Plankton blooms over the annual cycle shape trophic interactions under climate change

Understanding species phenology and temporal co‐occurrence across trophic levels is essential to assess anthropogenic impacts on ecological interactions. We analyzed 15 yr of monitoring data to identify trends and drivers of timing and magnitude of bloom‐forming phytoplankton and diverse zooplankton taxa in the central Baltic Sea. We show that the timings of phytoplankton blooms advance, whereas crustacean zooplankton seasonal timings remain constant. This increasing offset with the spring bloom is linked to the decline of Pseudocalanus, a key copepod sustaining pelagic fish production. The majority of copepod and cladoceran taxa, however, are co‐occurring with summer blooms. We also find new developing fall blooms, fueling secondary production later in the season. Our study highlights that response to climate change differs within and between functional groups, stressing the importance of investigating plankton phenologies over the entire annual cycle in pelagic systems.

consequences for trophic interaction if the rate of response varies across trophic levels (Ji et al. 2010;Thackeray et al. 2016).Shifts toward earlier spring phytoplankton blooms are one of the most conspicuous phenological events in mid to high latitude marine systems driven by increased light availability and earlier stratification onset (Edwards and Richardson 2004;Yamaguchi et al. 2022).Additional phytoplankton blooms over the year are common in coastal systems (Winder and Cloern 2010), and have intensified with global warming (Dai et al. 2023).Similarly, zooplankton exhibit multiple phenological patterns with peaks occurring over the seasons (Calbet et al. 2001;Winder and Varpe 2020).Yet, most of the plankton phenology studies focus on spring dynamics (Thackeray 2012;Yamaguchi et al. 2022;Vives et al. 2024) because of their substantial contribution to annual production and well-defined bloom periods (Behrenfeld and Boss 2014;Hjerne et al. 2019).However, to predict temporal mismatches across trophic levels, understanding plankton phenology over the annual cycle and their response to the changing environment is critical but often understudied in marine systems.
Zooplankton growth and development is driven by temperature (Heinle 1969;Gillooly et al. 2002).Their cycles, however, are governed by a diversity of life history traits that determine their adaptations to seasonality (Adrian et al. 2006;Winder and Varpe 2020) and likely their ability to respond to changing phytoplankton phenology (Winder and Schindler 2004).Fast-growing zooplankton, such as rotifers, are the first zooplankton to peak following the spring bloom.Their potential for rapid population growth can lead to fast responses to food availability even at low temperatures and modulation of phytoplankton peaks (Allan 1976;Steinberg and Landry 2017;Yamaguchi et al. 2022).Similarly, marine cladocerans often overwinter as resting stages but longer development times from egg to first reproduction leads to peaks later in the season (Möllmann et al. 2002).Copepods, the dominant metazoan group in marine systems, have considerably longer developmental times and more complex life cycles, and often reach their maximum abundances in summer (Corona et al. 2024).Zooplankton have developed various life history strategies to optimize food availability and growth, and development may resume from resting stages or overwintering adult stages (Hedberg et al. 2024), all of them affecting their response to climate change.
In this study, we investigate plankton phenology throughout the year in the Baltic Sea, a seasonal system that experienced rapid changes due to increasing anthropogenic stressors (Reusch et al. 2018).This brackish sea comprises a unique biodiversity sharing similarities with both freshwater and marine systems (HELCOM 2023).Here, two main phytoplankton blooms fuel the food web.Spring blooms, mainly comprised of diatoms and dinoflagellates, which have advanced with temperature warming (Hjerne et al. 2019).Summer blooms, dominated by pico-and filamentous cyanobacteria, have overall intensified over the last decades (Kahru et al. 2016).
Although cyanobacteria are assumed to be of low food quality and potentially toxic, they are the main source of primary production for dominating zooplankton taxa (Novotny et al. 2023).We ask how phenology of diverse zooplankton respond to the changing biotic and abiotic environment.We first present trends of abiotic factors and bloom dynamics throughout the seasons of diverse plankton taxonomic and functional groups at three stations in the central Baltic Sea over the last 15 yr, and then examine predictors of interannual variation in phenology of diverse rotifer, cladoceran and copepod species.

