127urn:lsid:arphahub.com:pub:15f175fc-9690-5d13-a2ea-114d8a2e74bdFrontiers of Biogeographybiogeography1948-6596Pensoft Publishers10.21425/fob.19.141935141935Research ArticlePiscesCommunity ecologyIsland biogeographyIsland ecologyRepublic of PalauCommunity diversity of marine fishes in complete and habitat islandsCarlsonBailey J.bcarlson4@ucmerced.eduhttps://orcid.org/0000-0001-8124-205X1MyersRobert F.2DawsonMichael Nhttps://orcid.org/0000-0001-7927-83951Quantitative & Systems Biology, School of Natural Sciences, University of California, Merced, 5200 North Lake Road, Merced, CA 95343, USASchool of Natural Sciences, University of California, MercedMercedUnited States of Americahttps://ror.org/00d9ah1059273 Silent Oak Circle, Wellington, FL 33411, USAUnaffiliatedWellingtonUnited States of America
Corresponding author: Bailey J. Carlson (bcarlson4@ucmerced.edu)
Editors Tom Matthews
20261008202619e141935545C1B81-569E-5272-AF29-0E42EC937A5D0912202428052026Bailey J. Carlson, Robert F. Myers, Michael N DawsonThis is an open access article distributed under the terms of the Creative Commons Attribution License (CC BY 4.0), which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.
Insular aquatic habitats have the potential to expand upon current theory by providing alterna/tive perspectives on island biogeographical processes. Here, we investigate whether marine lakes—seawater surrounded by land—show patterns expected for ‘complete’ (e.g., ‘true’ oceanic islands) or ‘habitat’ islands using species, phylogenetic, and functional diversity of fishes. We surveyed 22 locations across three location types: stratified lakes and mixed lakes (i.e., islands) and ocean (i.e., mainland). Stratified lake communities were depauperate in species, and distinct in the proportion of traits related to colonization, environmental persistence, and niche competition; they are akin to ‘complete’ islands. Mixed lake communities had higher species richness and redundancy in traits when compared to stratified lake communities, and substantial turnover in beta diversity of species, lineages, and traits when compared to the mainland ocean communities; they are akin to ‘habitat’ islands. While two categories of marine lakes—stratified and mixed—were defined a priori based on physical structure and roughly correspond to ‘complete’ and ‘habitat’ islands, communities indicate variation within, and possibly a gradation between location types, including overlapping confidence intervals and outlier cases along primary axes of variation. Such gradation in marine lakes may be attributable to the youth and small size of these marine islands (<20 ky and < 0.1 km2) relative to the ‘complete’ islands common to island biogeography theory (predominantly > 1 my and > 1 km2). Marine islands, therefore, may be defined from a biophysical perspective of communities as providing an emerging perspective with the potential to elucidate the community assembly of young islands generally.
Highlights
Fish communities in stratified lakes are distinguished from mixed lakes and ocean communities by their species, phylogenetic, and functional diversity.
Stratified lake communities share attributes of ‘true’ oceanic islands, such as being geographically isolated, species-poor, and potential hotspots of endemism, in contrast to mixed lake communities that exhibit similarities with ‘habitat’ islands, having high connectivity, being relatively species-rich, lacking endemicity, and having trait redundancy.
Stratified lakes do not share relationships with area and age as in mature and large islands; instead, they may illustrate the dynamics of small and young islands.
Island biogeography theory is relevant beyond classical terrestrial island types; by applying the theory to marine systems, we can better explore the processes shaping young island communities, which are historically understudied.
An island may be identified best by how organisms experience and respond to the geographical and environmental conditions.
functional diversityisland biogeographymarine biogeographymarine lakesphylogenetic diversityspecies diversityNational Science Foundation Dimensions of Biodiversity (OCE-1241255)
National Science Foundation Infrastructure Innovation for Biological Research (DBI-1927520)
David and Lucile Packard FoundationIntroduction
Islands have been fertile grounds for empirical and theoretical research, shaping and re-shaping our understanding of biogeographic processes, including in mainland environments (MacArthur and Wilson 1967; Brown 1971; Whittaker et al. 2017). ‘True’ islands, sensu MacArthur and Wilson (1967, pp. 114–115)—including oceanic islands, continental fragments, and continental shelf islands (Wallace 1880)—provided the setting for the pivotal Equilibrium Theory of Island Biogeography (MacArthur and Wilson 1967) and, as evidence accrued of differences among island types and of prevalent community disequilibrium, the need for more nuanced models (Heaney 2000). ‘True’ oceanic islands became the inspiration for the General Dynamic Model (Whittaker et al. 2008) and its subsequent expansion (Borregaard et al. 2016; Dawson 2016; Fernández-Palacios et al. 2016; Pinheiro et al. 2017; Ávila et al. 2019). Such studies of ‘true’ oceanic islands also expanded our understanding of other insular environments, considered ‘habitat’ islands—patches of habitable environment surrounded by a contrasting habitat matrix (Matthews 2021)—and facilitated applications of island theory (e.g., SLOSS [single large or several small] conservation decisions; Diamond 1975a; Zimmerman and Bierregaard 1986). However, island biogeography as a discipline remains uneven in its representation of insular settings that likely meet the fuller definition of ‘complete’ islands, sensu MacArthur and Wilson (1967, p. 122)—i.e., ‘true’ islands of “land surrounded by water” and also “water surrounded by land”—which may constrain what we can learn within the current framework.
Underrepresentation of the full diversity of islands has several causes. There is a relative cornucopia of terrestrial islands that unquestionably fit their eponymous definition (Weigelt et al. 2013; OED 2026); we humans, also being terrestrial, are naturally predisposed to study them (Dawson 2009, 2016). Thus, empirical work has been conducted primarily on the terrestrial biota of oceanic and continental islands (Darwin 1859; Wallace 1880; Diamond 1975b; Jacquet et al. 2017; Ferrari et al. 2023). In addition, there is an assumed distinction between such ‘true’ islands and other island-like habitats such as seamounts (McClain 2007), forest fragments, sky islands, hydrothermal vents (Matthews 2021), and island reefs (Pinheiro et al. 2017; Hachich et al. 2020). Moreover, those assumptions have at times been reinforced when, for example, marine biologists concluded that the evidence for island biogeographic patterns was ambiguous at best (Schoener 1974a, 1974b; Smith 1979; Thomson and Gilligan 2002), even though such conclusions were poorly substantiated (Dawson 2016). Consequently, the expansion of non-equilibrium island biogeography theory for ‘true’ islands (Heaney 2000; Whittaker et al. 2008; Borregaard et al. 2016) did not drive commensurate expansion in marine systems, which lacked a research community (Dawson 2016). These contingencies and the resulting dearth of studies of alternate island types mean that some attributes of islands are rare. For example, few ‘true’ islands naturally capture the early stages of community assembly (Thornton 2007), as do exceptionally young islands such as Anak Krakatau (Whittaker et al. 1989) and Surtsey (Baldursson and Ingadóttir 2007). However, another understudied category of island—marine lakes, i.e., seawater surrounded by land—provides a diversity of relatively young islands in which to explore island biogeographical patterns. At least 200 of these post-glacial habitats are found globally (Dawson et al. 2009). More, younger islands would allow us to explore whether, or the extent to which, they might blur the division of ‘true’ and ‘habitat’ islands if communities—lacking speciation, having minimal extinction, and still largely assembling—are not yet demarcated by endemic species. Understanding the biophysical context of species’ patterns across the full spectrum of islands, to compare with observed patterns from ‘true’ oceanic islands (e.g., island species-area relationships; Darlington 1957; MacArthur and Wilson 1967, pp. 41–56; Dawson and Hamner 2008), would therefore provide an opportunity to increase our understanding of community dynamics and elucidate the early stages in island biogeography theory more generally.
Another opportunity to expand on the historical understanding of species diversity on islands is to more routinely involve multiple measures of diversity (Santos et al. 2016; Schrader et al. 2021). The models and empirical studies of diversity used to advance island biogeography theory have focused on species richness and similar aspects of taxon, taxonomic, or species diversity (SD; MacArthur and Wilson 1967; Whittaker et al. 2008). The availability of traits and phylogenies now means we can consider relationships, beyond species richness, with area, isolation, age, and other additional variables, facilitating a greater understanding of similarities and differences between terrestrial and marine, and other islands. Incorporating traits and phylogenetic relationships in analyses (Cardillo et al. 2008; Whittaker et al. 2014; Jacquet et al. 2017; Ferrari et al. 2023; Schrader et al. 2023) allows us to move past theoretical assumptions of the neutrality of species (Hubbell 2001). Traits significant to the colonization, environmental persistence, and niche competition of species and whether traits in communities are deterministically or stochastically structured can be elucidated (Weiher and Keddy 1995; Villéger et al. 2017). We can discern lineage relationships between communities and the role of speciation with phylogenetic relatedness (Webb et al. 2002; Cavender-Bares et al. 2009). Investigating species (SD), phylogenetic (PD), and functional diversity (FD) in island communities will allow for a better understanding of community assembly and structure. For example, we may expect that oceanic islands differ from other ‘true’ islands in terms of their dynamics and biogeographical properties: oceanic islands are particularly notable for their isolation, species-poor communities derived from colonization from continental biotas, being difficult to reach but easier for immigrating species to invade, and greater potential for speciation and endemism (MacArthur and Wilson 1967, pp. 114–115, 122; Whittaker and Fernández-Palacios 2007, p. 46; Whittaker et al. 2017). In contrast, ‘habitat’ islands are typically harder to invade than ‘true’ oceanic islands due to their established communities and are surrounded by an environment that is distinct but not devoid of competitors (MacArthur and Wilson 1967, pp. 114–115, 122). These differences should leave signatures in SD, PD, and FD.
Marine lakes within the Palau archipelago (Hamner and Hamner 1998; Colin 2009) offer a special opportunity for exploring signatures of island dynamics: they are varied, young, isolated environments, and contain marine communities within which we can investigate SD, PD, and FD. These lakes formed within depressions in karst as the deglacial sea level rose, deeper depressions in the karst—followed by shallower depressions—flooded with seawater, dispersing marine life into these vacant habitats ~12–5 thousand years ago (Fig. 1; Hamner and Hamner 1998; Colin 2009). Marine lakes range in environment (salinity, oxygen concentration, pH) and geography (size, depth, isolation, tidal lag)—often classified as mixed or stratified based on the culmination of these attributes (Fig. 1; Hamner and Hamner 1998; Dawson et al. 2009)—with consequences for their species composition (Hamner and Hamner 1998). Prior studies of Palau’s marine lakes have focused on invertebrate and microbial communities (Meyerhof et al. 2016; Rapacciuolo et al. 2019); ecological patterns of fish communities have not yet been elucidated. Here, we explore how islands of seawater surrounded by land—marine lakes—compare to current island theory. We investigate SD, PD, and FD to assess how marine lake fish communities compare with ocean fish communities. We aim to examine four empirical questions. What diversity patterns do we observe in marine lake fish communities? With what environmental and geographical factors are these patterns correlated? Do marine lake fish communities exhibit typical relationships with geographic patterns akin to ‘true’ oceanic islands (i.e., area, distance from the mainland, and age) or do they show patterns more similar to ‘habitat’ islands? What do marine lakes tell us about island biogeography more generally?
