;Saenz-Agudelo et al., 2011. Anemonefish have been widely used as model species for this method, mainly because they are easily located and can be caught underwater through use of SCUBA. We used two anemonefishes as target organisms: The abundance at each host anemone of each anemonefish was surveyed in two different management zones. We explored their larval dispersal patterns along a 1.5 km stretch of coral reef that included the two zones, using genetic parentage analysis. Finally, based on the results of the parentage analysis, we assessed whether the numbers of recruitment from each zone and outside of the study area at each anemone, and the number of self-recruitment production per female differed between the zones. We hypothesized that lower abundance per anemone, recruitment number per anemone, and recruitment production per female anemonefish would be found in the MPA, in which predatory fishes are more abundant, while the host anemones are less abundant than in fishing zones.

| Study species and study site
The tomato anemonefish (Amphiprion frenatus) and the pink anemonefish (A. perideraion) are found from the eastern Indian Ocean to the western Pacific Ocean (Fautin & Allen, 1992). The two fish species generally do not share host anemones. The pelagic larval durations are 7-9 days for A. frenatus and 10-12 days for A. perideraion (Anto & Turingan, 2010;Thresher, Colin, & Bell, 1989).
We conducted a study at Laguindingan (LG) in northern Mindanao Island, the Philippines (Figure 1). The study site was situated on fringing reefs that faced the open sea. The study area included a coral reef area in two different management zones: one was a MPA, which has been maintained as a strict no-take zone since 2002 (Honda, Nakamura, Nakaoka, Uy, & Fortes, 2013), and the others were two fishing zones, which extends for 600 m from the boundary of the MPA in the east and west directions (Figure 1). In the fishing zones, snapper, emperor, grouper, and rabbit fish were common fishery targets, whereas anemonefish were not targeted for fishery or aquarium trade (author's personal communication). Based on the manta tow survey conducted at the study site in March 2013, the hard coral cover was higher in the MPA than in the two fishing zones, whereas coverage of dead coral (DC) was higher in the latter (Table S1). We selected potential predators of anemonefish by identifying fish species that had previously been reported in the literature to consume small reef fish.
We then compared the density of the potential predators in the MPA and fishing zones with the Wilcoxon rank sum test using the data previously collected at the study site (Table 1). Potential predators such as snapper (Lutjanus argentimaculatus, L. decussatus, L. fulviflamma, and L. fulvus) were significantly more abundant on coral reefs in the MPA than in the two fishing zones between March 2011 and September 2012 (K. Honda, unpublished data). The core home ranges of such predators, including snapper and emperor fish, were mostly restricted to areas within the MPA (Honda et al., 2016). The densities of Saurida gracilis (lizardfish) and Myripristis sp. 1 (squirrelfish) were also significantly higher in the MPA than in the east fishing zone, and a relatively higher density of other species was also found in the MPA, except for some wrasses and triggerfish between June and August 2011 (D. B. Recamara, unpublished data

| Field survey of anemonefish
In November 2012, we conducted a preliminary survey by snorkeling on coral reefs inside and outside the MPA to record the location of anemonefish and host anemones, because they were abundant only in such habitats (Figure 1; . A GPS device (Garmin eTrex 30) was used to determine the locations. On the basis of the location data, we investigated the distribution patterns of anemonefish on coral reefs at a depth of 0.5-15 m using SCUBA, from May to July in 2013 ( Figure 1). We counted the abundance of the target species at each anemone, recorded their total length (TL, mm), and measured long and short axial lengths (cm) of host sea anemones, using a ruler to estimate the habitat area as an oval by (long axial length) × (short axial length) × π/4 (Hattori, 1991;. Based on the measured fish size, we separately recorded adult (>30 mm TL) and F I G U R E 1 Study site at Laguindingan, northern Mindanao Island, the Philippines. Location of (a) the Laguindingan and (b) the study site within the box. (c, d) Map of study area on coral reefs (0.5-15 m depth), habitats, and MPA. Gray dots indicate all host anemones for (c) Amphiprion frenatus (n = 208) and (d) A. perideraion (n = 1,318) in the study area, and triangles on panel (c) indicate the deployment points of the acoustic Doppler current profiler (ADCP), electromagnetic current meter (EM), and water level logger (WL). The reef margin is located at a steep reef wall (i.e., drop-off), with a bottom depth along the wall of 20-30 m T A B L E 1 Mean individual number (±SD) of potential predators of anemonefish per 1,000 per m 2 that were observed by performing a visual fish census in the MPA and in two fishing zones at Laguindingan with the results of a Wilcoxon rank sum test examining differences in the individual numbers of the potential predators between the different management zones (MPA vs. fishing zone) for each species juvenile (≤30 mm TL) abundance levels of target species in accordance with the classification of juveniles for the parentage analysis below.

