The in situ mesocosm experiment was conducted at the Facility for the Study of Ocean Acidification Impacts of Xiamen University (FOANIC-XMU) located in the subtropical coastal region Wuyuan Bay of southern East China Sea (24.52° N, 117.18° E) from 9 October (day 0 relative to algal inoculation) to 14 November, 2019. Nine cylindrical and transparent thermoplastic polyurethane (TPU) mesocosm bags, each 3 m deep and 1.5 m in diameter, were filled with approximately 3000 L of in situ seawater that had been prefiltered (MU801-4T, Midea, China, pore size of 0.01 µm). The mesocosms were hooked to and secured within steel frames. Two pCO2 treatments were established to investigate the effects of ocean acidification on the in situ phytoplankton community: an ambient pCO2 treatment (AC, 410 µatm; 4 numbered bags) and a high pCO2 treatment representative of end-of-century projections (HC, 1000 µatm; 5 numbered bags). To adjust CO2 in seawater in the HC bags to the projected 1000 µatm in 2100s, approximately 11 L of CO2-saturated seawater was added to each HC bag. The AC and HC pCO2 levels were maintained by bubbling with ambient air and pre-mixed air-CO2 (1000 µatm CO2), respectively, at a rate of 5 Lmin-1 using a CO2 Enricher (CE-100B, Wuhan Ruihua Instrument & Equipment Ltd, China). After the carbonate system had been stabilized (leveled pH), 720 L of in situ seawater were filtered by a 180 µm mash to exclude large zooplankton, and each mesocosm bag was inoculated with 80 L of it to initiate the coastal microbial community. Samples were taken from each bag at a depth of 0.5 m at 9:00 a.m. by niskin bottles every 1–3 d for physical, chemical and biological analysis.

Solar light intensity was monitored every minute throughout the experimental period using a real-time solar irradiance monitoring device (EKO, Japan). Salinity, temperature and pH in each mesocosm were measured with a salinometer, a digital thermometer and a pH meter (Orino 2 STAR, Thermo Scientific, USA, calibrated with standard NBS buffer), respectively. Dissolved inorganic carbon (DIC) was sampled and measured using an Environmental Water Analyzer (Ma et al., 2018), and total alkalinity (TA) measured using an Automated Spectrophotometric Analyzer (Li et al., 2013). Other seawater carbonate chemistry parameters were calculated with CO2SYS software with known parameters of pCO2, salinity, pH, temperature, and nutrient concentration.

Nutrient samples of each bag were filtered through cellulose acetate membrane (0.22 µm, 47 mm), and the filtrate was divided into 2 subsamples; one was stored at -20 °C for the measurement of NO3- + NO2-, PO43-, and NH4+; another stored at 4 °C for the measurement of SiO32-. Measurement of NO3- + NO2-, PO43-, and SiO32- concentration was conducted using an auto-analyzer (AA3, Seal, Germany), NH4+ was measured with indophenol blue spectrophotometry using a spectrophotometer (Tri-223, Spectrum, China) at 25 °C. NO2- and NO3- were analyzed using the copper-cadmium reduction method, the standard concentrations used for the NO2- + NO3- calibration curve were 0, 1.04, 2.08, 4.16, 10.4, 20.8, and 41.6 µM, and those for NO2- calibration curve were 0, 0.04, 0.08, 0.16, 0.4, 0.8, and 1.6 µM (Dai et al., 2008). PO43- and SiO32- were measured using typical spectrophotometric method (Intergovernmental Oceanographic Commission, 1994), and the calibration curves of both parameters were prepared using standard concentrations of 0, 0.08, 0.16, 0.32, 0.8, 1.6, 3.2 µM and 0, 4, 8, 16, 40, 80, 160 µM, respectively.

Water samples of each mesocosm (100–1000 mL, depending on the biomass in the mesocosms) were filtered onto GF/F filter (Whatman, United States) by suction filter with low vacuum pressure (< 0.02 Mpa) and soaked in 5 mL pure methanol overnight. The extracts were centrifuged at 8000 × g and 4 °C for 10 min, then the absorption spectra of supernatants from 400 to 800 nm were measured using a UV-VIS spectrophotometer (DU 800, Beckman, USA). The Chl a concentration was calculated according to the following equation (Ritchie, 2006): Chla(µgL-1)=16.29×(A665-A750)-8.54×(A652-A750) where A750, A665, and A652 represents the absorbance of Chl a at wave length 750, 665, and 652 nm, respectively.

