Microgeographic differentiation in the genetic structure of catch bowl coral, Isopora cf. palifera (Scleractinia; Acroporidae), in Kenting National Park, Taiwan Authors: Shashank Keshavmurthy, Wai-Ling Fong, Savanna WenHua Chou, Chia-Ming Hsu, Chaolun Allen Chen Published: 2026 ABSTRACT With coral reefs increasingly threatened by rapid environmental changes, understanding genetic diversity at microgeographic scale is critical for assessing their capacity to respond to local stress regimes. Theory for continuous populations predicts that brooding corals with restricted dispersal should exhibit fine-scale genetic structure and isolation-by-distance, yet such patterns remain poorly resolved in marginal and environmentally extreme reef ecosystems. Here, we investigated the genetic structure of the catch bowl coral, Isopora cf. palifera, across 11 sites within ~ 14 km in Kenting National Park (KNP), southern Taiwan, a reefscape characterized by strong small-scale environmental heterogeneity, including chronic thermal influence from a nuclear power plant and tidally driven upwelling. We genotyped 466 colonies (six microsatellite loci yielding 302 unique multilocus genotypes) and sequenced nuclear PaxC 46/47-intron from 322 colonies of I. cf. palifera. Microsatellite data revealed strong genetic structure (K = 2, K = 5): principal coordinate analyses identified four geographic groupings, and Bayesian clustering (STRUCTURE) supported two major clusters separating Nanwan (plus Tantzei Bay) from the remaining coastal sites, with one site (Shiaowan) showing admixture. The PaxC marker resolved ten haplotypes, with H1 widespread, H2 concentrated along Nanwan, and H3 dominant at thermally influenced sites near the nuclear power plant outfall. Overall, populations showed high site differentiation, significant isolation-by-distance, and high self-recruitment (68–92%), indicating limited effective dispersal. A temporal comparison (2000–2015) at Tantzei Bay indicated stable genetic structure through time despite repeated regional disturbances. Generalized estimating equation (GEE) models showed that site-level seawater temperature was positively associated with both host haplotype composition (GEE; coefficient = 0.0479, p < 0.001) and Symbiodiniaceae genera (GEE; coefficient = 0.0462, p < 0.001, symbiont data from a previous work in KNP), suggesting non-random host-symbiont-environment associations at microgeographic scale. Together, these results indicate that I. cf. palifera in KNP exhibits pronounced fine-scale genetic structure consistent with restricted dispersal and possible microgeographic adaptation of the holobiont to local thermal regimes. While such structuring may enhance local resilience by maintaining diverse, site-specific host-symbiont combinations, it also implies limited scope for rescue via gene flow if future warming pushes populations beyond their adapted tolerances. Our findings underscore the importance of accounting for microgeographic genetic structure and local adaptation when designing management and conservation strategies for reefscape such as those in KNP.Supplementary InformationThe online version contains supplementary material available at 10.1038/s41598-026-48835-2. Introduction Understanding temporal and spatial patterns in genetic diversity is fundamental to inferring population dynamics, ecological-evolutionary processes, and ecosystem function1,2. Genetic diversity provides raw material that allows populations to respond to environmental change, including anthropogenic climate change, via selection on standing variation and new mutations. This diversity is jointly shaped by selection, drift and gene flow across generations, and is further modulated by ecological context, such as demographic structure, environmental heterogeneity, and species interactions. Consequently, the spatial distribution of individuals and the spatial structure of genetic variation are tightly linked and jointly determine a population’s evolutionary potential and ecological resilience3,4. Classical population-genetic theory shows that limited dispersal can create genetic structure even in the absence of population boundaries. In continuous habitats, this often appears as isolation by distance (IBD), wherein individuals that are spatially closer tend to be more genetically similar than those farther apart, so relatedness declines with geographic distance5. The strength of this pattern is shaped by the balance between local density and offspring dispersal distance, and continuous-space models are generally more appropriate than discrete island/deme models when habitat is spatially connected5,6. Within this context, fine-scale spatial genetic structure (FSGS) describes the non-random clustering of genotypes within local areas, often at distances on the order of typical dispersal distances. In this study, we use the term microgeographic structure to denote genetic structuring occurring at scales comparable to, or smaller than, typical parent-offspring dispersal, where even modest differences in dispersal, demography, or local conditions can produce notable genetic patterns7. Theory and empirical work further suggest that adaptive divergence can occur at these microgeographic scales when environmental heterogeneity coincides with restricted dispersal, reducing the homogenizing effect of gene flow7,8. This is particularly relevant for sessile foundation species, where local stress tolerance strongly influences ecosystem functioning. For such species, fine-scale gene flow and genetic structure determine not only adaptive potential, but also the spatial scale over which population dynamics, disturbance impacts, and recovery are coupled7,8. Coral reefs are a prime example of vulnerable ecosystems where such considerations are critical. Reef-building corals are experiencing unprecedented loss and degradation due to marine heatwaves, ocean acidification, and local anthropogenic stressors9–15. These pressures drive population declines, large-scale reef degradation, and even possible local extirpation. Yet, corals, as holobionts comprising the cnidarian host, Symbiodiniaceae, and associated microbiota, can respond through physiological acclimation or genetic adaptation (selection of genotypes over many generations), and phenotypic plasticity (which can occur within an individual’s lifespan) in both host and symbionts (reviewed in16. Recent work has documented local adaptation and acclimation at relatively fine spatial scales in some coral species, as well as strong associations between thermal environments, host genotypes, and symbiont communities17–26. In spatially continuous populations at demographic equilibrium, IBD theory predicts broadly similar within-site diversity across landscapes, while allele and haplotype frequencies change gradually with distance, reflecting neighbourhood size and local demography rather than discrete boundaries. In corals, brooding species with short pelagic larval durations often show strong differentiation over hundreds of metres to a few kilometres while maintaining moderate within-reef diversity; broadcast spawners tend to be less structured at comparable scales. Across regional to basin scales, population-genetic and -omic studies have revealed heterogeneous connectivity, cryptic lineages, and species-specific barriers to gene flow in scleractinian corals (reviewed in27. Many taxa show genetic breaks among reefs separated by tens to thousands of kilometres (e.g., Acropora millepora, Porites lutea28, Galaxea fascicularis29, Seriatopora hystrix30, and Montastraea cavernosa31, often implicating environmental and oceanographic factors32. Other taxa (e.g., Pocillopora verrucosa) show weak or inconsistent structure at similar scales33,34(but see35 for the taxonomical revision of this species). By contrast, far fewer studies have tested fine-scale (ranging from a few kilometres to tens of kilometres) structure within continuous reefscapes27,36–39, despite evidence that dispersal distances for some corals can be on the order of meters to hundreds of meters per generation39,40. Examples of fine-scale differentiation and IBD include Acropora hyacinthus41 and Pocillopora acuta along Phuket Island reefs42. Recent within-reef studies also report microscale kin clustering, clonal mosaics, and restricted effective dispersal, highlighting FSGS as a determinant of recovery and local resilience43,44. Nevertheless, the prevalence and strength of IBD and microgeographic structure in reef-building corals remain poorly resolved, particularly for brooding species with limited dispersal and for populations inhabiting marginal or environmentally extreme habitats. Fine-scale spatial heterogeneity may reflect neutral IBD processes, microgeographic divergence (trait or genetic differences among individuals within a dispersal neighbourhood), or local adaptation (adaptation of a population within the neighbourhood)7. While local adaptation is often facilitated by limited connectivity, strong selection can generate adaptive differentiation even under ongoing gene flow45. In corals, microgeographic differentiation can arise from environmental heterogeneity, localized selection, limited larval dispersal, and symbiotic interactions46. Distinguishing among these mechanisms is essential for interpreting fine-scale genetic mosaics