To systematically investigate the spatio-temporal dynamics of the phytoplankton community in tropical aquatic ecosystems, a comprehensive field survey was conducted across Hainan Province. The study area encompassed 22 representative drinking water sources, comprising 11 reservoirs and 11 rivers. Sampling campaigns were carried out during the dry season (December 2023) and the wet season (July 2024) to capture seasonal variations (Supplementary Figure S1; Supplementary Table S1). For each season, two to three water samples were collected from each source, totaling 128 samples across both seasons (Supplementary Table S2). Sampling was performed at a uniform depth of 0.5 m below the water surface. For qualitative sampling, a No. 25 plankton net (64 μm mesh) was towed slowly in a figure-eight (“∞”) pattern underwater for 3–5 min. Samples were fixed immediately after collection by adding 4% formalin. For quantitative sampling, a 5 L standard water sampler was used to collect water; after obtaining 1 L, 15 mL of Lugol’s solution was added on-site for fixation. Site information was meticulously recorded and samples were labeled accordingly. All samples were then transported to the laboratory under conditions that maintained low temperatures and protected them from light.

A 1L quantitative sample was transferred to a sedimentation cylinder and allowed to stand for 48 hours. After settling, the supernatant was removed with a small siphon, and the remaining concentrate was transferred to a beaker and reduced to a final volume of 50 mL. For taxonomic identification and enumeration, 0.1 mL of the concentrated sample was placed in a phytoplankton counting chamber. Species-level identification was performed with a Nikon E100 microscope using “Atlas of Freshwater Microorganisms” (Zhou and Chen, 2010) and “Descriptions of Freshwater Organisms in China” (Han and Shu, 1995) as the primary references. Three replicates were counted for each sample to ensure accuracy, and the average value was calculated for reliable results.

In this study, a portable multi-parameter water quality analyzer (ProQuatro YSI, USA) was utilized to measure environmental variables of the surface water at a constant depth of 0.5 m. The on-site measurements included physical indicators such as water temperature (WT), pH, dissolved oxygen (DO), Oxidation-Reduction Potential (ORP) and electrical conductivity (EC). Concurrently, 5 L water samples were collected, promptly stored, and transported to the laboratory for chemical analysis. For preservation, samples designated for chemical oxygen demand (COD), total nitrogen (TN), and total phosphorus (TP) were acidified with sulfuric acid to a pH of 2 or lower. Meanwhile, samples for five-day biochemical oxygen demand (BOD) were kept at 4°C in the dark. All chemical indicator analyses were conducted within 24 hours to ensure accuracy. The analysis methods for environmental variables employed in this study were as follows: COD was determined using the potassium dichromate method; BOD was measured via the dilution and seeding method; TP was detected using the ammonium molybdate spectrophotometric method; and Total Nitrogen (TN) was determined through the alkaline potassium persulfate digestion ultraviolet spectrophotometric method. These parameters were analyzed in strict accordance with the “Environmental Quality Standards for Surface Water” (GB 3838-2002) to ensure data accuracy and comparability (Yang et al., 2023). Supplementary Table S3 provides the seasonal average values of environmental factors for each lake.

Four indices were employed to assess the biodiversity of phytoplankton in tropical island drinking water sources: Four indices were employed to assess the biodiversity of phytoplankton in tropical island drinking water sources: the Shannon-Wiener diversity index (H′), the Simpson dominance index (DS), the Pielou evenness index (J) and the Margalef richness index (d) (Therriault and Kolasa, 2000; Yi et al., 2024). The corresponding formulas (Equations 1–4) are as follows:

In these formulas, S denotes the total number of phytoplankton species, while Pi indicates the proportion of the i-th species’ cell abundance relative to the total abundance. N represents the overall cell abundance of phytoplankton at the sampling sites.

To visualize compositional differences in the phytoplankton community, Principal Coordinates Analysis (PCoA) was performed. PERMANOVA (Anderson, 2017) was then applied to test for significant differences in community composition among habitats and seasons. Finally, β-diversity was quantified using the Jaccard dissimilarity index and decomposed into turnover and nestedness components to elucidate the underlying dynamic processes.

