The São Tomé-Príncipe Archipelago is located along the Cameroon Volcanic Line at 243 km off the continental West African coast in the Gulf of Guinea (Canterle et al., 2020) (Fig. 1; Table 1). It consists of two main islands, São Tomé and Príncipe, with a coastline extension of 185 km and total land area of 1042 km². Biogeographically, the archipelago is part of the Guinea Current Large Marine Ecosystem extending from Guinea Bissau to Angola within the Tropical East Atlantic (TEA) province (Fajemila and Langer, 2017). It is influenced by three incoming currents, namely the Gulf of Guinea Current from North, the Benguela Current from South, and the easterly flowing Equatorial Counter Current (Da Costa et al., 2022). The convergence of ocean currents causes seasonal equatorial upwelling and dominates regional productivity (Friedlander et al., 2014). Freshwater and particulate discharge from the main rivers (Ogooué and Congo) leads to great spatial-temporal variations in salinity and turbidity in coastal waters (Friedlander et al., 2014). Steep environmental gradients occur across the upper water column, with temperature reaching typical tropical ranges of 25–29 °C at surface while dropping to ∼ 20 °C below a constant thermocline at depths of 20–30 m, and likewise for salinity and nutrient content among other parameters (Maia et al., 2018). As constrained by regional hydrological conditions, no extensive matrix of true coral reefs exists in the Gulf of Guinea including STP, instead the benthic habitats are characterized by a mosaic of scattered rocky reefs colonized by various hard corals, turf algae, macroalgae, and sponges (Otero-Ferrer et al., 2020). Mangroves and seagrass beds are the other two major habitats along the islands' coast. In terms of substratum, a large proportion of the coastal area is covered by dark-colored volcanic sands and secondarily stable quartz sands and calcareous bioclastic sands.

The Zanzibar Archipelago is situated 49 km away from the Tanzania mainland along the East African coast (Painter, 2020) (Fig. 1; Table 1). The three main islands, Pemba, Unguja, and Mafia measure a total land area of 3002 km² and a coastline of 767 km. The archipelago belongs to the Somali Coastal Current Large Marine Ecosystem that stretches from Somalia to the northeastern coast of South Africa in the western Indian Ocean (WIO) province (Thissen and Langer, 2017). Major currents influencing this region include the westward-flowing South Equatorial Current and the northward-flowing East African Coastal Current (Painter, 2020). The tropical monsoonal climate regulates annual temperature variation between 25–29 °C with humid and dry seasons (Mahongo and Shaghude, 2014). The Zanzibar islands possess one of the largest reef areas along the coast of East Africa, but many local reefs are in early-middle stages of degradation because of increasing marine pollution and urbanization, especially in heavily populated Stone Town areas (Bravo et al., 2021; Grimsditch et al., 2009; Larsen et al., 2023). Shallow fringing reefs and deep fore reefs are the most common habitat, followed by vegetated sand flats and mangroves. The substratum is primarily calcareous bioclastic sands in fine to medium grain size with varying amounts of reef rubble, or otherwise fine quartz sands.

In total ten surface sediment samples were collected from the São Tomé and Príncipe islands in addition to one sample along continental coast in Libreville directly east of the archipelago (Fig. 1). Sample ID is coded with site abbreviations plus collection number. These samples cover a depth range of 0.5–30 m across the tidal and subtidal zones and represent the habitat types of marginal fringing reefs, sand flats, and mangroves. Samples were collected by scuba diving to scrape along the seabed and fill plastic containers with top 2 cm of the surface sediments, in order to avoid the loss of finer particles due to suspension. In the laboratory, sediments were washed through a 63 µm sieve, oven dried at 50 °C, and dry sieved over a 150 µm mesh sieve. Subfossil ostracods were picked from the > 150 µm size fraction and a single valve or a carapace was treated as one individual, which is the standard method in ostracod research (Boomer et al., 2003). Specimens preserved with soft parts (live) were counted together with the empty ones (dead) to make up the total, time-averaged assemblage, which is proven to effectively define benthic habitats (Tian et al., 2024a). In the next step, we integrated the census count of STP ostracods with previously published ostracod data from Zanzibar after rigorous taxonomic standardization (Tian et al., 2024a). Published foraminifera data from the two regions were integrated in the same way so that we build a trans-regional, multi-proxy, large-size dataset of benthic meiofaunas (Fajemila and Langer, 2017; Thissen and Langer, 2017).