and for conservation planning. Environmental variability is a major driver of microgeographic differentiation on reefs. Fine-scale gradients in temperature, light, sedimentation, and nutrients can alter performance and bleaching thresholds; prior exposure to higher temperatures can increase tolerance in some populations47,48. Dispersal dynamics also matter; although corals release many larvae, effective dispersal can be restricted by currents, oceanographic barriers, reproductive strategy, and larval behaviour49–51, generating structure over surprisingly short distances (e.g., < 10 km)52–54. Symbioses add an additional layer: different Symbiodiniaceae confer distinct thermal tolerances, and host–symbiont associations can be environment-dependent, enabling holobiont-level responses23,25,55–58. High-latitude, marginal, or otherwise extreme reef environments often exhibit strong environmental gradients over short distances (e.g., temperature, flow, turbidity, nutrient regimes), as well as recurrent disturbances, potentially resulting in low genetic connectivity and diversity under these unique conditions52,59,60. Extreme environments may further produce “chaotic” patchiness in differentiation due to episodic recruitment, bottlenecks, and spatially variable selection27,45,61,62. In these systems, it is especially important to distinguish whether observed fine-scale heterogeneity reflects neutral processes under IBD, local adaptation within the dispersal neighbourhood, or interactions between host genetics, symbiont communities, and environmental variation. In marginal or extreme environments, repeated disturbance, reduced recruitment, and smaller effective population sizes can reduce diversity and increase differentiation relative to core reefs, though this is not universal. Accordingly, we expected Isopora cf. palifera to exhibit (i) significant fine-scale genetic structure and IBD across our study area, consistent with a brooding life history and short effective dispersal, but (ii) generally moderate within-site diversity, with potential reductions at sites experiencing more extreme or variable conditions and disturbance. This pattern likely reflects the nature of the study site: Kenting National Park (KNP), located on the Hengchun Peninsula in southern Taiwan. The park comprises three geographic sub-regions: the west coast, the east coast, and Nanwan, a semi-enclosed bay. The distance between the east and west coasts is about 14 km. In addition, a nuclear power plant (NPP), which began operating in 1985, discharges heated seawater into Nanwan, creating thermal stress on reefs adjacent to the heated-water outlet (OL). Moreover, reefs in KNP have been exposed to human disturbances, including overfishing, habitat destruction, and sewage discharge, driven by population growth and poor management15,63,64. Combined with the effects of climate change, these pressures have contributed to a decline in live coral cover in KNP15,64,65. More broadly, fine-scale thermal mosaics are increasingly recognized as drivers of coral performance and adaptation. Work in systems such as Ofu Lagoon (American Samoa) indicates that exposure to more extreme or variable thermal regimes can promote higher thermal tolerance, and that heritable variation in thermal tolerance is widespread and linked to local thermal history and heatwave exposure20,47,66.We therefore expected haplotype frequencies in I. cf. palifera to covary with local temperature regimes at microgeographic scales, with certain host haplotypes enriched at warmer or more thermally variable sites that have experienced stronger or more frequent thermal stress. Host genetics, Symbiodiniaceae communities, and the environment also often show non-random associations across macro- and microgeographic scales. Symbiodiniaceae composition can track fine-scale thermal and depth gradients, and joint host–symbiont analyses suggest that both host population structure and environmental history can shape symbiont community structure and host–symbiont co-structure. In some systems, host lineages are consistently associated with distinct dominant symbiont genera (e.g., Cladocopium spp. versus Durusdinium spp.) along gradients, consistent with selection on host–symbiont combinations. For I. cf. palifera in Kenting, prior work24,67–69 reports seasonal and interannual shifts between thermally sensitive Cladocopium sp. (C3) and more tolerant Durusdinium sp. (D1a) in response to local temperature variation. Building on this, we hypothesised that (i) host haplotype composition would be non-randomly associated with Symbiodiniaceae taxa, and (ii) host haplotype–symbiont combinations would be enriched at warmer or more thermally variable sites. Testing these expectations allows us to assess whether microgeographic genetic patterns primarily reflect neutral dispersal and demography or also carry a signature of holobiont-level responses to local thermal environments. In this study, we leveraged the environmentally heterogeneous and disturbed reef system of Kenting National Park (KNP) (southern Taiwan) to examine spatial and temporal population structure in the catch bowl coral I. cf. palifera. This brooding species67 is expected to have short effective dispersal distances, with fecundity and recruitment strongly influenced by local disturbance regimes68. Although I. cf. palifera shows flexibility in Symbiodiniaceae associations and can acclimate to temperature variation via shifts in symbiont communities and host physiology24,69–71, the fine-scale genetic structure of the host in this region remains poorly characterised72, and it is unknown whether host populations exhibit microgeographic structure and potential local adaptation comparable to that observed in their symbionts. We therefore investigated the temporal and spatial genetic structure of I. cf. palifera across multiple sites spanning ~ 14 km, a scale likely comparable to or smaller than the dispersal neighbourhood. Specifically, we (i) quantified FSGS and tested for IBD across this continuous reefscape, (ii) assessed temporal stability of genetic structure across sampling periods spanning major disturbance events, and (iii) evaluated whether observed microgeographic differentiation matches neutral IBD expectations or indicates additional influences of environmental heterogeneity and host–symbiont dynamics. Results Seawater temperature Monthly means of daily average temperature range at NPP-OL, affected by thermal effluent from the nuclear power plant, were 25.5 ± 0.9–30.6 ± 1.5 °C. The same at TZB (central Nanwan) was 23.3 ± 0.6–30.0 ± 1.2 °C, and LK was 24.5 ± 0.6–27.8 ± 1.0 °C, respectively. (Fig. 122–24;. The reef near NPP-OL not only showed the highest sea surface temperature (SST) (32.2 ± 1.1 °C on July 26, 2007; Hsu et al., 2012), but also the greatest daily fluctuation (max. △T = 7.7 °C, mean △T ± SD = 2.73 ± 1.23 °C). Daily SST fluctuation differed significantly across all sites, with the lowest recorded at WLT (0.92 ± 0.43 °C) (Tukey’s post-hoc test, p < 0.001). Fig. 1Isopora cf. palifera population structure across sites in 2015. Pie diagrams show the haplotype distribution (PaxC) at different locations. Horizontal bar plot at the bottom shows the results of STRUCTURE (microsatellites) at K = 2. Locations are represented by a thin line in the bar chart, with labels coloured according to sampling year (see Key). Orange line = 2000, Blue line = 2006, Light blue line = 2009 and Black line = 2015 respectively. The coloured segments in the structure plot indicate two clusters (cluster 1 in yellow – sites in the central KNP; cluster 2 in red- sites along west and east coasts). Blue to red gradient on the map represents seawater temperature differences based on data obtained from HOBO© pendant loggers installed at NPP-OL, TZB, SW, and SJW at Nanwan and LK along the east coast. Solid and dotted red lines are the flow direction of currents in Nanwan. Isopora cf. palifera population structure across sites in 2015. Pie diagrams show the haplotype distribution (PaxC) at different locations. Horizontal bar plot at the bottom shows the results of STRUCTURE (microsatellites) at K = 2. Locations are represented by a thin line in the bar chart, with labels coloured according to sampling year (see Key). Orange line = 2000, Blue line = 2006, Light blue line = 2009 and Black line = 2015 respectively. The coloured segments in the structure plot indicate two clusters (cluster 1 in yellow – sites in the central KNP; cluster 2 in red- sites along west and east coasts). Blue to red gradient on the map represents seawater temperature differences based on data obtained from HOBO© pendant loggers installed at NPP-OL, TZB, SW, and SJW at Nanwan and LK along the east coast. Solid and dotted red lines are the flow direction of currents in Nanwan. Multi-locus genotyping and testing of Hardy-Weinberg equilibrium All 466 samples were successfully amplified at the six microsatellite loci (Table S1). Genetic analysis yielded 302 unique multi-locus genotypes (MLGs), indicating 18% clonality (Table 1). Hardy-Weinberg equilibrium (HWE) tests showed that all six loci were largely in HWE, with only one of 66 population-locus pairwise comparisons (IP235 in STZ) showing significant deviation after Bonferroni correction, likely due to null alleles (MICROCHECKER results, Table