Environmental drivers of plankton communities in coastal waters: the central role of temperature revealed by co-occurrence networks

To identify the environmental factors that drive plankton community composition and structure in coastal waters, a shallow northwestern Mediterranean lagoon was monitored from winter to spring in two contrasting years. The campaign was based on high-frequency recordings of hydrological and meteorological parameters and weekly samplings of nutrients and the plankton community. The collected data allowed the construction of correlation networks, which revealed that water temperature was the most important factor governing community composition, structure and succession at different trophic levels, suggesting its ubiquitous food web control. Temperature favoured phytoplanktonic agellates (Cryptophyceae, Chrysophycea, and Chlorophyceae) and ciliates during winter and early spring. In contrast, it favoured diatoms, dinoagellates, phytoplankton < 6 µm and aloricate Choreotrichida during spring. The secondary drivers were light, which inuenced phytoplankton, and wind, which may regulate turbidity and the nutrient supply from land or sediment, thus affecting benthic species such as Nitzschia sp. and Uronema sp. or salinity-tolerant species such as Prorocentrum sp. The central role of temperature in structuring the co-occurrence network suggests that future global warming could deeply modify plankton communities in shallow coastal zones, affecting whole-food web functioning.


Introduction
Environmental forcing factors play a central role in driving plankton community composition and dynamics in marine and freshwater ecosystems. At a global scale, along latitudinal gradients, species distribution and community composition depend on abiotic conditions, such as temperature, light, and nutrients 1 . On the other hand, at the local level, food web structure is more affected by biotic processes such as predation, competition, population growth and behaviour 2 , which are constrained by environmental conditions. However, in highly dynamic systems subject to intense environmental stressors, physico-chemical forcing factors may play a predominant role in shaping communities 3,4 . Furthermore, the plankton community's response to environmental forcing factors in these systems is challenging to determine, as these factors can be in uenced by various elements and are often linked together 5 . For example, alongshore wind in coastal ecosystems triggers deep, cool and nutrient-rich water upwelling, thus in uencing plankton communities 6,7 . Consequently, the conjunction of wind direction and speed, column mixing, water temperature and nutrient concentration changes explains the plankton response during these events. Therefore, it is essential to study multiple environmental parameters together in particularly highly dynamic systems, as they can be tightly linked together.
Shallow coastal waters, including coastal lagoons, estuaries, seagrass beds, and coral reefs, are highly dynamic and often exposed to extreme environmental events. Community composition and structure in these zones are driven by environmental forcing factors, mainly because the zones occupy the interface between land and sea 8 . These factors in uence plankton communities directly or indirectly. For instance, river runoff transports nutrients and terrigenous organic matter, in uences water turbidity, and modulates phytoplankton and bacterial production 9,10 . Seawater currents or tides recirculate nutrients, which provide critical elements for the food web and enrich local communities of offshore organisms [11][12][13] . These water inputs and precipitation induce signi cant salinity variations that affect the plankton community 14,15 .
Moreover, the shallow depth of these coastal waters makes them particularly sensitive to the action of wind and temperature variations. This in uences the stability of the whole water column (mixing and strati cation) and indirectly alters community dynamics through sediment and nutrient resuspension and turbidity increases [16][17][18] . Furthermore, coastal ecosystems exhibit relatively low inertia, and due to their shallow nature, environmental changes can be rapid 3,19 . However, few studies have investigated the environmental drivers structuring the plankton community composition and structure in shallow coastal waters, preventing a deep understanding of plankton food web functioning.
Correlation analyses are used to build networks mapping the relationships between species [20][21][22] . The analysis of these networks is practical for formulating hypotheses about the most critical mechanisms underlying community assembly. For example, it may suggest the prevalence of top-down forces or a shift from grazing chain-dominated food webs to systems where the microbial loop prevails 22 . In addition to being routinely applied to model planktonic food webs, network analysis is also a powerful tool for studying the effects of abiotic forcing factors on plankton food webs 23,24 . It can, for example, shed light on the role of complex physico-chemical changes responsible for shifts in planktonic food webs triggered by environmental forcing, for example, anthropogenic hydrology alterations in natural coastal lagoons 24 .