Spearman rank correlation was used to construct phytoplankton co-occurrence networks for different habitats (rivers and reservoirs) and seasons (dry and wet). Only interspecific associations with strong correlations (|R|≥0.6) (Barberán et al., 2014) that were statistically significant (p< 0.05) (Liang et al., 2023) were retained. In each network, nodes represented phytoplankton species, with node size reflecting species abundance, while edges indicated significant correlations. Networks were visualized in Gephi (Bastian et al., 2009). To describe network topology, the number of nodes, number of edges, modularity index, clustering coefficient, average path length, graph density, and average degree were calculated. Additionally, 1,000 Erdős–Rényi random networks matched to the observed networks by node and edge counts were generated to benchmark these topological metrics.

To assess the relative roles of dispersal and niche processes across phases, the Dispersal–Niche Continuum Index (DNCI) framework was applied. Pairwise DNCI values were then compared among phase combinations to identify whether dispersal or niche processes dominated community assembly across hydrological seasons and habitats.

Statistical analyses were performed using R (v.4.4.1) and SPSS (v.27.0). To compare environmental parameters and α-diversity indices among seasons, the Kruskal–Wallis H nonparametric test was applied in SPSS. To detect compositional differences, Principal Coordinates Analysis (PCoA) based on Bray–Curtis distances and Permutational Multivariate Analysis of Variance (PERMANOVA; 999 permutations) were performed using the vegan package (v.2.5-6) (Oksanen et al., 2007) in R. Furthermore, β-diversity was decomposed into turnover and nestedness components using the betapart package (v.1.5.6) (Baselga and Orme, 2012). For network analysis, networks were constructed and topological parameters computed with the igraph package (v.2.1.4) (Csardi and Nepusz, 2005), while empirical and random networks were compared using the Z-test (Zhao et al., 2016). Finally, community assembly mechanism was investigated with the DNCImper package (Vilmi et al., 2020) by applying the DNCI_multigroup function to calculate pairwise DNCI values.

The physicochemical characteristics of drinking water sources demonstrate distinct dynamic patterns, varying both seasonally (dry vs. wet season) and spatially (river vs. reservoir). Most physical and chemical factors, except salinity (SAL), exhibited significant spatio-temporal variations in drinking water sources during the study period (Kruskal-Wallis H test, p< 0.05), as shown in Supplementary Figure S2. In river water, the temperature rose markedly from the dry season (25.38 ± 1.81°C) to the wet season (29.99 ± 2.46°C) (p<0.001). Reservoirs exhibited a similar seasonal warming pattern, with significantly higher water temperatures observed during the wet season compared to the dry season (p < 0.001). Nutrient and organic matter analyses revealed that TN (1.24 ± 0.35 mg/L), TP (0.12 ± 0.06 mg/L), and COD (15.03 ± 4.02 mg/L) in rivers during the wet season were significantly higher than those observed in reservoirs during the wet season, as well as in rivers during the dry season (p < 0.001). During the wet season, the reservoir’s pH (8.65 ± 0.42) and DO (8.17 ± 0.74 mg/L) were significantly higher compared to both the river during the same season and the reservoir in the dry season (p < 0.001). Notably, ORP and BOD also exhibited significant differences between the reservoir’s dry and wet season (p < 0.001).

A total of 228 phytoplankton taxa (including varieties and forms) were identified through classical morphological analysis. These species distributed across 99 genera within 8 phyla. Chlorophyta emerged as the most diverse, comprising 102 species across 41 genera, representing 44.7% of the total species. Bacillariophyta followed with 62 species in 26 genera, and Cyanophyta with 40 species in 20 genera. In contrast, Xanthophyta had the fewest, with just one species in a single genus (Figure 1A). The cell density for each phylum and the overall density were illustrated in Supplementary Figure S3. The overall distribution of phytoplankton density varied from 14.12 to 65.72×10⁶ cells/L. Peak density occurred in reservoir samples during the wet season, while the lowest density was recorded in the dry season. Among the different groups, Cyanophyta was the most dominant, with density ranging from 4.39 to 44.32×10⁶ cells/L. Chlorophyta followed, with density varying between 2.81 and 13.73×10⁶ cells/L. In contrast, Xanthophyta exhibited the lowest density, ranging from 0 to 0.43×10⁴ cells/L. The total phytoplankton biomass ranged from 3.51 to 12.38 mg/L (Supplementary Figure S4). Biomass peaked in reservoir samples during the wet season and reached its minimum in the dry season. Among phyla, Chlorophyta dominated, with density from 1.54 to 6.49 mg/L. Bacillariophyta was the second most abundant group, ranging from 1.57 to 2.67 mg/L.