S3). Significant linkage disequilibrium was found at three sites: TZB (IP5 and IP235; IP271 and IP297; IP291 and IP297), STZ (IP291 and IP297), and LK (IP181 and IP271; IP181 and IP297). Genetic diversity All six microsatellite loci were polymorphic except for IP5 in STZ, with variability ranging from two alleles (IP5) to 10 alleles (IP271) (Table S3). The average number of alleles per locus (Na) varied from 3.667 in SSJ to 5.167 in TZB. Since allele count is influenced by sample size, the average number of effective alleles per locus (Ne) was also calculated, ranging from 1.862 in STZ to 2.987 in SJW. Observed heterozygosity (Ho) ranged from 0.333 in STZ to 0.609 in SJW, while expected heterozygosity (He) ranged from 0.389 in STZ to 0.641 in SJW. The inbreeding coefficient (FIS) ranged from − 0.113 in LK to 0.159 in STZ. Except for SSJ and LK, all sites had positive FIS values, indicating inbreeding or undetected null alleles; however, only the FIS of STZ showed a significant departure from HWE, with homozygote excess after Bonferroni correction (Table S2). Population differentiation and genetic structure Significant population pairwise FST values (p < 0.001) were observed for all population pairs (Table 1). FST values ranged from 0.033 to 0.209, with the minimum value detected between MBT and NPP-OL and the maximum value detected between DBS and LK. Pairwise G’ST (Hedrick) values had a similar pattern to that of FST. Significant population pairwise G’ST values (p < 0.001) were observed for all population pairs (Table 5). G’ST values ranged from 0.060 to 0.671, with the minimum value detected between STZ and NPP-OL and the maximum value detected between DBS and LK. Both FST and G’ST values were quite high, as expected for coral species with a low dispersal ability. Among MBT, STZ, and NPP-OL, the three smallest FST and G’ST values were possibly due to short geographic distances among these three populations. Table 1Pairwise FST (below diagonal) and Hedrick’s G’ST (above diagonal) values (microsatellite) of population differentiation of Isopora cf. palifera populations in the Kenting National Park.SSJDBSMBTSTZNPP-OLTZBSWSJWLKJLSCFBSSJ0.3150.4220.4740.3960.3850.4100.4320.5160.3530.401DBS0.1100.4800.5220.5030.4570.4810.5310.6710.4580.616MBT0.1580.1630.0660.0650.2400.0980.3880.3900.3720.489STZ0.1900.1920.0340.0600.3590.2130.4170.4530.4940.553NPP-OL0.1540.1760.0330.0340.2640.1060.3510.3630.3610.448TZB0.1380.1470.0910.1440.1040.2120.3490.4600.3810.375SW0.1420.1500.0410.0880.0470.0760.2580.2610.2110.320SJW0.1270.1380.1190.1390.1130.1010.0760.4400.4090.283LK0.1770.2090.1420.1780.1380.1580.0910.1250.2900.281JLS0.1270.1440.1350.1900.1370.1310.0770.1150.1020.158CFB0.1270.1690.1570.1920.1500.1150.0980.0740.0880.056All pairwise FST comparisons were statistically significant (permutation testing, p < 0.001). Table 1Pairwise FST (below diagonal) and Hedrick’s G’ST (above diagonal) values (microsatellite) of population differentiation of Isopora cf. palifera populations in the Kenting National Park.SSJDBSMBTSTZNPP-OLTZBSWSJWLKJLSCFBSSJ0.3150.4220.4740.3960.3850.4100.4320.5160.3530.401DBS0.1100.4800.5220.5030.4570.4810.5310.6710.4580.616MBT0.1580.1630.0660.0650.2400.0980.3880.3900.3720.489STZ0.1900.1920.0340.0600.3590.2130.4170.4530.4940.553NPP-OL0.1540.1760.0330.0340.2640.1060.3510.3630.3610.448TZB0.1380.1470.0910.1440.1040.2120.3490.4600.3810.375SW0.1420.1500.0410.0880.0470.0760.2580.2610.2110.320SJW0.1270.1380.1190.1390.1130.1010.0760.4400.4090.283LK0.1770.2090.1420.1780.1380.1580.0910.1250.2900.281JLS0.1270.1440.1350.1900.1370.1310.0770.1150.1020.158CFB0.1270.1690.1570.1920.1500.1150.0980.0740.0880.056All pairwise FST comparisons were statistically significant (permutation testing, p < 0.001). Pairwise FST (below diagonal) and Hedrick’s G’ST (above diagonal) values (microsatellite) of population differentiation of Isopora cf. palifera populations in the Kenting National Park. All pairwise FST comparisons were statistically significant (permutation testing, p < 0.001). Principal Coordinate Analyses (PCoA) based on FST and G’ST values revealed four genetic clusters corresponding to geographic sub-regions of KNP. For the FST-based PCoA (Fig. 2A), the first coordinate (32% of variance) separated the western Nanwan populations (MBT, STZ, and NPP-OL) from others, with these three populations clustering closely. The clusters were grouped by first coordinate (32% of variance)- (1) TZB, SW, and SJW, (2) MBT, STZ, and NPP-OL and by the first coordinate (21.9% variance)- (3) LK, JLS, and CFB (4) SSJ and DBS; respectively. The G’ST-based PCoA (Fig. 2B) showed a similar clustering pattern. A Mantel test supported significant isolation-by-distance (R² = 0.1675; P < 0.05), with geographic distance explaining 16.8% of the genetic variation (Fig. 3). Fig. 2Principal Coordinate Analysis (PCoA) constructed based on FST matrix (A), and G’ST (Hedrick) matrix (B) calculated by GenAlEx. Patterns displayed for first two axes of variation; the first two axes explain 32.0% and 21.9% of total variance for FST, and 29.6% and 24.8% of total variance for G’st, respectively. Each point represents a site. Principal Coordinate Analysis (PCoA) constructed based on FST matrix (A), and G’ST (Hedrick) matrix (B) calculated by GenAlEx. Patterns displayed for first two axes of variation; the first two axes explain 32.0% and 21.9% of total variance for FST, and 29.6% and 24.8% of total variance for G’st, respectively. Each point represents a site. AMOVA indicated that 74.4% of allelic variation was within individuals, 5.7% among individuals within populations, and 19.9% among populations, with a global FST value of 0.199. Four sample groupings were tested, with the highest variation (11.33%, FCT = 0.113, P < 0.001) observed among four geographic clusters, supporting PCoA results (Table 2; Fig. 2A, B). Despite strong differentiation (FST and G’ST, Fig. 2A, B) and regional variation (AMOVA), STRUCTURE analysis suggested two genetic clusters - K = 2 (however see K = 5 which does appear to show a clear split based on five subregions; (1) west coast (SSJ and DBS), (2) west Nanwan Bay (MBT to OL), (3) centre Nanwan Bay (TZB), (4) east Nanwan Bay (SW and SJW; with some admixture), and (5) east coast (LK, JLS, and CFB). (See Supplementary File 2 for STRUCUTRE plots of other K-values and Delta K, Fig S2). Nanwan, MBT, STZ, NPP-OL, and TZB mainly aligned with cluster 1, while SSJ, DBS, SJW, LK, JLS, and CFB aligned with cluster 2. SW showed admixture from both clusters (Fig. 1). No evidence of recent bottleneck events was detected in the BOTTLENECK results (Table S2). Fig. 3Results from Mantel test for isolation-by-distance in Isopora cf. palifera from 11 sampling sites. Geographic distance explained 16.75% of the variation in linear FST values across all sampling sites (R2 = 0.1675, p = 0.013). Results from Mantel test for isolation-by-distance in Isopora cf. palifera from 11 sampling sites. Geographic distance explained 16.75% of the variation in linear FST values across all sampling sites (R2 = 0.1675, p = 0.013). Table 2Hierarchical AMOVA (microsatellite) of Isopora cf. palifera samples from Kenting National Park (KNP).ComparisondfSum of squaresVariance componentFixation indicesPercentage of variation No groupingsAmong populations10224.8040.38390FST = 0.1990219.90***Among individuals within populations291481.4950.10960FIS = 0.070945.68***Within individuals302433.5001.43543FIT = 0.2558474.42***Grouped by structure results (SSJ, DBS, SJW, LK, JLS, CFB) (MBT, STZ, NPP-OL TZB, SW)Among groups162.3850.14491FCT = 0.072657.26***Among populations9162.4190.30685FSC = 0.1658815.38***Within populations593914.9951.54299FST = 0.2264777.35***Grouped by 2 regions (SSJ, DBS, LK, JLS, CFB) (MBT, STZ, NPP-OL TZB, SW, SJW)Among groups153.9850.13068FCT = 0.065516.55***Among populations9170.8190.32119FSC = 0.1722916.10***Within populations593914.9951.54299FST = 0.2265177.35***Grouped by 3 regions (SSJ, DBS) (MBT, STZ, NPP-OL, TZB, SW, SJW) (LK, JLS, CFB)Among groups294.6330.22727FCT = 0.1118511.18***Among populations8130.1710.26172FSC = 0.1450212.88***Within populations593914.9951.54299FST = 0.2406575.94***Grouped by 4 regions (SSJ, DBS) (MBT, STZ, NPP-OL) (TZB, SW, SJW) (LK, JLS, CFB)Among groups3134.5840.22381FCT = 0.1133111.33***Among populations790.2200.20844FSC = 0.1190110.55***Within populations593914.9951.54299FST = 0.2188378.12***Two regions: Inside Nanwan and outside Nanwan.Three regions: Western Coast, Nanwan, and Eastern Coast.Four regions: Western Coast, Nanwan West, Nanwan East, and Eastern Coast.***p < 0.001. Table 2Hierarchical AMOVA (microsatellite) of Isopora cf. palifera samples from Kenting National Park (KNP).ComparisondfSum of squaresVariance componentFixation indicesPercentage of variation No groupingsAmong populations10224.8040.38390FST = 0.1990219.90***Among individuals within populations291481.4950.10960FIS = 0.070945.68***Within individuals302433.5001.43543FIT = 0.2558474.42***Grouped by structure results (SSJ, DBS, SJW, LK, JLS, CFB) (MBT, STZ, NPP-OL TZB, SW)Among groups162.3850.14491FCT = 0.072657.26***Among populations9162.4190.30685FSC = 0.1658815.38***Within populations593914.9951.54299FST = 0.2264777.35***Grouped by 2 regions (SSJ, DBS, LK, JLS, CFB) (MBT, STZ, NPP-OL TZB, SW, SJW)Among groups153.9850.13068FCT = 0.065516.55***Among populations9170.8190.32119FSC = 0.1722916.10***Within populations593914.9951.54299FST = 0.2265177.35***Grouped by 3 regions (SSJ, DBS) (MBT, STZ, NPP-OL, TZB, SW, SJW) (LK, JLS, CFB)Among groups294.6330.22727FCT = 0.1118511.18***Among populations8130.1710.26172FSC = 0.1450212.88***Within populations593914.9951.54299FST = 0.2406575.94***Grouped by 4 regions (SSJ, DBS) (MBT, STZ, NPP-OL) (TZB, SW, SJW) (LK, JLS, CFB)Among groups3134.5840.22381FCT = 0.1133111.33***Among populations790.2200.20844FSC = 0.1190110.55***Within populations593914.9951.54299FST = 0.2188378.12***Two regions: Inside Nanwan and outside Nanwan.Three regions: Western Coast, Nanwan, and Eastern Coast.Four regions: Western Coast, Nanwan West, Nanwan East, and Eastern Coast.