The present study's objective was to investigate environmental drivers associated with changes in plankton community composition and structure in a highly dynamic system, Thau Lagoon, a shallow coastal lagoon along the northwestern Mediterranean coast. Correlation networks were constructed using organism abundances and environmental metrics in different seasons. Environmental multiparameter monitoring and weekly samplings for planktonic abundance quanti cation were carried out from winter to late spring during two consecutive years, 2015 and 2016. These two years were very distinct and presented different characteristics. While 2015 was a typical year in terms of climate, with a high thermic amplitude from winter to spring, 2016 was characterised by anomalously high water temperatures during winter, with the warmest winter ever recorded in southern France (http://www.meteofrance.fr/climatpasse-et-futur/bilans-climatiques/bilan-2016/hiver#:~:text=Sur%20l'ensemble%20de%20la,aequo%20(%2B%201.8%20%C2%B0C). These particular climatic characteristics of 2016 were already proposed as the reason for a shift towards the dominance of smaller phytoplankton in the community 19 . These conditions might also have favoured an increase in the number of biotic interactions among smaller organisms 22 , even if the links between environmental variables and plankton community composition were not explicitly investigated. Thus, the present study will, for the rst time, quantitively link multiple environmental variables to plankton community composition and structure in a characteristic shallow coastal lagoon.

Results
Temporal dynamics of planktonic organisms Daily mean Chl a uorescence was used as an index of phytoplankton biomass, allowing us to distinguish three phytoplankton bloom periods that occurred in winter, early spring and late spring in 2015 ( Fig. 1; for details, see Trombetta et al., 2019). For 2016, only one long spring bloom was found. The dynamics of various planktonic groups in Thau Lagoon are shown in Fig. 1. These dynamics showed that phytoplankton < 6 µm, Bacillariophyceae, and auto/mixotrophic Dinophyceae mainly peaked during spring in both years. Cryptophyceae, Chrysophyceae, Chlorophyceae, and other phytoplankton groups mainly peaked during winter and early spring in both years. The abundance of viruses and bacteria slightly increased from winter and reached its highest level from early spring to late spring in both years.
The abundance of heterotrophic nano agellates (HFs) was always high in 2015 and peaked in winter, early spring and late spring, while their abundance was low in winter 2016 and slightly increased to reach a maximum value in late spring. Aloricate ciliates and tintinnids exhibited differences between 2015 and 2016. Choreotrichida were the only taxa peaking during late spring of both years. Oligotrichidae, Prorodontida, Cyclotrichiida, Codonellidae, Metacylididae, Tintinnidae, and Codonellopsidae peaked in winter or early spring in 2015 but had low abundance during late spring. These groups, except Prostomatea and Metacylididae, peaked from spring to mid/late spring in 2016. The abundance of mesozooplankton larvae and heterotrophic Dinophyceae, analysed only in 2016, was low from winter to early spring and peaked towards the end of spring. Plankton community composition and environmental drivers Non-metric multidimensional scaling (nMDS) illustrated the changes in plankton community composition along sampling dates, together with the ordination of environmental parameters (Fig. 2). A progressive transition in community composition occurred in both years from winter to spring, resulting in clear separation of the bloom and non-bloom periods. During both years, winter was characterised by low (air and water) temperatures, light (ultraviolet B radiation (UVBR) and photosynthetically active radiation (PAR)) and nutrients (  Environmental and plankton community composition relationships Table 1 and Fig. 3 show the strongest relationships between plankton groups and the most important environmental factors. Temperature was the factor most connected to taxonomic clusters, with seven and ten signi cant correlations in 2015 (Fig. 3A) and 2016 (Fig. 3B), respectively. These correlations involved all types of groups, from viruses to large zooplankton. During both years, temperature was associated with phytoplankton < 6 µm, Bacillariophyceae, Choreotrichida and Codonellidae. PAR and UVBR were correlated with different taxonomic groups but mainly with phytoplankton. During both years, PAR was correlated with auto/mixotrophic Dinophyceae, while UVBR was correlated with Bacillariophyceae and phytoplankton < 6 µm. In both years, the nutrients were correlated exclusively with heterotrophic groups, and SiO 2 was linked with Chlorophyceae in 2015 only.
