Phytoplankton species are the most widespread and quantitatively relevant primary producers in aquatic ecosystems (Basset et al., 2008). A significant increase in phytoplankton biomass, known as a phytoplankton bloom, indicates an imbalance between phytoplankton growth and loss processes (Carstensen et al., 2007). It is widely accepted that global climate change and eutrophication have intensified phytoplankton blooms in lakes and reservoirs around the world (Ho et al., 2019; Griffith and Gobler, 2020; Liu et al, 2021). China has been severely threatened by water body eutrophication and frequent harmful phytoplankton bloom events during the past decade (Huang et al., 2020), which is one of the most challenging environmental problems (Huisman et al., 2018).

Reservoirs have become important frameworks within ecological studies of assemblage organization (Leitão et al., 2003), mainly due to the abiotic instability generated in their water bodies. Compared to lakes, reservoirs are subject to a greater degree of artificial management. Therefore, reservoirs can serve as a valuable water body type to study the ecological response of ecosystems to artificial interventions. Globally, reservoirs are inhabited by phytoplankton assemblages dominated by cyanobacteria such as Microcystis and Dolichospermum, or dinoflagellates such as Ceratium, forming either mixed or single blooms (see the literature cited by Bordet et al., 2017). Intense algal blooms can cause severe changes in the aquatic ecosystems (Reid et al., 2019), posing a significant threat to aquatic organisms and even humans using these water sources (Olokotum et al., 2020). Reservoir impoundment alters turbulence regimes, allowing the development of phytoplankton species that cannot thrive in fast-flowing waters, but gradually reach large populations in frequently stratified water columns with sufficient nutrients and light conditions (Bordet et al., 2017; Zeng et al., 2023). As the largest reservoir for water conservancy and hydropower projects in China, the Three Gorges Reservoir (TGR) has attracted widespread attention for its ecological impacts. Since the impoundment of the reservoir, the water level has risen significantly, the flow velocity has slowed down greatly, the water exchange has weakened, the artificial regulation has counteracted seasonal hydrological fluctuations (Ding et al., 2022; Ou et al., 2022), and the self-purification capacity of the water body has decreased (Liu et al., 2013). As a result, many tributary bays of the TGR have suffered from phytoplankton blooms of various sizes every year. In summer and early autumn (the low-water level period in the Three Gorges Reservoir), the light and water temperature conditions are good, which are favorable for algal growth, and algal blooms, especially cyanobacterial blooms, are prone to occur (Wang et al., 2011; Zhou et al., 2022). For example, a severe cyanobacterial bloom occurred in 2008 in Xiangxi Bay, a typical tributary of the Three Gorges Reservoir (Ministry of Environmental Protection of China, 2014), which seriously affected the safety of the water quality in the Three Gorges Reservoir area.

Research on phytoplankton bloom has primarily focused on the total amount of phytoplankton and the dominant species. However, the total algal abundance or biomass provides limited insight into the processes that govern their spatial and temporal distribution (Zhao et al., 2017). During blooms dominated by different species and at different stages of bloom development, the structure of phytoplankton communities can undergo significant changes with profound impacts on water quality. In addition, weekly sampling is essential for effective studies of phytoplankton blooms. This frequency is consistent with the timescale of population responses, as the generation time of phytoplankton varies from hours to a few days (Reynolds, 2006). While weekly dynamics of phytoplankton communities have been reported in Xiangxi Bay (Wang et al., 2011), changes in phytoplankton communities composition at higher sampling resolutions during the bloom process have not been documented. Therefore, more fine-scale and structured information on phytoplankton communities is urgently needed to fully understand the conceptual mechanisms driving bloom development.

The present study aimed to analyze phytoplankton communities during the bloom process using daily sampling data. Bloom stages were broadly defined based on total and relative abundance and biomass, as well as α diversity. We expected to identify differences in phytoplankton community structure within and between different bloom stages, and to explore the combined effects of environmental factors through a multifactorial investigation. The influence of environmental factors on the variation in phytoplankton community composition among sampling days (β diversity) was evaluated, and significant environmental factors were identified. In addition, β diversity was decomposed into turnover and richness changes to assess their relative importance. Co-occurrence networks were constructed to compare the internal structure of the phytoplankton community in different bloom stages.