***p < 0.001. Hierarchical AMOVA (microsatellite) of Isopora cf. palifera samples from Kenting National Park (KNP). Two regions: Inside Nanwan and outside Nanwan. Three regions: Western Coast, Nanwan, and Eastern Coast. Four regions: Western Coast, Nanwan West, Nanwan East, and Eastern Coast. ***p < 0.001. Inferred levels of self-replenishment and recent migration High self-replenishment levels (68–92%) were inferred across all 11 populations (Table S3), suggesting that each population is primarily sustained by its own recruits rather than by migration from other populations. Migration rates per generation between most population pairs were below 10%, except for NPP-OL to MBT (m = 25%), NPP-OL to STZ (m = 24%), NPP-OL to SW (m = 18%), and CFB to JLS (m = 18%). Temporal variation of population genetic structure Population genetic structure analysis across years indicated two primary genetic clusters (Fig. 1). In the STRUCTURE plot, individuals from the same site across years were grouped for clear temporal comparison. Most individuals at MBT, STZ, NPP-OL, and TZB were assigned to cluster 1 (yellow), while most at LK belonged to cluster 2 (red). SW and SJW showed admixture of both clusters (Fig. 1). Additionally, PCoA based on pairwise FST values showed that populations from different years clustered together, indicating no temporal variation in host genetic structure (Fig. S1). Population genetics of Isopora cf. palifera based on Pax-C46/47 intron DNA sequences A total of 322 Pax-C46/47 introns of I. cf. palifera were successfully amplified and sequenced from samples across 11 sites in KNP (Table 3A). Nucleotide diversity (π) per site ranged from 0.00083 (JLS) to 0.00223 (STZ) (Table 3A). Ten haplotypes (H1–H10) were identified, with haplotype diversity ranging from 0.26625 (DBS) to 0.61756 (MBT) (Table 3A, Fig. 1). Haplotype 1 was found at all sites and dominant at SSJ (west of KNP), CFB (east of KNP), and JLS and LK (south of KNP and east of Nanwan). Haplotype 2 appeared at sites along Nanwan and was absent at SSJ and CFB. At sites influenced by NPP-OL, haplotype 3 was dominant. In Nanwan, western sites were dominated by haplotypes 1 and 3, while eastern sites were dominated by haplotypes 1 and 2 (Fig. 1). FST values ranged from 0.00668 (p < 0.01) to 0.72205 (p < 0.0001), with the minimum value detected between STZ and NPP-OL and maximum value detected between DBD and CFB. Values near zero were found between TZB-SJW (−0.001, p < 0.01), SW-SJW (−0.00306, p < 0.01) towards central and east of Nanwan, and between LK-JLS (−0.00899, p < 0.01) along the east coast. The same was true for STZ and MBT (−0.00248, p < 0.01) towards the west of Nanwan (Table 3B). AMOVA indicated that 78.71% of the variation was within populations and 21.28% among populations, with a global FST value of 0.21287 (Table 4). Four sample groupings based on PaxC results were tested, with the highest variation and highest FCT (24.21%, FCT = 0.24129, P < 0.001) observed among four geographic clusters, indicating differentiation and regional variation (AMOVA). Grouping according to PaxC was as follows: Nanwan, MBT, STZ, NPP-OL were grouped together, while TZB, SW and SJW were grouped together. The other combined group along the east and west coast was SSJ, LK, CFB and JLS. Finally, DBS in the west coast did not fall into any group due to the composition of haplotypes. The pattern of genetic differentiation between microsatellite and PaxC46/47 intron markers was similar, with a significant (p = 0.0021) linear correlation (R2 =0.1642) of pairwise FST values between these two markers (Fig. S3). Effect of location and seawater temperature on coral hosts and their symbiont communities At a microgeographic scale within KNP, both coral host genetic composition and associated Symbiodiniaceae communities varied among sites and appeared to track spatial differences in seawater temperature (SST) (Symbiodiniaceae from previously published datasets24,70,71 Site-to-site SST variation was significantly associated with shifts in dominant host haplotypes and in the relative dominance of stress-tolerant Symbiodiniaceae types (GEE regression: host haplotype, coef = 0.0479, p < 0.001; Symbiodiniaceae, coef = 0.0462, p < 0.001). Table 3Sampling sites, sample size (n), number of haplotypes (nh), haplotype diversity (h ± SD), and nucleotide diversity (π ± SD) (A); and pairwise Fst based on Pax-C46/47 intron sequence data of Isopora cf. palifera populations in the Kenting reef (B).(A) Site n n h hπCFB24100DBS2740.266250.00141JLS1840.300000.00083LK2730.351500.00099MBT3240.617560.00202NPP-OL3130.512430.00151SJW2630.527900.00168SSJ3840.262810.00111STZ3350.555710.00223SW2930.455190.00124TZB3720.503520.00134Total32280.522580.00169(B)SSJDBSMBTSTZNPP-OLTZBSWSJWLKJLSCFB SSJ0 DBS0.60952***0 MBT0.0975***0.42313***0 STZ0.05701***0.43448***−0.00248*0 NPP-OL0.07984***0.43523***0.01386*0.00668*0 TZB0.3259***0.23537***0.199***0.19488***0.13584***0 SW0.18769***0.37888***0.12276***0.1044***0.04521**0.03365**0 SJW0.23612***0.27601***0.14146***0.1329***0.07702**−0.001*−0.00306*0 LK0.05301***0.55243***0.12368***0.07829***0.05565***0.2149***0.07685**0.12235***0 JLS0.05073**0.55324***0.10721***0.06153***0.03825**0.20462***0.06568**0.11484***−0.00899*0 CFB0.04859***0.72205***0.18907***0.11172***0.14051***0.40192***0.25843***0.30443***0.07944***0.07341***0* = <0.5, ** = < 0.001, *** = < 0.0001 Table 3Sampling sites, sample size (n), number of haplotypes (nh), haplotype diversity (h ± SD), and nucleotide diversity (π ± SD) (A); and pairwise Fst based on Pax-C46/47 intron sequence data of Isopora cf. palifera populations in the Kenting reef (B).(A) Site n n h hπCFB24100DBS2740.266250.00141JLS1840.300000.00083LK2730.351500.00099MBT3240.617560.00202NPP-OL3130.512430.00151SJW2630.527900.00168SSJ3840.262810.00111STZ3350.555710.00223SW2930.455190.00124TZB3720.503520.00134Total32280.522580.00169(B)SSJDBSMBTSTZNPP-OLTZBSWSJWLKJLSCFB SSJ0 DBS0.60952***0 MBT0.0975***0.42313***0 STZ0.05701***0.43448***−0.00248*0 NPP-OL0.07984***0.43523***0.01386*0.00668*0 TZB0.3259***0.23537***0.199***0.19488***0.13584***0 SW0.18769***0.37888***0.12276***0.1044***0.04521**0.03365**0 SJW0.23612***0.27601***0.14146***0.1329***0.07702**−0.001*−0.00306*0 LK0.05301***0.55243***0.12368***0.07829***0.05565***0.2149***0.07685**0.12235***0 JLS0.05073**0.55324***0.10721***0.06153***0.03825**0.20462***0.06568**0.11484***−0.00899*0 CFB0.04859***0.72205***0.18907***0.11172***0.14051***0.40192***0.25843***0.30443***0.07944***0.07341***0* = <0.5, ** = < 0.001, *** = < 0.0001 Sampling sites, sample size (n), number of haplotypes (nh), haplotype diversity (h ± SD), and nucleotide diversity (π ± SD) (A); and pairwise Fst based on Pax-C46/47 intron sequence data of Isopora cf. palifera populations in the Kenting reef (B). * = <0.5, ** = < 0.001, *** = < 0.0001 Table 4Hierarchical AMOVA (PaxC) of Isopora cf. palifera samples from Kenting National Park (KNP).ComparisondfSum of squaresVariance componentFixation indicesPercentage of variation No groupingsAmong populations1043.7370.06910FST = 0.2128721.28661***Within populations631161.2250.2555178.71339***Grouped by PaxC results (SSJ, LK, JLS, CFB) (MBT, STZ, NPP-OL) (TZB, SW, SJW) (DBS)Among groups339.2590.08259FCT = 0.2412924.21***Among populations within groups73.4780.00418Fsc = 0.016091.22086***Within populations631161.2250.25551FST = 0.2535074.64965***Grouped by 2 regions (SSJ, DBS, LK, JLS, CFB) (MBT, STZ, NPP-OL TZB, SW, SJW)Among groups11.514−0.01016FCT = 10.03175−3.17460***Among populations within groups941.2240.07455Fsc = 0.2258723.30411Within populations631161.2250.25551FST = 0.2013079.87049***Grouped by 3 regions (SSJ, DBS) (MBT, STZ, NPP-OL, TZB, SW, SJW) (LK, JLS, CFB)Among groups27.635−0.00320FCT = −0.00991−0.99085***Among populations within groups835.1030.07113Fsc = 0.2177621.99127Within populations631161.2250.25551FST = 0.2100078.99958***Grouped by 4 regions (SSJ, DBS) (MBT, STZ, NPP-OL) (TZB, SW, SJW) (LK, JLS, CFB)Among groups317.0810.01180FCT = 0.036123.61183***Among populations within groups725.6570.05942Fsc = 0.1886718.18584Within populations631161.2250.25551FST = 0.2179878.20233***Two regions: Inside Nanwan and outside Nanwan.Three regions: Western Coast, Nanwan, and Eastern Coast.Four regions: Western Coast, Nanwan West, Nanwan East, and Eastern Coast.