Experimental design and study area

A 53-day experiment was conducted in the mesotrophic environment of Raunefjord near Bergen, Norway (60°15^’55^” N, 5°12^’21^” E) during the post-spring-bloom period (May-July 2022). This setting was selected, as mesotrophic conditions are typical for many temperate shelf and fjord systems, which are potential targets for nearshore OAE.

A gradient-based design was used in which TA was elevated in increments of 150 µmol per kilogram of seawater, resulting in five ΔTA levels: 0, 150, 300, 450, and 600 μmol kg^-1 (S1 Fig) in order to test four different alkalinity levels and stay well below commonly discussed safety boundaries of ΔTA < 1000 µmol kg^-1; pH < 9 [62]. The selected range also provides adequate resolution to detect non-linear responses and exceeds natural short-term TA variability, ensuring statistical detectability of treatment effects.

To remain within the logistical limits of large volume mesocosms, one mesocosm was assigned to each ΔTA level. This choice sacrifices treatment-level replication but provides higher gradient resolution and more closely reflects the spatial heterogeneity expected in real-world OAE deployments. This design aimed to identify safe thresholds for OAE applications with minimal ecological disruption to the planktonic community [63–66]. Furthermore, this study compared the effects of two mineral-based OAE strategies, calcium- and silicate-based applications.

Ten mesocosms, each 22 m long (∼21.5 m water column plus a 0.5 m sediment trap) with a 2 m diameter (∼ 60,000 L volume), were deployed [63]. On day 0, each mesocosm was filled with fjord water, retaining the natural planktonic community but excluding organisms > 3 mm. This truncation followed KOSMOS protocols and was applied to reduce chance-driven differences among mesocosms caused by the random enclosure of rare large predators (e.g., fish, jellyfish), which can introduce strong between-mesocosm variability and obscure treatment effects [67]. Predation pressure was instead standardized by the controlled addition of 95 laboratory-reared Atlantic herring (Clupea harengus) larvae per mesocosm. Mixing was passive and maintained fjord-typical vertical structure (including episodic stratification), driven primarily by wave-induced surface turbulence and convective adjustment to ambient temperature changes. Light followed ambient diel conditions. Five mesocosms were assigned to calcium-based OAE (Ca-based) treatments and five to silicate-based OAE (Si-based) treatments. On day 6, TA in both treatments was increased along a 0–600 μmol kg^-1 gradient by adding varying amounts of sodium hydroxide (NaOH) dissolved in 20 L of Milli-Q water.

In Ca-based mesocosms, calcium chloride (CaCl₂) was co-added in proportions equal to half the TA enhancement (i.e., Δ[Ca²⁺] = ½ ΔTA; equations 1−2). In Si-based mesocosms, magnesium chloride (MgCl₂) was added similarly (Δ[Mg²⁺] = ½ ΔTA). To ensure that ΔTA remained the sole driver and to avoid conflating carbonate chemistry with micronutrients or particle effects, sodium metasilicate (Na₂SiO₃) was added at 75 µmol L^-1 to all Si-based treatments, rather than scaled with ΔTA. A Na₂SiO₃ gradient would (i) create a co-gradient in Si with independent ecological effects, (ii) approach metasilicate solubility and promote colloid formation at higher doses, and (iii) increase the risk of CaCO₃ precipitation (equation 3−6). Because each mole of Na₂SiO₃ contributes two equivalents of TA, the Si‐control mesocosm received 150 µmol kg^-1 HCl to counterbalance the added TA [41,68]. All reagents were pre-dissolved and added using a vertically hauled, specially designed, and commonly used mesocosm-mixing and manipulation device [67]. First post-OAE manipulation samples were collected during the regular sampling on the following day. We acknowledge that potential slightly higher carbonate chemistry perturbations within the 1^st 24-hour period upon alkalinity enhancement, than those reported here.

(1)

(2)

(3)

(4)

(5)

(6)

The phytoplankton community was in a post-bloom stage and nutrient-limited from the start of the 53-day experiment. To alleviate the progressive nutrient limitation, establish a clear contrast between an initial nutrient-depleted phase and a subsequent nutrient-repleted phase, macronutrients were added to all mesocosms on days 26 and 28. The resulting dissolved nutrient concentrations are shown in S2 Fig.

Additional details on the study area, mesocosm infrastructure, carbonate chemistry, and the controlled addition of herring larvae are provided by Goldenberg et al. [68], Ferderer et al. [41], and Marín-Samper et al. [69].

All permits, including site access permits, were obtained in accordance with national legislation; no additional permits were required for this study.