To evaluate the biogeography of STP and more generally the TEA, we compiled species occurrence data of the Recent and sub-Recent shallow-marine ostracods for five tropical-subtropical provinces in the Atlantic Ocean through extensive literature search (Table 2; See Table S1 in the Supplement and supplementary data for the complete species occurrence list and literature cited). The geographic delineation of each province follows Floeter et al. (2008) and Le Lœuff and Von Cosel (1998). Ostracod fauna of each province was compared for the number of shared species as indication of biogeographic link. The compiled species list is not exhaustive and many areas are poorly studied for ostracods, which leads to unavoidable sampling bias. Nevertheless, it is believed that such compilation gives an overview of the biogeographic relationship among provinces and hints at the underlying evolutionary process.

For ostracod diversity measures, we used Hill numbers (i.e., the effective number of equally abundant species) parameterized by a diversity order q (Chao et al., 2014a). The order q determines how much weight is given to the relative abundance of species. Specifically, the Hill number (0D) reduces to species richness for q= 0; the Hill number (1D) measures the diversity of the abundant species for q= 1; and the Hill number (2D) measures the diversity of dominant species for q= 2 (Chao et al., 2014a). To tackle the problem of unequal sample efforts among ostracod assemblages and datasets, we standardized the Hill numbers with rarefaction or extrapolation to the largest sample completeness possible for alpha diversity across samples (82.5 %) and gamma diversity across regions (96.7 %) (Chao et al., 2014b). Multiplicative beta diversity was calculated as gamma diversity divided by alpha per sample, which quantifies the extent of among-assemblage differentiation in faunal composition (Chao et al., 2023). The mean and 95 % confidence intervals of the Hill numbers were estimated by bootstrap resampling with 100 repetitions. Species evenness was computed based on the slope of the Hill number profiles as a function of order q (Chao and Ricotta, 2019). In an even assemblage, the species richness and number of abundant and dominant species are similar, resulting in a more gentle slope. In contrast, an uneven assemblage is dominated by one or a few species, leading to a steeper slope. The species evenness or the normalized slopes of Hill number profiles were computed at orders q= 1 (1E) and q= 2 (2E).

We used the Generalized Additive Mixed-effect Model (GAMM) to investigate effective environmental controls of ostracod alpha diversity in each region. The environmental variables used in GAMM include habitat topographic type, algae coverage, and sediment type as habitat factors, along with human impact, water depth, and distance to land as physical factors. Each of these factors has been shown to influence the diversity distribution of benthic organisms (Otero-Ferrer et al., 2020; Porriños et al., 2024). All variables were measured on site except for distance to land, which was computed based on the distance between sampling coordinates and shoreline data from https://www.naturalearthdata.com (last access: 7 October 2025). The GAMM used penalized cubic regression spline smooths with restricted maximum likelihood (REML) method (Wood, 2024). In addition to water depth and distance to land as numeric variables, habitat type (fore reef, fringing reef, back reef, sand flat, and mangrove) and sediment type (bioclastic sand, fine-grained sand, and volcanic sand) were treated as categorical variables. Algae coverage and human impact were handled as ordinal variables with three levels (i.e., low, medium, and high). GAMM estimated the fixed effects of these environmental factors on the diversity patterns. It also incorporated a random factor of island by smooths, as penalized regression terms. The evaluation of random effects helped to distinguish whether differences in biodiversity are due to site-specific conditions or are more uniformly affected by fixed factors (Wood, 2024). The comparison and ranking of GAMM models were based on AICc, which is an adjustment of the standard Akaike Information Criterion (AIC) incorporating a correction for small sample sizes. Relative model support was measured by the Akaike Weights (weight) (Anderson et al., 2000), with a higher value denoting a better fit to the data for a given number of model parameters. The parameter estimates were averaged across all candidate models weighted by their relative support. This approach accounted for uncertainty in model selection and provided appropriate confidence intervals (Anderson et al., 2000). The relative importance of a predictor variable was then determined by summing the Akaike weights of all the models in the candidate set in which that specific predictor variable occurred. The summarized top model was validated for normality and homogeneity. The spatial autocorrelation in model residues was examined by Moran's I statistic with a permutation test. No significant spatial autocorrelation was detected in the top model.