***p < 0.001. Table 4Hierarchical AMOVA (PaxC) of Isopora cf. palifera samples from Kenting National Park (KNP).ComparisondfSum of squaresVariance componentFixation indicesPercentage of variation No groupingsAmong populations1043.7370.06910FST = 0.2128721.28661***Within populations631161.2250.2555178.71339***Grouped by PaxC results (SSJ, LK, JLS, CFB) (MBT, STZ, NPP-OL) (TZB, SW, SJW) (DBS)Among groups339.2590.08259FCT = 0.2412924.21***Among populations within groups73.4780.00418Fsc = 0.016091.22086***Within populations631161.2250.25551FST = 0.2535074.64965***Grouped by 2 regions (SSJ, DBS, LK, JLS, CFB) (MBT, STZ, NPP-OL TZB, SW, SJW)Among groups11.514−0.01016FCT = 10.03175−3.17460***Among populations within groups941.2240.07455Fsc = 0.2258723.30411Within populations631161.2250.25551FST = 0.2013079.87049***Grouped by 3 regions (SSJ, DBS) (MBT, STZ, NPP-OL, TZB, SW, SJW) (LK, JLS, CFB)Among groups27.635−0.00320FCT = −0.00991−0.99085***Among populations within groups835.1030.07113Fsc = 0.2177621.99127Within populations631161.2250.25551FST = 0.2100078.99958***Grouped by 4 regions (SSJ, DBS) (MBT, STZ, NPP-OL) (TZB, SW, SJW) (LK, JLS, CFB)Among groups317.0810.01180FCT = 0.036123.61183***Among populations within groups725.6570.05942Fsc = 0.1886718.18584Within populations631161.2250.25551FST = 0.2179878.20233***Two regions: Inside Nanwan and outside Nanwan.Three regions: Western Coast, Nanwan, and Eastern Coast.Four regions: Western Coast, Nanwan West, Nanwan East, and Eastern Coast.***p < 0.001. Hierarchical AMOVA (PaxC) of Isopora cf. palifera samples from Kenting National Park (KNP). Two regions: Inside Nanwan and outside Nanwan. Three regions: Western Coast, Nanwan, and Eastern Coast. Four regions: Western Coast, Nanwan West, Nanwan East, and Eastern Coast. ***p < 0.001. Discussion This study examined the spatial and temporal genetic structure of the catch bowl coral Isopora cf. palifera across 11 sites in Kenting National Park (KNP) at a microgeographic scale (~ 0.5–14 km). Consistent with expectations for isolation-by-distance (IBD) in populations with spatially restricted dispersal, we detected clear fine-scale genetic structure despite the small geographic extent. From 466 samples, we identified 302 unique multilocus genotypes, revealing substantial asexual reproduction in Isopora cf. palifera, while PaxC 46/47 intron sequences resolved 10 haplotypes. Both marker types showed significant population structure and broadly concordant patterns, including a significant correlation between pairwise FST estimates. Principal coordinate analyses (based on FST and GST) identified four geographic groupings: (1) western Nanwan (MBT, STZ, NPP-OL), (2) central-eastern Nanwan (TZB, SW, SJW), (3) southeastern coastal sites (LK, JLS, CFB), and (4) western coastal sites (SSJ, DBS). Bayesian clustering (STRUCTURE) further simplified this into two major genetic clusters (K = 2), largely separating Nanwan plus TZB from the remaining coastal sites, with SW showing admixture. On the other hand, results from K = 5 showed five clusters matching the geographic division of KNP: 1) west coast, 2) west Nanwan, (3) centre Nanwan, (4) east Nanwan (with some admixture), and (5) east coast, (Fig S2). However, the results for K = 5 require more careful analysis to better assess the presence of five clusters. Overall, populations exhibited strong among-site differentiation but relatively low within-site variation (5.75% within vs. 19.9% among populations), consistent with restricted dispersal interacting with environmental heterogeneity to generate pronounced structure over < 20 km. Local factors—including the nuclear power plant (NPP) thermal plume and oceanographic processes such as tidally driven upwelling and eddies—likely contribute to this pattern (see70,73. Genetic differentiation, isolation-by-distance, and self-replenishment Under a continuous-space framework, a brooding coral with short effective dispersal distances is expected to show strong IBD and microgeographic structure, while maintaining moderate within-site diversity unless demographic history or disturbances reduce effective population size. Our results broadly match these expectations but reveal notable site-specific deviations. First, genetic structure at TZB remained stable across time (2000, 2006, 2009, 2015) (Fig. 1), despite repeated regional disturbances (see15,70. Individuals from different years at TZB clustered together in both STRUCTURE and PCoA, indicating temporal stability in local demography and dispersal. However, this pattern should be interpreted cautiously. The sampled colonies ranged from approximately 10 cm to > 100 cm in diameter, indicating that multiple size classes were represented; however, colony size cannot be reliably translated into age in scleractinian corals because modular growth, partial mortality, fragmentation, and asexual propagation can decouple colony size from age74,75. More broadly, generation time in reef-building corals is difficult to define, but demographic studies suggest that it often spans several decades, commonly around 30–40 years76, although faster-growing branching taxa may have shorter generation times of approximately 5–15 years77. Species-specific estimates are not available for I. cf. palifera, but given its morphology, the 15-year sampling interval may represent only a limited number of generations, perhaps on the order of 2–3. If so, the absence of clear temporal differentiation at TZB is not unexpected, especially in a species with overlapping generations and restricted dispersal. In addition, although disturbances such as typhoons and bleaching can cause substantial mortality in KNP corals, our results suggest that these events did not produce a strong enough demographic shift at TZB to generate detectable temporal genetic change over the sampling period, possibly because of survival of local colonies, self-recruitment, or insufficient time for divergence to accumulate. This suggests that neighbourhood sizes at TZB may be sufficiently large to buffer stochastic turnover. In contrast, STZ exhibited the lowest genetic diversity (Table S2) and a significantly positive FIS, with inbreeding signals similar to those in MBT and NPP-OL in western Nanwan, implying smaller effective neighbourhood sizes and greater vulnerability to drift and demographic fluctuations. These sites may therefore face higher risk of genetic erosion and elevated extinction risk under continued environmental change78. Second, significant pairwise differentiation and a clear IBD signal at small spatial scales indicate limited gene flow among most sites, consistent with rapid larval settlement near parent colonies in a brooder. The reproductive mode of I. cf. palifera – brooding with short pelagic larval duration – might be contributing to reduced neighbourhood size and strengthened fine-scale structure (see38,39. Life-history traits can therefore influence genetic structure through demographic mechanisms79. Results also showed significant differentiation among populations (Tables 5 and 7), and isolation-by-distance on a small geographic scale (Fig. 3), again supporting restricted dispersal. Unlike many marine taxa with high connectivity, I. cf. palifera, likely has limited gene flow due to short larval dispersal and rapid settlement80. Our migration analysis reinforces this interpretation, showing that several sites (MBT, STZ, SW) appear to be demographically connected among populations81. Populations at MBT, STZ, and SW appear to be replenished by NPP-OL (a potential primary source), and JLS by CFB, with migration rates > 10% supporting their classification within the same management unit. Other populations showed lower dispersal rates (< 10%, Table 8), suggesting self-sustainability and demographic isolation. Low recruitment can further increase inbreeding in I. cf. palifera populations, making them vulnerable to disturbance-driven mortality and slow recovery. Importantly, the results did not show bottleneck signals (Table S4), suggesting that the observed structure is not merely a transient consequence of recent demographic collapses, but instead reflects longer-term processes of restricted-dispersal, drift, and spatially variable recruitment. The combination of strong microgeographic differentiation, modest within-site diversity, and a mix of self-replenishment with limited source–sink connections aligns with expectations for brooders in heterogeneous reefscapes (e.g., Porites astreoides showing lower allelic richness82. Microgeographic differentiation and local adaptation The pronounced differentiation of I. cf. palifera populations over such short distances aligns with broader evidence for microgeographic structure and adaptation in sessile marine organisms (see4,83,84. In KNP, physical oceanography and disturbance regimes vary markedly between the coasts within Nanwan. Moreover, Nanwan is influenced by tidally induced upwelling, distinct tidal dynamics, and transient cyclonic eddies, which modify temperature regimes and larval transport70,85–87(see also Fig. 1). In particular, the NPP-OL site and adjacent areas experience large and rapid temperature fluctuations due to upwelling and thermal discharge (see51, while western KNP is comparatively less affected70. Nanwan is also more frequently and directly impacted by typhoons than the east and west KNP coasts70,88, adding further heterogeneity in disturbance history. The two-cluster population structure aligns well with these physical features (Fig. 1). Eddy dynamics can retain larvae and reduce circulation boundaries, shrinking effective dispersal neighbourhoods and making geographic distance a weak proxy