Key elements of the study design. The locations of a. Palau in the Western Pacific, marked by the yellow dot on the globe, and the 22 sampling locations across the islands of b. Ngerchaol and Ngeruktabel; c. Koror and Ulebsechel; d. Ngeruktabel and Ongael, and e. Mecherchar. See Suppl. material 1: fig. S2 for a detailed map with each sampling location labeled with the corresponding three-letter code. The locations are color-coded according to location type—6 ocean locations (red), 8 mixed lakes (light blue), and 8 stratified lakes (green); f. A schematic showing variation in the environment and geography of marine lakes. The white horizontal line represents a thermocline, and the pink horizontal line represents a chemocline; ‰ indicates approximate salinity; the blue lines represent channels connecting lakes to the ocean and possibly to each other; and the brown at the bottom of lakes represents sediment accumulation. Stratified lakes typically have a thermocline and chemocline, which may coincide; salinity differences between layers; weak connections to the ocean and possibly other lakes; and more sediment due to older age and stratification. Mixed lakes have salinity concentrations similar to the ocean, strong connections to the ocean, and vary in depth, impacting the mode of lake mixing, either by tidal flux, wind, or both; they are typically isoclinal. On the left side of the schematic, numbers to the left of the white vertical line represent years before present (k = thousands), and numbers on the right represent estimated depth in meters, illustrating an approximate formation timeline and corresponding depth (Smittenberg et al. 2011; Sachs et al. 2018); g. Several fish species surveyed: (top left) Acentrogobiusjanthinopterus; (top right) Atherinomorusduodecimalis; (middle left) Pomacanthusxanthometopon; (middle right) Lutjanuskasmira and Gnathodentexaurolineatus; (bottom left) Sphaeramiaorbicularis; and (bottom right) Plectorhinchuschaetodonoides, Heniochuschrysostomus, and Sargocentronspiniferum; h. Several surveyed marine lakes pictured: (top) Ongael (OLO, Ongael; mixed; ~13000 m2); (center and within the terrestrial island) Clear (CLM, Mecherchar; stratified; ~54000 m2) and Big Crocodile (BCM, Mecherchar; stratified; ~77000 m2); and (bottom) Ongeim’l Tketau (OTM, Mecherchar; stratified; ~62000 m2). Photos in g & h are courtesy of Robert F. Myers & Patrick L. Colin, respectively.
https://binary.pensoft.net/fig/1741692Materials and methodsData collection and compilation
We conducted 22 fish surveys in August 2013 at three categories of location type: six ocean locations, eight mixed lakes, and eight stratified lakes (Fig. 1a–f; Suppl. material 1: tables S1, S2). Surveys at each location followed a protocol similar to the Conservation International Rapid Assessment Program, as outlined by Allen and Erdmann (2013a, 2013b), for marine fishes. Led by an experienced ichthyologist familiar with Palau fish species composition (RFM), we conducted standardized surveys at each location following pre-determined routes that assessed both the range of habitable space of the water column and environments at each location. Each survey was completed within three hours or once the location was circumnavigated, whichever came first. The surveys were complemented, compared, and integrated with prior fish survey data from Palau to ensure as complete a record as possible for each of the locations. Due to the ocean having less clear boundaries than the marine lakes (which are circumscribed by karst), sampling locations in the ocean were lagoons and coastal regions of similar extent near or outside several of the marine lakes surveyed. Fishes that could not be identified to species in the field were photographed using a Nikon D90 with a 28–80 mm lens, and those unable to be photographed clearly or that needed further study were collected for later identification when possible. All fishes were identified to species, except ten, which could only be identified to genus. Together, these 249 identified species constituted our local species list.
We compiled a reference species list of 1724 fishes that inhabit the upper 200 meters of the ocean surrounding Palau, based on Myers (1999, with revisions and additional species added; RFM unpublished data). This reference species list contained all fishes identified to species from the local list. We used FishBase (Froese and Pauly 2024) via ‘rfishbase’ v4.1.2 (Boettiger et al. 2012) within R v4.4.3 (R Core Team 2025; used for all analyses hereafter) to validate species names, with the ‘validate_names’ function, to standardize the local and reference species lists for analyses of species (SD), phylogenetic (PD), and functional diversity (FD). FishBase validated 1675 names. Our reference list contained 49 species that did not have records on FishBase, of which 43 identifications were only to genus.
Environment and geography
From 2011–2015, we measured salinity (ppt), dissolved oxygen concentration (mg/L dO2), pH, and temperature (°C) at one-meter depth intervals in all lakes and a representative subset of three ocean locations (IBK, Iwayama Bay, Koror; NCN, Ngerchaol Cove, Ngerchaol; & RCA, Risong Cove, Ulebsechel), annually, using a DS5 or Quanta Hydrolab environmental sensor (Suppl. material 1: table S3). There were five exceptions: Big Crocodile Lake (BCM, Mecherchar) was not measured in 2012; Heliofungia Lake (HLO, Ongael) and Ongael Lake (OLO, Ongael) were not measured in 2015; Long Lake (LLN, Ngeruktabel) was measured only in 2012 and 2013; and Iwayama Bay (IBK, Koror) was measured in 2017. We excluded environmental measurements below depths with ≤ 0.7 mg/L dO2, at which fishes from the reference list cannot survive (but see Utne-Palm et al. 2010), and at each location, we calculated the mean of each variable at each depth across years and then calculated the median of the means across depths. Median values were used to limit the effect of outliers. To address potential multicollinearity of variables, we calculated variance inflation factors (VIFs) using ‘car’ v3.1–3 (Fox and Weisberg 2019). We excluded the environmental measure of pH, which had the highest VIF, and after its removal, all VIFs were < 2 for downstream analyses. We scaled and calculated the Euclidean distances for each of the remaining variables.
We also calculated total volume and volume-above-chemocline (m3; which are the same for locations lacking a chemocline); minimum, mean, and median distances to the ocean (m); surface area (m2); log area (m2); perimeter (m); maximum depth (m); tidal lag time (minutes); and tidal efficiency (ratio of ocean tide and location tide amplitudes; Suppl. material 1: tables S4, S5). Area, perimeter, and volume were measured using imagery from 2019–2021 (Google Earth, version 7.3.6.9345 2019–2021). The surface area of each location was log10-transformed (log area). Bathymetric data (provided by Thomas Stieglitz and Herwig Stibor) were collected every two meters using a kayak-towed echo sounder with integrated GPS and AHRS (attitude and heading reference system) to calculate volume for 16 locations. Google Earth was then used to calculate the volume for these 16 locations by fitting geometric shapes to the observed dimensions of the locations. The similarity of bathymetric and Google Earth volumes (R2 = 0.99) permitted us to infer volume for six locations that lacked bathymetric data. Marine lake minimum, mean, and median distances from the ocean were calculated using geographic information systems (provided by Thomas Stieglitz and Herwig Stibor). The tidal range in the ocean and at each location was measured using Onset HOBO water level loggers (#U20-001-04) and used to calculate the tidal efficiency, according to Ayers and Vacher (1986), and tidal lag time (except for Long Lake, LLN, Ngeruktabel; provided by Thomas Stieglitz and Herwig Stibor). We assessed the collinearity of geographical variables using VIFs with ‘car’ v3.1–3 (Fox and Weisberg 2019). Using a stepwise approach, we eliminated one variable at a time with the highest VIF until all remaining variables had VIFs < 2, which were minimum distance to the ocean, maximum depth, and log area. We scaled and calculated the Euclidean distance for each of these geographical variables.
Species
With the location type grouping (ocean, mixed lake, and stratified lake), we created a Venn diagram to illustrate species richness (SRic) and the number of shared species between location types using ‘ggvenn’ v0.1.10 (Yan 2023). The SRic for each location and location type was used for alpha and beta SD analyses.
Phylogeny
A synthetic, dated phylogeny—containing 1627 taxa from our reference list—was downloaded from the Open Tree of Life v13.4 (Redelings et al. 2019) using CHRONOSYNTH (McTavish and Sanchez-Reyes 2022), utilizing previous phylogenetic topologies and dating sources (Suppl. material 1: table S23). We edited species names in the phylogeny using ‘rfishbase’ v4.1.2 (Boettiger et al. 2012) to align taxonomic labels in the different datasets. Of the 97 species absent from the tree, 81 species were bound by genera, and 16 species were added at the subfamily or family level using the ‘phytools’ v2.0–3 (Revell 2012) functions ‘add.species.to.genus’ and ‘bind.tip’. The phylogeny was used for alpha and beta PD analyses.
Traits
For species in the reference list, we downloaded trait data from FishBase (Froese and Pauly 2024) using the ‘rfishbase’ v4.1.2 (Boettiger et al. 2012) functions ‘species’, ‘morphology’, ‘estimate’, ‘ecology’, and ‘reproduction’. We filtered our data to keep only traits that were at least 90% complete for all species in our reference list. We kept traits shown to be ecologically and morphologically relevant to fishes (Villéger et al. 2010; Mouillot et al. 2014; Kültz et al. 2015; Nay et al. 2015; Price et al. 2015; Pecuchet et al. 2016; Villéger et al. 2017; Cattano et al. 2018; Beukhof et al. 2019; Larouche et al. 2020; Facey et al. 2023; Froese and Pauly 2024). The traits (with the FishBase term in parentheses) from this dataset were body shape (BodyShapeI), habitat association (DemersPelag), operculum presence (OperculumPresent), the minimum number of dorsal spines (DorsalSpinesMin), the maximum number of dorsal spines (DorsalSpinesMax), maximum total length (MaxLengthTL; cm), trophic level (Troph), modeled minimum depth (DepthMin; meters), modeled maximum depth (DepthMax; meters), modeled minimum temperature (TempPrefMin; °C), modeled maximum temperature (TempPrefMax; °C), feeding pathway (FeedingPath), ability to inhabit freshwater (Fresh), ability to inhabit brackish water (Brack), and ability to inhabit saltwater (Salt). The FishBase description for each trait can be found in Suppl. material 1: table S6. We calculated the mean number of dorsal spines (DorsalSpinesMean) from the minimum and maximum and used the mean trait value for downstream analyses. We combined the ability to inhabit fresh, brackish, and saltwater into a single trait, water habitat preference (WaterPref), for the trait analyses. The trait matrix did not include any reproductive traits that were at least 90% complete for all species, so we included three reproductive traits (missing 44–57% of trait data) in the trait matrix—egg care (RepGuild1), egg strategy (RepGuild2), and parental care (ParentalCare)—given their relevance to colonization (Bradbury et al. 2008; Patzner 2008).
If species lacked trait data in any category or for fishes identified only to genus, first, the mean (for numerical traits) or modal (for categorical traits) trait value for the genus was substituted in place for the missing value, and second, if missing values were still present, then the mean or modal trait value for the family was substituted in place for the missing value. 1,132 (66%) species had at least one missing trait value, and a total of 3,387 (13%) trait values were missing before average genus data were substituted. Of those species with a missing trait value that was not replaced by the average genus value, 723 (42%) species had at least one missing trait value, and a total of 1829 (7%) missing trait values remained before average family data were substituted. After substituting in average genera and family trait data, only 297 (1%) missing trait values remained. We replaced additional missing values of reproductive strategies using Patzner (2008), if a single reproductive strategy was present for the family. The final matrix contained only 41 (0.2%) missing trait values for the 15 traits of the 1724 species. The final trait matrix for the 249 surveyed species included only 3 (<0.1%) missing trait values for the 15 traits in the matrix.
We identified traits with near-zero variance using the ‘caret’ v7.0–1 (Kuhn 2008) function ‘nearZeroVar’ and calculated the VIFs of each functional trait using ‘car’ v3.1–3 (Fox and Weisberg 2019) to prevent multicollinearity between traits. We removed habitat association (DemersPelag) due to near-zero variance and egg care (RepGuild1) because it was an alias of egg strategy (RepGuild2). The remaining traits in the matrix had VIFs < 2 and were kept separate to assess the influence of individual traits in alpha and beta FD analyses.
The trait matrix was transformed using Gower distances (Gower 1971) with the ‘BAT’ v2.9.6 (Cardoso et al. 2024) function ‘gower’, followed by hierarchical clustering with R v.4.4.3 (R Core Team 2025) using ‘hclust’, and transformed into a ‘phylo’ object using the ‘ape’ v5.8–1 (Paradis and Schliep 2019) function ‘as.phylo’ for downstream alpha and beta FD analyses.
Species, phylogenetic, and functional diversityAlpha diversity
We used species counts from the presence-absence surveys to calculate alpha diversity of each community as species richness (SRic) and log10-transformed species richness (log-SRic). The phylogeny and trait ‘phylo’ object were used to calculate phylogenetic richness (Faith’s PD; Faith 1992) and functional richness (Petchey and Gaston’s FD; Petchey and Gaston 2002), respectively, using the ‘picante’ v1.8.2 (Kembel et al. 2010) function ‘ses.pd’, which is a tree or dendrogram-based method calculating the minimum distance of edges connecting all taxa in the community. The standard effect sizes (z-scores) of alpha phylogenetic (αPD-z) and functional diversity richness (αFD-z) were calculated using the ‘taxa.labels’ null model with 999 runs in ‘picante’ v1.8.2 (Kembel et al. 2010). To identify outliers by location type of SRic, αPD-z, and αFD-z, we used the interquartile range (IQR) method (Tukey 1977) to calculate the 1st (Q1) and 3rd quartiles (Q3) and multiplied the IQR by 1.5, which was subtracted from Q1 to set the lower bound and added to Q3 to set the upper bound. αPD-z and αFD-z were used for downstream analyses because they did not violate Shapiro-Wilk’s normality test nor Levene’s test for homogeneity of variance; using αPD-z and αFD-z avoided the influence of species richness. We assessed the relationship of log-SRic, αPD-z, and αFD-z between the location type grouping (ocean, mixed lake, and stratified lake) using a negative binomial generalized linear model (NBGLM), with the ‘MASS’ v7.3.65 (Venables and Ripley 2002) function ‘glm.nb’, for log-SRic and analyses of variance (ANOVAs), using the function ‘aov’ (R Core Team 2025), and Tukey’s Honestly Significant Difference (HSD) tests for pairwise comparisons of αPD-z and αFD-z.
We assessed the relationships of log-SRic, αPD-z, and αFD-z with environmental and geographical variables using linear models, NBGLMs for analyses including log-SRic, and independent analyses of covariance (ANCOVAs) for analyses including αPD-z or αFD-z, including Bonferroni correction of p-values when the same dataset of location types was used. With the NBGLMs and ANCOVAS, we tested for an interaction between location types and each environmental and geographical variable influencing log-SRic, αPD-z, or αFD-z. If the interaction was not significant, we performed independent additive NBGLM with log-SRic and ANCOVAs with αPD-z or αFD-z with each environmental and geographical variable, grouped by the location type, to assess the variation explained by the covariate and grouping variable, with Bonferroni correction of p-values. We also assessed the relationship of variance for log-SRic, αPD-z, and αFD-z within mixed lake and stratified lake communities independently, with each environmental and geographical variable using univariate NBGLMs or ANOVAs.