Algal blooms have emerged as a significant environmental problem in inland waters (Hou et al., 2022) and pose a significant threat to public health and aquatic ecosystems worldwide (Brooks et al., 2016). Daily sampling provides a more detailed view of the phytoplankton bloom process compared to the broader dynamics captured by weekly sampling. Our previous weekly study (Wang et al., 2011) missed the peaks of both Microcystis aeruginosa on July 2^nd and Peridiniopsis niei on September 15^th, despite overlapping study periods. Inaccurate time scales can lead to misinterpretation of key factors influencing algal community composition and confound studies of growth dynamics (Gupta et al., 2023). There is an urgent need for fine-scale and structured information on phytoplankton communities to fully understand the conceptual mechanisms driving bloom development, which will provide technical support for watershed management. Different studies use different definitions of the biomass threshold for identifying bloom occurrence. For example, Tett (1987) defines exceptional blooms as those that exceed a chlorophyll threshold of 100 µg/L. In practice, a chlorophyll a level of 15 µg/L is often used as a standard for algaecide application (Hudnell et al., 2010). In our study, the peak chlorophyll a levels in stages 1 and 4 reached 70.80 µg/L and 706.44 µg/L, respectively (unpublished data). However, it is not appropriate to compare two types of phytoplankton blooms, Microcystis and Peridiniopsis blooms, based on chlorophyll a concentrations alone. The Microcystis bloom had higher density but lower biomass and chlorophyll a concentration and lasted longer. In contrast, the Peridiniopsis bloom had lower density but higher biomass and chlorophyll a concentration, with its clear advantage lasting one to two days. Consideration of phytoplankton community structure provides a more insightful framework for studying bloom dynamics.

Species dominance in the environment is typically not attributed to the influence of a single environmental factor, but is often determined by complex nonlinear relationships involving many environmental variables. Wind- driven hydrodynamic conditions are often an important factor in determining the extent of phytoplankton blooms (Wu et al., 2015). Light and the hydrodynamic conditions also play a critical role in the development and maintenance of blooms (Wang et al., 2020; Summers and Ryder, 2023). However, in this study, the influence of meteorological and hydrological factors was less than that of physiochemical factors ( Figure 5 ). Excessive nutrient enrichment in water bodies undoubtedly promotes the growth of harmful cyanobacteria (Nowicka-Krawczyk et al., 2022). Throughout the study period, the density of Microcystis aeruginosa showed a stronger correlation with NH₄-N ( Figure 4 ). The Microcystis bloom created conditions that could further increase its dominance, such as higher pH and lower Zeu ( Table 1 ). These conditions favor the development of cyanobacterial blooms (Amorim and Moura, 2021). In addition to being photoautotrophic, cyanobacteria are known to use several ecophysiological strategies to outcompete eukaryotic algae and ensure their mass dominance in the phytoplankton. These strategies include the presence of gas vesicle aggregates, rapid cell division, and the use of specific metabolic pathways, among others (Li et al., 2016). Summers and Ryder (2023) concluded that prolonged stratification reduces competition, decreases diversity, and leads to cyanobacteria-dominated blooms. In our study, stages 1 and 2 had relatively higher RWCS, but there was a drastic decrease in the abundance of Microcystis, due in part to reductions in TN and NH₄-N. This suggests that nutrient availability may contribute more significantly than stratification to the transformation of the phytoplankton community towards a reduction of Microcystis blooms. Furthermore, the influence of rainfall could not be excluded, as rainfall occurred on July 1^st (11.6 mm/day) and 2^nd (0.3 mm/day).

Freshwater dinoflagellates, including Peridiniopsis, are found in many aquatic ecosystems worldwide and can sometimes lead to harmful blooms that adversely affect water quality (Viner-Mozzini et al., 2003; Yatigammana et al., 2011). Dinoflagellates possess several abilities, such as tolerance to high irradiance, vertical migration within the water column to optimize photosynthesis and growth, and the secretion of toxins, which allow them to compete, thrive, and dominate within the phytoplankton community (Viner-Mozzini et al., 2003; Regel et al., 2004). In our study, Peridiniopsis niei density showed a stronger correlation with TP and PO₄-P, highlighting the importance of phosphorus in maintaining their dominance. Wu et al. (2017) also showed that high phosphorus increased the photosynthetic efficiency of Peridiniopsis and stimulated its growth. However, certain environmental factors, such as lower RWCS in stage 4, did not provide suitable conditions to maintain their competitive advantage. The dominance of Peridiniopsis niei persisted for one to two days, possibly due to the rainfall events that occurred on September 15^th (0.1 mm/day). A similar phenomenon was reported by Zhou et al. (2011), who attributed the decrease in phytoplankton biomass and the disappearance of blooms to heavy rainfall.