| Field collection of genetic samples
During the field survey from May to July in 2013, we also collected genetic samples of target anemonefish. We targeted a pair of the two largest fish and single largest fish in each habitat whose total length (TL) was longer than the minimum mature size of each species (≥80 mm for A. frenatus female and ≥46 mm for its male, ≥57 mm for A. perideraion female and >39 mm for its male; Hattori, 1991Hattori, , 2000 as "breeders." Individuals <30 mm TL (≤30 mm TL) were also targeted as "juveniles" for both species (Berumen et al., 2012). We captured anemonefish using hand-nets and clove oil, and then we measured their TL to the nearest mm underwater. Biodegradable colored tape was placed near to an anemone where an anemonefish had been collected to mark the position. Anemonefish were fin clipped using scissors and then released back to the same host sea anemone. Fish that were too small to be fin clipped (<30 mm) were collected. All of the samples were stored in 95% EtOH and brought back to the laboratory for the subsequent genetic analysis.
A 30 mm Amphiprion was estimated to be approximately 3-4 months old (Ochi, 1986); therefore, we regarded all juveniles as being no more

| Genetic analysis
All of the individuals were genotyped using 14 microsatellite loci for A. frenatus and 15 loci for A. perideraion (Table S2) The allele frequencies, observed and expected heterozygosity, deviation from Hardy-Weinberg equilibrium, and the frequency of null alleles were calculated using CERVUS v. 3.0.7 (Kalinowski, Taper, & Marshall, 2007). For each species, we assessed genetic differentiation between the MPA and fishing zones using F statistics via AMOVA in GenAlEx v. 6.5 (Peakall & Smouse, 2012).
We conducted a genetic parentage analysis to identify selfrecruits for each target species using the program COLONY v. 2.0.5.0 (Jones & Wang, 2010). This program implements a full-likelihood method of parentage analysis and defines the a priori probability that the true parent is present in the samples. COLONY is robust to uncertainty in the sampling rate of true parentage and has been shown to outperform other programs with less than 20 polymorphic loci (Harrison, Saenz-Agudelo, Planes, Jones, & Berumen, 2013 where S was the number of settlers assigned to breeders in the study site (self-recruits), and I was the number of settlers not assigned to the breeders (immigrants). Based on the assignments, we classified (1) recruits (juveniles) that migrated from the fishing zones, (2) those from the MPA, and (3) those from outside of the study area ( Figure 2b).
We calculated the direct distance and direction (east-west) between the origin and destination anemone of self-recruits to generate an observed dispersal distance and direction.
To assess the information sufficiency of our markers for accurate reconstruction of parental assignment, we used the simulation module in COLONY (Wang, 2013). The module simulates juvenile genotypes with a predefined parentage and sibship structure, based on a given marker number, allele frequencies, and an assumed mating matrix.
It then returns a metric of the accuracy of parentage assignments (Muralidhar, De Sá, Haddad, & Zamudio, 2014). We used identical parameters to the original COLONY run to simulate juveniles at the study site and to determine the confidence in our parentage assignments.

| Directions in dispersal tracks of self-recruits
To examine whether the self-recruits of A. frenatus and A. perideraion had unidirectional dispersal patterns, we tested for differences in the proportion of juveniles traveling east or west along the coast. If the predominant currents had unidirectional patterns along the coast and their influence was crucial for the dispersal, the larval dispersal patterns should reflect the dominant current flows.
Self-recruitment rate = S S + I

| Current measurement
To assess the general circulation patterns around the study site, we in burst mode (i.e., one measurement every 1 s for 300 s at 120-min intervals), and the average of these measurements was used for each 120-min interval.