To evaluate faunal variation among ostracod assemblages, we conducted hierarchical cluster analysis based on Ward's minimum variance and Hill-number-based dissimilarity indices (1-CqN), where N indicates the number of assemblages for comparison and order q determines the indices' weight on relative abundance (Chao et al., 2014a). Depending on the parameter q, the Sørensen (q= 0), Horn (q= 1), and Morisita–Horn (q= 2) indices measure compositional dissimilarity in terms of species presence-absence, abundant species, and dominant species, respectively (Chao et al., 2014a). The optimal number of clusters was determined at which the average silhouette width is highest, indicating cohesion within a cluster and separation between clusters. We identified the top 10 indicator species of each cluster based on the Indicator Value (IndVal), as defined by (Cáceres and Legendre, 2009; Dufrêne and Legendre, 1997). Specifically, the IndVal was calculated as the square root of the product of relative abundance (i.e., specificity) and relative frequency (i.e., fidelity) of a species present in the defined group. The faunal composition of each sample was illustrated with a heat map showing the abundance (species count after applying a fourth root transformation) of the top 10 indicator species of each cluster, with the relationship between species determined by Hellinger distances. To examine faunal compositional changes across environmental gradients in each region, we performed non-metric multidimensional scaling (nMDS) based on the Sørensen, Horn, and Morisita–Horn dissimilarity indices and calculated the correlations between environmental variables and MDS axes with permutation tests. We also used Distance-based Redundancy Analysis (dbRDA) to measure the variance explained by each environmental variable and thus identify significant determinants of faunal structure.

All statistical analyses were replicated on the foraminifera data in exactly the same manner. Samples of low abundance (< 50 individuals) were excluded from all quantitative analyses. All analyses were implemented in RStudio (Posit team, 2025) using tidyverse, iNEXT, mgcv, MuMIn, hillR, indicspecies, pheatmap, and vegan packages.

First of all, ostracods and foraminifera concordantly show exceptionally high diversity in Zanzibar in contrast to low diversity in STP at both local (alpha) and regional (gamma) scales. This mirrors broader patterns of species richness in the WIO versus TEA based on other benthic groups (Cowman et al., 2017; Tittensor et al., 2010), confirming the usefulness of meiofaunal proxies in macroecological studies. The impoverishment of STP fauna is likely due to environmental and historical constraints acting in conjunction. In the TEA along the west African coast, cold boundary currents and seasonal coastal upwelling restrict the geographic range of true tropical segments (Da Costa et al., 2022) and also inhibit the growth of large coral reefs where high benthic diversity is usually located (Friedlander et al., 2014). Moreover, the coastline is mostly straight, and the continental shelf is very narrow and dominated with monotonic sandy and muddy bottoms (Friedlander et al., 2014; Polidoro et al., 2017). In fact, the TEA has the smallest shelf area of the world's main tropical regions (Polidoro et al., 2017). Small habitat areas may not support high biodiversity as predicted by the species-area relationship (Losos and Schluter, 2000), and low habitat complexity limits the potential of allopatric speciation (Bellwood et al., 2012). Over the evolutionary history of the TEA, increasing isolation had made it difficult to receive species from other regions, especially the biodiversity hotspots in Caribbean and broadly Indo-Pacific realm (Floeter et al., 2008; Le Lœuff and Von Cosel, 1998). All these environmental and biogeographic factors disfavor the origin and preservation of high biodiversity, and together they may account for the depauperation of the TEA province and subsequently the low diversity in STP, since the oceanic islands are integral components of larger biogeographic regions.