Methods
Sampling was conducted within the Swedish National Marine Monitoring Programme at three stations in the central Baltic Sea (Fig. 1a) from 2007 to 2022 following a common approach (HELCOM 2017).At the northern-most station BY31, sampling occurred monthly during winter (October-February), weekly during the spring bloom (March and April) and biweekly during the remaining season, and at BY15 and BY5 once a month targeting the same week of the year (Supporting Information Fig. S1).For phytoplankton, integrated water samples were collected between 0 and 20 m deep at BY31 and between 0 and 10 m at BY15, BY5, and during the winter at BY31.Mesozooplankton were collected with a WP-2 net (90 μm) between 0 and 30 m deep, except Pseudocalanus elongatus that was collected between 30 and 60 m.Temperature and salinity were recorded over the entire water column using a CTD instrument, and are highest at the southern-most station and decrease following a south-to-north latitudinal gradient (Fig. 1b).
Data were retrieved from the Swedish Meteorological and Hydrological Institute data archive platform (SMHI 2023).Data processing and analyses were performed in R (R Core Team 2021) and visualized using ggplot2 (Wickham 2016).All parameters were linearly weekly interpolated from June 2007 to December 2022 using zoo (Zeileis and Grothendieck 2005).Produced datasets and scripts are publicly available (Jan et al. 2023).
Biomass (μg C L À1 ) of the total autotrophic and mixotrophic nano-and microorganisms between weeks 5 and 22 (February-May) were considered as "spring bloom" and between weeks 23 and 37 (June-mid-September) as "summer bloom" after inspection of average biomass over the full timeseries (Fig. 1c).Diatoms, dinoflagellates, cyanobacteria, and the mixotrophic ciliate Mesodinium rubrum were selected for taxa specific analyses (Supporting Information Table S1).Diatoms were separated as early (during the spring bloom) and late (after the summer bloom) diatoms.Dinoflagellates were selected during the spring bloom to capture the main bloom (Fig. 1c).Mesozooplankton biomass (μg L À1 ) was estimated from abundance data and established individual body masses (Supporting Information Table S2; HELCOM 2017).We selected the dominating taxa, including copepodite and adult stages of the copepods Acartia spp.(i.e., Acartia tonsa, Acartia bifilosa, and Acartia longiremis), Temora longicornis, Centropages hamatus, and P. elongatus that also formed the total copepod biomass, and the rotifer Synchaeta spp.(Fig. 1d).We also selected the cladocerans Bosmina coregoni and Evadne nordmanni that formed the total cladoceran biomass at BY31 only because low sampling resolution prohibited detections of clear seasonal patterns at the other stations.Further analyses were conducted at the species level, except for Acartia.Henceforth, zooplankton taxa are referred to by their genus names.
Interpolated data from 2008 onwards were used for the phenology metrics calculation.We used the center of gravity of the biomass (B) between the weeks (w) 2 and 51 to calculate seasonal timings (T) for phytoplankton blooms and zooplankton peaks (Eq.1).This approach is sensitive to change in seasonal cycle and allows to analyze multimodal dynamics by splitting the year into smaller periods (Edwards and Richardson 2004).Peak initiations and terminations were based on the 25 th and 75 th percentile of the cumulative seasonal biomasses, respectively (Ji et al. 2010), and peak magnitudes (M) were calculated as the average biomass between peak initiation (w i ) and peak termination (w t ) (Eq. 2).Monotonic trends within stations were assessed with the Mann-Kendall test using Kendall (McLeod 2022) and quantified with the Sen's slope (s) using mblm (Komsta Lukasz 2019) for plankton phenology and seasonal temperature and salinity between 2008 and 2022.
To assess effects of sampling frequency on phenological trends, we subsampled the data at BY31 with structured monthly sampling and compared phenological metrics, as described above, from the subsampled dataset with the full dataset.
We assessed drivers of interannual variability for zooplankton seasonal peak timings and magnitudes using generalized linear models with a Gaussian error distribution.Predictor and response variables were standardized using their z-scores among stations to reduce location effects.We used spring integrated water temperatures between 0 and 20 m depth (T C) and timings of the total phytoplankton spring (SpBT) and summer (SuBT) blooms as predictor variables for interannual variability in zooplankton peak timings.For zooplankton peak magnitudes (ln-transformed) we used integrated spring water salinity between 0 and 20 m depth (SAL) and T C, total summer phytoplankton bloom magnitude (SuBM, ln-transformed) and the trophic mismatch index (TMI).These predictor variables were selected because they are assumed to be the main abiotic and biotic drivers for zooplankton biomass (Vuorinen et al. 1998;Möllmann et al. 2000;Sommer et al. 2012;Thackeray 2012).TMI was calculated by subtracting the standardized zooplankton timings from the standardized spring bloom timings.For cladocerans, we used one predictor variable per model to avoid overfitting as they were only assessed at BY31.Seasonal peak timings were tested against T C and peak magnitudes (ln-transformed) against SuBM.For all models, the goodness of fit was estimated using the coefficient of determination R 2 .