| Statistical analysis
We first tested whether the (1) total, (2) juvenile (≤30 mm TL), and  Figure 2b) as well as the numbers of self-recruit production per female differed between the two zones to examine the effects of zone on recruitment and recruit production. As in the above analysis, we used the number of recruits per anemone to rule out the confounding effects of anemone density in each zone. We used a GLMM with a Poisson error distribution and treated zone, anemonefish species, and the interaction term as fixed factors and location as a random factor. To account for variations in the habitat size (sea anemone) and female size (female TL), the log of habitat size and female size were included as the offset terms in the predictors for the numbers of recruits per anemone and that of self-recruit productions per female, respectively. We did not use the degree of saturation as an offset term for the number of recruits per anemone for the same reason we did not use juvenile abundance. An LRT was performed to determine the significance of the fixed factors.

| Distribution patterns of anemonefish
From May to July 2013, we found a total of 462 individuals of Amphiprion frenatus and 208 individuals of its host sea anemone F I G U R E 2 Illustrations of measurements of (a) anemonefish abundance per anemone and (b) number of recruits (juveniles) from different zones per anemone. Origins of recruits (i.e., MPA, fishing zones, and outside of the study site) were determined by parentage analysis species, and 1,205 individuals of A. perideraion and 1,318 individuals of its host sea anemone species (including anemone individuals without anemonefish) at the study site. The occurrences of both anemonefishes were observed in depth ranges of 1.0-12.0 m. Both anemonefish densities were higher in the western and eastern fishing zones (0.42 and 0.36 fish/100 m 2 for A. frenatus; 1.43 and 0.80 fish/100 m 2 for A. perideraion) than in the MPA (0.23 fish/100 m 2 for A. frenatus; 0.42 fish/100 m 2 for A. perideraion). The host anemone densities were also higher in the two fishing zones (0.19 and 0.16 anemone/100 m 2 for A. frenatus; 1.13 and 1.18 anemone/100 m 2 for A. perideraion) than in the MPA (0.10 anemone/100 m 2 for A. frenatus; 0.42 anemone/100 m 2 for A. perideraion). The total individual numbers of both anemonefish species and the host sea anemones also showed the same pattern (Table 2).

| Variation in abundance per sea anemone between zones and between species
The total abundance level per sea anemone of both anemonefishes was significantly different between the two zones, while the adult abundance per anemone was not significantly different (Table 3). The total abundance per anemone of both anemonefishes was significantly lower in the MPA than in the two fishing zones (Figure 3). The total and adult abundance level per anemone was significantly different between the anemonefish species. The abundance of A. perideraion was lower than that of A. frenatus (Table 3 and Figure 3). The interaction term between zone and species was significant for juvenile abundance per sea anemone. The juvenile abundance per sea anemone of A. perideraion was significantly lower in the MPA than in the two fishing zones (p = .018), whereas that of A. frenatus was not significantly different between the two zones (p = .093). The estimates of all abundance levels (total, juvenile, and adult abundances) per anemone were lower in the MPA than in the two fishing zones for both species (Figure 3).

| Summary statistics of genetic analysis
The 14 and 15 markers were polymorphic for A. frenatus and A. perideraion, respectively. The average number of alleles per locus was 13.00 and 14.73, ranging from 3 to 32 and from 6 to 26, for A. frenatus and A. perideraion, respectively (Table S2)

| Parentage analysis
The parentage analysis revealed that 19 A. frenatus and 46 A. perideraion juveniles were assigned to breeders within the study area, indicating that the percentages of self-recruitment were 15.2% (19 self-recruits/125 total juveniles) for A. frenatus

| Variation in numbers of recruits per sea anemone and self-recruit production per female between zones and between species
The numbers of recruits (juveniles) from the fishing zones and from outside of the study area per anemone were significantly different between the two zones and between the two species (Table 4).
For both species, the number of recruits from the fishing zones and from outside of the study area was significantly lower at each anemone in the MPA than at each anemone in the fishing zones ( Figure 5). The number of recruit of A. perideraion was lower than that of A. frenatus. However, the number of recruits from the MPA per anemone was not different between the two zones or between the two species. In addition, the number of self-recruit productions per female was not different between the two zones or between the two species (Table 4).