Analysis of community composition identified 49 phytoplankton species as consistently present across all spatial and temporal conditions. Habitat specificity comparison revealed the greatest overlap of phytoplankton species, totaling 90 species, between rivers and reservoirs during the wet season. Conversely, the overlap was minimal, with only 65 species, between reservoirs in the dry and wet seasons (Supplementary Figure S5). In terms of community dynamics, phytoplankton abundance in rivers and reservoirs was similar during the dry season but differed significantly in the wet season. During the wet season, the relative abundances of Bacillariophyta and Chlorophyta decreased, while Cyanophyta increased significantly, becoming the dominant group (Figure 1B, Supplementary Figure S6). Spatio-temporal analysis of species richness showed that the Nandu River had the highest phytoplankton richness during the dry season, with 46 species, while the Songtao Reservoir recorded the lowest, with 19 species. In the wet season, the Wanning Reservoir exhibited the highest richness, with 54 species, whereas the Longtang River had the lowest, with 18 species (Figures 1C, D).

The analysis of the α-diversity within the phytoplankton community demonstrated distinct spatio-temporal distribution patterns. Spatially, α-diversity was greater in river habitats compared to reservoir habitats (Figures 2A–D). During the dry season, the river exhibited higher Shannon-Wiener and Pielou indices than the reservoir, though these differences were not significant. In contrast, the river showed significantly higher Margalef and Simpson indices. During the wet season, all α-diversity indices were significantly greater in the river than in the reservoir. Temporally, during the wet season, reservoirs exhibited a significantly higher Shannon-Wiener index than during the dry season (p < 0.01). Conversely, the Pielou index was significantly elevated in the dry season compared to the wet season (p < 0.05). Similarly, the Margalef index in rivers was notably higher during the dry season than in the wet season (p < 0.001), while the Simpson index showed no significant temporal variation (Figures 2A–D). Further analysis using PCoA (Figure 2E) and PERMANOVA (Supplementary Table S4) confirmed significant differences in the phytoplankton community composition across various spatio-temporal conditions (R² = 0.11, F = 5.13, P = 0.001).The phytoplankton community showed consistently moderate β-diversity across environments and seasons, with values of 0.55 for rivers in the dry season, 0.57 for reservoirs in the dry season, 0.59 for rivers in the wet season, and 0.57 for reservoirs in the wet season. This narrow range (0.55–0.59) indicates that species turnover remained relatively stable across the examined spatial and temporal scales. A closer examination of the β-diversity components indicated that the turnover processes played a significantly larger role in community differences than the nestedness process, contributing 86.4%, 76.4%, 84.0%, and 87.9%, respectively (Figure 2F; Supplementary Figure S7).

The Mantel test’s statistical analysis revealed notable differences in how the phytoplankton community structure and diversity relate to environmental factors across seasons and habitats. In the dry season, river environments showed a significant negative correlation between TN and community structure (r< 0.25, p < 0.05). Conversely, in reservoirs, pH demonstrated a strong positive correlation (0.25< r< 0.05, p < 0.01), while ORP and latitude (Lat) exhibited weak positive correlations. During the wet season, DO in rivers was positively correlated with the community structure, whereas ORP, TP, and TN showed negative correlations. In reservoirs, temperature was positively correlated with BOD, and ORP was negatively correlated (Figure 3A). Phytoplankton α-diversity generally demonstrated a negative correlation with temperature and BOD, while showing a positive correlation with pH and ORP. Notably, the Margalef richness index uniquely responded to temperature and COD. Overall, β-diversity exhibited negative correlations with ORP, TP, and pH, but was positively correlated with Lat. A significant negative correlation was observed between the turnover component of β-diversity and TDS, SAL and TP. In contrast, the nestedness component showed no significant association with any water quality parameters (Figure 3B).