Temperature was also the factor most connected to ESD clusters, with seven signi cant correlations in both 2015 and 2016 ( Fig. 3C and 3D). In both years, it was associated with all phytoplankton ESD clusters (except the > 25 µm cluster in 2016). Temperature was also associated with large tintinnids (> 80 µm) and aloricate ciliates (28-50 µm) in 2015 and with tintinnids (< 45 µm and > 80 µm) in 2016. PAR and UVBR were also correlated with a large diversity of ESD clusters depending on the year. Correlation networks of environmental factors and planktonic organisms The number of signi cant correlations linking each environmental factor to planktonic taxa is summarised in Table   During blooms, correlation networks of both years showed the central position of temperature ( Fig. 3B and 3D), which was connected to both autotrophs and heterotrophs. PAR and UVBR were also highly connected but mainly with heterotrophic organisms, especially aloricate ciliates and tintinnids, in the nonbloom periods of both years. During the bloom of 2015, the SiO 2 concentration was negatively linked to four phytoplankton taxa, including two Bacillariophyceae (Chaetoceros sp1. and Pseudo-nitzschia sp.) (Fig. 4B), while during the 2016 bloom, it showed negative relationships with metazoans (rotifers and Phoronidae larvae) and the aloricate ciliate Strombidium sp4. (Fig. 4D).
In the non-bloom correlation networks, wind direction (2015) and speed (2016) exhibited various relationships with diverse planktonic taxa ( Fig. 3A and 3C). Cyanobacteria were correlated with these factors in both years. In the non-bloom of 2016, wind speed shared many neighbours with NO 3 − , NO 2 − and PO 4 3− , but the sign of these connections was the opposite of that for nutrients (Fig. 4C). Turbidity was among the most correlated parameters in the non-bloom periods of both years. During the nonbloom period of 2015 (Fig. 4A), turbidity was mostly connected to phytoplankton, while in the same period of 2016 (Fig. 4C), it was mainly linked to heterotrophs. Bacteria were positively correlated with turbidity in both years (HNA in 2015 and LNA in 2016). In the non-bloom period of 2015, water temperature, pressure and depth were closely and positively connected to common taxa, including phytoplankton, heterotrophic nano agellates and aloricate ciliates.
Spearman's correlation networks between environmental parameters are shown in Supplementary Fig. 6.

Discussion
Trombetta et al. (2019, 2020) 19,22 found that in Thau Lagoon, water temperature triggers phytoplankton blooms and suggested that interannual warming favours interactions among small organisms and trophic cascades. The ndings of the present study suggest for the rst time that water temperature, rising from winter to spring, is the most important environmental factor regulating the composition, succession (Figs. 2 and 3; Table 1) and species dynamics ( Fig. 4; Table 2) of the whole planktonic food web in a shallow coastal lagoon. Such regulation encompasses various trophic levels, from phytoplankton, viruses and bacteria to heterotrophic nano agellates, ciliates and mesozooplankton. The presence of signi cant correlations between water temperature and organisms belonging to various trophic levels suggests extensive impacts on the entire plankton food web. Although these correlationbased results cannot describe causal mechanisms, they clarify the pervasive impact that temperature has on the structuring of the planktonic food web. Temperature affects organisms in several ways: directly, by acting on biotic processes such as growth and predation rates 25,26 , or indirectly, through abiotic processes such as mixing and strati cation 27 .