| DISCUSSION
Although the physical barriers to genetic or demographic connectivity have often been examined in the sea, marine ecologists have rarely focused on biological barriers (Marshall et al., 2010). Hypothetically, the higher density of predators and lower density of habitat in MPAs T A B L E 3 Results of GLMMs testing the effect of zone (fishing zone = 0 or MPA = 1) and species (Amphiprion frenatus = 0 or A. perideraion = 1) on total, juvenile, and adult abundance levels per anemone, accounting for the effects of habitat size by offset term. A coefficient of the interaction term is shown only when it was significant (p < .05) F I G U R E 3 Effects of zone (MPA and fishing zones) and species (Amphiprion frenatus and A. perideraion) on the total, juvenile, and adult abundance levels per anemone. The abundance levels in figures indicate values estimated by the coefficients of explanatory variables (Table 3) when habitat sizes were fixed to their averages. Each figure represents estimates and 95% confidence intervals (error bars). An * indicates a significant difference (p < .05) between the MPA and the two fishing zones for each species than in fishing zones may cause lower numbers of recruitment of anemonefishes into the former. In the present study, we found all the following were lower in the MPA than they were in the fishing zones: the abundance per sea anemone of two anemonefishes and the number of recruits from the fishing zones and outside of the study site per sea anemone.
An important point to consider is whether these results were largely due to biological factors, including top-down control and habitat heterogeneity, or to other physical factors. One physical factor could be that ocean circulations around the study site may have resulted in lower number of recruits and thus a lower abundance per habitat for anemonefish in the MPA than in the fishing zones.
In this study, strong currents to the northeast were observed on the inner reef, but dispersal trajectories did not follow such oceanographic patterns. We also documented frequent reversals of current direction in the west and east on the outer reefs, as well as larvae dispersing in both directions in similar proportions. Therefore, although we cannot exclude the possibility that directions of larval movement followed such oceanographic circulations, it is unlikely that the current reversals caused lower settlement intensity of larvae into the MPA, which was located in the middle position of the two fishing zones.  (Connel, 1998;Holbrook & Schmitt, 2002Holmes & McCormick, 2006;A. Nanami, personal communication). It is possible that the higher density of these species in the MPA contributes to top-down control of recruits of anemonefish there. Elliott, Elliott, and Mariscal (1995) have reported high predation mortality during settlement for anemonefish. Buston (2003b) has found that the smallest individuals (lowest social rank) of anemonefish in a habitat tend to be evicted by larger ones (higher social rank) and to be subject to predation, even after settlement. Moreover, Dixson (2012) has found that settling juvenile anemonefish select lowpredation-risk habitat over high-risk habitat using olfactory cues, thus suggesting that larvae may avoid MPAs, where predators are abundant. Although it is rare to detect top-down control in coral reef MPAs (e.g., Babcock et al., 2010;Emslie et al., 2015), previous studies have reported a lower density of small coral reef fishes in no-take zones or MPAs than in fishing zones, mainly due to trophic interactions (e.g., Boaden & Kingsford, 2015;Graham, Evans, & Russ, 2003). These two studies also found no significant differences in the habitat structure (e.g., live coral cover and structural complexity) between the zones, and these conditions in the absence T A B L E 4 Results of GLMMs testing the effect of zone (fishing zone = 0 or MPA = 1) and species (Amphiprion frenatus = 0 or A. perideraion = 1) on the number of recruits from the fishing zones per anemone, those from the MPA per anemone, and those from outside of the study area per anemone, as well as the number of self-recruit productions per female. The effects of habitat and female sizes were accounted by the offset terms for the numbers of recruits and self-recruit production, respectively. Coefficients of the interaction term are not shown because they were not significant (p > .05) of confounding effects of habitat characteristics may enable the detection of top-down impacts on prey fish in an MPA. Unlike other damselfish species, anemonefish inhabit anemones. In addition, we used the anemonefish abundance at each anemone (abundance per anemone) rather than the anemonefish density per unit area to compare between the two zones, because the anemonefish density was strongly correlated with the anemone density. We think that our comparisons ruled out the confounding effects of anemone density by focusing on the anemonefish abundance per anemone.