Beta diversity
Total, richness, and replacement beta species (βSD), phylogenetic (βPD), and functional (βFD) diversity were calculated from the species counts, phylogeny, and trait ‘phylo’ object using the ‘BAT’ v2.9.6 (Cardoso et al. 2024) function ‘beta’ with the default Jaccard family of beta diversity measures (Jaccard 1912). With total βSD, βPD, and βFD, we calculated multivariate homogeneity with a test of dispersion, using the ‘vegan’ v2.6–10 (Oksanen et al. 2022) function ‘betadisper’, for within-group variance (Anderson 2006) of location types—determining significance with an ANOVA and a Tukey’s HSD test for pairwise comparisons. The location type within-group variance was tested to determine homoscedasticity of groups and for outliers using the IQR method (Tukey 1977). A permutation multivariate analysis of variance (PERMANOVA; Anderson 2017) was used to determine the between-group variation of location types for total, richness, and replacement beta diversity and pairwise comparisons, with Bonferroni correction of p-values, using the ‘vegan’ v2.6–10 (Oksanen et al. 2022) function ‘adonis2’ and the ‘pairwiseAdonis’ v0.4.1 (Martinez Arbizu 2020) function ‘pairwise.adonis’. The hierarchical clustering of total βSD, βPD, and βFD was assessed within a dendrogram using ‘hclust’ (R Core Team 2025). We ascertained whether the Jaccard distances of total βSD, βPD, and βFD were correlated with the scaled Euclidean distances of environmental and geographical variables using univariate Mantel tests (Mantel 1967), with Bonferroni correction of p-values, using the ‘vegan’ v2.6–10 (Oksanen et al. 2022) function ‘mantel’. Total, richness, and replacement βSD, βPD, and βFD by location type were visualized using non-metric multidimensional scaling (NMDS) ordination (Kruskal 1964) with the ‘vegan’ v2.6–10 (Oksanen et al. 2022) function ‘metaMDS’ with a maximum of 500 random starts and 500 iterations. We tested for outliers in the NMDS1 and NMDS2 axes of total βSD, βPD, and βFD by location type using the IQR method (Tukey 1977). The function ‘envfit’ in ‘vegan’ v2.6–10 (Oksanen et al. 2022) was used to identify measured environmental and geographical variables significantly correlated, after Bonferroni correction of p-values using the function ‘p.adjust.envfit’ (https://www.davidzeleny.net/anadat-r/doku.php/en:start), with the NMDS ordinations of total βSD, βPD, and βFD.
Individual trait analyses
Differences in categorical trait proportions of communities by location type (ocean, mixed lake, and stratified lake) were evaluated using the function ‘fisher.test’ to conduct a Fisher’s exact test (Fisher 1922) for all location types and pairwise comparisons, with Bonferroni correction of p-values. Differences in numerical trait distributions of communities by location type were evaluated using the function ‘dunn.test’ to perform a Kruskal-Wallis test (Kruskal and Wallis 1952) for all location types and a Dunn test (Dunn 1964) for pairwise comparisons, with Bonferroni correction of p-values. The total beta functional diversity for each categorical and numerical trait was calculated using the ‘BAT’ v2.9.6 (Cardoso et al. 2024) function ‘beta’, and a PERMANOVA was performed to identify which of those traits were significantly different between location types using the ‘vegan’ v2.6–10 (Oksanen et al. 2022) function ‘adonis2’. We also calculated pairwise comparisons of each trait’s total beta functional diversity between location types, with Bonferroni correction of p-values, using ‘pairwiseAdonis’ v0.4.1 (Martinez Arbizu 2020). With the subset of traits identified as significantly different between location types, we calculated total βFD using ‘beta’ and analyzed between-group variation of location types and pairwise comparisons, with Bonferroni p-value correction, using a PERMANOVA with ‘adonis2’ and ‘pairwise.adonis’. We used an NMDS ordination, using ‘metaMDS’, to visualize the location type patterns of total βFD with the subset of traits dataset. All visualizations in the methods used ggplot2 v3.5.1 (Wickham 2016).
ResultsData collection and compilation
The variation between the reference species list and the communities in the ocean and marine lakes was greater than the variation between the ocean and marine lake communities (Suppl. material 1: fig. S25). The reference species list was therefore excluded from downstream analyses.
Species
The surveys recorded a total species richness (SRic) of 249 fishes out of the 1724 species from the reference pool from Palauan waters. Ocean and mixed lake communities shared similarly high species richness (166 and 185 species, respectively), while stratified lake communities had low species richness (30 species), except for two locations: TLN (T-Lake, Ngermeuangel) and HLM (Hot Water Lake, Mecherchar; Fig. 2a); the richness of HLM (18 species) was an outlier from other stratified lake communities and TLN (17 species) was more similar in species richness to HLM than to the other stratified lakes (mean = 4.17, sd = 1.33, Fig. 2a). The richness of ocean communities (166) was similar to mixed lake communities when the sample size of locations was downsampled to six mixed lake communities (152 ± 17; Suppl. material 1: fig. S3).
Alpha diversity violin plots of a. species richness; b. phylogenetic richness z-scores, and c. functional richness z-scores, with communities grouped by location type, reveal depauperate richness of stratified lake communities and the random distribution of traits and phylogenetic relationships in marine lake communities. Colors show location types: ocean (red), mixed lake (light blue), and stratified lake (green). Purple points in the violin plots are outliers identified with the interquartile range method. Asterisks represent significant differences. Stratified lake communities are significantly different from ocean and mixed lake communities for species richness, and functional richness z-scores of stratified lake communities are significantly different from those of ocean communities.
https://binary.pensoft.net/fig/1741693Species, phylogenetic, and functional diversityAlpha diversity
The alpha diversity of communities differed significantly by location type for log-SRic (df = 2,19; LRT = 47.83; p < 0.001), but for the post-hoc pairwise comparisons, only stratified lake communities were significantly different from ocean (df = 1,12; LRT = 39.33; p < 0.001) and mixed lake communities (df = 1,14; LRT = 37.53; p < 0.001; Suppl. material 1: table S7). The alpha phylogenetic diversity richness z-scores (αPD-z) did not significantly differ between location types (df = 2,19; F = 0.89; p = 0.43; Fig. 2b; Suppl. material 1: table S7), nor for any of the post-hoc pairwise comparisons: ocean and mixed lake communities (q = 1.67, p = 0.48), ocean and stratified lake communities (q = 1.67, p = 0.48), and mixed lake and stratified lake communities (q < 0.01, p = 1). Mixed lake and stratified lake communities had similar αPD-z values resembling stochastic phylogenetic relatedness in each community (z-scores ≤ 2 or ≥ -2), while ocean communities had the highest (LCN, Long Lake Cove, Ngeruktabel; z = 0.64) and lowest (IBK, Iwayama Bay, Koror; z=-3.54) αPD-z values. The low αPD-z scores (≤-2) illustrate clustered phylogenetic relatedness of species at that location. The community of LCN was identified as an outlier in αPD-z for ocean communities, having a similar z-score to mixed lake and stratified lake communities. Removing LCN from post-hoc analyses did lead to increased q-values but not significant differences between ocean communities and mixed lake (q = 3.13, p = 0.12) or stratified lake communities (q = 3.13, p = 0.12). The alpha functional diversity richness z-scores (αFD-z) showed significant differentiation between all location types (df = 2,19; F = 4.88; p = 0.02), but only the post-hoc pairwise comparison between ocean and stratified lake communities was significant (q = 4.30, p = 0.02; Fig. 2c; Suppl. material 1: table S7). The post-hoc pairwise comparison between ocean and mixed lake communities was nearly significant for αFD-z (q = 3.30, p = 0.08). When HLO—identified as an outlier for mixed lake communities—was removed from post-hoc analyses, the differentiation between ocean and mixed lake communities increased and was significant (q = 5.07, p = 0.008), identifying HLO as similar to ocean communities.
The species-area relationship investigated for log-SRic and log area, using a linear model, revealed the slope (z) of ocean locations was 0.03, 0.35 for mixed lakes, and < 0.001 for stratified lakes.
Five of the negative binomial generalized linear models (NBGLMs) and five analyses of covariance (ANCOVAs), testing for interactions between log-SRic, αPD-z, and αFD-z of ocean, mixed lake, and stratified lake communities and their pairwise comparisons with the six environmental and geographical variables, were significant or nearly significant after Bonferroni correction (Suppl. material 1: table S8). The significant interactions identified using NBGLMs were for log-SRic and oxygen (all location types: df = 2,13; LRT = 16.54; p = 0.001; ocean & stratified: df = 1,7; LRT = 27.16; p < 0.001; mixed & stratified: df = 1,12; LRT = 11.72; p = 0.003) and log area (all location types: df = 2,16; LRT = 6.32; p = 0.02; ocean & mixed: df = 1,10; LRT = 7.91; p = 0.02). The significant interactions identified using ANCOVAs involved max depth with αPD-z (df = 2,16; F = 7.08; p = 0.04) and αFD-z (df = 2,16; F = 6.03; p = 0.08) and for the pairwise comparisons between ocean and stratified lake communities for αPD-z (df = 1,10; F = 14.07; p = 0.03) and mixed and stratified lake communities for αFD-z (df = 1,12; F = 10.34; p = 0.06). A significant interaction between location type and distance from the ocean was observed for mixed and stratified communities for αFD-z (df = 1,12; F = 14.07; p = 0.03).
Of the remaining tests with nonsignificant interactions, the resulting NBGLMS and ANCOVAs with log-SRic, αPD-z, and αFD-z of all location types and pairwise comparisons showed variation in their relationships with salinity, oxygen, minimum distance to the ocean, max depth, and log area (Suppl. material 1: table S8). For all location type comparisons, αPD-z had no significant ANCOVAs, while log-SRic was significantly correlated with salinity (df = 1,15; LRT = 11.96; p = 0.003), and αFD-z was significantly correlated with oxygen (df = 1,15; F = 14.52; p = 0.01; Suppl. material 1: table S8, figs S4, S5). For the pairwise comparisons of ocean and mixed lake communities, log-SRic showed significant correlations with oxygen (df = 1,8; LRT = 7.38; p = 0.04) and max depth (df = 1,11; LRT = 7.77; p = 0.03), αPD-z also showed significant correlations with oxygen (df = 1,8; F = 12.22; p = 0.05) and max depth (df = 1,11; F = 16.93; p = 0.01), and αFD-z showed significant correlations with oxygen (df = 1,8; F = 13.63; p = 0.04), minimum distance to the ocean (df = 1,11; F = 10.35; p = 0.05), and max depth (df = 1,11; F = 12.47; p = 0.03). The pairwise comparisons of ocean and stratified lake communities, showed the same patterns as the comparisons including all location types, with αPD-z having no significant relationships, while log-SRic was correlated with salinity (df = 1,8; LRT = 17.47; p < 0.001), and αFD-z was correlated with oxygen (df = 1,8; F = 22.86; p = 0.008). The pairwise comparisons of mixed lake and stratified lake communities had no significant environmental or geographical correlations with αPD-z and αFD-z, but log-SRic showed significant correlations with salinity (df = 1,13; LRT = 10.76; p < 0.001). Log-SRic was nearly significantly correlated with minimum distance to the ocean for all location types and pairwise comparisons, except for ocean and mixed lake communities (Suppl. material 1: table S8). αFD-z was nearly significantly correlated with log area for all location types and pairwise comparisons, except mixed and stratified lake communities. Within mixed lake communities αPD-z and αFD-z had no significant correlations with environmental or geographic variables but log-SRic was significantly correlated with log area (df = 1,6; LRT = 27.60; p < 0.001). Log-SRic within stratified lake communities was significantly correlated with salinity (df = 1,6; LRT = 18.84; p < 0.001), oxygen (df = 1,6; LRT = 10.09; p = 0.004), and minimum distance to the ocean (df = 1,6; LRT = 6.61; p = 0.03), and αFD-z was significantly correlated with salinity (df = 1,6; F = 12.45; p = 0.04).