Water temperature directly in uences the metabolism of organisms. Speci cally, it accelerates the metabolism and consequently the growth rates of organisms 25 . In the present study, in most of the cases, water temperature was linked to taxa by positive correlations (Fig. 4), meaning that their abundance increased when water temperature increased. For instance, the phytoplankton species Chaetoceros spp. and Pseudo-nitzschia sp. mainly appeared during spring blooms of both years and were positively correlated with water temperature (Fig. 4). In Thau Lagoon and in Mediterranean coastal waters, Chaetoceros spp. and Pseudo-nitzschia sp. are the main Bacillariophyceae (diatom) components of spring phytoplankton communities, blooming between 12°C and 14°C and persisting at high abundance even when the water temperature rises above 20°C 19,28,29 . On the other hand, the direct metabolic response is more complex than a simple rise in abundance driven by a temperature increase and might also be due to the thermal optimum of taxa and temporal thermal niches realised 30,31 . In fact, water temperature in uences the composition and succession of the plankton community from winter to spring ( Fig. 2 and Table 1), including a large diversity of taxonomic and ESD clusters, for both heterotrophic and autotrophic groups (Fig. 3). The aloricate ciliates belonging to Choreotrichida and Codonellidae peaked on several dates from winter to spring (Fig. 1). During the 2016 bloom, the aloricate ciliates Leegardiella sp. and Tintinnopsis angulata exhibited a negative correlation with temperature, while Lohmaniella sp. and Tintinnopsis corniger displayed a positive connection with this factor. At low temperature, the abundance of Leegardiella sp. and T. angulata was high, while Lohmaniella sp. and T. corniger increased in abundance with increasing temperature, suggesting that they have different thermal niches.
Furthermore, water temperature can have a direct effect on biotic interactions 26 . There is evidence of the in uence of temperature on biotic interactions of the planktonic food web, such as predation 32,33 , competition 34,35 , mutualism 36 and parasitism 37,38 . In our case study, temperature modi cations might have in uenced organismal interactions and played an important role in the succession of the plankton community. The modi cation of grazing rates due to water temperature increases 25,39 was pointed out several times as a major actor modifying the plankton community composition in mesocosm experiments where the temperature was manipulated 33,40 .
Water temperature also modi es the abiotic environment and has indirect effects on plankton. Water temperature variations regulate vertical water transport and induce mixing during cold events or strati cation during heat events, even in shallow costal lagoons 3,41 . As an example, Nitzschia sp., often classi ed as a benthic Bacillariophyceae 42 , was negatively correlated with water temperature during the 2015 bloom (Fig. 4B), suggesting that colder temperatures affecting the mixing of the water column might cause resuspension of Nitzschia sp., increasing its abundance. On the other hand, higher water temperatures could have strengthened the strati cation by modifying the physical and chemical conditions of the water column (e.g., oxygen or nutrient depletion or changes in salinity or the daily light dose) and have triggered shifts in plankton assemblages 27,30 .
Light was correlated signi cantly with the nMDS axis (Fig. 2), thus representing (after water temperature) one of the most important drivers of the structure of the entire plankton community. In Thau Lagoon, light was suggested as non-limiting for phytoplankton bloom initiation 19 . The present study suggested that light does not in uence phytoplankton abundance (Fig. 4, low number of Spearman's correlations between phytoplankton and light) but instead plays an important role in the composition and succession of the phytoplankton community (Fig. 3, high number of Mantel's correlations between phytoplankton and light). Small phytoplankton cells are more e cient at utilising low light intensity; due to the smaller packaging effect, they are less penalised by self-shading 43,44 than larger phytoplankton. Consistent with these observations, smaller cells, such as those of Cryptophyceae, Chrysophyceae and Chlorophyceae, peaked during winter and early spring, when light intensity was lower. However, for cryptophytes, which are known to be mixotrophic, it cannot be excluded that their high abundance during the winter was due to a shift towards the heterotrophic mode due to the low light intensity available for photosynthesis, causing them to feed on bacteria 45 . Larger cells, such as those of Bacillariophyceae and Dinophyceae, increased instead during late spring (Fig. 1). Phytoplankton succession can also be due to photoacclimation 46 . The daily dose of incident light is generally high, and from winter to spring, it does not represent a limiting factor in shallow Mediterranean coastal sites 19 . Consequently, phytoplankton succession may be in uenced by the capacity to acclimate to different light conditions through the production of photosynthetic or photoprotective accessory pigments rather than being limited by light availability.