Another potential mechanism is that the lower density of the host sea anemones in the MPA than in the fishing zones may have caused lower recruitment per anemone of anemonefish in the MPA. Some studies using field and laboratory experiments have found that settling juvenile anemonefish strongly prefer particular host species of anemone (Dixson, 2012;Dixson et al., 2008;Elliott et al.,1995). Therefore, a higher host anemone density in the fishing zones may attract more recruits of target anemonefishes and result in higher number of recruits per anemone than in the MPA. The anemones are probably present at a higher density in the fishing zones because of less live coral cover that competes with sea anemones for space. The low coral cover in the fishing zone is thought to be due to past destructive fishing at the study site. Overall, top-down control and/or a lower host anemone density in the MPA are probable factors that account for the observed differences in the abundance and recruitment patterns of two anemonefishes between the zones. Although our study could not determine which factor caused the different recruitment patterns between the two zones, we think that the dissimilarity in the marine environment between the two zones is a biological barrier to the connectivity of anemonefish. Anemonefish at low latitudes are known to reproduce year round (Buston & Elith, 2011;Ross, 1978). The spatial patterns of recruitment of anemonefish strongly determine the abundance patterns of residents (Elliott & Mariscal, 2001;Schmitt & Holbrook, 2000). Therefore, at our study site, the different abundance patterns of anemonefishes between the two zones may be attributed to persistent lower recruitment into the MPA throughout the year.
Although the number of recruits from the fishing zones and outside of the study site per anemone differed between the zones, those from the MPA per anemone and the number of self-recruit production per female did not differ between them. The effect of zone was not detected for the number of recruits from the MPA, probably because of the low statistical power for the small number of the recruits from the MPA. The negative effects of predators on reproductive success and egg survival are known to be crucial in some marine organisms (Nakaoka, 2000;Richardson et al., 2011) but do not appear to be the crucial for the self-recruitment production of anemonefish. Anemonefish are generally protected by sea anemones, and the males care for the eggs until hatching (Buston & Elith, 2011;Mariscal, 1970); therefore, the presence of predators may not strongly affect reproduction and/or egg survival. In addition, variation in the habitat density between the two zones was unlikely to have caused a difference in the reproductive success of the anemonefish. Therefore, top-down effects and/or a lower habitat density may be less influential before larval hatching for the demographic connectivity of anemonefish.
F I G U R E 5 Effect of zone (MPA and fishing zone) and species (Amphiprion frenatus and A. perideraion) on the number of recruits (juveniles) from the fishing zones per anemone, those from the MPA per anemone, those from the outside of study area per anemone, and the number of self-recruit production per female. The number of recruits and self-recruit production indicates values estimated by the coefficients of explanatory variables (Table 4) when the habitat sizes and female sizes were fixed to their averages, respectively. Each figure represents estimates and 95% confidence intervals (error bars). An * indicates a significant difference (p < .05) between the MPA and the two fishing zones for each species Our results indicate that the dispersal distance of self-recruits was <1.3 km for the target anemonefishes. Although the small sampling area along a 1.5 km stretch of reef may be the primary cause of the dispersal distance, this result is in agreement with previous studies that showed short dispersal distances for anemonefish (e.g., Buston, Jones, Planes, & Thorrold, 2012;Jones et al., 2005;Planes et al., 2009). Our study also found low self-recruitment rates for both anemonefishes (15.2% and 14.2%). Because our study site is situated in a continuous coral reef, adjacent coral reefs outside of the study area may provide a large number of larvae for the study site, resulting in a large proportion of immigrants among the recruits, as predicted by Pinsky, Palumbi, Andréfouët, and Purkis (2012).
Although the results of our study were based on a single location with one MPA in a single season, there can be biological barriers to larval dispersal between other MPAs and fishing zones, because environmental dissimilarity between MPAs and fishing zones is often observed (e.g., Babcock et al., 2010;Graham et al., 2003;McCook et al., 2010;Mumby et al., 2006Mumby et al., , 2007Shears & Babcock, 2002;White 1986). In addition, such a biological barrier may be present not only for anemonefish but also for other coral reef fishes. In tropical coastal areas, aquarium fishing activities significantly impact the local population of small coral reef fish (e.g., anemonefish, the Banggai cardinalfish, and mandarinfish; Shuman, Hodgson, & Ambrose, 2005;Vagelli, 2008;reviewed in Thornhill, 2012). If the protection of such fishes is particularly necessary, the establishment of a buffer zone, where fishing for only large predators is permitted and protection of their specific habitats is prioritized, could be an option for weakening biological barriers to recruitment of such species in MPAs because it frees them from both strong top-down and aquarium fishing pressure while enhancing their habitat quality. Networks of MPAs have been widely established to enhance connection among MPAs (e.g., Bonin et al., 2016;Harrison et al., 2012;Horigue, Aliño, White, & Pressey, 2012;Planes et al., 2009). Our results highlight the importance of biological barriers as a factor that affects connectivity, and this provides important knowledge to aid in the optimization of such networks.

CONFLICT OF INTEREST
None declared.

DATA ACCESSIBILITY
Raw field data and microsatellite genotypes are available in DRYAD