Despite the morphological and physiological differences between cyanobacteria and dinoflagellates, reservoirs provide ecological conditions conducive to their success. The possession of aerotopes or flagella allows individuals to dominate in newly created regimes with less turbulence (Bordet et al., 2017). Although TP reduction is widely used as a primary strategy to mitigate the severity of harmful algal blooms (Pick, 2016), cases have shown that TP decreases can be accompanied by increases in chlorophyll a concentrations, suggesting that other abiotic factors may also influence the observed trends (Huang et al., 2020). The study by Nowicka-Krawczyk et al. (2022) showed that the composition of cyanobacterial blooms is strongly influenced by the nitrogen/phosphorus (N/P) ratio, with N/P explaining 13.3% of the variation and PO₄-P explaining 10.5%. In hypereutrophic waters (TP > 0.1 mg/L), phytoplankton biomass was found to increase with additions of both phosphorus and nitrogen (Filstrup and Downing, 2017). Therefore, it is critical to consider the TN : TP ratio for these systems (Summers and Ryder, 2023). Some studies have considered N/P as a factor influencing phytoplankton structure, rather than just absolute N and P concentrations (Liu et al., 2024; Yang et al., 2020). In this study, the maximum concentrations of TP reached 0.13 and 0.30 mg/L in stages 1 and 4, respectively. Therefore, the ratio of TN : TP should be considered. We correlated TN, TP, and TN : TP with the biomass proportion of Microcystis aeruginosa and Peridiniopsis niei for the sampling days when either Microcystis aeruginosa or Peridiniopsis niei was present (see Figure 9 ). The results showed that a high TN concentration, low TP concentration, and high TN : TP ratio corresponded to a high biomass proportion of Microcystis aeruginosa. Conversely, a low TN concentration, high TP concentration, and low TN : TP ratio corresponded to a high biomass proportion of Peridiniopsis niei. This contrasting response of Microcystis aeruginosa and Peridiniopsis niei suggests that TN : TP plays an important role in maintaining the competitive advantages of different dominant bloom species.

The bloom-forming taxa significantly influenced the composition and structure of the phytoplankton. Microcystis aeruginosa blooms significantly reduced phytoplankton species richness, Shannon diversity, and evenness, reaching the lowest levels observed (refer to Figure 3 and Table 2 ), consistent with the findings of Escalas et al. (2019). Similarly, blooms of Peridiniopsis niei also resulted in reduced phytoplankton species richness, Shannon diversity, and evenness, consistent with recent findings by Amorim and Moura (2021). In addition, intense blooms of cyanobacteria and dinophyta are known to cause fish kills by depleting oxygen concentrations during biomass decomposition (Meichtry de Zaburlín et al., 2016), leading to significant disruptions in aquatic food webs, ecosystem functioning, and ecosystem service provision (Sukenik et al., 2015). However, throughout the entire study period, stage 3 exhibited low total density and total biomass, with the highest species richness (as indicated by the highest nodes in Table 2 ), Shannon diversity, and evenness, as well as the lowest dominance index. The successful growth of Peridiniopsis niei after stage 3, despite the initial low phytoplankton density, indicates its resilience under reduced competition from other phytoplankton.

The β diversity of phytoplankton can serve as an excellent model for elucidating the mechanisms that shape aquatic communities in reservoirs (de Moura et al., 2022). Studies examining replacement rates and variations in phytoplankton richness or abundance can be interpreted as indicators of ecosystem function, given the significant dependence of reservoir trophic webs on phytoplankton dynamics (Znachor et al., 2020). In this study, β diversities based on abundance data were higher than those based on presence/absence data. Repl had a greater effect on β diversity differences based on presence/absence data. This may indicate that during blooms dominated by different species and at different stages of bloom development, certain species had replaced others, resulting in variations in β diversity. Conversely, RichDiff had a greater effect on β diversity differences based on species abundance data. This suggests the presence of species with significant differences in abundance between samples, particularly dominant phytoplankton bloom species, and that these differences in abundance were one of the main drivers of β diversity differences. These results indicate that β diversity differences throughout the phytoplankton bloom process included significant species turnover and abundance gradients, both of which jointly influenced the differences in phytoplankton species composition on different sampling days. Szabo et al. (2019) suggested that both abundance and binary data sets should be considered when assessing ecological uniqueness, as they can shed light on different ecological interpretation patterns, and de Moura et al. (2022) also identified some differences between data types.