Beta diversity
When grouping by location type, total beta species (βSD: df = 2,19; F = 0.13; p = 0.88), phylogenetic (βPD: df = 2,19; F = 2.89; p = 0.08), and functional diversity (βFD: df = 2,19; F = 0.47; p = 0.64) were homoscedastic, and no outliers were found (Suppl. material 1: table S9, figs S6a, S12a, S18a). Total and richness βSD, βPD, and βFD PERMANOVAs were significant for all location types (total βSD: df = 2,19; F = 3.27; R2 = 0.26; p = 0.001 & richness βSD: df = 2,19; F = 10.12; R2 = 0.52; p = 0.001; total βPD: df = 2,19; F = 6.16; R2 = 0.39; p = 0.001 & richness βPD: df = 2,19; F = 16.81; R2 = 0.64; p = 0.001; and total βFD: df = 2,19; F = 4.54; R2 = 0.32; p = 0.001 & richness βFD: df = 2,19; F = 11.28; R2 = 0.54; p = 0.001). In pairwise comparison PERMANOVAs, total and richness βSD, βPD, and βFD of stratified lake communities differed significantly from ocean (total βSD: df = 1,12; F = 3.91; R2 = 0.25; p = 0.003 & richness βSD: df = 1,12; F = 14.09; R2 = 0.54; p = 0.009; total βPD: df = 1,12; F = 8.36; R2 = 0.41, p = 0.006 & richness βPD: df = 1,12; F = 27.02; R2 = 0.69; p = 0.003; and total βFD: df = 1,12; F = 5.54; R2 = 0.32; p = 0.003 & richness βFD: df = 1,12; F = 12.99; R2 = 0.52; p = 0.006) and mixed lake communities (total βSD: df = 1,14; F = 4.16; R2 = 0.23; p = 0.003 & richness βSD: df = 1,14; F = 14.15; R2 = 0.50; p = 0.006; total βPD: df = 1,14; F = 9.42; R2 = 0.4; p = 0.003 & richness βPD: df = 1,14; F = 23.17; R2 = 0.62; p = 0.003; and total βFD: df = 1,14; F = 6.24; R2 = 0.31; p = 0.009 & richness βFD: df = 1,14; F = 17.09; R2 = 0.55; p = 0.003), but not for ocean and mixed lake community comparisons (total βSD: df = 1,12; F = 1.58; R2 = 0.12; p = 0.14 & richness βSD: df = 1,12; F=-0.32; R2=-0.03; p = 1; total βPD: df = 1,12; F = 1.47; R2 = 0.11; p = 0.33 & richness βPD: df = 1,12; F=-0.07; R2=-0.01; p = 1; and total βFD: df = 1,12; F = 1.62; R2 = 0.12; p = 0.18 & richness βFD: df = 1,12; F=-0.21; R2=-0.02; p = 1; Fig. 3; Suppl. material 1: tables S10, S13, S16, figs S7b, S13b, S19b). However, the PERMANOVAs for replacement βSD and βFD were only significantly different for the pairwise comparison of ocean and mixed lake communities (replacement βSD: df = 1,12; F = 3.44; R2 = 0.22; p = 0.02 & replacement βFD: df = 1,12; F = 3.58; R2 = 0.23; p = 0.02), and nearly significant for replacement βPD (df = 1,12; F = 2.94; R2 = 0.20; p = 0.06; Suppl. material 1: tables S10, S13, S16, figs S7a, S13a, S19a).
The non-metric multidimensional scaling ordinations, based on the default Jaccard family of beta diversity measures, of total beta a. species; b. phylogenetic, and c. functional diversity highlight the distinct communities of the stratified lakes and similarities and overlap of ocean and mixed lake communities. Locations are identified by type: ocean (red), mixed lake (light blue), and stratified lake (green). Ellipses are 95% confidence intervals. Each non-metric multidimensional scaling ordination is overlaid with the significant or nearly significantly correlated environmental variable, median salinity (S), in yellow. The correlation strength may not match the length of the vector; instead, see Suppl. material 1: tables S11, S14, S21.
https://binary.pensoft.net/fig/1741694
The NMDS ordinations of total βSD, βPD, and βFD had low stress values (Fig. 3). Overlap of 95% confidence intervals of nearly all location types was observed for total βSD, βPD, and βFD NMDS ordinations. The exceptions were: the confidence intervals of ocean communities and stratified lake communities did not overlap each other’s points for total βSD, βPD, and βFD; the confidence interval of mixed lake communities did not overlap any points of stratified lake communities for total βSD and βFD; and the confidence interval of stratified lake communities did not overlap any points of mixed lake communities for total βPD. The overlap of 90% confidence intervals within the NMDS ordinations was less, with only ocean and mixed lake communities overlapping for total βSD, βPD, and βFD, in addition to the confidence interval of mixed lake communities overlapping points of stratified lake communities for total βPD and the confidence interval of stratified lake communities overlapping points of mixed lake communities for total βFD (Suppl. material 1: figs S6b, S12b, S18b). The NMDS ordinations of replacement βSD, βPD, and βFD had high stress, making them unreliable for visualization, while richness βSD, βPD, and βFD had low stress and a 95% confidence interval overlap of ocean and mixed lake communities, plus two stratified lake communities, TLN (T-Lake, Ngermeuangel) and HLM (Hot Water Lake, Mecherchar), clustering near these communities and within the 95% confidence intervals of both communities for the NMDS ordinations of βSD and βFD (Suppl. material 1: figs S7, S13, S19).
The NMDS ordinations and dendrograms of total βSD, βPD, and βFD highlighted two ocean communities, LCN (Long Lake Cove, Ngeruktabel) and OCO (Ongael Lake Cove, Ongael), as separate from the other ocean communities and the stratified lake communities, HLM (Hot Water Lake, Mecherchar) and TLN (T-Lake, Ngermeuangel), separate from other stratified lake communities (Fig. 3; Suppl. material 1: figs S8, S9, S14, S15, S20, S21, S24). LCN and OCO were identified as outliers on the NMDS2 axis of the ordination for total βSD and βFD, while only LCN was identified as an outlier for the NMDS2 axis of the total βPD (Suppl. material 1: figs S8b, S14b, S20b). When we removed LCN from pairwise PERMANOVA calculations, the magnitude and significance of differences between ocean and mixed lake communities increased (total βSD: df = 1,11; F = 2.14; R2 = 0.16; p = 0.02; total βPD: df = 1,11; F = 2.03; R2 = 0.16; p = 0.06; and total βFD: df = 1,11; F = 2.25; R2 = 0.17; p = 0.04). No overlap of the ocean confidence interval over mixed lake communities remained in the NMDS ordinations, when LCN was excluded (Suppl. material 1: figs S9a, S15a, S21a). However, removing OCO from the PERMANOVA analyses had a limited effect on the relationship of ocean and mixed lake communities pairwise comparisons (total βSD: df = 1,11; F = 1.58; R2 = 0.13; p = 0.17; total βPD: df = 1,11; F = 1.55; R2 = 0.12; p = 0.39; and total βFD: df = 1,11; F = 1.60; R2 = 0.13; p = 0.24). TLN and HLM were not identified as outliers in either of the axes for the NMDS ordinations but clustered within the 90% confidence interval of mixed lake communities for total βPD (Suppl. material 1: fig. S12b) and clustered with mixed lake communities in the dendrograms of total βSD and βFD, separate from the other stratified lake communities (Suppl. material 1: fig. S24a, c). When TLN and HLM were removed from PERMANOVA analyses, the exclusivity of stratified lake communities from ocean (total βSD: df = 1,10; F = 4.15; R2 = 0.29; p = 0.012; total βPD: df = 1,10; F = 9.19; R2 = 0.48; p = 0.009; and total βFD: df = 1,10; F = 6.50; R2 = 0.39; p = 0.009) and mixed lake communities increased (total βSD: df = 1,12; F = 4.73; R2 = 0.28; p = 0.006; total βPD: df = 1,12; F = 10.50; R2 = 0.47; p = 0.003; and total βFD: df = 1,12; F = 7.55; R2 = 0.39; p = 0.006; Suppl. material 1: figs S9b, S15b, S21b).
Salinity was the only variable of the included environmental or geographical factors that was significantly or nearly significantly correlated with the NMDS ordinations and Jaccard distances of total βSD, βPD, and βFD for all location types and many (15 of 18 analyses) of the pairwise comparisons (Suppl. material 1: tables S11, S12, S14, S15, S21 & S22). The NMDS of total βSD showed a significant correlation with salinity for the pairwise comparison of ocean and mixed lake communities (R2 = 0.78, p = 0.03) and nearly significant correlations for all location types (R2 = 0.74, p = 0.09) and the pairwise comparison of mixed lake and stratified lake communities (R2 = 0.75, p = 0.08). The Mantel test identified Jaccard distances of βSD with significant correlations with salinity for all location types (r = 0.67, p = 0.01) and the pairwise comparisons of stratified lake and ocean communities (r = 0.55, p = 0.05) and mixed lake and stratified lake communities (r = 0.67, p = 0.08). The NMDS of βPD showed nearly significant correlations with salinity for all location types (R2 = 0.70, p = 0.06) and the pairwise comparisons of ocean and mixed lake communities (R2 = 0.73, p = 0.08) and mixed lake and stratified lake communities (R2 = 0.68, p = 0.08). The Jaccard distances of βPD were identified as significantly correlated with salinity for all location types (r = 0.51, p = 0.02) and all pairwise comparisons, including ocean and mixed lake communities (r = 0.46, p = 0.05), ocean and stratified lake communities (r = 0.48, p = 0.04), and mixed and stratified lake communities (r = 0.46, p = 0.02). Salinity was significantly correlated with the NMDS of βFD for all location types (R2 = 0.79, p = 0.04) and all the pairwise comparisons: ocean and mixed lake communities (R2 = 0.78, p = 0.04), ocean and stratified lake communities (R2 = 0.78, p = 0.05), and mixed lake and stratified lake communities (R2 = 0.79, p = 0.02). The Jaccard distances of βFD were significantly correlated with all location types (r = 0.69, p = 0.01) and the pairwise comparisons of ocean and stratified lake communities (r = 0.57, p = 0.02) and mixed lake and stratified lake communities (r = 0.68, p = 0.02) and nearly significantly correlated with the pairwise comparison of ocean and mixed lake communities (r = 0.46, p = 0.07).
Max depth and oxygen were also identified as significantly or nearly significantly correlated, but with only the pairwise comparisons of ocean and mixed lake communities for total βSD, βPD, and βFD (Suppl. material 1: tables S11, S12, S14, S15, S21, & S22). Max depth was significantly correlated with the NMDS of βPD (R2 = 0.67, p = 0.01) and βFD (R2 = 0.56, p = 0.05) and Jaccard distances of βPD (r = 0.31, p = 0.04) and nearly significantly correlated with the NMDS of βSD (R2 = 0.57, p = 0.07). Oxygen was significantly correlated with the Jaccard distances of βPD (r = 0.55, p = 0.02) and βFD (r = 0.55, p = 0.04) and nearly significantly correlated with the Jaccard distances of βSD (r = 0.53, p = 0.10) and the NMDS of βPD (R2 = 0.80, p = 0.06).
Mixed lake communities showed significant or nearly significant correlations with several geographical features for total βSD, βPD, and βFD (Suppl. material 1: tables S11, S12, S14, S15, S21, S22). The NMDS of βSD for mixed lake communities was significantly correlated with log area (R2 = 0.83, p = 0.03) and nearly significantly correlated with max depth (R2 = 0.83, p = 0.10), while the Jaccard distances were significantly correlated with max depth (r = 0.68, p = 0.01). The Jaccard distances of βPD were nearly significantly correlated with max depth (r = 0.69, p = 0.06) and log area (r = 0.40, p = 0.10). The NMDS ordination of βFD was nearly significantly correlated with oxygen (R2 = 0.87, p = 0.02), and the Jaccard distances of βFD were nearly significantly correlated with max depth (r = 0.63, p = 0.06).
Stratified lake communities showed no significant or nearly significant correlations with any environmental or geographical variables for total βSD, βPD, and βFD (Suppl. material 1: tables S11, S12, S14, S15, S21, S22).
Individual trait analyses
The proportions of three categorical traits and distributions of three numerical traits significantly differed between all location types and for at least one pairwise comparison (Suppl. material 1: tables S17, S18). Between all location types and pairwise comparisons, ocean and mixed lake communities, ocean and stratified lake communities, and mixed and stratified lake communities significantly differed in habitat water preference (WaterPref; Fig. 5c; all location types: p < 0.001; ocean & mixed: p < 0.001; ocean & stratified: p < 0.001; mixed & stratified: p < 0.001), trophic level (Troph; Fig. 4a; all location types: df = 2,19; H = 29.74; p < 0.001; ocean & mixed: df = 1,12; z = 4.36; p < 0.001; ocean & stratified: df = 1,12; z=-4.44; p < 0.001; mixed & stratified: df = 1,14; z=-2.19; p < 0.001), and modeled minimum depth (DepthMin; Suppl. material 1: fig. S23f; all location types: df = 2,19, H = 41.23, p < 0.001; ocean & mixed: df = 1,12; z=-2.34; p = 0.03; ocean & stratified: df = 1,12; z = 6.41; p < 0.001; mixed & stratified: df = 1,14; z = 5.29; p < 0.001). The mean number of dorsal spines (DorsalSpinesMean; Fig. 4b) significantly differed between all location types (df = 2,19; H = 16.84; p < 0.001) and the pairwise comparisons of ocean and stratified lake communities (df = 1,12; z = 4.10; p < 0.001) and mixed lake and stratified lake communities (df = 1,14; z = 3.48; p < 0.001), and parental care (ParentalCare; Fig. 4d) was nearly significant for the same comparisons (all location types: p = 0.11; ocean & stratified: p = 0.06; mixed stratified: p = 0.08). Body shape (BodyShapeI; Suppl. material 1: fig. S23c) and egg strategy (RepGuild2; Fig. 4c) significantly differed between all location types (Body shape: p = 0.004 & Egg strategy: p = 0.04) and for the pairwise comparison of ocean and stratified lake communities (Body shape: p = 0.004 & Egg strategy: p = 0.04), and operculum presence (OperculumPresent; Suppl. material 1: fig. S23a) was nearly significantly different for all location types (p = 0.06) and the pairwise comparison of ocean and stratified lake communities (p = 0.07). Egg strategy was also marginally non-significant for the pairwise comparison between mixed lake and stratified lake communities (p = 0.08).