The tight connection between light and water temperature may explain the relevance of light in correlations with heterotrophic taxa. The fact that water temperature, PAR and UVBR shared links with many nodes in common supports this hypothesis (Figs. 3 and 4). During the bloom of 2015, UVBR exhibited positive correlations with planktonic taxa, and more than half of these taxa were correlated with water temperature. Air anticyclones generally increase light and air temperature 47 , consequently raising water temperature and affecting the plankton community. However, this cannot be the only explanation, as variables describing light conditions were sometimes far from water temperature in the correlation network and did not share any links with common taxa with this factor. UVBR affected the composition and taxon abundance dynamics of the plankton community in non-univocal way. The relationships with taxa varied in sign and changed according to species sensitivity and period. The positive correlations between prey taxa (lower trophic level) and UVBR may re ect harmful effects on predators. Such negative impacts of UVBR reduce predation pressure and result in positive effects on prey, as demonstrated previously 48, 49 . In contrast, UVBR could indirectly trigger positive effects on phytoplankton due to photochemically induced breakdown of dissolved organic matter, which releases nutrients and enhances phytoplankton growth 50 .
The non-bloom periods during winter and early spring were characterised by high nutrient concentrations ( Fig. 2). During both years, the concentrations of PO 4 3− , NO 3 − , NO 2 − and SiO 2 correlated well with various clusters (Fig. 3) and taxon abundances (Table 1 and Fig. 4

). A previous study suggested that Thau
Lagoon is a nitrogen-and phosphorous-limited system 51 (Fig. 3B and 3D). This result suggested enhanced activity of the microbial loop, as either the excretion of heterotrophic agellates and Cyclotrichia or organic matter release due to viral lysis might have triggered bacterial nitri cation. This process consists of bacteria using NH 4 to produce NO 2 − and then NO 3 − , with rapid assimilation of the last compound by phytoplankton 53 . 2016 was an unusually warm year 19 , and water temperature could also have accelerated nutrient remineralisation. The present results are in accordance with those of a previous study suggesting that the warm conditions of 2016 in Thau Lagoon favoured the co-occurrence of smaller taxa, including heterotrophic nano agellates, viruses and bacteria 22 .
Wind and turbidity were positively correlated ( Supplementary Fig. 6) and were among the six environmental nodes most connected to plankton taxa (Figs. 4A and 4C). During the non-bloom periods, wind (direction in 2015 and speed in 2016) and turbidity were closely related to nutrient concentrations (Fig. 2). The nMDS results illustrated consistency between wind, turbidity and nutrients. They suggested that wind could have been responsible for sediment resuspension and inputs of nutrients from the sediment to the water column. Wind frequently in uences the Thau Lagoon ecosystem by increasing turbidity and contributing to inputs of nutrients through resuspension 16,17 . These processes were also described for other coastal ecosystems 18,54 . The supply of nutrients from sediment resuspension is fairly constant and su cient to ensure phytoplankton growth in Thau Lagoon 16, 19 . Here, sediment (and potentially nutrients) resuspension through wind seems to be more important during non-bloom periods in winter. However, weekly correlation analysis did not reveal any links between wind and nutrient concentrations ( Supplementary Fig. 6) 19 . The absence of signi cant relationships between wind and nutrients may be due to the weekly samplings, which may be inadequate for studying nutrient dynamics in systems characterised by regular resuspension and phytoplankton uptake 19 . Moreover, multiple and simultaneous mechanisms, such as precipitation and discharge from rivers, may interact to change the concentration of nutrients in the lagoon. On the other hand, wind can in uence the plankton community through resuspension of benthic organisms in the water column 55,56 . The benthic Bacillariophyceae Nitzschia sp. 42 and the benthic aloricate ciliate Uronema sp. 57 were positively correlated with wind speed in the non-bloom period of 2016, suggesting that their abundance increased because of wind resuspension.