for connectivity. We hypothesize that larvae from MBT, STZ, NPP-OL, TZB, and SW may be entrained and recirculated within Nanwan eddies, while the cyclonic eddy on Nanwan’s western side acts as a partial barrier to exchange with other parts of KNP. Eddy swirling speeds (~ 50 cm/s) could still transport larvae among nearby sites within the eddy path, reinforcing within-region connectivity while limiting cross-boundary exchange. Such dynamics would help explain why oceanography and reproductive mode can outweigh simple distance effects in determining genetic differentiation and connectivity89–91. Habitat degradation and fragmentation at microgeographic scales can erode variation and increase divergence among populations92. Our inference contrasts with systems where small-scale differentiation is weak; for example93, reported no differentiation in Pocillopora damicornis at a patch reef in Kaneohe Bay, Hawaii, attributed to high clonality and inbreeding. In I. cf. palifera, restricted dispersal and environmental heterogeneity likely promote genetic patchiness45. Microclimate variability on reefs may further promote microgeographic adaptation under restricted gene flow7,46. Sites in KNP experience different disturbances15,70, affecting genetic diversity and connectivity. Reefs in KNP have experienced multiple bleaching events (1998, 2007, and 2010)15,70, with severe impacts at NPP-OL94. Such disturbances could contribute to differentiation over time95,96 although disturbance effects are not necessarily generalizable97. Since microsatellite marker are largely neutral, they are not expected to directly capture adaptive differentiation; however, the strong microgeographic structure they reveal sets the stage for selection to act differently across sites. Restricted dispersal and self-replenishment increase the potential for local adaptation by limiting homogenizing gene flow, although this same isolation can also make populations more vulnerable to extreme events that exceed local adaptive capacity. Evidence from other KNP coral species – reciprocal transplant (e.g., Platygyra verweyi98, Leptoria phyrgia99 and common-garden experiments demonstrating site- and season-specific differences in growth, thermal tolerance, and Symbiodiniaceae communities100–102– supports the idea that local adaptation can evolve at similar spatial scales in this reef system. Local adaptation to perturbations has also been reported elsewhere. For example, Porites lobata populations inhabiting polluted nearshore sites were strongly differentiated from those in nearby, less polluted offshore sites, and reciprocal transplant experiments showed local adaptation103. Our findings suggest that I. cf. palifera may be following a similar trajectory, although targeted experiments would be required to confirm adaptive divergence among sites. Coral taxa such as I. cf. palifera may exhibit microgeographic spatial differentiation due to their environmental conditions. While fine-scale environmental adaptation is recognized in coral reef organisms, the mechanisms remain unclear. Some species may offset high selection costs by producing large offspring or by exhibiting settlement behaviours that favour microgeographic adaptation, potentially reducing the role of post-settlement filtering. Alternatively, weak effective gene flow among fine-scale habitats may indicate strong environmental filtering, as evidenced by within-cohort allele frequency shifts at adaptive loci. Studies on Antarctic sea urchins104 and the coral Acropora hyacinthus92,105 have shown genetic differentiation at small spatial scales, indicative of chaotic genetic patchiness (see106. Haplotype structure, temperature regimes, and Symbiodiniaceae associations We hypothesised that host haplotype composition would covary with local temperature regimes and Symbiodiniaceae communities at microgeographic scales. Under neutrality within an IBD framework, haplotype frequencies should shift gradually with distance; however, thermally extreme or highly variable sites might show distinct haplotype compositions if host genotypes differ in tolerance. PaxC haplotypes were strongly structured spatially and broadly aligned with known environmental gradients. Sites with more extreme or variable regimes—such as NPP-OL and nearby sites (STZ, MBT) influenced by upwelling and thermal discharge, and sites such as TZB and SW affected by freshwater runoff and elevated temperatures—showed distinct haplotype profiles relative to more moderate sites. While these data do not establish causality, the non-random association between haplotype composition and site-specific thermal conditions is consistent with selection shaping haplotype distributions, especially given repeated thermal disturbances in KNP. Host–symbiont patterns further support this interpretation. Prior work in KNP indicates that I. cf. palifera at NPP-OL and nearby sites is predominantly associated with stress-tolerant Durusdinium spp., whereas colonies at cooler or less variable sites more frequently host Cladocopium spp.; Durusdinium dominance is also reported at TZB and SW where temperatures are elevated by freshwater runoff24,70. These host–symbiont–environment associations suggest that particular host haplotypes and symbiont taxa may form holobiont configurations suited to specific local conditions. Previous studies have suggested long-term acclimatization via symbiont shuffling in I. cf. palifera24,70, and broader Symbiodiniaceae differentiation among multiple coral genera across KNP70, consistent with persistent environmental sorting. Our results, therefore, support the view that microgeographic differentiation in I. cf. palifera may not be purely neutral but partly shaped by selection on holobiont traits under contrasting thermal regimes. However, this inference is limited by the use of a single nuclear locus (PaxC) and by reliance on previously published Symbiodiniaceae data rather than concurrent sampling. Future studies should integrate genome-wide host markers with contemporaneous symbiont profiling and fine-scale environmental measurements to partition drift, dispersal, and selection more rigorously. AMOVA results from both microsatellites and PaxC supported geographic division within KNP, though the groupings differed (microsatellite grouping based on107, PaxC grouping influenced by DBS due to distinct haplotype composition relative to other west/east coast sites). The significant correlation between microsatellite and PaxC FST values (Fig. S3) suggests a shared underlying demographic or evolutionary process shaping differentiation among sites, likely linked to the environmental setting. Microgeographic differentiation, structured by location-specific combinations of common and unique host haplotypes and Symbiodiniaceae, likely reflects interactions among biological and oceanographic drivers, including thermal variability in Nanwan. Both host and symbiont may also respond via trait shifts and epigenetic mechanisms under microgeographic environmental contrasts (see25. Figure 4 summarizes haplotype distributions and Symbiodiniaceae ratios in 2015 across the 11 sites (Symbiodiniaceae data from24,70,71. Fig. 4Heat map of haplotype distribution and ratio of Symbiodiniaceae in 2015 across 11 sites at KNP (A) and percentage of Cladocopium spp. and Durusdinium spp. from the corals sampled from the 11 sites (B). Notably. these findings cannot be generalized across KNP corals because species differ in life-history traits and population dynamics. For example, P. verweyi in KNP showed weak host differentiation across similar sites, while Symbiodiniaceae composition exhibited patterns resembling those observed here22. Heat map of haplotype distribution and ratio of Symbiodiniaceae in 2015 across 11 sites at KNP (A) and percentage of Cladocopium spp. and Durusdinium spp. from the corals sampled from the 11 sites (B). Notably. these findings cannot be generalized across KNP corals because species differ in life-history traits and population dynamics. For example, P. verweyi in KNP showed weak host differentiation across similar sites, while Symbiodiniaceae composition exhibited patterns resembling those observed here22. Microgeographic structure, resilience, and vulnerability under climate change The microgeographic differentiation documented here has dual implications for persistence under climate change. On one hand, spatially structured diversity across host genotypes and Symbiodiniaceae communities may preserve adaptive potential. Populations persisting in chronically warm or highly variable environments (e.g., near the NPP or upwelling-affected Nanwan) may harbour stress-tolerant holobiont combinations, potentially acting as sources of tolerant propagules if connectivity permits. On the other hand, strong structure and self-replenishment imply that many populations have limited capacity to import adaptive variation once conditions exceed local tolerance. With low effective dispersal and small neighbourhood sizes, recovery and recolonization may be slow, particularly if disturbances affect multiple sites simultaneously. From a management perspective, I. cf. palifera in KNP should not be treated as a single homogeneous unit. Eastern and western KNP, and distinct sectors within Nanwan, likely represent partially independent demographic and evolutionary