In contrast to an explicit, unambiguous pattern of regional differences in alpha and gamma diversity, beta diversity of STP and Zanzibar shows inconsistent variations across orders q for ostracods and foraminifera, which implies some ecological distinction between two groups and different mechanisms underlying their community assembly (Fig. 2). For ostracods, across the regional diversity gradient, it appears that the richness of local assemblage approximately scales to the size of regional species pool (i.e., similar levels of beta diversity at q= 0 in two regions), indicating that the ostracod communities are regionally enriched (DeVantier et al., 2020). Strong evidence for regional enrichment of biodiversity has been found in corals and marine epifaunal invertebrates in general, and it is widely acknowledged that local diversity is shaped by processes operating on larger spatial scales (DeVantier et al., 2020; Witman et al., 2004). At order q= 2, however, elevated beta diversity of dominant species in Zanzibar may reflect a higher habitat diversity there, as the dominant species are usually well-adapted to certain habitats. The transitions from mature to marginal reefs and eventually to non-reefs across a large environmental gradient correspond to fundamental shifts in the ostracod composition in terms of well-adapted dominant species. Unexpectedly for foraminifera across the two regions, beta diversity is higher in STP at orders q= 0 and q= 1 with a comparatively large regional species pool yet low local richness. Instead of a saturation effect on local assemblages, i.e., biotic interactions limit the number of species that may coexist locally (DeVantier et al., 2020), it is more likely that certain environmental filtering determines the spatial patterns of species occurrence as many species have narrow and specific habitat ranges in STP (Fajemila and Langer, 2017). For example, Ammonia cf. A. aoteana is exclusively found in the sand flat of P4 while Glabratellina sp.1 is specific to the sand flat of P2 (Fig. 6B). Strict environmental structuring of foraminifera assemblages for rare and abundant species thus translates to profound changes in species composition across sites, which collectively add up to a comparatively large regional species pool in STP. The scenario in Zanzibar is quite the opposite that low beta diversity at orders q= 0 and q= 1 is accounted by homogenous species distributions among most of the reefal habitats. Finally at order q= 2, STP and Zanzibar foraminifera reach similar levels of beta diversity with each cluster dominated by few well-adapted species (e.g., Neorotalia calcar in C2, Pararotalia cf. P. nipponica in C4, and Amphistegina gibbosa in C5) (Fig. 6B) (Fajemila and Langer, 2017; Thissen and Langer, 2017). Comparing the beta diversity patterns of the two taxonomic groups, our results tentatively indicate that the local-regional relation of biodiversity may be twisted by environmental conditions for different organisms depending on their ecology.

The GAMM modelling reveals that environmental effects on local diversity vary between the two organisms and between two regions (Table 3). In Zanzibar, ostracod local diversity is regulated by habitat type as the fore and fringing reefs with high topographic complexity support more diverse assemblages compared to marginal reefs, mangroves, and featureless soft bottom habitats (Fig. 5A). Diversity is especially high on some fringing reefs with medium algae coverage (e.g., RNs and CBs) (Fig. 3A), likely because interlaced hard corals, macro algae, and turf algae offer diverse and heterogenous microhabitats to accommodate different ecological groups (Tian et al., 2024a). It is intriguing that the local diversity of foraminifera in Zanzibar responds nonlinearly to the algae factor, with the lowest diversity at medium coverage (e.g., MAs and RNPs) (Figs. 4A and 5B). In fact, many sites with medium algae coverage can be characterized as transitional environments between hard and soft bottoms, where the sediments are homogenous, medium-grained bioclastic sands. These sites are idiosyncratic in terms of not only diversity but also composition (e.g., the MAs form its own foraminifera cluster C2 and have very low evenness) (Fig. 6B). Neither reefal nor phytal foraminifera flourish in such environments and the assemblages are dominated by a few opportunistic species, so that local diversity records the lowest level. This is in sharp contrast to the ostracod pattern that various ecological groups overlap in algae-covered reefal habitats to achieve high diversity, which conforms to the classic intermediate disturbance hypothesis (Townsend et al., 1997; Viljur et al., 2022). A possible explanation for foraminifera being a contrarian is that they may have high ecological specialization and specific habitat requirements, which make them avoid transitional algae-covered environments. Then, in STP, the local diversity of both groups does not evidently follow any environmental regulations tested (Fig. 5). Diversity distribution is essentially uniform for ostracods while site-specific for foraminifera. As the seascape is predominantly sandy and muddy bottoms along the west African coast (Friedlander et al., 2014), small fringing reefs in STP may provide one of the few hard substrates in this region for reef organisms yet they lack the topographic complexity required to host high diversity like true tropical coral reef ecosystems. Consequently, there are no locally diverse benthic assemblages in STP.