The distribution (a, b) of biotic numerical trait values, trophic level and mean number of dorsal spines, and proportion (c, d) of biotic categorical trait values, egg strategy and parental care, illustrate the unique composition of traits within the Palau stratified lake communities when compared to ocean and mixed lake communities. Traits are plotted by location type with the x-axis representing: O, ocean (red); M, mixed lake (light blue); and S, stratified lake (green) fish communities. The scale for categorical trait proportions is from 0 to 1 for each location type.
The abiotic distribution of a, b. numerical trait values, modeled minimum and maximum temperature, and proportion of the c. categorical trait value, water habitat preference, identify unique assemblages of traits within the Palau stratified lake communities, when compared to ocean and mixed lake communities. The type of water fishes inhabit could be salt, brackish, freshwater, a combination of any two, or all three. Traits are plotted by location type, and the x-axis represents: O, ocean (red); M, mixed lake (light blue); and S, stratified lake (green) fish communities.
https://binary.pensoft.net/fig/1741696
The subset of traits identified from the individual PERMANOVAs as significantly different—after Bonferroni correction—between all location types and pairwise comparisons between mixed lake and stratified lake communities were trophic level (all location types: df = 2,19; F = 6.34; R2 = 0.40; p = 0.03; mixed & stratified: df = 1,14; F = 7.70; R2 = 0.35; p = 0.02), mean number of dorsal spines (all location types: df = 2,19; F = 5.64; R2 = 0.37; p = 0.04; mixed & stratified: df = 1,14; F = 6.53; R2 = 0.32; p = 0.01), parental care (all location types: df = 2,19; F = 8.67; R2 = 0.48; p = 0.01; mixed & stratified: df = 1,14; F = 13.77; R2 = 0.50; p = 0.006), and modeled minimum temperature (TempPrefMin; °C; Fig. 5a; all location types: df = 2,19; F = 9.03; R2 = 0.49; p = 0.01; mixed & stratified: df = 1,14; F = 12.23; R2 = 0.47; p = 0.006; Suppl. material 1: table S19). The pairwise comparison for ocean and stratified lake communities was also significant for parental care (df = 1,12; F = 10.21; R2 = 0.46; p = 0.006) and modeled minimum temperature (df = 1,12; F = 8.79; R2 = 0.42; p = 0.003). The pairwise comparisons between ocean and mixed lake communities for each trait were not significant, but trophic level was nearly significant (df = 1,12; F = 4.72; R2 = 0.28; p = 0.07). Additionally, pairwise comparisons between mixed lake and stratified lake communities were significantly different for egg strategy (df = 1,14; F = 5.27; R2 = 0.29; p = 0.02), water habitat preference (df = 1,14; F = 4.81; R2 = 0.26; p = 0.01), modeled maximum temperature (TempPrefMax; °C; Fig. 5b; df = 1,14; F = 6.03; R2 = 0.30; p = 0.002), and maximum total length (MaxLengthTL; cm; df = 1,14; F = 4.79; R2 = 0.26; p = 0.01; Suppl. material 1: fig. S23b, table S20). The pairwise comparison between ocean and stratified lake communities for maximum total length was also significant (df = 1,12; F = 2.82; R2 = 0.19; p = 0.01). Only including the traits that were significantly different between location types—trophic level, the mean number of dorsal spines, parental care, and modeled minimum temperature—in the total βFD calculation resulted in the PERMANOVA results of total βFD increasing the strength of differences between location types (df = 2,19; F = 5.58; R2 = 0.37; p = 0.001) compared to total βFD using all traits. When calculating pairwise comparisons, only analyses including stratified lake communities with ocean (df = 1,12; F = 6.58; R2 = 0.35; p = 0.003) and mixed lake communities (df = 1,14; F = 7.84; R2 = 0.36; p = 0.003) increased in strength. The pairwise comparison between ocean and mixed lake communities decreased in strength (df = 1,12; F = 1.33; R2 = 0.11; p = 0.39), when compared to total βFD using all traits. The NMDS ordination showed substantial clustering of ocean and mixed lake communities and separation of stratified lake communities (Suppl. material 1: fig. S22).
Discussion
Although a diverse set of islands—from ‘true’ to ‘habitat’, aquatic and terrestrial—were considered by the Equilibrium Theory of Island Biogeography (MacArt h ur and Wilson 1967), subsequent studies have tended to emphasize terrestrial 'true' islands: oceanic islands, continental fragments, and continental shelf islands (Wallace 1880; Whittaker et al. 2017; Matthews and Triantis 2021). Our goal here has been to explore one category of underrepresented islands—marine lakes—to consider how they compare with the patterns of existing island theory. We are particularly interested in whether marine lakes align more with ‘complete’—especially ‘true’ oceanic islands—or with ‘habitat’ islands (MacArthur and Wilson 1967, pp. 114–115, 122) and, therefore, to address the theoretical and empirical implications of studying young islands for understanding island biogeography generally.
Stratified lakes and their fish communities share some characteristics with ‘complete’ islands, while mixed lakes are more similar to ‘habitat’ islands. Stratified lakes resemble ‘complete’ islands with their depauperate communities; communities having distinct trait compositions related to colonization (i.e., parental care and egg strategy), environmental persistence (i.e., water habitat preference), and niche competition (i.e., trophic level and body shape); and unique beta diversity when compared to ocean communities; though they do not match expectations exactly, given the lack of a species-area relationship. In contrast, mixed lakes are akin to ‘habitat’ islands (sensuMacArthur and Wilson 1967, pp. 114–115, 122), given their communities having similar total beta species (βSD), phylogenetic (βPD), and functional diversity (βFD) and richness beta diversity (βSD, βPD, and βFD) but differences in replacement beta diversity (βSD, βPD, and βFD) when compared to ocean communities and greater species richness and redundancy of traits than stratified lake communities.
Between all location types, species richness (SRic, log-SRic) and alpha functional diversity richness z-scores (αFD-z) are significantly different, but alpha phylogenetic diversity richness z-scores (αPD-z) are not. Log-SRic is correlated with salinity and nearly significantly correlated with minimum distance to the ocean for all location type comparisons, except the pairwise comparison of ocean and mixed lake locations. αFD-z is correlated with the oxygen concentration and nearly significantly correlated with log area between all location types and pairwise comparisons, except for mixed lake and stratified lake communities. Analyses with αPD-z only show significant correlations with oxygen concentration and max depth, which occur between ocean and mixed lake communities. In contrast, between location types, total beta diversity (βSD, βPD, and βFD) is significantly different; the analyses, including all location types and pairwise comparisons, are similarly significantly or marginally non-significantly correlated with salinity. Despite the differences in community composition, environment, and geography between location types, some fish communities reduce the variation between the stratified lakes and mixed lakes. The differences of species (SD), phylogenetic (PD), and functional diversity (FD), and environmental variation of the stratified lake and mixed lake communities illustrate a diverse set of ‘complete’ and ‘habitat’ islands with the potential to expand our understanding of island communities. In the following sections, we explore fish community SD, PD, and FD patterns and correlations with environmental and geographical factors to clarify their implications in the context of island biogeography theory.
Diversity patterns in marine lake communitiesDeterministic or stochastic assemblies?
Patterns of SD, PD, FD, and the proportion and distribution of traits, indicate how deterministic and stochastic processes shape marine lake communities, when compared to ocean communities. Though our surveys are based on incidence data, the significant differences of log-SRic correlated with salinity between the mixed lake and stratified lake communities are nonetheless consistent with deterministic patterns distinguishing marine lake communities based on location type. In contrast, the stochastic αPD-z and αFD-z distinguish both mixed and stratified lakes together from the ocean, with substantial differences in αFD-z between the marine lakes and the ocean. Yet the correlation of αFD-z of all location types and the pairwise comparison of ocean and stratified lake communities with oxygen, and the nearly significant differences in αFD-z between ocean and mixed lake communities being correlated with oxygen, max depth, and minimum distance to the ocean, reveal some deterministic patterns of traits based on location type. The deterministic pattern is identified by the linear relationship of αFD-z with these environmental and geographical variables. αPD-z does not show significant differences between location types, but a linear relationship between ocean and mixed lake communities correlated with oxygen and max depth is observed. Teasing apart the deterministic and stochastic patterns in mixed lake and stratified lake communities might clarify if community diversity is related to priority effects and historical contingency (Fukami 2015), in which species entered marine lakes randomly upon formation and competitively excluded later colonizers, or masking intransitive competition, in which indirect competition is caused by variation in the competitive ability of organisms across resources and at different life stages (Gallien 2017). Differentiating between these processes will require additional temporal data, for example, from a long-term observational time-series or sediment cores, and/or experiment.
Considering beta diversity (βSD, βPD, and βFD), dispersal and selection are prominent in explaining the variation between location types (Vellend 2010; Vellend 2016). The mixed lake communities are distinct from ocean communities in terms of the replacement of βSD and βFD and are nearly significantly different for the replacement of βPD, but are not significantly different for other beta diversity measures (total and richness βSD, βPD, or βFD). Likewise, beta diversity (βSD, βPD, and βFD) correlations of ocean and mixed lake communities with oxygen and max depth, alongside the turnover between these two location types, implicate them as additional influences differentiating community assemblages. In contrast, differences in beta diversity (total and richness βSD, βPD, and βFD), and the proportion and distribution of traits distinguish stratified lake communities from the ocean and mixed lake communities, highlighting the overall uniqueness of these environments. Specific trait values—parental care, egg strategy, trophic level, minimum and maximum temperature preferences, minimum preferred depth, body shape, mean number of dorsal spines, maximum length, and water habitat preference—common in stratified lakes could provide an advantage in colonization, environmental persistence, and niche competition. Functional diversity analyses indicate significant differences in traits that facilitate dispersal, establishment, and persistence; support exaptation via source habitats; and indicate changes in trophic webs and related morphological phenotypes between stratified lakes and the other location types. Differences between communities are supported by consistent correlations of salinity with beta diversity (βSD, βPD, and βFD), indicating its (or a co-correlate’s) influence on communities between location types. PD and FD of marine lake communities reveal patterns not observed when only species-level taxonomy is considered.
However, the SD, PD, and FD of some communities diminish the distinction between ocean and mixed lake communities and between mixed lake and stratified lake communities. The similarity in SD, PD, and FD of several mixed lake communities with the ocean communities appears in part due to the blurring effect of the outlier ocean location, LCN (Long Lake Cove, Ngeruktabel), sharing similar diversity patterns with mixed lake communities. This effect is likely driven by environmental and geographical characteristics: LCN is next to a channel with mangroves that loses overlying water during low tide, whereas most of the other ocean locations are primarily coral reefs and permanently subtidal. When LCN is removed from total beta diversity (βSD, βPD, and βFD) analyses, the differentiation between ocean and mixed lake communities increases, however, both communities are still more similar to each other than to stratified lake communities.
While stratified lake communities are generally distinct from other location types, two stratified lakes show affinities toward mixed lakes. TLN (T-Lake, Ngermeuangel) and HLM (Hot Water Lake, Mecherchar) are different from other stratified lakes in their environmental and geographical features and are similar to ocean and mixed lake communities concerning most SD, PD, and FD results. TLN is intermittently stratified with a chemocline only at the very bottom (Hamner and Hamner 1998; Suppl. material 1: fig. S26); it is the shallowest stratified lake and closest to the ocean. HLM is stratified primarily by temperature and salinity, with a 1.2 m high by 1.8 m wide tunnel at a depth of 25 meters connected to the ocean, and occasional dysoxic or anoxic conditions at the bottom (Hamner and Hamner 1998). Fish communities in TLN and HLM represent intermediate communities between stratified and mixed lakes, emphasizing the possibility of a continuum of communities between ‘complete’ and ‘habitat’ islands.