The phytoplankton co-occurrence networks constructed for four distinct habitat conditions exhibited entirely positive connections and highly modular structures (Figures 4A–D). The analysis of the network’s topological structure revealed that the river ecosystem’s complexity, including average degree and graph density, was lower compared to the reservoir ecosystem. Notably, the reservoir network during the wet season demonstrated the greatest stability, characterized by a clustering coefficient of 0.701 and a modularity of 0.810, whereas its stability in the dry season was comparatively weaker (Supplementary Table S5). In terms of habitat connectivity, node-sharing analysis revealed that the river and reservoir shared the most nodes (49 species) during the wet season, but shared the fewest (only 23 species) during the dry season (Supplementary Figure S8B). Regarding species composition, the dominant nodes were primarily composed of Chlorophyta, Bacillariophyta and Cyanophyta. Seasonal variations significantly impacted this composition: during the wet season, the proportion of Chlorophyta nodes was substantially higher compared to the dry season, whereas Bacillariophyta exhibited an inverse pattern. Interestingly, Cyanophyta represented the smallest proportion in the river network during the wet season (Supplementary Figure S8A).

Most notably, although Euglenophyta, Chrysophyta and Cryptophyta occupied a relatively small proportion of network nodes—reflecting lower species richness—they exhibited exceptionally high average connectivity within specific habitats. This topological feature implies that they may function as “keystone taxa,” playing a crucial role in maintaining community stability (Supplementary Table S6). Specifically, Euglenophyta demonstrated the strongest connectivity in the dry season river (Dry-HL) network (average degree: 6.00). Conversely, Cryptophyta were most active in the dry season reservoir (Dry-SK) network (average degree: 4.33). Furthermore, Chrysophyta maintained high connectivity across both habitats during the wet season (Wet-HL: 5.00; Wet-SK: 5.50). These findings suggest that these rare or low-abundance groups exert a disproportionate influence on community assembly and succession through intense interspecific interactions, particularly in response to complex environmental fluctuations.

The Dispersal-Niche Continuum Index (DNCI) analysis revealed significant spatio-temporal heterogeneity in community assembly mechanism. All pairwise comparisons exhibited significantly negative DNCI values (Table 1), indicating the predominance of dispersal-driven stochastic processes in shaping tropical island phytoplankton community. Specifically, intense hydrological connectivity during the wet season facilitated the highest level of passive dispersal between the reservoir and river (DNCI = -28.81). In contrast, reduced water levels during the dry season significantly attenuated this dispersal effect (-13.42), leading to a relative resurgence in the influence of environmental filtering. These findings were further corroborated by null model simulations, where the “Niche + Dispersal” model yielded the lowest E-values and provided the optimal fit to the empirical data, outperforming the pure niche model (Figure 5). Collectively, these results confirm the dominant role of dispersal processes in this ecosystem, with environmental selection playing a subsidiary role under specific hydrological conditions.

Our results showed that phytoplankton species in the studied tropical island drinking water sources were diverse, predominantly comprising Chlorophyta and Bacillariophyta (Figure 1A). This aligned with observations by Kumar et al. (2020) and Wagaw et al. (2021) regarding tropical lakes and reservoirs. During the dry season, the abundance of phytoplankton in rivers and reservoirs of tropical island water sources was similar, with Bacillariophyta, Cyanophyta, and Chlorophyta being dominant (Figure 1B). Similar phytoplankton abundance composition have been reported in studies by Mohammed et al. (2023) and Moura et al. (2013). During the wet season, characterized by higher temperature, Cyanophyta’s abundance proportion increased, contrasting with the declining trend of Bacillariophyta (Figure 1B, Supplementary Figure S6). This seasonal succession pattern aligns closely with previous studies. In long-term monitoring of Prashar Lake and Hongmen Reservoir, Jindal et al. (2014) and Liu et al. (2021a) reported that high temperatures and abundant nutrients (such as phosphate) during the wet season (or summer) drove Cyanophyta outbreaks and dominance, whereas low temperatures limited their growth. Meanwhile, rainfall-induced turbidity reduces light penetration, inhibiting diatom growth and diminishing their abundance (Duong et al., 2019; Lin et al., 2011).