Results
Monthly mean water temperatures and salinity in the upper 20 m depth showed a consistent tendency of increase across stations and season over the last 15 yr, albeit effects were significant mainly in summer and fall, likely due to the relatively short time series (Fig. 2; Supporting Information Fig. S2).We find consistency in phenological metrics calculated with the full dataset and the subsampled dataset in BY31 (Spearman's ρ ≥ 0.81), but trends showed lower significance levels at monthly sampling frequencies (Supporting Information Fig. S3).As such, we report phenological trends from BY31 here and refer to stations BY15 and BY5 in the Supporting Information.
Spring blooms, mainly comprised of dinoflagellates and to lesser extent diatoms, occurred on average in mid-April at all stations (mean AE SD, week 16 AE 1; Fig. 3a) and advanced by about 1.5 weeks per decade at BY31 (s = À0.16,p = 0.011; Fig. 4), while a trend was less clear at the other stations (Supporting Information Fig. S4).Earlier spring blooms are driven by dinoflagellates, which, on average occurred in week 16 AE 1 (Fig. 3b) and advanced by 0.72 week per decade (s = À0.07,p = 0.04; Fig. 4).Diatom blooms preceded the total spring blooms by about 2 weeks, with no change in magnitudes and timings at all locations.Blooms of Mesodinium occurred about 1.5 months after the phytoplankton spring blooms with no change in timings and magnitudes (Figs.3b, 4).A second prominent phytoplankton bloom occurred in mid-July (week 29 AE 1), dominated by cyanobacteria (i.e., Nostocales and Synechococcales) that contributed on average to more than half of the summer phytoplankton biomass (Figs. 1c,3b).At BY31, the summer bloom initiation advanced by about 1 week per decade (s = À0.09,p = 0.027) and the peak magnitude increased (s = 0.13, p = 0.023; Fig. 4).Conversely, peak magnitudes showed a tendency of decrease at the other stations, driven by Synechococcales in the southern-most station (s = À0.13,p = 0.048; Supporting Information Fig. S4).Additional diatom blooms were observed in November (week 45 AE 1; Fig. 3b) when water temperature decreased (Fig. 1b), nutrient concentration increased and the stratification weakened (Supporting Information Fig. S5), although with on average relatively low magnitudes across stations (Supporting Information Table S3).Late diatoms showed a tendency to advance across stations, but this was only significant at BY15 (s = À0.06,p = 0.037; Supporting Information Fig. S4).Moreover, at BY31, fall diatom peak magnitudes were low in the beginning of the study period and increased significantly over the last years (p = 0.001; Fig. 4).
For zooplankton, the rotifer Synchaeta was the first to peak after the spring blooms with seasonal peak timings on average in mid-June (week 25 AE 4; Fig. 3) showing no significant trend despite a negative Sen's slope (Fig. 4).All crustacean species co-occurred with the summer blooms, with steady seasonal timings (Figs. 3, 4; Supporting Information Fig. S4).Total copepod and taxa specific peak magnitudes stayed unchanged, except for Pseudocalanus that declined at all stations (s ≤ À0.06, p ≤ 0.05; Fig. 4; Supporting Information Fig. S4).The cladoceran Bosmina was the only zooplankton showing significant increasing peak magnitudes over the study period (s = 0.1, p < 0.001; Fig. 4).