units, some of which are potentially locally adapted to specific regimes. Conservation and restoration should therefore incorporate fine-scale connectivity and local environmental matching—for example, recognizing source populations (e.g., NPP-OL, CFB) that contribute disproportionately to replenishment, and prioritizing protection of sites harbouring unique haplotypes or holobiont combinations associated with extreme or variable conditions. Finally, while our focus here is on I. cf. palifera, the contrast with other KNP taxa such as P. verweyi, which shows weak host differentiation but similar Symbiodiniaceae patterns, underscores that microgeographic structure and local adaptation are strongly species-specific, depending on life-history traits, dispersal, and interactions with symbionts. Broadly, integrating fine-scale population genetics, host-symbiont eco-physiology, and environmental variability will be essential for understanding and managing coral populations in marginal and rapidly changing reef systems. Implications for persistence under climate change Microgeographic differentiation observed in I. cf. palifera might aid local adaptation within populations46 and maintain reservoirs of genetic diversity that enhance resilience. Persistence near the nuclear power plant—a chronically warm environment—may mirror “natural laboratories” such as the Red Sea108 and American Samoa back reefs36, where corals persist under elevated thermal stress. In addition, GEE regression analyses combining previous Symbiodiniaceae datasets with this coral in KNP indicate that both host and symbiont are influenced by SST variation among sites. Although our data do not include all perturbations, environmental differences among sites likely contribute to the microgeographic differentiation observed in both host and Symbiodiniaceae, a pattern not commonly reported in corals. However, microgeographic differentiation can also increase vulnerability when disturbances exceed localized adaptive capacity. At NPP-OL and nearby sites, I. cf. palifera may already be near its thermal tolerance limit, increasing risk under continued climate change (see99), and additional warming and marine heatwaves could cause local extirpations. If temperature anomalies intensify, current signals of local adaptation via microgeographic differentiation may not be sufficient to ensure persistence. Understanding the resilience of KNP corals, including I. cf. palifera, remains essential for guiding management and conservation actions under accelerating climate change. Methods Field study permissions All the samples used in the experiment were collected through permission granted by the Kenting National Park headquarters (permit numbers 1040008112 and 1040002080) and the Taiwan Power Company for allowing us to work at the Nuclear Power Plant in Kenting. Study sites and seawater temperature monitoring Kenting National Park (KNP) encompasses a well-developed fringing reef in southern Taiwan (Fig. 1). The environmental settings vary along Nanwan in the centre and along the west and east coasts of the Hengchun Peninsula (reviewed in70 (Fig. 1). In addition, the nuclear power plant (NPP) in Nanwan has been discharging warm water from its cooling system continuously since 1984 (2.0–3.0 °C higher than the ambient temperature), resulting in chronic disturbances to the reef adjacent to its outlet (OL)16,22–24,63,70,98. The nuclear power plant (NPP) is located on the west side of Nanwan, a semi-enclosed bay facing the Bashi Channel that spans 14 km. KNP can be divided into four main geographic sub-regions based on environmental settings as well as coral communities, comprised of the west coast of Hengchun Peninsula (WHP) facing the Taiwan Strait, east coast of Hengchun Peninsula (EHP) adjacent to the Pacific Ocean, Maobitou (MBT) to the west of Nanwan, and Eulanbi (ELB) to the east of Nanwan, facing the Bashi Channel that connects the Pacific Ocean and Taiwan Strait107. Fringing reefs with over 300 described reef-building coral species have developed along the coast of the peninsula, including at Nanwan109. These reefs are influenced by a branch of the Kuroshio Current, the monsoon-induced South China Sea Surface Current70, and a tide-induced upwelling and eddies73,85,87 that not only shape coral communities but also reduce seawater temperature and create fluctuating thermal environments. Over time, these have favored thermal-resistant corals, particularly those close to the thermal effluent of a nuclear power plant outlet (NPP-OL) west of Nanwan that was built in 1979 and began operation in 1984 (reviewed in70. Eleven study sites were selected along the KNP; Siashuijue (SSJ), Dingbaisha (DBS) – West coast of KNP, Maobitou (MBT), Siatanzai (STZ), Nuclear Power Plant Outlet (NPP-OL), Tanziwan (TZW), Shiaowan (SW), Siangjiaowan (SJW) – in Nanwan and Longkeng (LK), Jialeshuei (JLS), and Chufengbi (CFB) – East coast of KNP (Fig. 1; Table 5). To visualize spatial sea surface temperature (SST) variation at Nanwan, data from Hobo© pendant data loggers (Onset Corporation) deposited at the 3 m depth at five sites —NPP-OL, TZB, SW, and SJW at Nanwan and LK along the east coast — to monitor SST variability at the sites along KNP were used. From 2006 to 2009 and 2015–2016, seawater temperatures, were monitored at 60-min intervals using in situ temperature loggers (Hobo Pendant, Onset Corp., Cape Cod, Massachusetts, USA, with a precision of ± 0.5 °C). Data from71 were used to characterize temporal SSTs in 2000–2001 for TZB. Daily SST variability (△T = daily maximum-minimum) was calculated for every site and compared using one-way ANOVA followed by Tukey’s post-hoc test. The temperature data was overlaid onto the map of KNP using QGIS software (QGIS.org, 2022). Coral sampling The encrusting growth form of Isopora cf. palifera, which occurred from the lower intertidal zone to the depth of 3 m, was collected in July and August 2015 at 11 sites along the coastline of Kenting National Park (KNP) in Southern Taiwan (Fig. 1). Information on collecting sites, abbreviations, sub-regions, and sample numbers are listed in Table 5. Isopora cf. palifera is usually described as colonies having thick branches with multiple axial corallites opening along the tips of the branch, and sometimes colonies are encrusting on the reef edge or submerged reefs with strong currents110. These two growth forms of I. cf. palifera are likely representing distinct lineages in our study sites with unique environmental settings, as recent studies have shown that the diversity of Acropora spp., the con-familial genus of Isopora, is higher than currently recognized and widespread nominal species may represent multiple distinct species111. Therefore, we employ provisional qualifiers (cf. = confer) in species names throughout this paper to address this uncertainty, as this approach reflects tentative identifications while awaiting further taxonomic confirmation. A total of 368 samples of Isopora cf. palifera were randomly collected from individual colonies in 2015 using a hammer and chisel at least one meter apart to minimize the likelihood of sampling clones. Coral fragments were then preserved in 95% ethanol for laboratory experiments. In addition, samples from 81 colonies in 2009 from 7 of the 11 sites: MBT, STZ, NPP-OL, TZB, SW, SJW, and LK and samples from 5 to 12 colonies collected from TZB in 2000 and 2006, respectively (Table 1), and included in the analysis. Table 5Site information, including name, abbreviation, sub-region, and number of samples collected at 11 sites in 2000, 2006, 2009, and 2015.SiteAbbreviationSub-regionNumber of samples2000200620092015SiashuijueSSJWestern coast39DingbaishaDBSWestern coast36MaobitouMBTNanwan West1038SiatanzaiSTZNanwan West1234Nuclear power plant outletNPP-OLNanwan West1239Tantzei BayTZBNanwan East5121347ShiaowanSWNanwan East1130ShinjaowanSJWNanwan East1329LongkenLKEastern coast1029JialeshuiJLSEastern coast18ChufengbiCFBEastern coast29Total (each year)51281368 Table 5Site information, including name, abbreviation, sub-region, and number of samples collected at 11 sites in 2000, 2006, 2009, and 2015.SiteAbbreviationSub-regionNumber of samples2000200620092015SiashuijueSSJWestern coast39DingbaishaDBSWestern coast36MaobitouMBTNanwan West1038SiatanzaiSTZNanwan West1234Nuclear power plant outletNPP-OLNanwan West1239Tantzei BayTZBNanwan East5121347ShiaowanSWNanwan East1130ShinjaowanSJWNanwan East1329LongkenLKEastern coast1029JialeshuiJLSEastern coast18ChufengbiCFBEastern coast29Total (each year)51281368 Site information, including name, abbreviation, sub-region, and number of samples collected at 11 sites in 2000, 2006, 2009, and 2015. Genomic DNA extraction Genomic DNA was extracted using a high-salt method modified by112. About 30 mg of coral tissue was cut using sterilized scissors and homogenized in 200 µL lysis buffer (0.25 M Tris-Boric pH 8, 0.05 M EDTA pH 8, 2% sodium dodecylsulfate (SDS), and 0.1 M NaCl). Tissues were lysed with 10 µL of 10 mg/mL proteinase E overnight in a 55˚C water bath, after which 210 µL of 7 M NaCl (NaCl saturated H2O) was added to each sample. Samples were vortexed for 30 s at maximum speed and centrifuged at 12,000 rpm for 30 min at room temperature. The supernatant was transferred in fresh tubes. An equal volume of 100% isopropanol (stored at −20˚C) was