For the analysis of particulate organic carbon (POC) and nitrogen (PON) across two particle size fractions, water samples of known volume were first filtered through a 20 µm mash to obtain subsamples containing particle organic matters smaller than 20 µm. Particles larger than 20 µm (retained on 20 µm mesh) were backwashed using an equal volume of prefiltered (0.22 µm) in situ seawater, yielding in subsamples containing particulate organic matters larger than 20 µm. All subsamples were then filtered on pre-combusted (450 °C, 6 h) GF/F filter (Whatman) and stored at -20 °C until analysis. Before analyses, all filters were fumed over pure HCl for 12 h and dried at 60 °C to remove inorganic carbon, then packed in tin cups and measured with a CHNS elemental analyzer (Vario EL cube Elementar, Germany).

Before sunrise, 120 mL of water samples from each mesocosm were collected and dispensed into six 25 mL borosilicate bottles (three bottles for 12 h incubations, and three for 24 h incubations). For each culture duration of each mesocosm, two bottles were illuminated under natural light and one bottle was wrapped tightly in aluminum foil as a dark control. After incubation, cells were filtered onto the GF/F filters (Whatman) under dim light and stored at -20 °C. Before measurement, filters were placed individually in 20 mL scintillation vials and exposed to HCl fumes overnight, dried at 60 °C for over 6 h to remove any unincorporated H14CO3-. The incorporated 14C by algae was counted with a liquid scintillation counter (Beckman, LS6500, Germany) in the presence of 5 mL scintillation cocktail (Hisafe 3, Perkin-Elmer, United States). Nighttime respiratory carbon loss was calculated as the difference between carbon fixation over 12 h (daytime primary production) and 24 h (daily net primary production).

Water samples (500–2000 mL) from each mesocosm were collected into polyethylene bottles and fixed with 1.5 % acidic lugol's iodine. The samples were statically placed for 2–3 d and concentrated into 50 mL subsamples in the centrifuge tube using siphons within 3 d, then examined with a microscopy (Nikon Eclipse Ns2) and a plankton counting chamber to assess phytoplankton abundance and diversity based on the morphological characteristics as previous described (Hasle and Syvertsen, 1997; Steidinger and Jangen, 1997; Yang and Liu, 2018). To distinguish whether dinoflagellates are autotrophic or heterotrophic, we observed living algal cells in the unstained water samples under the microscope for cell transparency and the presence of chloroplasts.

An aliquot of 100 µL for each mesocosm was loaded onto a counting chamber for microscopic enumeration. In each aliquot, the count was deemed valid only when the total number of cells exceeded 200; otherwise, the subsample volume for microscopy was increased to achieve sufficient counts. For samples collected during the exponential growth phase that exhibited excessively high cell densities, appropriate dilution with 0.22 µm-filtered, sterilized seawater was performed prior to counting (State Oceanic Administration, 2005).

The data were all expressed by the mean and standard deviation (means ± SD) and plotted by Origin 2024. Independent-samples t test was conducted to check the significant effects of increased pCO2 at the level of p < 0.05 using SPSS 19. To evaluate α-diversity, Shannon diversity index was calculated based on the relative abundances of phytoplankton taxa using the estimate R and diversity functions from the vegan package (version 2.6-4) in R (Version 4.2.2). Shannon index incorporates both species diversity and evenness. Patterns of physiological parameters over time were emphasizing using generalized additive models (GAMs) and constructed using the `mgcv' package in R to analyze changes in physiology through the experiment.

Throughout the experiment, most days were sunny, with daytime mean PAR (12 h-average photosynthetic active radiation) ranging from 200 to 850 µmolphotonsm-2s-1 (Fig. S1 in the Supplement). The environmental temperatures decreased gradually from 26.7 ± 0.05 °C at day 0 (9 October) to 21.1 ± 0.28 °C at day 38 (14 November) (Fig. 1a). Significant differences in pHNBS and pCO2 between HC and AC were maintained throughout most of the experimental period, while there's no significant difference in total alkalinity (TA) between the two pCO2 treatments (p = 4.02 × 10⁻⁸, 2.87 × 10⁻¹², 0.549, respectively. Figures 1b–d and S5a–c).