Sampling procedure, carbonate chemistry, nutrients, and chlorophyll a measurements

Sampling was carried out every second day throughout the experiment, followed by a consistent sequence of CTD profiling, water collection, and plankton net tows. Vertical profiles of temperatures, salinity, pH, dissolved oxygen, chlorophyll a, and photosynthetically active radiation (PAR) were obtained with a hand-held self-logging CTD probe (CTD60M, Sea and Sun Technologies, Trappenkamp, Germany). Immediately afterwards, depth-integrated water samples were collected with an integrating water sampler (IWS, HYDRO-BIOS, Kiel, Germany), providing a column-averaged sample from each mesocosm. Subsamples were directly taken from the IWS for parameters susceptible to gas exchange or contamination, namely dissolved inorganic nutrients (NO₃⁻, NO₂⁻, NH₄, Si(OH)₄, PO₄³⁻), dissolved inorganic carbon (DIC), and total alkalinity (TA). For nutrient analysis, triplicate subsamples were taken to account for technical variability. These were passed through 33 mm, 0.45 µm PES membrane syringe filters (Filtropur S. SARSTEDT, Nümbrecht, Germany) within 1–2 h of collection, stored in the dark at ambient temperature, and analysed within 6 h. Concentrations of nitrate, nitrite, phosphate, and silicate were determined spectrophotometrically following Hansen and Koroleff [70], whereas ammonium was measured fluorometrically with a 10-AU fluorometer (Turner Designs, San Jose, CA, USA) according to Holmes et al. [71].

The remaining IWS water was transferred into 10 L plastic canisters, transported to the laboratory, and stored at in situ water temperature in a temperature-controlled room in the dark until processing later the same day. From these canisters, subsamples were taken for total particulate carbon (TPC), particulate organic carbon (POC), nitrogen (PON), total particulate phosphorus (TPP), biogenic silica (BSi), chlorophyll a, phytoplankton pigments, and microscopic counts of phytoplankton and microzooplankton.

For carbonate system characterization, TA, DIC, and total-scale seawater pH were measured, and the remaining carbonate parameters were calculated with CO2SYS using in situ temperature and salinity obtained from CTD with constants from Luecker et al. [72]. TA and pH were sampled every second day. TA was determined by potentiometric two-step titration (Metrohm 862 Compact Titrosampler, 0.05 M HCl, Aquatrode Plus Pt1000, 907 Titrando) following Chen et al. [73]. Seawater pH samples were equilibrated to 25 °C in a thermostatic bath and measured spectrophotometrically at 25 °C with a VARIAN Cary 100 and a 10 cm cuvette [74]. DIC was measured on day 9 with an AIRICA system (MARIANDA, Kiel, Germany) coupled to a LI-7000 differential gas analyser (LI-COR Biosciences GmbH, Bad Homburg, Germany) at room temperature within 12 h of sampling [75,76].

For chlorophyll a, 500−1000 ml subsamples were filtered onto GF/F glass fibre filters (Whatman Merck, Darmstadt, Germany), under low-light conditions. Filters were placed in plastic vials, frozen at −80°C, and analysed fluorometrically on the following day according to Welschmeyer [77].

Larvacean abundance

Larvacean abundances were estimated along with the entire zooplankton community from samples, which were collected every four days using an Apstein net (mesh size: 55 µm, diameter: 17 cm). The net was vertically hauled to a fixed depth of 21.5 m, filtering a total volume of ~469 L per tow. The collected samples were transferred to 500 mL jars using filtered seawater and stored in a cooler during transport to the laboratory. In the laboratory, samples were size fractionated into three categories using mesh filters: small (55–200 µm), intermediate (200–500 µm), and large (>500 µm) zooplankton. Samples were preserved in 70% ethanol. Taxonomic identification and enumeration were carried out on-site using a Leica S9i stereomicroscope.

House production and prey clearance rates estimation

On day 36, O. dioica specimens were collected from each mesocosm for laboratory incubation experiments to estimate house production and prey clearance rates. Surface water from each mesocosm was transferred into 5 L transparent PP beakers to individually pick O. dioica within their mucus houses using a wide-mouthed glass pipette. Larvaceans that showed visible gonads were excluded to avoid reproduction during incubations.

For house production rate measurements, individual larvaceans were transferred into 100 mL glass bottles pre-filled with 55 µm filtered seawater from the corresponding mesocosm. For prey clearance measurement, two 250 ml incubation bottles were prepared per mesocosm: one containing only 55 µm filtered seawater served as a control for phytoplankton growth in the absence of grazers, and the other containing three O. dioica individuals of similar size. All glass bottles were closed without trapping air bubbles, time was noted as the incubation start time-point, and the bottles were immersed in in situ seawater in a cooling container and transported to the laboratory. In the laboratory, bottles were mounted on a plankton wheel and incubated for ~24 hours in a temperature-controlled room set at 11°C, mimicking in situ conditions. At the end of the incubation, larvaceans were gently removed.