The dbRDA analysis demonstrates that environmental controls over faunal composition are highly concordant for the two organisms in two regions, with habitat type being of overwhelming importance (Table 4). Reefal and non-reefal assemblages are fundamentally divergent as they show widest separation in cluster and nMDS analysis (Figs. 6–7). Among various types of reefal habitats, algae coverage further differentiates local assemblage compositions, as the pristine Pemba reefs dominated by hard corals (ostracod cluster C3) have lower abundance of phytal species as compared to all other moderately algae-covered reefs. The effects of human impact are weaker and mostly apparent at the highest level of disturbance, as the Stone Town sites (ostracod cluster C4) are characterized by some Trachyleberididae genera (e.g., Actinocythereis) but their ecological significance is not well understood in this case (Tian et al., 2024a). Thus, the delineation of ostracod clusters directly and specifically reflects the interacting effects of habitat, algae, and human factors. With regard to foraminifera, apart from aforementioned environmental drivers, distance to land also explains a minor proportion of faunal variance in Zanzibar, as indicated by the dominance of opportunistic taxa (e.g., Ammonia convexa) in nearshore lagoonal and mangrove sites (foraminifera cluster C3) (Thissen and Langer, 2017).

Despite distinct taxonomic compositions of the STP and Zanzibar faunas at species and even genus level, there is a marked inter-regional convergence in the ecological structure of ostracod reefal and non-reefal assemblages in terms of the proportions of coral, phytal, bottom-dwelling, and euryhaline taxa (Fig. 8). The fringing reef cluster (C5) in STP and pristine fore reef cluster (C3) in Zanzibar both show highest abundance of coral affiliated taxa, while the other two reefal clusters with algae cover (C1 and C4) in Zanzibar have comparatively more phytal taxa besides coral taxa. Regarding the non-reefal clusters, C6 in STP is dominated by bottom-dwelling and secondarily phytal taxa, in addition to a small proportion of euryhaline and almost no coral taxa. C2 in Zanzibar is instead comprised of primarily phytal taxa followed by euryhaline and lastly bottom-dwelling taxa. The discrepancy between non-reefal clusters C6 and C2 is likely caused by low vegetation coverage in the sand flat habitats in STP, where sea floor is mostly bare, so that the bottom-dwelling taxa thrive while the phytal taxa are rare. The exclusive occurrence of euryhaline taxa in the non-reefal intertidal clusters indicates salinity variations there, in contrast to normal marine conditions in deeper subtidal environments. Our trans-regional comparison clearly demonstrates a persistent correspondence between ostracod ecological structure and environmental character. Specifically, the relative abundances of coral-affiliated, phytal, and bottom-dwelling taxa seem to vary predictably along a benthic community gradient depending on the percentage coverage of hard corals, algae and bare sands. Although more studies from other regions are imperative to test the generality of this pattern, the findings presented here are considered illuminating. Ostracods can potentially be an indicator of reef condition to track the degradation from coral- to algae-dominated states (Tian et al., 2024a), if a quantitative correlation is established between ostracod ecological composition and benthic community in tropical reefs and reef-associated habitats.