Following a sharp increase in phytoplankton biomass from day 4 to day 8 (Fig. 3a), the pHNBS in the HC and AC mesocosms increased and peaked at 8.24 ± 0.16 and 8.56 ± 0.14, respectively (Fig. 1b). Correspondingly, the pCO2 value dropped to the lowest points of 238.48 ± 49.02 µatm (HC) and 82.82 ± 32.88 µatm (AC) (Fig. 1c). Then, as the phytoplankton biomass decreased after day 8, pHNBS gradually declined and pCO2 increased, both stabilizing at relatively constant levels from day 18 until the end of experiment. There were no obvious temporal changes observed in total alkalinity (TA) throughout the experiment (Fig. 1d).

The initial nutrient concentrations reflected the eutrophic condition in the coastal seawater (NO3- + NO2-: 27 µM, PO43-: 1.4 µM). In the mesocosm bags, nutrient concentrations declined dramatically in the early phase (up to day 8, Fig. 2). The NO3- + NO2- and NO3- concentrations decreased sharply to nearly 0 by day 8 (Fig. 2a and b). In contrast, the NO2- concentration experienced a slight increase on day 2, then declined to nearly 0 by day 8 and remained at a low level until the end of experiment in both HC and AC mesocosms (Fig. 2c). Under HC condition, the NO3- concentration decreased more rapidly than that under the AC until day 20, although the difference was not significant (p = 0.423, Fig. S5d). Both NH4+ and PO43- concentrations dropped to nearly 0 after 4 d and remain relatively stable thereafter. There were no significant difference observed between HC and AC mesocosms (p = 0.579 and 0.631, respectively, Figs. 2d, e and S5e, f). The SiO32- concentration decreased from 41.81 ± 0.48 µM in HC and 42.88 ± 0.91 µM in AC on day 0 to a minimum of 4.62 ± 3.82 µM and 7.79 ± 5.52 µM by day 8, respectively. Thereafter, SiO32- concentrations in both HC and AC mesocosms gradually increased until the end of experiment (p = 0.343, Figs. 2f and S5g).

Phytoplankton biomass, indirectly indicated by Chl a concentration, increased to peak at 35.88 ± 3.25 µgL-1 in the HC and 36.54 ± 4.88 µgL-1 in the AC mesocosms on day 8, and then gradually decreased to 1.12 ± 0.43 µgL-1 in HC by day 14 and 0.58 ± 0.31 µgL-1 in AC by day 16, followed by a slight increase by the end of experiment (Fig. 3a). Based on the natural logarithm (ln⁡) scale of Chl a concentration, the phytoplankton growth kinetics under the two pCO2 treatments showed the following phases (Fig. 3b): the exponential phase (from day 0 to day 5), the stationary phase (from day 6 to day 10), the decline phase (from day 11 to day 16), and a second exponential phase from day 17 to day 24 in the HC and to day 20 in the AC mesocosms. Then, phytoplankton assemblages in the HC mesocosms entered a second stationary phase until the end of experiment, while in the AC ones, they entered a decline phase until day 29, followed by a slight increase on day 32.

Throughout the experiment, the elevated pCO2 resulted in higher average value of Chl a concentration at most sampling times, although the differences were not statistically significant (p = 0.142, Fig. S5h).

The primary production and night-respiratory per water volume showed patterns similar to those of phytoplankton biomass (indicated by Chl a concentration) (Fig. 4a–c). They reached their maximal values on day 6, which corresponded to the end of exponential phase. As the phytoplankton communities entered the stationary phase, daytime (12 h) primary production, daily (24 h) net primary production and nighttime respiration per water volume progressively decreased, and then slightly increased again when the phytoplankton communities underwent the second exponential phase. The elevated pCO2 increased both daytime and daily net primary production during the middle phase of the experiment, although the positive effect on 24 h primary production tended to decline by the end of experiment (p = 0.038 and 0.012, Fig. S6a and b). The nighttime respiration of phytoplankton was suppressed before day 8 and enhanced thereafter under the elevated pCO2, though no significant difference was observed (p = 0.444, Fig. S6c).