Increasing evidence suggests that the ability of microbial systems to support ecological functions and resist environmental pressures is an emergent property resulting from interactions among many taxa (Escalas et al., 2019). Network-based approaches provide an integrated representation of interactions among different taxa within microbial communities (Deng et al., 2012). However, studies on the structure of phytoplankton networks are still limited (Carey et al., 2017; Tarafdar et al., 2023). Each phytoplankton network constructed in this study represents the covariation in phytoplankton density over time for each stage, with similar responses to environmental changes or interactions among phytoplankton species (Rottjers and Faust, 2018). The co-occurrence networks of stages 1 and 4 had relatively high average degree and graph density, low connected components and average path length. This combination of features may reflect a relatively dense and evenly connected network, where connections between nodes are tight and the entire network is a relatively cohesive unit. Such a network structure may indicate efficient information propagation, shorter path lengths, and higher network accessibility under certain circumstances. Although composed of fewer nodes and links for stage 1 in Table 2 , the Microcystis-dominated networks could be considered as more complex than the others, as they have a higher graph density and a higher average number of links (degree) per node. Similar results were reported in the study of Escalas et al. (2019). Furthermore, the co-occurrence networks in stage 3 were relatively special, with the lowest average degree and graph density, but the highest number of nodes, edges, modularity, connected components, and average path length. This indicated that there were some sparsely connected small-world structures within the community, resulting in a lower clustering coefficient for nodes within the community; however, the overall community division remained compact. It has been reported that networks with small-world structures tend to be more robust to perturbations and random loss of species (Peura et al., 2015). A slightly more positive correlation at stage 3 also suggests that high cooperative interactions among species may result in a more stable community structure and thus more resilient to environmental changes (Tarafdar et al., 2023).

The study analyzed the dynamics of the phytoplankton community during a low-water level period (June to September) when blooms of Microcystis aeruginosa and Peridiniopsis niei occurred sequentially in a large subtropical reservoir bay. The entire bloom process was clearly delineated into four stages based on daily changes in total and relative phytoplankton abundance and biomass, and α diversity including Shannon, evenness and dominance. Microcystis aeruginosa maintained significant dominance for almost a month in stage 1, with low Shannon and evenness but a high dominance index. This was followed by stage 2, where total density and biomass decreased dramatically, but Microcystis aeruginosa still accounted for some proportion. Phytoplankton total density and biomass continued to decrease in stage 3, which was characterized by the highest Shannon and evenness but the lowest dominance index. Peridiniopsis niei bloomed in stage 4, but its dominant advantages lasted only one to two days. Total phytoplankton density and Microcystis aeruginosa density showed a stronger correlation with NH₄-N, while total phytoplankton biomass and Peridiniopsis niei density showed a stronger correlation with TP and PO₄-P. The TN : TP ratio could be considered as an important indicator to determine whether Microcystis aeruginosa or Peridiniopsis niei dominated the phytoplankton community. Precipitation partially contributed to their significant decrease or disappearance. Variation partitioning analyses (VPA) indicated that physical and chemical factors and their interactions explained more variation in phytoplankton data than meteorological and hydrological factors. DSi, NH₄-N, pH, WT, Zeu, and RWCS were identified as key environmental factors associated with phytoplankton community variation using the db-RDA technique. The β diversity of the phytoplankton community was assessed and further partitioned into richness difference (RichDiff) and replacement (Repl). The results showed an increase in average β diversity with stage progression, i.e., β diversity in stage 1< stages 1-2< 1-3< 1-4, 2< 2-3< 2-4, and 3< 3-4. In all stages, Repl contributed more to the variation than RichDiff with presence/absence data, while RichDiff contributed more to the variation than Repl with abundance data. Co-occurrence network analysis revealed that stage 1 had the highest average degree, graph density, and average clustering coefficient, along with the lowest connected components and average path length, indicating the most complex structure. However, stage 3 had the lowest average degree, average weighted degree, graph density, along with the highest modularity, connected component, and average path length, suggesting that the network for stage 3 was relatively sparse, despite maintaining a compact overall community division.

The raw data supporting the conclusions of this article will be made available by the authors, without undue reservation.

LT: Data curation, Investigation, Writing – original draft. LW: Formal analysis, Methodology, Visualization, Writing – review & editing. QC: Conceptualization, Project administration, Writing – review & editing.