Spatially, rivers consistently exhibited higher α-diversity than reservoirs, regardless of the season. This disparity is likely attributed to differences in habitat heterogeneity and hydrological dynamics. Rivers are characterized by high spatial heterogeneity—including velocity gradients, substrate diversity, and fluctuating nutrient inputs—which create varied niches that support a wide range of generalist species (Sriswasdi et al., 2017). In contrast, the relatively static and homogeneous environment of reservoirs tends to promote competitive exclusion, resulting in the dominance of a few specialist species and a consequent reduction in overall diversity (Clavel et al., 2011; Therriault and Kolasa, 2000). Temporally, divergent patterns were observed between species diversity and evenness. The Shannon-Wiener diversity index peaked during the wet season, likely due to rainfall-induced nutrient influxes (N and P) that compromise water quality but stimulate the proliferation of tolerant taxa, such as Chlorophyta and Cyanophyta (Barçante et al., 2020). Conversely, Pielou’s evenness exhibited a significantly declining trend during the wet season (p < 0.05). This suggests that while nutrient enrichment increases species richness, the concurrent rise in water temperatures favors the rapid expansion of specific dominant species. These dominant taxa outcompete others, thereby reducing community evenness despite the higher overall richness (Vallina et al., 2017; Yang et al., 2021). The spatio-temporal pattern of α-diversity is fundamentally shaped by the tropical monsoon marine climate of the study area. Characterized by distinct dry and wet seasons, this climate features concentrated precipitation during the summer, resulting in significant fluctuations in water levels and temperatures. Such environmental variability directly drives the seasonal succession of the phytoplankton community (Wentzky et al., 2020). Although specific rainfall data were not collected in this study, prior research has established that precipitation regimes exert a strong regulatory effect on phytoplankton distribution. Notably, heavy rainfall events typically promote habitat homogenization and accelerate the physical dispersion of plankton in tropical water bodies (Brasil et al., 2016). Collectively, these climatic and hydrological factors elucidate the pronounced spatio-temporal differences observed in the phytoplankton community structure.

In examining the β-diversity of drinking water sources on tropical islands across various spatio-temporal scales, the turnover component predominates over the nestedness component. This indicates that differences in species composition are primarily due to spatial replacements, a pattern consistent with previous findings in studies of β-diversity among phytoplankton communities in different biological groups and regions (Bottero et al., 2020; Maloufi et al., 2016). The prominence of the turnover component in this study area may result from factors such as habitat loss, geographical isolation, dispersal limitations, environmental filtering, and competition (Gutiérrez-Cánovas et al., 2013). The fragmented distribution of drinking water sources on tropical islands has resulted in considerable geographical isolation between rivers and reservoirs (Supplementary Figure S1). This isolation fosters a distinct microhabitat pattern by enhancing habitat condition heterogeneity, including variations in hydrological conditions and localized anthropogenic disturbances, while simultaneously restricting biological diffusion (Li et al., 2018, 2024).

Water temperature, nutrients and pH significantly impact the structure of the phytoplankton community. Among these factors, water temperature is crucial for influencing community structure and succession (Angeler, 2013; Legendre, 2014; Su et al., 2017). It directly affects phytoplankton metabolism, growth, and reproduction by regulating enzyme activity (Pulsifer and Laws, 2021), particularly impacting the succession of cyanobacteria and diatoms. In tropical island environments, this study highlights that seasonal water temperature variations are a key driver of the phytoplankton community succession, notably the alternating dominance between cyanobacteria and diatoms. During the wet season, favorable water temperatures promote cyanobacteria growth, leading to a marked increase in their abundance. Previous studies indicate that cyanobacteria exhibit enhanced competitiveness and reproductive ability when water temperatures exceed 20°C, achieving optimal growth between 25°C and 35°C, thus becoming the dominant group (Li et al., 2019; Rhee and Gotham, 1981). During the dry season, river water temperatures are relatively low. There is a significant negative correlation between Bacillariophyta cell density and water temperature (p< 0.05), aligning with diatoms’ cold-stenothermal ecological characteristics (Abirhire et al., 2015). In tropical regions, although the dry season’s lower water temperatures are still higher than temperate winters, they sufficiently inhibit thermophilic Cyanophyta and Chlorophyta, thereby giving diatoms a competitive advantage. Diatoms thrive in low-temperature, well-mixed water environments (Firsova et al., 2023), making them more likely to proliferate in rivers during the dry season.

Nitrogen input increased the nitrogen content in the water body but did not affect the α-diversity of phytoplankton. This could be because, although nitrogen input boosted phytoplankton quantity to some extent, it did not significantly alter community composition (Kafouris et al., 2019). Nitrogen input can influence the growth rate and competitiveness of phytoplankton (Nwankwegu et al., 2020). While some phytoplankton may grow faster due to nitrogen input, their competitive relationships with other species might remain stable.