Functional diversity (FD)
Traits present in a community indicate how or whether a species can disperse to and persist in an environment (McGill et al. 2006). Fishes in ocean, mixed lake, and stratified lake communities differ in the representation of reproductive characteristics, morphologies, diets, and preferred habitats. Stratified lake communities have greater proportions and distributions of fishes with parental investment reproductive strategies that may confer benefits to eggs/larvae; more elongated and less short and deep body shapes; higher median trophic levels, with most fishes clustered around the median rather than distributed uniformly; fewer dorsal spines; euryhaline tolerance instead of only stenohaline tolerance; and lower temperature tolerances when compared to ocean and mixed lake communities. Mixed lake communities also have a greater proportion of euryhaline-tolerant fishes, higher median trophic levels, and significant differences in the lower distribution of minimum depth preferences when compared to ocean communities. Despite the inference that αFD-z is stochastic for marine lake communities, beta diversity and several traits differ predictably in richness, replacement, proportion, and distribution across location types. Trait differences between communities in different location types have the potential to explain what habitats fishes occupy based on: [1] colonization, having larvae disperse into, establish, and maintain a population in the habitat (Bradbury et al. 2008; Patzner 2008); [2] environmental persistence, maintaining homeostasis in various combinations of low to high salinity, pH, oxygen, and temperature (Kültz et al. 2015; Nay et al. 2015; Pecuchet et al. 2016; Cattano et al. 2018; Beukhof et al. 2019); and [3] niche competition, acquiring suitable food, habitat, and space and avoiding predation (Price et al. 2015; Villéger et al. 2017; Larouche et al. 2020). All three trait categories are relevant to the marine lakes, which differ in their environmental and geographical factors, impacting the fish communities residing in them.
The trait distributions of the ocean and mixed lake communities reveal mostly similar communities with redundancy in traits within each location type, but significant differences in replacement βFD, trait proportions of habitat water preference, distribution of trophic level, and distribution of modeled minimum depth between the two location types. The difference in diversity between these location types is significantly correlated with salinity, oxygen, log area, and max depth, suggesting a relationship between the environment and geography influencing community trait differences. The correlation with salinity is consistent with communities differing significantly in the proportion of species showing a habitat water preference, with mixed lake communities having higher proportions of euryhaline fishes inhabiting slightly lower salinities. The differences in minimum depth preference may relate to what environment a species inhabits and whether it can swim through the tunnels connecting the mixed lakes to the ocean and the correlated differences in max depth, but could be unrelated. Higher trophic level distribution in mixed lake communities does not clearly link to the environmental or geographical factors and is possibly due to how FishBase reports average trait data for species, rather than reporting location-specific trait data for populations. Additional research, including intrapopulation variation in traits, may elucidate the processes influencing these patterns.
The stratified lake communities are functionally sparse yet distinct as a group. Though the stratified lakes are farther from the ocean, fishes possess a greater proportion of reproductive strategies geared toward higher parental investment: displaying parental care and caring for eggs as brooders, nesters, or clutch tenders. These strategies are often associated with smaller brood sizes, larger egg size, and enhanced larval survival, increasing the colonization potential of larvae (Facey et al. 2023, pp. 400–401) and reducing the probability of larvae encountering the lethal hydrogen sulfide layer of most stratified lakes. That communities of fishes in stratified lakes, far inland relative to mixed lakes, have a greater proportion of fishes with reproductive strategies related to parental investment and shorter pelagic dispersal is reminiscent of the Paradox of Rockall, wherein a brooding gastropod populates a distant island (Johannesson 1988), indicating the primary determinant of population success may be the ability to persist. Though we only investigate the egg strategy and parental care of species, these life history strategies have been shown to influence survival of larvae (Shine 1978; Gross and Sargent 1985; Garrido et al. 2015; Dani and Kodandaramaiah 2017; Benun Sutton and Wilson 2019; Facey et al. 2023, pp. 400–401), potentially supporting fishes in successfully colonizing the stratified lakes despite the distance inland.
Several other traits were identified that support the distinctness of the stratified lake communities. The difference in mean number of dorsal spines and body shape, which have ties to defensive strategies countering predation (Price et al. 2015) and locomotion (Claverie and Wainwright 2014) may support survival and maneuverability amidst the available habitats of mangrove and rocky crevices found in these environments, or could be nonessential to the niche space they occupy in the stratified lakes. The increased frequency of shallow minimum depth preference in the stratified lakes supports fishes colonizing the stratified lakes from adjacent coastal environments. A greater proportion of fishes exhibiting euryhalinity in the stratified lakes, given the lower salinity of stratified lakes, and correlation between communities and salinity, suggest those coastal environments include estuaries and mangroves, highlighting the osmoregulatory ability of fishes as an important adaptation (Kültz et al. 2015) and an exaptation to inhabiting stratified lakes. A higher mean trophic level with reduced variance in stratified lakes may reflect greater dietary niche breadth by prey generalists in low biodiversity environments (Hayden et al. 2019); species in more biodiverse environments, like the ocean and mixed lakes, exhibit a narrower dietary niche breadth (Hayden et al. 2019). Whether temperature preferences play an important role is unclear; the warmer stratified lake environments are counter-intuitively characterized by fishes with lower modeled minimum and maximum temperatures than other location types. Conducting additional community surveys of marine lakes, determining the abundance of each species, and measuring intrapopulation variation in traits might reveal additional changes to community trait compositions, their proportions and distributions, and correlations with environmental or geographical factors in marine lakes, further illuminating the cause (s) of community trait differences. Including multiple traits related to colonization, environmental persistence, and niche competition has allowed for the discovery of traits unique to location types and trait combinations distinct within communities.
Environmental and geographical correlations with diversity
Species richness (SRic) of island communities has long been known to be driven by area, isolation, and age (Darlington 1957; Diamond 1975b; Hachich et al. 2015; Santos et al. 2016); studies exploring phylogenetic and functional diversity have identified similar significant correlations (Cardillo et al. 2008; Whittaker et al. 2008; Whittaker et al. 2014; Jacquet et al. 2017; Ferrari et al. 2023; Schrader et al. 2023). The species-area relationship slope value (z) of the well-connected and less-disturbed mixed lake communities is similar to that of islands (Rosenzweig 1995, p. 277), and substantial correlations of log-SRic and beta diversity (βSD, βPD, and βFD) with log area and max depth, the proxy we used for age, suggest mixed lakes are similar to island environments. In contrast, the stratified lake communities do not show a similar species-area relationship slope value to islands, nor do they exhibit correlations with log area or age. Although analyses including only stratified lake communities do show a significant correlation between log-SRic and minimum distance to the ocean, and nearly significant correlations for pairwise comparisons involving log-SRic of stratified lake communities and the other location types with minimum distance to the ocean. Additionally, pairwise comparisons of stratified lake communities against ocean and mixed lake communities show significant differences and substantial correlations between log-SRic, αFD-z, and beta diversity (βSD, βPD, and βFD) and salinity. The uniqueness of stratified lakes and the differences in environmental and geographical factors influencing relationships of stratified lakes with other location types may be due to their young age, relatively small size, isolation, or environmental disruptions (stratification). The youngest age of other marine islands studied is ~1 million years (Hachich et al. 2015), whereas the oldest stratified lakes formed post-Last Glacial Maximum (Dawson et al. 2009). In addition to age, the smallest size that other marine islands encompass is > 1 km2 (Hachich et al. 2015), which means even these small marine islands are likely less affected by disturbances than the still smaller marine lakes, paralleling the case for large versus small terrestrial islands (Neufeld et al. 2017). Stratified lakes are < 0.1 km2 (Colin 2009, pp. 242–243)—possibly susceptible to the small-island effect outlined by Lomolino and Weiser (2001)—and experience increased isolation with more environmental disruptions than mixed lakes. Greater distance from the mainland and smaller size negatively impact island biodiversity (MacArthur and Wilson 1967, pp. 22,181; Matthews and Triantis 2021), and environmental disturbances also disproportionately impact small terrestrial islands (Morrison 2002). The lack of a relationship in stratified lakes between log-SRic, αPD-z, αFD-z, and beta diversity (βSD, βPD, and βFD) with area and age, makes stratified lakes a potentially informative outlier, indicating a rarely studied early stage of island formation, relative to other ‘complete’ islands.
Island biogeographyComplete or habitat islands?
Do marine lakes behave more like ‘complete’ or ‘habitat’ islands? ‘True’ oceanic island communities are (1) species-poor when compared to continental communities, colonized by species with initial high dispersive abilities (Whittaker et al. 2017); (2) hotspots for speciation and endemism (Whittaker and Fernández-Palacios 2007, p. 46); (3) vulnerable to habitat changes and geological impacts (Whittaker et al. 2008); (4) isolated by an uninhabitable matrix surrounding them (Matthews 2021); and (5) inhabited by a lower diversity of competitors, making them easier to colonize (MacArthur and Wilson 1967, pp. 114–115, 122). Stratified lake communities likewise (1) are depauperate compared to the mainland ocean communities but species are equipped to colonize the available niche space; (2) have no endemic fishes existing only in stratified lakes, but have the potential for speciation as evidenced by other endemic invertebrate sub-species and populations that have been described (Dawson 2005); (3) have experienced and continue to experience habitat changes with stratification, infilling with sediment occurring in millimeters per year (Dawson et al. 2009), and climate perturbations disturbing populations (Dawson et al. 2001); (4) are surrounded by an uninhabitable matrix except for seawater that percolates through fissures and small tunnels in the carbonate substrate (Hamner and Hamner 1998; Dawson et al. 2009; Dawson 2016); and (5) individually are inhabited by only a select group of species with overall low PD and FD. Thus, stratified lake communities share many characteristics with ‘true’ oceanic island communities; their affinity to oceanic islands is closer than to other ‘true’ islands because both marine lakes and oceanic islands depend on dispersal for colonization, rather than inheriting previously established communities, as do continental fragments and continental shelf islands.
In contrast, mixed lake communities (1) are species-rich, filled with an overflow from nearby mainland ocean communities, similar to MacArthur & Wilson’s description for ‘habitat’ islands (1967; pp. 114–115); (2) have no endemic fishes specific to mixed lakes but one mixed lake does contain an endemic sub-species (Dawson and Hamner 2005); (3) are anticipated to experience minimal disturbance given the uniform mixing and normal physics and chemistry of their habitat (Hamner and Hamner 1998; Colin 2009, p. 244); (4) are connected by short habitable tunnels accessible to the ocean (Colin 2009, p. 255), a nonmaximal nor negligible contrast of habitat, as outlined by Matthews (2021); and (5) have greater diversity, challenging the establishment of would-be colonizers. Mixed lake communities seem to correspond to ‘habitat’ island communities.
Expanding island biogeography
What do marine lakes tell us about island biogeography? Most of the fish community patterns liken stratified lakes to ‘complete’ islands and mixed lakes to ‘habitat’ islands, however, several marine lakes blur this distinction. The stratified and mixed lakes may show how young islands develop and change over time and do not always follow the community relationships with area, isolation, and age. Instead, these young communities are primarily influenced by environmental factors and the traits species possess, which relate to colonization (i.e., parental care and egg strategy), environmental persistence (i.e., water habitat preference), and niche competition (i.e., trophic level and body shape). Viewing marine lakes as a gradation from ‘habitat’ to ‘complete’ islands, with overlap between location types, may be more beneficial to expanding our current understanding of islands. A common framework that accommodates ‘complete’ and ‘habitat’ islands, including marine lakes as young islands, may tell us about island biogeography generally. In addition, a biophysical perspective allows us to see that a location may be an island for one species but not for another (Adler and Levins 1994; Dawson and Hamner 2008; Dawson 2016). An island may be identified best by how organisms experience and respond to the geographical and environmental conditions (Manning et al. 2004; Matthews 2021). With this approach, the application of island biogeography to communities in marine islands has the potential to synthesize and advance island biogeography theory (Dawson 2006, 2016; Dawson et al. 2009).
Acknowledgments
We thank the following individuals for their support in bringing this work to fruition. G. Rapacciuolo provided extensive discussion, feedback, statistical insight, and R scripts. M. Carlson, M. DeBiasse, S. Fellows, K. Higgins, A. Klovrza, L. Paggeot, T. Perry, L. Schiebelhut, & N. Wutthituntisil provided feedback on draft manuscripts. I.R. da Silva & D. Toews provided enriching discussions and statistical insights. P. Colin, L. Bell, and staff at Coral Reef Research Foundation (CRRF) provided support in Palau, environmental data, and marine lake photographs. T. Stieglitz & H. Stibor provided lake bathymetry and tidal propagation information. We are grateful for the support provided by the funding agencies and institutions.