Occasional links connecting depth and salinity to community composition (Figs. 2 and 3; Table 1) and taxon abundance ( Fig. 4; Table 2) were observed. Depth and salinity displayed mutual positive correlations but did not match any particular period. In Thau Lagoon, the positive correlations with salinity and depth depend on seawater input through the main channel connecting the lagoon to the Mediterranean Sea. An increase in depth is generally associated with an increase in salinity due to marine water being pushed into the lagoon by southern winds. This mechanism could have two distinct effects on the plankton community composition. First, seawater input could have brought offshore taxa into the lagoon, directly modifying the food web. This phenomenon is common in marine lagoons or coastal waters subject to tides, and the composition of the plankton community depends on the balance between imports and exports 8,58 . However, in coastal zones with low tides, such as Thau Lagoon, tidal water transport is limited and often masked by other forcing factors (i.e., wind, sea currents, river inputs, or topographical constraints such as channels and natural or arti cial dykes) 59,60 . The transport of plankton into the lagoon is due to currents or wind pushing marine water from the Mediterranean Sea rather than tidal action. Such factors increase depth (water level) and salinity in the lagoon. Second, salinity increases due to marine water transport could have in uenced the plankton community through physiological effects. Salinity exposes sensitive organisms to osmotic stress and promotes the replacement of salinity-sensitive species by salinity-tolerant taxa 61 . Prorocentrum sp. is known to be a salinity-tolerant taxon 61,62 and was found to be positively correlated with salinity during the bloom period of 2016 (Fig. 4D). In estuaries and coastal ecosystems subject to constant changes, salinity is an important factor in uencing plankton community structure 58,63 . In terms of salinity, Thau Lagoon is relatively stable for a coastal site, and important variations are limited to strong rains, evaporation or water inputs 9,19 . The mean water residence time at the study site is approximately 50 days 64 , and the effect of salinity is therefore limited to occasional events.
The present paper highlighted that water temperature exerts stronger impacts than other environmental factors on the plankton community in a shallow coastal zone. This factor governed the composition, succession and structure of diverse plankton groups, species and trophic levels, suggesting its ubiquitous role in food web control. In such shallow systems moderated by water temperature, global warming and the increased frequency of extreme events such as heat waves could potentially deeply modify plankton communities, and as pointed out in the present study, it might potentially occur at every trophic level, thus affecting whole-food web functioning

Study site and monitoring design
Monitoring took place in Thau Lagoon ( Supplementary Fig. 1), a productive marine lagoon located on the and is connected to the Mediterranean Sea by two main channels. Thau Lagoon is a mesotrophic, phosphorus-and nitrogen-limited system 51 with a turnover rate of 2% (50 days). Salinity ranges from 34 to 38 PSU, and high-amplitude temperature uctuations occur throughout the year, ranging from 4°C in the winter to 30°C in the summer 19,65 . The lagoon is frequently exposed to high wind speeds throughout the year 17,19 . Thau Lagoon hosts economically important activities, mainly oyster farms representing 10% of French production. nutrients and all organisms was carried out in the morning between 09 am and 10 am. Some of the data used in the present study were already published in previous manuscripts 19,22 . Thus, details on acquisition methods can also be found in these manuscripts, as speci ed below.