Nutrient concentrations in water bodies are crucial for phytoplankton growth and reproduction, with their spatial and temporal variations greatly affecting phytoplankton density and community structure evolution (Szymczak-Żyła et al., 2017). During the wet season, total phosphorus concentrations in reservoirs and rivers increase significantly, coinciding with marked rises in the abundance and density of Cyanophyta and Chlorophyta. This spatio-temporal model underscores the critical role of phosphorus in the dynamics of cyanobacterial populations. As noted by Carvalho et al. (2013), phosphorus concentration is a determining factor for the potential environmental capacity to support cyanobacterial growth. The elevated phosphorus levels during the wet season likely alleviate nutrient limitations experienced in the dry season, thereby enhancing the carrying capacity of aquatic systems for cyanobacteria. This shift strengthens the competitive advantage and dominant position of cyanobacteria within the phytoplankton community. Additional research indicates that, under certain conditions, the stimulatory effect of nutrients—particularly phosphorus—on cyanobacterial growth may surpass the impact of rising water temperatures (Rigosi et al., 2014). These findings underscore phosphorus as a crucial regulator of the phytoplankton community composition and dynamics. During the wet season, elevated phosphorus levels not only boost phytoplankton density but also facilitate the dominance of Cyanophyta in community succession. In tropical island inland freshwater wetlands, where phosphorus is a limiting factor, increased dissolved phosphorus significantly enhances the absorption and utilization efficiency of various algae, especially Cyanophyta and Chlorophyta (Brookes and Carey, 2011; Wu et al., 2020). While some research indicates that phosphorus input can enhance phytoplankton species diversity (Jiang and Nakano, 2022; Xu et al., 2022), our study revealed no significant correlation between elevated phosphorus levels in water and the α-diversity index of the phytoplankton community. This lack of correlation can be attributed to phosphorus input not altering the community’s composition and structure. Further analysis suggests several mechanisms behind this phenomenon. Firstly, the phytoplankton community possesses adaptability that allows it to thrive in high-phosphorus environments. It achieve this by adjusting its metabolic pathways and growth patterns in response to increased phosphorus input (Stephens et al., 2024). In environments with elevated phosphorus levels, organisms can enhance phosphorus absorption and utilization to sustain normal growth and reproduction. This phenomenon is partly due to phosphorus’s role in stimulating the robust growth of submerged macrophytes, which indirectly modulate the phytoplankton community (Muylaert et al., 2010). Submerged macrophytes impact phytoplankton in two primary ways. Firstly, they release chemical substances that inhibit phytoplankton growth and development (Pakdel et al., 2013). Secondly, they offer habitats for zooplankton, thereby intensifying predation pressure on phytoplankton (Vilas et al., 2018).

In the reservoir during the dry season, phytoplankton abundance correlated strongly and positively with pH, a pattern observed in other tropical reservoirs. Silva et al. (2014) reported that pH is a primary regulator of tropical reservoir phytoplankton, and multiple regression analysis identified pH as one of the most important parameters explaining variation in phytoplankton abundance (Mostafa et al., 2023). By contrast, in the river habitat during the wet season, phytoplankton abundance correlated significantly and negatively with pH, consistent with Chakraborty (2009). One plausible mechanism is that pH alters the bioavailability of nutrients and trace metals, thereby constraining phytoplankton physiology (Parthasarathi et al., 2011). During the wet season, strong river flow amplifies environmental fluctuations (for example, acidification or rapid pH shifts), which can stress species that lack tolerance to such changes. The phytoplankton α-diversity indices (Shannon, Pielou, Simpson, Margalef) in this study all correlated significantly and positively with pH, whereas β-diversity correlated significantly and negatively with pH. Thus, higher pH—particularly in the reservoir during the wet season—appears to increase local species richness and evenness while concurrently increasing compositional similarity among sampling sites. Mechanistically, elevated pH likely improves the carbonate chemistry of the water and enhances silicate dissolution (Koester et al., 2016), thereby creating more favorable growth conditions for groups such as Chlorophyta and Bacillariophyta (Lv et al., 2024; Zhang et al., 2019) and supporting higher α-diversity. At the same time, this pH-driven improvement in habitat suitability may partly homogenize community composition across habitats, reduce species turnover, and lower β-diversity. These findings align with previous studies that identified pH as a key driver of phytoplankton diversity (Lin et al., 2023; Sun et al., 2022) and reported that high pH often coincides with high diversity (Lv et al., 2024).