Our data suggests high endemism of STP ostracods since many undescribed species (52 out of 90) are found here. However, undersampling of the TEA ostracods complicates firm conclusions and chances are that some of our species may actually have undocumented distributions in coastal West Africa and thus are not true STP endemics. Outside of the TEA, STP ostracods show weak biogeographic connectivity with other tropical Atlantic regions as indicated by the low number of common species (Table 2; Fig. 9). In the west, the mid-Atlantic Barrier (deep and wide Atlantic itself) effectively isolates the TEA from West Atlantic but is occasionally permeable through the easterly flowing Equatorial Counter Currents (Fajemila and Langer, 2017; Floeter et al., 2008; Witte, 1993). Indeed, there are many characteristic West Atlantic genera found in STP, including Neocaudites, Puriana, and Cativella (Coimbra et al., 2004; Omatsola, 1972). The STP ostracod fauna is slightly more similar to the Southwestern Atlantic fauna at species level despite the Northwestern Atlantic being much more speciose as a biodiversity hotspot, which is likely due to the difference in geographic distance (∼ 3500 km from the Brazil and ∼ 8696 from the Caribbean) (Da Costa et al., 2022). In the north, subtropical species from the Mediterranean and Northwest African coast may be brought southward by the Canary Current and further dispersed by the Gulf of Guinea Current (Le Lœuff and Von Cosel, 1998). STP ostracods therefore hold certain similarity with the Northeastern Atlantic fauna. Interestingly, with the major biogeographic barrier of the Benguela Current in the south, STP shares 11 species in common with Zanzibar but only 3 with the Southwest African coast, for which there are three possible explanations. First, foraminifera evidence shows that the Benguela barrier could be breached when warm water Agulhas eddies pinched off into the South Atlantic during interglacial (i.e., the Agulhas leakage hypothesis) (Gordon, 2003). Some amphisteginid foraminifera from the warm WIO managed to colonize the Atlantic in this way, but their populations in colder Southwest Africa became locally extinct during glacial intervals and eventually only the TEA populations survived until today (Fajemila and Langer, 2017). We suggest that ostracods may undergo similar processes in history. For example, Kotoracythere inconspicua and Keijia demissa as two common species in STP and Zanzibar are known to originate during late Miocene in the Indo-Pacific and display pan-tropical distributions today (Coimbra et al., 1999; Sridhar et al., 2007). It is likely that they took the dispersal route from the Indo-Pacific to TEA through South Africa, since the Tethyan Seaway had already been closed by late Miocene. Second, some common species in STP and Zanzibar are Tethyan relicts, Paracytheridea tschoppi for instance (Coimbra et al., 1999). This species dispersed into the Atlantic and Indo-Pacific before the closure of the Tethyan Seaway and experienced a long period of evolutionary stasis (Coimbra et al., 1999). Lastly, low faunal overlap between STP and Southwest Africa could be caused by sampling bias, but the relatively large species pool of the Southwest Africa province makes this explanation less plausible. In summary, although our biogeographic study here is not corrected for uneven sampling efforts and patchy geographic coverage in each province, it builds a reasonably solid basic framework of ostracod provinciality in the tropical Atlantic. An exceedingly high level of endemism underscores the predominant role of STP as a diversity reservoir instead of a steppingstone, as the soft and hard barriers strongly inhibit dispersal in all directions (Witte, 1993). Furthermore, ostracods have low dispersal capacity because they do not have a planktic larvae stage, unlike foraminifera and many other benthic groups (Danielopol et al., 1994; Yasuhara et al., 2017); yet some species still achieve transoceanic and even cosmopolitan distribution being passively carried by floating algae, migratory birds, and vessels (Danielopol et al., 1994). The coexistence of wide-ranging species together with many endemic species in the oceanic islands of STP invokes further surveys to understand ostracod provinciality and dispersal vectors, in face of increasing anthropogenic transport.