added to each sample and mixed gently for 5 min. Samples were incubated at −20˚C for 2 h. After that, samples were centrifuged at 15,000 rpm for 30 min at room temperature. The pellet was washed with 70% ethanol three times, dried, and re-suspended in 50 to 100 µL TE buffer 1X. Microsatellite genotyping and population genetics analysis A total of 466 I. cf. palifera DNA samples were genotyped using six previously published microsatellite loci72. PCR was performed and followed with electrophoresis on Gel-Scan 3000 (Corbett Robotics, Australia) using the same protocol described for the microsatellite primer validation in72. Identical multi-locus genotypes (MLGs) from the same population were removed from the dataset using the R-package poppr113, leaving only unique MLGs for subsequent analyses. ARLEQUIN v.3.5114 software was used to determine if populations were in Hardy-Weinberg equilibrium (HWE) (1,000,000 steps in Markov chain and 100,000 dememorization steps) and loci within the populations are in linkage disequilibrium (LD) (10000 permutations). P-values for multiple testing were adjusted using Bonferroni correction115. The presence of allele dropout and null alleles were detected using MICROCHECKER v.2.2.3116. GenAlEx v.6.502117 was used to calculate the following indices of population genetic diversity: the number of alleles per locus (Na), effective number of alleles (Ne), number of private alleles, observed heterozygosities (HO), and expected heterozygosities (HE). Inbreeding coefficient (FIS) was calculated using ARLEQUIN v.3.5, with Bonferroni correction for P-values. Population-level pairwise FST118 and G’ST119 comparisons were performed using GenAlEx v.6.502, and statistical significance was evaluated by permutation testing (9,999 permutations). Pairwise FST and G’ST were applied to generate Principal Coordinate Analyses (PCoA) plots to visualize population differentiation. A Mantel test was performed using the isolation-by-distance model in GenAlEx v.6.502 to test correlations between genetic and geographic matrices among populations. Statistical significance was evaluated by permutation testing (9,999 permutations). Geographic distances were measured in Google Earth Pro v.7.1.5 using set points, and the shortest marine geographic distances between points were measured. Analysis of molecular variance (AMOVA) was performed by ARLEQUIN v.3.5 to examine the extent of population structuring using four sets of hierarchical AMOVA. There were four hypotheses of how samples in different locations group together. First, samples in different locations were divided into two groups: inside Nanwan Bay and outside Nanwan Bay. Second, samples were divided into three groups: Eastern Coast, Western Coast, and inside Nanwan Bay. Third, samples were divided into four groups: Eastern Coast, Western Coast, East Nanwan Bay, and West Nanwan Bay. Finally, another AMOVA grouped locations according to the clusters identified by STRUCTURE (see after). The population genetic structure of I. cf. palifera samples collected in 2015 was estimated using a Bayesian clustering approach implemented in STRUCTURE v.2.3.4120. Admixture model and allele frequency correlation were used. Values of the number of genetic clusters (K) from 1 to 10 were tested by running three replicate simulations per K with 1,000,000 Markov chain Monte Carlo repetitions and 100,000 burn-in iterations. The most likely value of K was determined using the method of121 implemented in the web-based STRUCTURE HARVESTER (Earl 2012). Ten replicate simulations were run using the most likely value of K determined, the results of the ten runs were merged using Full Search method in CLUMMP v.1.1.2122, and then visualized using DISTRUCT v.1.1123. Moreover, I. cf. palifera samples collected in different years (2000, 2006, 2009, and 2015) were all assembled for population genetic structure analysis following the abovementioned procedures. Recent demographic decline or expansion was determined using BOTTLENECK v.1.1.02124. For populations that have experienced recent reductions in effective population size, allelic diversity was reduced faster than heterozygosity. This program compares the observed heterozygosity with the heterozygosity expected from the observed allele number. A heterozygosity excess indicates a recent bottleneck event. The Wilcoxon’s test with 10,000 iterations was used to test for significant deviations from population mutation-drift equilibrium, under a two-phase model (TPM) with 95% single-step mutations and TPM variance of 12. To determine the rate of contemporary migration over recent generations, a Bayesian method implemented in BayesAss 3.0125 was used. The software was run for 10,000,000 MCMC iterations, with 1,000,000 iterations for burn-in for homogeneity with the rest. We ran two identical maximum likelihood analyses with different starting seeds to assess whether the chains had sampled the parameter space properly. Additional parameters for final maximum likelihood runs included 20 short chains in increments of 20 and a sampling of 500; five long chains in increments of 20 and a sampling of 1,000 with a burn-in of 100; and the default heating scheme. Bayesian analyses were run with 5 long-chains, in increments of 20 and a sampling of 100,000, with a burn-in of 20,000. Additional Bayesian parameters included an update frequency of 0.4, a Metropolis-Hastings sampling algorithm for both θ and M, uniform priors for θ of 0 to 4, which seemed reasonable according to the mutation rate of these loci, and the effective population size determined by126, assuming a uniform prior distribution of M from 0 to 8. Population genetic analysis using PaxC-46/47 intron The PaxC-46/47 intron DNA sequence was amplified according to127. Briefly, 50–200 ng of template DNA was amplified in a 50 µL reaction using 1X PCR buffer, 1.5 mM MgCl2, 0.2 mM dNTPs, 0.08 µM primers, and 0.2 µL Taq polymerase (Invitrogen, USA). PCR amplification cycles consisted of 2 min at 95 °C, followed by 35 cycles at 94 °C for 45 s, 55 °C for 45 s, and 72 °C for 1.5 min, with a final extension at 72 °C for 5 min. PCR products were purified using PCR Clean Up-M (Viogene, Taiwan), and analysed in both directions using an ABI 3730 XL DNA Sequencer (PE/ABI, USA). All haplotype sequences obtained in this study were submitted to GenBank and can be accessed through the GenBank database with accession numbers PZ232838 - PZ233158. Raw sequences were inspected, trimmed, and assembled with Geneious version R10 (Biomatters, New Zealand). The cleaned consensus sequence of each sample was imported to MEGA X128 for sequence alignment using the built-in MUSCLE algorithm (371 bp). A 371 bp PaxC-46/47 intron sequence was used for further analyses. Haplotypes were reconstructed from the sequenced genotypes using Phase runs of 10,000 iterations in DnaSP v.6.12129. Genetic diversity was evaluated for the number of haplotypes (nh), haplotype diversity (h), and nucleotide diversity (π) for each sampling site using DnaSP v.6.12. Pairwise FST comparisons were calculated between sampling sites, with statistical significance assessed by performing 10,000 permutations using Arlequin v.3.5.2130. The statistical significance of FST estimates was subsequently adjusted by Benjamini and Hochberg’s FDR correction131. AMOVA was performed to examine the population differentiation of Isopora cf. palifera based on the PaxC-46/47 intron using the same grouping hypotheses derived from microsatellite markers. To reveal the trend of genetic differentiation between microsatellites and the PaxC-46/47 intron, FST values derived from both markers were analysed by performing linear regression of correlation. Effect of location and seawater temperature on coral hosts and their symbiont communities Generalized Estimating Equations (GEE) were performed using haplotype and Symbiodiniaceae genera data (for the samples obtained in 2015) to see if there was any effect of location and seawater temperature. Symbiodiniaceae community data were sourced from previously published work24,70,71 corresponding to the year 2015 (same as the coral samples used in this study) and were not newly generated here. The novelty of the present study lies in integrating these published Symbiodiniaceae datasets with host haplotype composition and site-level SST variation within a unified statistical framework to test microgeographic associations. Both host haplotype and Symbiodiniaceae genera (Cladocopium sp. or Durusdinium sp.) were converted into numbers to make the analysis easier (see Supplementary file 1). Only five sites (NPP-OL, TZB, SW, SJW, and LK) were included in the analysis because there were no continuous seawater temperature data for other locations. The analysis used the mean seawater temperature from the summer of 2015 (July-September). All the statistical analysis has been carried out by Python (3.13.0, Python Software Foundation, 2016) and statsmodels package132 for analysing GEE or using GraphPad Prism (v.10.4.1 (532) for MacOS; GraphPad Software, San Diego, California USA, www.graphpad.com). The code and GEE results output are provided in Supplementary file 2. Supplementary Information Below is the link to the electronic supplementary material. Supplementary Material 1 Supplementary Material 1 Supplementary Material 2 Supplementary Material 2