Primary productivity per Chl a increased sharply on day 4, and decreased to the lowest values on day 8. On day 12, both daytime and 24 h primary productivity in the HC increased drastically and then remained relatively stable until the end of experiment (Fig. 4d and e). In contrast, two additional peaks were observed in the AC mosocosms on days 16 and 26. The elevated pCO2 appeared to have enhanced primary productivity from day 2 to day 20, though these effects were not statistically significant (p = 0.946 for daytime and p = 0.985 for 24 h, Figs. 4d, e and S6d, e).

Nighttime respiration per µg Chl a initially increased on day 4, then decreased to nearly zero in both the HC and AC mesocosms on day 8 and remained relatively stable till the end of experiment. The elevated pCO2 had a negative effect on phytoplankton respiration before day 12, but increased it thereafter, though no significant difference was observed between the HC and AC treatments (p = 0.834, Figs. 4f and S6f).

A total of 47 genera identified microscopically include 33 genera of diatoms, 7 of dinoflagellates, 2 of cyanobacteria, 2 of chlorophyta, 2 of cryptophyta and 1 of euglenophyta. In all mesocosms, the dominant species included Cerataulina pelagica, Eucampia cornuta, Guinardia delicatula, Leptocylindrus danicus, Skeletonema costatum, Protoperidinium sp., Gyrodinium spirale, Cryptophyta sp. and Pyramimonas sp. (Fig. S4).

Phytoplankton communities underwent dynamic succession in the mesocosms (Fig. 5). Diatoms (mainly Cerataulina pelagica) dominated the phytoplankton communities during the early and middle stages of the experiment, as indicated by the similar temporal trends in total phytoplankton and diatom cell counts compared with Chl a concentration (Figs. 5a, b and S4a). There was no significant difference in diatom density between the HC and AC mesocosms (p = 0.259, Fig. S7a), although the average value was lower in the former than in the latter treatment. Autotrophic dinoflagellates began to emerge on day 8 and rapidly declined on day 12 in both HC and AC enclosures (Fig. 5c). Except for days 6 to 18, the elevated pCO2 increased the biomass of autotrophic dinoflagellates, though the difference was insignificant (p = 0.505, Fig. S7b). Hetero-dinoflagellates began to emerge on day 6, with their abundance peaked on day 12 in the AC and on day 14 in the HC mesocosms, then decreased by day 22. The elevated pCO2 did not result in any significant change in terms of their cell numbers (p = 0.785, Figs. 5d and S7c). On day 26, the biomass of hetero-dinoflagellates increased again in the HC treatment, while it remained constant in the AC treatment (p = 0.729, Independent-samples t test).

The biomass of small taxa (Cyanobacteria, Chlorophyta, Cryptophyta and Euglenophyta) started to increase on day 8, the HC treatment significantly increased the total biomass of these small phytoplankton species thereafter (p = 0.019, Figs. 5e, S4h, i and S7d). From day 22, when diatoms biomass decreased to the lowest level, the temporal variation in small taxa biomass became the main factor controlling overall phytoplankton dynamics (Fig. 5e and S4h, i). Accordingly, the positive effect of HC on the small phytoplankton species led to an earlier transition of phytoplankton from the large diatoms and dinoflagellate (mainly fall within the micro size fraction) to the smaller ones (Fig. 6a and b). This accelerated transition in the HC treatment was also evidenced by higher concentration of POC and PON in the < 20 µm fraction and lower concentration in the > 20 µm fraction (Figs. S2 and S3).

In both HC and AC mesocosms, Shannon diversity index decreased sharply from day 2, reaching the lowest values on day 8 in AC mesocosms and on day 10 in HC mesocosms (Fig. S8a). Before day 22, Shannon diversity index increased under elevated pCO2, whereas it is lowered under elevated pCO2 level since day 24, although the differences were not statistically significant (p = 0.161, Fig. S8b).