Discussion
Using a high taxonomy resolution dataset over the entire annual cycle, we find that spring phytoplankton blooms are occurring earlier, while zooplankton seasonal timings stayed stable over the past 15 yr.This created an increasing temporal mismatch between mesozooplankton and the spring phytoplankton bloom that is related to interannual variability of rotifer biomass and to declines of the deep-dwelling copepod Pseudocalanus.A large majority of zooplankton taxa co-occur with summer phytoplankton blooms that exhibited different dynamics across locations and favored cladoceran production.This highlights the complexity of seasonal plankton dynamics and the importance of considering the full productive season in phenology studies.
In contrast to phytoplankton, seasonal timings of all zooplankton species were remarkable steady over the investigated period, causing an increasing temporal mismatch with the spring bloom.This offset and warmer temperature explain the decline of Pseudocalanus, a key prey species for planktivorous fish in the central Baltic Sea (Möllmann et al. 2003;Novotny et al. 2022), that may be related to earlier sedimenting spring phytoplankton, causing a mismatch with their reproduction and growth of nauplii that commence in spring (Supporting Information Fig. S6; Renz et al. 2007).The relationship between Pseudocalanus biomass and the temporal mismatch with the spring bloom adds a new hypothesis explaining these declines that were previously linked to decreased salinity and increased predation pressure (Ojaveer et al. 1998;Möllmann et al. 2000).Similarly, Synchaeta magnitudes are dependent on the timings of the spring bloom as the mismatch explains interannual variability of their peak biomass.However, Synchaeta occur earlier in years with warmer spring temperature, suggesting that this rotifer is shifting in synchrony with spring phytoplankton blooms, despite a non-significant trend in timing.This is supported by stable Synchaeta densities with a tendency of increase over the past 15 yr.For Synchaeta, a longer sampling period is likely needed before the trends become statistically significant (Stegman et al. 2017).This highlights that phenological drivers differ among and within plankton functional groups with implication for trophic coupling and food web dynamics.
The "earlier when warmer" response gives to the rotifers the opportunity to match shifting spring blooms, whereas the temperature response varied among crustacean plankton.Acartia was the only copepod to respond to temperature.This genus is comprised of three species that peak at different time throughout the summer (Winder and Varpe 2020) and warmer spring temperature may change their relative biomass importance.However, seasonal timings of the remaining copepods were stable and unrelated to the observed range of interannual variability in temperature, suggesting that their seasonal dynamics are controlled by different life history strategies (Adrian et al. 2006;Winder and Varpe 2020).Some zooplankton, including Temora and Pseudocalanus, spend the winter as active adult or late juvenile stages (Renz and Hirche 2006;Dutz et al. 2010), while others produce overwintering resting eggs (Katajisto et al. 1998;Möllmann et al. 2002) whose emergence is often triggered by photoperiod and temperature.Despite sharing the same overwintering strategy, Temora and Pseudocalanus responses to shifting spring blooms varied, which may be explained by their voltinism.Pseudocalanus require an entire year to complete their life cycle (Renz et al. 2007), thus are likely more reliant on spring blooms for successful recruitment.In contrast, Temora develop several generations throughout the productive season and have more opportunities to balance a mismatch with the spring bloom (Dutz et al. 2010).This indicates that Temora and crustacean zooplankton emerging from resting stages are less dependent on spring blooms and rely on summer primary production.Moreover, our results suggest that zooplankton phenology is strongly regulated by intrinsic biological drivers.
Interestingly, spring blooms shift uniformly, suggesting that their durations and magnitudes are mainly constrained by nutrient concentration (Supporting Information Fig. S5; Sommer et al. 2012).The temporal decoupling between crustacean plankton and the spring bloom can be explained by the fact that spring blooms occur at temperature below 5 C, restricting growth of crustacean zooplankton, which have relatively long and, especially copepods, complex life cycles (Huntley and Lopez 1992;Lee et al. 2003).This observation is in line with copepod peak timings in the North Sea (Edwards and Richardson 2004), and particularly the decoupling between Calanus finmarchicus and the spring bloom (Carlotti and Radach 1996).Conversely, our results confirm temporal co-occurrence of crustacean zooplankton with summer phytoplankton blooms and agree with the supporting role of summer phytoplankton for secondary production via the microbial food web or direct consumption (Sommer et al. 2012;Novotny et al. 2023;Serandour et al. 2023).We further find newly occurring diatom blooms in fall in the northern-most station, indicating that fall blooms have shifted northward, extending the productive season, likely favoring secondary production and even fish spawning later in the year (Reygondeau et al. 2015).Surprisingly, fall diatom blooms are occurring earlier, likely because increasing fall temperatures lower the rate at which the mixed layer depth deepens, promoting nutrient input from deep-water layers and phytoplankton growth (Findlay et al. 2006).
Although the Baltic Sea is a unique system with low salinity, plankton seasonality and species composition are similar to other temperate marine systems.The earlier appearance of spring blooms in the Baltic Sea is consistent with observations in the North Atlantic Ocean (Zhai et al. 2011;Henson et al. 2018) and predictions in latitude > 40 N (Asch et al. 2019;Yamaguchi et al. 2022).Similarly, the spatially contrasted dynamics of summer phytoplankton blooms with earlier occurrence, and increased magnitude only in the northern-most station are in line with reported trends in the Baltic Sea and other systems (Cloern et al. 2016), and suggest that summer blooms are driven by local environmental conditions (Wasmund et al. 2011;Griffiths et al. 2020).Moreover, the discrepancy of phenological significant shifts across the central Baltic Sea is partly explained by differences in sampling frequencies (Stegman et al. 2017).High temporal sampling resolution in the northern-most station with weekly sampling during the spring bloom and biweekly thereafter, allowed the detection of clear spring, summer and fall dynamics, whereas monthly sampling identified similar patterns but often not statistically detectable.
Climate projections on plankton phenology typically assume one zooplankton group that synchronously follows the spring phytoplankton dynamics (Henson et al. 2018;Asch et al. 2019;Yamaguchi et al. 2022).However, our findings challenge this assumption and show that multiple pulses of primary production support secondary production over the seasons and that different temperature responses among zooplankton taxa need to be considered for projecting trophic interactions.Including the role of microzooplankton in counterbalancing trophic mismatch will be important directions for linking phytoplankton shifts and zooplankton performance.Our results imply that climate change indirectly affect zooplankton production via shifts in primary production dynamics, but the consequence of trophic mismatch varies depending on the species-specific adaptations to the seasonal cycle.To broaden our understanding of the ecological consequences of climate change in pelagic systems, our study highlights that plankton phenologies need to be considered over the entire annual cycle.