ReferencesAdler GH, Levins R (1994) The island syndrome in rodent populations. The Quarterly Review of Biology 69: 473–490. https://doi.org/10.1086/418744Allen GR, Erdmann MV (2013a) Coral reef fishes of Timor-Leste. In: Erdmann MV, Mohan C (Eds) A rapid marine biological assessment of Timor-Leste (32–82). Conservation International. https://doi.org/10.1896/054.066.0103Allen GR, Erdmann MV (2013b) Reef fish of Bali, Indonesia. In: Mustika PLK, Ratha IMJ, Purwanto S (Eds) Bali marine rapid assessment program 2011 (15–68). Conservation International. https://doi.org/10.1896/054.064.0104Anderson MJ (2006) Distance-based tests for homogeneity of multivariate dispersions. Biometrics 62: 245–253. https://doi.org/10.1111/j.1541-0420.2005.00440.xAnderson MJ (2017) Permutational multivariate analysis of variance (PERMANOVA). In: Balakrishnan N, Colton T, Everitt B, Piegorsch W, Ruggeri F, Teugels JL (Eds) Wiley StatsRef: Statistics reference online (1–15). John Wiley & Sons, Ltd. https://doi.org/10.1002/9781118445112.stat07841Ávila SP, Melo C, Berning B, Sá N, Quartau R, Rijsdijk KF, Ramalho RS, Cordeiro R, De Sá NC, Pimentel A, Baptista L, Medeiros A, Gil A, Johnson ME (2019) Towards a ‘Sea-Level Sensitive’ dynamic model: impact of island ontogeny and glacio-eustasy on global patterns of marine island biogeography. Biological Reviews 94: 1116–1142. https://doi.org/10.1111/brv.12492Ayers JF, Vacher HL (1986) Hydrogeology of an atoll island: a conceptual model from detailed study of a Micronesian example. Groundwater 24: 185–198. https://doi.org/10.1111/j.1745-6584.1986.tb00994.xBaldursson S, Ingadóttir Á (2007) Nomination of Surtsey for the UNESCO World Heritage List. Icelandic Institute of Natural History, Reykjavík.Benun Sutton F, Wilson AB (2019) Where are all the moms? External fertilization predicts the rise of male parental care in bony fishes. Evolution 73: 2451–2460. https://doi.org/10.1111/evo.13846Beukhof E, Frelat R, Pecuchet L, Maureaud A, Dencker TS, Sólmundsson J, Punzón A, Primicerio R, Hidalgo M, Möllmann C, Lindegren M (2019) Marine fish traits follow fast-slow continuum across oceans. Scientific Reports 9: 17878. https://doi.org/10.1038/s41598-019-53998-2Boettiger C, Lang DT, Wainwright PC (2012) rfishbase: exploring, manipulating and visualizing FishBase data from R. Journal of Fish Biology 81: 2030–2039. https://doi.org/10.1111/j.1095-8649.2012.03464.xBorregaard MK, Matthews TJ, Whittaker RJ (2016) Towards a unified theory of island biogeography. Global Ecology and Biogeography 25: 805–816. https://doi.org/10.1111/geb.12348Bradbury IR, Laurel B, Snelgrove PVR, Bentzen P, Campana SE (2008) Global patterns in marine dispersal estimates: the influence of geography, taxonomic category and life history. Proceedings of the Royal Society B 275: 1803–1809. http://doi.org/10.1098/rspb.2008.0216Brown JH (1971) Mammals on mountaintops: nonequilibrium insular biogeography. The American Naturalist 105: 467–478. https://doi.org/10.1086/282738Cardillo M, Gittleman JL, Purvis A (2008) Global patterns in the phylogenetic structure of island mammal assemblages. Proceedings of the Royal Society B 275: 1549–1556. http://doi.org/10.1098/rspb.2008.0262Cardoso P, Mammola S, Rigal F, Carvalho J (2024) BAT: biodiversity assessment tools. R package version 2.9.6. https://CRAN.R-project.org/package = BATCattano C, Claudet J, Domenici P, Milazzo M (2018) Living in a high CO2 world: a global meta-analysis shows multiple trait-mediated fish responses to ocean acidification. Ecological Monographs 88: 320–335. https://doi.org/10.1002/ecm.1297Cavender-Bares J, Kozak KH, Fine PVA, Kembel SW (2009) The merging of community ecology and phylogenetic biology. Ecology Letters 12: 693–715. https://doi.org/10.1111/j.1461-0248.2009.01314.xClaverie T, Wainwright PC (2014) A morphospace for reef fishes: elongation is the dominant axis of body shape evolution. PLoS ONE 9: e112732. https://doi.org/10.1371/journal.pone.0112732Colin P (2009) Marine environments of Palau. Indo-Pacific Press. https://coralreefpalau.org/wp-content/uploads/2017/04/Colin-PL-2009-Marine-Environments-of-Palau.pdfDani KG, Kodandaramaiah U (2017) Plant and animal reproductive strategies: lessons from offspring size and number tradeoffs. Frontiers in Ecology and Evolution 5: 38. https://doi.org/10.3389/fevo.2017.00038Darlington PJ (1957) Zoogeography: the geographical distribution of animals. John Wiley & Sons, Inc.Darwin C (1859) On the origin of species by means of natural selection, or the preservation of favoured races in the struggle for life. John Murray Press. https://doi.org/10.5962/bhl.title.68064Dawson MN (2005) Five new subspecies of Mastigias (Scyphozoa: Rhizostomeae: Mastigiidae) from marine lakes, Palau, Micronesia. Journal of the Marine Biological Association of the United Kingdom 85: 679–694. https://doi.org/10.1017/S0025315405011604Dawson MN (2006) Island evolution in marine lakes. Journal of the Marine Biological Association Global Marine Environment 3: 10–13.Dawson MN (2009) perspective: Trans-realm biogeography: an immergent interface. Frontiers of Biogeography 1: 62–70. https://doi.org/10.21425/F5FBG12268Dawson MN (2016) Island and island-like marine environments. Global Ecology and Biogeography 25: 831–846. https://doi.org/10.1111/geb.12314Dawson MN, Hamner WM (2005) Rapid evolutionary radiation of marine zooplankton in peripheral environments. Proceedings of the National Academy of Sciences of the USA 102: 9235–9240. https://doi.org/10.1073/pnas.0503635102Dawson MN, Hamner WM (2008) A biophysical perspective on dispersal and the geography of evolution in marine and terrestrial systems. Journal of the Royal Society Interface 5: 135–150. https://doi.org/10.1098/rsif.2007.1089Dawson MN, Martin LE, Penland LK (2001) Jellyfish swarms, tourists, and the Christ-child. Hydrobiologia 451: 131–144. https://doi.org/10.1023/A:1011868925383Dawson MN, Martin LE, Bell LJ, Patris S (2009) Marine lakes. In: Gillespie RG, Clague DA (Eds) Encyclopedia of islands (603–607). University of California Press. https://doi.org/10.1525/9780520943728-143Diamond JM (1975a) The island dilemma: lessons of modern biogeographic studies for the design of natural reserves. Biological Conservation 7: 129–146. https://doi.org/10.1016/0006-3207(75)90052-XDiamond JM (1975b) Assembly of species communities. In: Cody MM, Diamond JM (Eds) Ecology and the evolution of communities (342–444). Belknap Press of Harvard University Press.Dunn OJ (1964) Multiple comparisons using rank sums. Technometrics 6: 241–252. https://doi.org/10.2307/1266041Facey DE, Bowen BW, Collette BB, Helfman GS (2023) The diversity of fishes: biology, evolution, and ecology (3rd edn.). John Wiley & Sons, Inc.Faith DP (1992) Conservation evaluation and phylogenetic diversity. Biological Conservation 61: 1–10. https://doi.org/10.1016/0006-3207(92)91201-3Fernández-Palacios JM, Rijsdijk KF, Norder SJ, Otto R, de Nascimento L, Fernández-Lugo S, Tjørve E, Whittaker RJ (2016) Towards a glacial-sensitive model of island biogeography. Global Ecology and Biogeography 25: 817–830. https://doi.org/10.1111/geb.12320Ferrari DS, Floeter SR, Leprieur F, Quimbayo JP (2023) A trait-based approach to marine island biogeography. Journal of Biogeography 50: 528–538. https://doi.org/10.1111/jbi.14549Fisher RA (1922) On the interpretation of χ2 from contingency tables, and the calculation of p. Journal of the Royal Statistical Society 85: 87–94. https://doi.org/10.2307/2340521Fox J, Weisberg S (2019) An R companion to applied regression (3rd edn.). Sage. https://www.john-fox.ca/Companion/Froese R, Pauly D [Eds] (2024) FishBase. World Wide Web Electronic Publication. www.fishbase.org [version (10/2024)]Fukami T (2015) Historical contingency in community assembly: integrating niches, species pools, and priority effects. Annual Review of Ecology, Evolution, & Systematics 46: 1–23. https://doi.org/10.1146/annurev-ecolsys-110411-160340Gallien L (2017) Intransitive competition and its effects on community functional diversity. Oikos 126: 615–623. https://doi.org/10.1111/oik.04033Garrido S, Santos AM, Ferreira S, Teodósio MA, Cotano U, Irigoien X, Peck MA, Saiz E, Ré P (2015) Born small, die young: intrinsic, size-selective mortality in marine larval fish. Scientific Reports 5: 17065. https://doi.org/10.1038/srep17065Google Earth, version 7.3.6.9345 (2019–2021) Google Earth. https://www.google.com/earth/versions/Gower JC (1971) A general coefficient of similarity and some of its properties. Biometrics 27: 857–871. https://doi.org/10.2307/2528823Gross MR, Sargent RC (1985) The evolution of male and female parental care in fishes. Integrative and Comparative Biology 25: 807–822. https://doi.org/10.1093/icb/25.3.807Hachich NF, Bonsall MB, Arraut EM, Barneche DR, Lewinsohn TM, Floeter SR (2015) Island biogeography: patterns of marine shallow-water organisms in the Atlantic Ocean. Journal of Biogeography 42: 1871–1882. https://doi.org/10.1111/jbi.12560Hachich NF, Silva Ferrari D, Quimbayo JP, Pinheiro HT, Floeter SR (2020) Island biogeography of marine shallow-water organisms. In: Goldstein MI, DellaSala DA (Eds) Encyclopedia of the world’s biomes. Elsevier, 61–75. https://doi.org/10.1016/B978-0-12-409548-9.11947-5Hamner WM, Hamner PP (1998) Stratified marine lakes of Palau (Western Caroline Islands). Physical Geography 19: 175–220. https://doi.org/10.1080/02723646.1998.10642647Hayden B, Palomares MLD, Smith BE, Poelen JH (2019) Biological and environmental drivers of trophic ecology in marine fishes - a global perspective. Scientific Reports 9: 11415. https://doi.org/10.1038/s41598-019-47618-2Heaney LR (2000) Dynamic disequilibrium: a long-term, large-scale perspective on the equilibrium model of island biogeography. Global Ecology and Biogeography 9: 59–74. https://doi.org/10.1046/j.1365-2699.2000.00163.xHubbell SP (2001) The unified neutral theory of biodiversity and biogeography. Princeton University Press.Jaccard P (1912) The distribution of flora in the alpine zone. New Phytologist 11: 37–50. https://doi.org/10.1111/j.1469-8137.1912.tb05611.xJacquet C, Mouillot D, Kulbicki M, Gravel D (2017) Extensions of island biogeography theory predict the scaling of functional trait composition with habitat area and isolation. Ecology Letters 20: 135–146. https://doi.org/10.1111/ele.12716Johannesson K (1988) The paradox of Rockall: why is a brooding gastropod (Littorinasaxatilis) more widespread than one having a planktonic larval dispersal stage (L.littorea)? Marine Biology 99: 507–513. https://doi.org/10.1007/BF00392558Kembel SW, Cowan PD, Helmus MR, Cornwell WK, Morlon H, Ackerly DD, Blomberg SP, Webb CO (2010) Picante: R tools for integrating phylogenies and ecology. Bioinformatics 26: 1463–1464. https://doi.org/10.1093/bioinformatics/btq166Kruskal JB (1964) Multidimensional scaling by optimizing goodness of fit to a nonmetric hypothesis. Psychometrika 29: 1–27. https://doi.org/10.1007/BF02289565Kruskal WH, Wallis WA (1952) Use of ranks in one-criterion variance analysis. Journal of the American Statistical Association 47: 583–621. https://doi.org/10.1080/01621459.1952.10483441Kültz D, Podrabsky JE, Stillman JH, Tomanek L (2015) Physiological mechanisms used by fish to cope with salinity stress. Journal of Experimental Biology 18: 1907–1914. https://doi.org/10.1242/jeb.118695Kuhn M (2008) Building predictive models in R using the caret package. Journal of Statistical Software 28: 1–26. https://doi.org/10.18637/jss.v028.i05Larouche O, Benton B, Corn KA, Friedman ST, Gross D, Iwan M, Kessler B, Martinez CM, Rodriguez S, Whelpley H, Wainwright PC, Price SA (2020) Reef-associated fishes have more maneuverable body shapes at a macroevolutionary scale. Coral Reefs 39: 1427–1439. https://doi.org/10.1007/s00338-020-01976-wLomolino MV, Weiser MD (2001) Towards a more general species–area relationship: diversity on all islands, great and small. Journal of Biogeography 28: 431–445. https://doi.org/10.1046/j.1365-2699.2001.00550.xMacArthur RH, Wilson EO (1967) The theory of island biogeography. Princeton University Press. https://doi.org/10.2307/1796430Manning