Hydrological and meteorological data, including air and water temperature, wind speed and direction, PAR The atmospheric pressure, humidity, precipitation, depth, large heterotrophic Dinophyceae, rotifer, mesozooplankton larva, and copepod data presented in this study were not published elsewhere.
Hydrological, meteorological, nutrient, and chlorophyll a measurements

Plankton identi cation and abundance
The diversity and abundances of (1) viruses; (2) non-pigmented planktonic cells < 1 µm, including archaea, heterotrophic bacteria, and chemosynthetic bacteria (hereafter called "bacteria"); (3) phytoplankton; (4) HFs; (5) aloricate ciliates and tintinnids; (6) large heterotrophic Dinophyceae; and (7) metazooplankton were determined. For each of these groups (except large heterotrophic Dinophyceae and metazooplankton), the methods of abundance estimation were described in Trombetta et al.  Table 3 summarises all data used in this manuscript, with the methods used and if already published elsewhere. Sampled with a Niskin bottle and analysed using an epi uorescence microscope (Olympus AX-70) 22 Category of data Type of data Type of instrument References Heterotrophic nano agellates (4 groups) 22 Bacteria (2 groups) Sampled with a Niskin bottle and analysed using a ow cytometer (FACSCalibur, Becton Dickinson) 22 Phytoplankton < 6 µm (4 groups) Sampled with a Niskin bottle and analysed using optical microscopy (Olympus IX-70) 19,22 Phytoplankton > 6 µm (42 groups/taxa) 19,22 Ciliates (57 groups/taxa) 22 Clustering of plankton Planktonic groups/taxa/species were clustered on the basis of two descriptors: (1) ESD and (2) taxonomy. Clustering enabled statistical analysis of speci c groups of interest and comparisons of their differential responses to the various environmental parameters.
First, planktonic groups/taxa/species were clustered into groups according to ESD. Such a clustering method was applied in a previous study, and the clusters identi ed and used below can be found in  22 considered the same periods and found different food web network structures depending on the period (bloom or non-bloom). Consequently, the same bloom and non-bloom dataset separation was used in the present study to investigate the environmental drivers associated with changes in plankton community composition and structure during these productive periods.
nMDS ordination was applied to classify planktonic taxa based on temporal and environmental similarities. Community composition from winter to spring, along the axis of dates, and during productive periods was investigated for the years 2015 and 2016. To determine which environmental factors drive plankton community assembly, the environmental variables were ordinated, and their correlation scores with nMDS axes were calculated using the env t function in R (vegan package, version 2.4-2).
Mantel's test was used to identify the signi cant correlations between environmental factors and plankton community composition according to each ESD and taxonomic cluster. This study aimed to identify the potential role of environmental variables in driving plankton community composition during the bloom and non-bloom periods in both years. Correlations were calculated between two matrices: (1) one of environmental parameters and (2) one of abundances for the ESD or taxonomic clusters. Mantel's tests were performed on each environmental parameter and ESD/taxonomic cluster pair. Only environmental factors signi cantly correlated with the nMDS axes were used in Mantel's test.
Spearman's rank correlation was applied to identify signi cant mutual changes between environmental factors and the abundance of taxa in the bloom and non-bloom periods in both years. Correlation tests were carried out for each environmental parameter and taxon abundance pair. A Monte Carlo resampling procedure was applied, and 9,999 iterations were used to quantify the p-values.
Outcomes of Mantel's and Spearman's tests were visualised through networks to clarify the emergence of patterns at the scale of the entire plankton community. A network is a spatial representation of associations, marked by lines (called 'edges') linking two entities (called 'nodes'). The nodes are the environmental factors, taxa, and ESD or taxonomic clusters, and the edges represent signi cant correlations (p-value < 0.05).

Declarations
ACKNOWLEDGMENTS Figure 1 Dynamics of Chl a and plankton abundances in 2015 (left) and 2016 (right). The green background represents bloom periods, and the white background represents non-bloom periods.