Fig. 1 .
Fig. 1.Seasonal dynamics of abiotic parameters and plankton biomass in the central Baltic Sea from 2008 to 2022.(a) Sampling stations in the central Baltic Sea.(b) Average monthly temperature and salinity from 0 to 20 m depth; vertical bars represent interannual monthly range (min-max).(c) Average annual biomass for major phytoplankton taxonomic groups and Mesodinium, and total phytoplankton biomass.(d) Average seasonal biomass for dominant zooplankton genera and total zooplankton biomass.Data in (c) and (d) are weekly interpolated and smoothed using a kernel density estimate.Dashed lines denote time periods for spring, summer, and fall phytoplankton blooms.

Fig. 2 .
Fig. 2. Interannual temperature and salinity dynamics and trends at BY31 from 2008 to 2022.(a) Average values (z-scores) and (b) Sen's slopes for winter (January, February, March), spring (April, May, June), summer (July, August, September), and fall (October, November, December).The significance levels are based on the Mann-Kendall test and error bars represent the 95% confidence interval.

Fig. 3 .
Fig. 3. Plankton succession in the central Baltic Sea from 2008 to 2022.Seasonal timing, peak magnitude and bloom duration of (a) total spring and summer phytoplankton and total copepod and cladoceran biomass, and (b) dominating phytoplankton and zooplankton taxa.Point positions (x-axis) correspond to the average peak timing (week), point sizes to the average peak magnitude (zooplankton: μg L À1 ; phytoplankton: μg C L À1 ) and the horizontal bars to the average initiation and termination (week) of their respective growing seasons.

Fig. 4 .
Fig. 4. Interannual plankton phenologies of major taxonomic groups at BY31 from 2008 to 2022.Interannual variability of phytoplankton and zooplankton seasonal timings and peak magnitudes (z-scores) for (a) phytoplankton and (b) zooplankton.(c) Sen's slopes of plankton peak initiation, peak timing, peak termination, and peak magnitude.The significance levels are based on the Mann-Kendall test and error bars represent the 95% confidence interval.Taxa on the x-axis are ordered according to the timing of their seasonal peak.

Fig. 5 .
Fig. 5. Effects of abiotic and biotic drivers for zooplankton peak timings and peak magnitudes.Effect size is indicated by the standardized regression coefficients AE standard errors of the model output.Predictor variables are the z-scores of spring temperature (T C), spring salinity (SAL), peak timings of the phytoplankton spring bloom (SpBT) and summer bloom (SuBT), the trophic mismatch index (TMI), and summer bloom peak magnitudes (SuBM).Note that for Bosmina and Evadne only one predictor variable is used due to a limited sample size.