Ecological network analysis reveals that positive interactions dominate among phytoplankton, aligning with previous findings on the phytoplankton community in China’s drinking water (Zhang et al., 2023). These co-occurrence networks exhibit notable differences across various spatial and temporal scales, primarily due to the heterogeneity of nutrient supply environments. Research indicates that nutrient supply conditions significantly influence the topological structure and scale of these networks (Hillebrand and Lehmpfuhl, 2011; Hu et al., 2017). In reservoir drinking water source areas, the unique geographical isolation coupled with oligotrophic conditions imposes dual constraints on the phytoplankton community: it limits both species growth and reproduction rates while significantly reducing species migration between habitats (Hayes et al., 2015). River ecosystems, characterized by significant flow fluctuations and high water velocities, enhance water connectivity, facilitating the migration and diffusion of phytoplankton. In these dynamic systems, the formation of the phytoplankton community primarily relies on spatial diffusion rather than local species interactions. This results in weaker interspecies associations and a simpler network structure (Chen et al., 2019). In contrast, reservoirs provide a stable environment where the phytoplankton community forms mainly through interactions among local species, leading to a more complex network structure. During the wet season, reservoir water temperatures are significantly higher than in the dry season (p< 0.05), which can substantially increase the metabolic efficiency of the phytoplankton community. This, in turn, accelerates the consumption rate of water nutrients. Under such environmental pressures, significant niche differentiation may occur between sensitive taxa with high nutrient requirements and generalist taxa (Sam, 2020). As niche separation increases, species repulsion within the network rises, leading to reduced clustering and modularity. This phenomenon elucidates the weakening of species interaction intensity and the decline in stability of the co-occurrence network during the dry season. Key taxa are crucial in regulating the phytoplankton community interaction networks, playing a fundamental role in sustaining community structure and function (Liu et al., 2022). Notably, global phytoplankton network analysis reveals that most key species are rare taxa with low abundance and infrequent occurrence. Research findings demonstrate that in the four co-occurrence networks, taxa with the highest average connectivity are Euglenophyta, Cryptophyta, and Dinophyta, rather than the more common Chlorophyta, Cyanophyta, and Bacillariophyta. This finding reinforces the notion that rare species play crucial, irreplaceable roles in sustaining the stability of the microbial community (Barabási, 2009). Research on eukaryotic plankton in reservoirs suggests that the interaction between rare and abundant taxa enhances the plankton community’s resilience to environmental changes. This cooperation acts as a buffer, protecting the plankton network from disturbances and thereby supporting the ecosystem’s overall function and stability (Xue et al., 2018).

The assembly of the phytoplankton community in drinking water sources on tropical islands is primarily influenced by dispersal, particularly between reservoirs and rivers during the wet season. Consistent with previous research indicating that random processes predominantly influence the assembly of the phytoplankton community (Yi et al., 2021; Zhao et al., 2021), this phenomenon can be elucidated through two primary mechanisms. First, the river, serving as the main water source for the reservoir, continuously introduces terrestrial and upstream phytoplankton propagules due to its relatively swift flow. This process results in a sustained exogenous inoculation of the reservoir community, which underpins the observed similarity between the two communities (Lu et al., 2020). Second, heavy rainfall and other events during the study period can significantly increase river runoff, potentially leading to reservoir overflow or active flood discharge. Such occurrences establish a rapid and efficient hydrological connection (Song et al., 2020). Although a reservoir typically functions as a relatively hydrodynamic closed system under normal conditions, the combination of continuous inoculation and intermittent high-intensity exchanges effectively facilitates the cross-water propagation of phytoplankton propagules. Consequently, this interaction demonstrates a high degree of diffusion homogeneity at the community level (Liu et al., 2021b). Furthermore, prior studies highlight that both stochastic and deterministic processes in ecological community construction exhibit significant scale dependence. At the local scale, deterministic processes, particularly homogeneous selection, typically dominate community assembly (Gu et al., 2021; Pearman et al., 2022). In contrast, as spatial scale increases, dispersal limitation intensifies, and stochastic processes gain prominence (Niño-García et al., 2016). Consequently, in the drinking-water source rivers and reservoirs of tropical islands, the phytoplankton community assembly is primarily influenced by stochastic processes. This is likely due to the expansive spatial scale of the study area, which amplifies dispersal limitation and diminishes the influence of deterministic processes.