AD, Lindenmayer DB, Nix HA (2004) Continua and umwelt: novel perspectives on viewing landscapes. Oikos 104: 621–628. https://doi.org/10.1111/j.0030-1299.2004.12813.xMantel N (1967) The detection of disease clustering and a generalized regression approach. Cancer Research 27: 209–220.Martinez Arbizu P (2020) pairwiseAdonis: pairwise multilevel comparison using adonis. R package version 0.4Matthews TJ (2021) On the biogeography of habitat islands: the importance of matrix effects, noncore species, and source-sink dynamics. The Quarterly Review of Biology 96: 73–104. https://doi.org/10.1086/714482Matthews TJ, Triantis K (2021) Island biogeography. Current Biology 31: 1201–1207. https://doi.org/10.1016/j.cub.2021.07.033McClain CR (2007) Seamounts: identity crisis or split personality? Journal of Biogeography 34: 2001–2008. https://doi.org/10.1111/j.1365-2699.2007.01783.xMcGill BJ, Enquist BJ, Weiher E, Westoby M (2006) Rebuilding community ecology from functional traits. Trends in Ecology & Evolution 21: 178–185. https://doi.org/10.1016/j.tree.2006.02.002McTavish EJ, Sanchez-Reyes LL (2022) Chronosnyth. https://github.com/OpenTreeOfLife/chronosynthMeyerhof MS, Wilson JM, Dawson MN, Beman MJ (2016) Microbial community diversity, structure and assembly across oxygen gradients in meromictic marine lakes, Palau. Environmental Microbiology 18: 4907–4919. https://doi.org/10.1111/1462-2920.13416Morrison LW (2002) Determinants of plant species richness on small Bahamian islands. Journal of Biogeography 29: 931–941. https://doi.org/10.1046/j.1365-2699.2002.00730.xMouillot D, Villéger S, Parravicini V, Kulbicki M, Ernesto J, Bender M, Chabanet P, Floeter SR, Friedlander A, Vigliola L, Bellwood DR (2014) Functional over-redundancy and high functional vulnerability in global fish faunas on tropical reefs. Proceedings of the National Academy of Sciences of the USA 111: 13757–13762. https://doi.org/10.1073/pnas.1317625111Myers RF (1999) Micronesian reef fishes: a comprehensive guide to the coral reef fishes of Micronesia (3rd edn.). Coral Graphics.Nay TJ, Johansen JL, Habary A, Steffensen JF, Rummer JL (2015) Behavioural thermoregulation in a temperature-sensitive coral reef fish, the five-lined cardinalfish (Cheilodipterusquinquelineatus). Coral Reefs 34: 1261–1265. https://doi.org/10.1007/s00338-015-1353-4Neufeld CJ, Starko SK, Burns C (2017) Disturbance and diversity in a continental archipelago: a mechanistic framework linking area, height, and exposure. Ecosphere 8: e01957. https://doi.org/10.1002/ecs2.1957OED (2026) Oxford English Dictionary: Island. https://www.oed.com/dictionary/island_n?tab = meaning_and_use#1795 [Accessed 15 April 2026]Oksanen J, Simpson G, Blanchet F, Kindt R, Legendre P, Minchin P, O’Hara R, Solymos P, Stevens M, Szoecs E, Wagner H, Barbour M, Bedward M, Bolker B, Borcard D, Carvalho G, Chirico M, De Caceres M, Durand S, Evangelista H, FitzJohn R, Friendly M, Furneaux B, Hannigan G, Hill M, Lahti L, McGlinn D, Ouellette M, Ribeiro Cunha E, Smith T, Stier A, Ter Braak C, Weedon J (2022) vegan: community ecology package. R package version 2.6–4. https://CRAN.R-project.org/package = veganParadis E, Schliep K (2019) ape 5.0: an environment for modern phylogenetics and evolutionary analyses in R. Bioinformatics 35: 526–528. https://doi.org/10.1093/bioinformatics/bty633Patzner RA (2008) Reproductive strategies of fish. In: Rocha MJ, Arukwe A, Kapoor BG (Eds) Fish reproduction (311–350). Taylor & Francis Group.Pecuchet L, Törnroos A, Lindegren M (2016) Patterns and drivers of fish community assembly in a large marine ecosystem. Marine Ecology Progress Series 546: 239–248. https://doi.org/10.3354/meps11613Petchey OL, Gaston KJ (2002) Functional diversity (FD), species richness and community composition. Ecology Letters 5: 402–411. https://doi.org/10.1046/j.1461-0248.2002.00339.xPinheiro HT, Bernardi G, Simon T, Joyeux JC, Macieira RM, Gasparini JL, Rocha C, Rocha LA (2017) Island biogeography of marine organisms. Nature 549: 82–85. https://doi.org/10.1038/nature23680Price SA, Friedman ST, Wainwright PC (2015) How predation shaped fish: the impact of fin spines on body form evolution across teleosts. Proceedings of the Royal Society B 282: 20151428. http://doi.org/10.1098/rspb.2015.1428R Core Team (2025) R: A language and environment for statistical computing. R Foundation for Statistical Computing. https://www.R-project.org/Rapacciuolo G, Beman JM, Schiebelhut LM, Dawson MN (2019) Microbes and macro-invertebrates show parallel β-diversity but contrasting α-diversity patterns in a marine natural experiment. Proceedings of the Royal Society B 286: 20190999. https://doi.org/10.1098/rspb.2019.0999Redelings B, Sanchez-Reyes LL, Cranston KA, Allman J, Holder MT, McTavish EJ (2019) Open tree of life synthetic tree (12.3). Zenodo. https://doi.org/10.5281/zenodo.3937742Revell LJ (2012) phytools: an R package for phylogenetic comparative biology (and other things). Methods in Ecology and Evolution 3: 217–223. https://doi.org/10.1111/j.2041-210X.2011.00169.xRosenzweig ML (1995) Species diversity in space and time. Cambridge University Press. https://doi.org/10.1017/CBO9780511623387Sachs JP, Blois JL, McGee T, Wolhowe M, Haberle S, Clark G, Atahan P (2018) Southward shift of the Pacific ITCZ during the Holocene. Paleoceanography and Paleoclimatology 33: 1383–1395. https://doi.org/10.1029/2018PA003469Santos AMC, Field R, Ricklefs RE (2016) New directions in island biogeography. Global Ecology and Biogeography 25: 751–768. https://doi.org/10.1111/geb.12477Schoener A (1974a) Colonization curves for planar marine islands. Ecology 55: 818–827. https://doi.org/10.2307/1934417Schoener A (1974b) Experimental zoogeography: colonization of marine mini-islands. The American Naturalist 108: 715–738. https://doi.org/10.1086/282951Schrader J, Wright IJ, Kreft H, Westoby M (2021) A roadmap to plant functional island biogeography. Biological Reviews 96: 2851–2870. https://doi.org/10.1111/brv.12782Schrader J, Wright IJ, Kreft H, Weigelt P, Andrew SC, Abbott I, Westoby M (2023) ETIB-T: an equilibrium theory of island biogeography for plant traits. Journal of Biogeography 50: 223–234. https://doi.org/10.1111/jbi.14526Shine R (1978) Propagule size and parental care: the “safe harbor” hypothesis. Journal of Theoretical Biology 75: 417–424. https://doi.org/10.1016/0022-5193(78)90353-3Smith GB (1979) Relationship of eastern Gulf of Mexico reef-fish communities to the species equilibrium theory of insular biogeography. Journal of Biogeography 6: 49–61. https://doi.org/10.2307/3038151Smittenberg RH, Saenger C, Dawson MN, Sachs JP (2011) Compound-specific D/H ratios of the marine lakes of Palau as proxies for West Pacific warm pool hydrologic variability. Quaternary Science Reviews 30: 921–933. https://doi.org/10.1016/j.quascirev.2011.01.012Thomson DA, Gilligan MR (2002) Rocky-shore fishes. In: Case TJ, Cody ML, Ezcurra E (Eds) A new island biogeography of the Sea of Cortés (154–180). Oxford University Press. https://doi.org/10.1093/oso/9780195133462.003.0013Thornton IWB (2007) Island colonization: the origin and development of island communities (New TR, Ed.). Cambridge University Press. https://doi.org/10.1017/CBO9780511618710Tukey JW (1977) Exploratory data analysis. Addison-Wesley.Utne-Palm AC, Salvanes AGV, Currie B, Kaartvedt S, Nilsson GE, Braithwaite VA, Stecyk JAW, Hundt M, Flynn B, Sandvik GK, Klevjer TA, Sweetman AK, Brüchert V, Pittman K, Peard KR, Lunde IG, Strandabø RAU, Gibbons MJ (2010) Trophic structure and community stability in an overfished ecosystem. Science 329: 333–336. https://doi.org/10.1126/science.1190708Vellend M (2010) Conceptual synthesis in community ecology. The Quarterly Review of Biology 85: 183-206. https://doi.org/10.1086/652373Vellend M (2016) The theory of ecological communities (Monographs in Population Biology 57). Princeton University Press. https://doi.org/10.1515/9781400883790Venables WN, Ripley BD (2002) Modern applied statistics with S (4th edn.). Springer. https://www.stats.ox.ac.uk/pub/MASS4/Villéger S, Miranda JR, Hernández DF, Mouillot D (2010) Contrasting changes in taxonomic vs. functional diversity of tropical fish communities after habitat degradation. Ecological Applications 20: 1512–1522. https://doi.org/10.1890/09-1310.1Villéger S, Brosse S, Mouchet M, Mouillot D, Vanni MJ (2017) Functional ecology of fish: current approaches and future challenges. Aquatic Sciences 79: 783–801. https://doi.org/10.1007/s00027-017-0546-zWallace AR (1880) Island life. or, the phenomena and causes of insular faunas and floras, including a revision and attempted solution of the problem of geological climates. Macmillan and Co. https://doi.org/10.5962/bhl.title.98129Webb CO, Ackerly DD, McPeek MA, Donoghue MJ (2002) Phylogenies and community ecology. Annual Review of Ecology, Evolution, and Systematics 33: 475–505. https://doi.org/10.1146/annurev.ecolsys.33.010802.150448Weigelt P, Jetz W, Kreft H (2013) Bioclimatic and physical characterization of the world’s islands. Proceedings of the National Academy of Sciences of the USA 110: 15307–15312. https://doi.org/10.1073/pnas.1306309110Weiher E, Keddy PA (1995) Assembly rules, null models, and trait dispersion: new questions from old patterns. Oikos 74: 159–164. https://doi.org/10.2307/3545686Whittaker RJ, Bush MB, Richards K (1989) Plant recolonization and vegetation succession on the Krakatau Islands, Indonesia. Ecological Monographs 59: 59–123. https://doi.org/10.2307/2937282Whittaker RJ, Fernández-Palacios JM (2007) Island biogeography: ecology, evolution, and conservation. Oxford University Press. https://doi.org/10.1093/oso/9780198566113.001.0001Whittaker RJ, Triantis KA, Ladle RJ (2008) A general dynamic theory of oceanic island biogeography. Journal of Biogeography 35: 977–994. https://doi.org/10.1111/j.1365-2699.2008.01892.xWhittaker RJ, Rigal F, Borges PAV, Cardoso P, Terzopoulou S, Casanoves F, Pla L, Guilhaumon F, Ladle RJ, Triantis KA (2014) Functional biogeography of oceanic islands and the scaling of functional diversity in the Azores. Proceedings of the National Academy of Sciences of the USA 111: 13709–13714. https://doi.org/10.1073/pnas.1218036111Whittaker RJ, Fernández-Palacios JM, Matthews TJ, Borregaard MK, Triantis KA (2017) Island biogeography: taking the long view of nature’s laboratories. Science 357: eaam8326. https://doi.org/10.1126/science.aam8326Wickham H (2016) ggplot2: Elegant graphics for data analysis (3rd Ed.). Springer.Yan L (2023) ggvenn: draw venn diagram by ‘ggplot2’. R package version 0.1.10. https://CRAN.R-project.org/package = ggvennZimmerman BL, Bierregaard RO (1986) Relevance of the equilibrium theory of island biogeography and species-area relations to conservation with a case from Amazonia. Journal of Biogeography 13: 133–143. https://doi.org/10.2307/2844988Additional informationConflict of interest
The authors have declared that no competing interests exist.
Ethical statement
No ethical statement was reported.
Artificial Intelligence (AI) use
The authors accept full responsibility for the content of the manuscript, including the disclosure of any use of AI. Artificial intelligence of ChatGPT with GPT-4 was used to refine ggplot2 code for differentiating points of the trait plots in R v.4.4.3 (R Core Team 2025).
Funding
This work was conducted under permits from the Ministry of Natural Resources, Environment, and Tourism (RE-13-11) and Koror State Government (#13-233), Palau. The project was funded by the National Science Foundation programs in Dimensions of Biodiversity (OCE-1241255) and Infrastructure Innovation for Biological Research (DBI-1927520) and by the David & Lucile Packard Foundation (#2012-38207).
Author contributions
Bailey J. Carlson: Formal analysis, Investigation, Data Curation, Writing - Original draft, Writing - Review and Editing, Visualization. Rob F. Myers: Methodology, Data Curation, Writing - Review and Editing. Michael N Dawson: Conceptualization, Methodology, Data Curation, Supervision, Project administration, Funding Acquisition, Writing - Original draft, Writing - Review and Editing.