For the house production estimates, the removal of the animal within its house represented the termination time-point and was used to calculate house production rates throughout the experiment, expressed per hour. All O. dioica were individually retrieved and examined under a stereomicroscope (ZEISS Stemi 508) to measure their trunk lengths. Afterward, the incubation bottle content (100 mL) was inspected under a microscope to count all discarded houses.

For clearance rate assays, paired control and grazed bottles were run per mesocosm; both bottles were filled simultaneously from the same mesocosm water (collected using an integrated water sampler), ensuring identical initial prey concentrations. Rather than measuring initial concentrations from each bottle, we used concurrent mesocosm flow‐cytometry data, an acceptable proxy given the shared source water, to establish starting prey levels. At the end of each ~24 h incubation (11°C), phytoplankton counts were determined on a flow cytometer (BD Accuri C6). At 11°C, net phytoplankton growth typically does not exceed 0.1 d^-1, which over a 24 h period corresponds to a < 10% change in control counts [78], so any growth‐related change in the control is small relative to grazing losses. Accordingly, we estimated apparent clearance rates using Frost’s [79] equation without a growth-correction [80], attributing the difference in final prey concentration between control and grazing bottles, assuming that any deviation from the control reflects net grazing activity.

Data analysis

We delineated the study period into three phases based on the timing of OAE application and nutrient additions. Phase 0 (Days 1–5) represented the pre-OAE application period. Phase I (Days 7–25) captured the initial responses under nutrient limitation following the OAE application. Phase II (Days 27–53) encompassed the post-nutrient addition period, including the subsequent repletion and eventual return to depletion. To assess top-down controls on O. dioica, we quantified the abundances of large zooplankton (>500 µm, e.g., copepods, chaetognaths, and hydromedusae) that are known predators of O. dioica, its eggs, and discarded houses [81–84].

To ensure statistical independence, daily observations were averaged within each phase, yielding one mean value per response variable per mesocosm per phase. Thus, the mesocosm was the experimental unit for all analyses. This gradient design improves our ability to detect thresholds in ΔTA responses, while maintaining ecological realism in large volume mesocosm systems [63–65].

The phase-averaged responses were analysed with Type III analysis of covariance (ANCOVA) with ΔTA as a continuous predictor (from 0 to 600 μmol kg^-1), mineral type as a categorical predictor (silicate vs calcium), and their interaction (ΔTA × mineral). To explores potential indirect effects through trophic interactions, we computed Pearson´s correlations between O. diocia metrics and their prey and predator groups.

Variables exhibiting right skew were log₁₀(x + 1) transformed before analysis; normality and homoscedasticity were confirmed by residuals vs. fitted and Q‐Q plots. All statistical tests used α = 0.05, and all analyses were performed in R (v4.4.0, R Core Team 2023).

Resilience of Oikopleura dioica to OAE

Our 53‐day mesocosm study demonstrated that the larvacean O. dioica sustained stable population densities and key physiological functions (house production, prey clearance) under non-CO₂-equilibrated OAE conditions across a ΔTA gradient up to 600 µmol kg^-1 (ΔpH 0.7). Despite notable temporal variability, no consistent effect of alkalinity enhancement was detected. Although overall responses were similar across Ca- and Si-based treatments, we observed a transient effect of ΔTA-mineral interactions on the larger larvacean size fraction. Marín‐Samper et al. [69] reported delayed phytoplankton bloom onset and reduced community production in Ca‐enhanced mesocosms. Ferderer et al. [41] found enhanced diatom silicification under Si‐based OAE. Collectively, these studies highlight that alkalinity source, beyond its direct contribution to TA, can modulate ecological outcomes by modulating the phytoplankton community. In addition, we observed correlations between larvacean abundance and both picoplankton and predator groups; these must be viewed in light of the mesocosm’s overlapping biotic and abiotic processes. Our study was not designed to disentangle food web interactions from direct chemical effects. Nevertheless, collectively, these results underscore the resilience of O. dioica to the tested OAE scenarios, contributing to our understanding of zooplankton responses to this mCDR approach. Due to our mesocosm set-up, we cannot differentiate direct pH/TA effects (e.g., acid-base disruption) from indirect, trophic-mediated responses (e.g., shifts in prey abundance or quality, predator composition). Quantifying each pathway independently was therefore beyond this study’s scope, but this design reflects real-world OAE deployments, where chemical perturbations and ecological interactions co-occur. To delineate our interpretation clearly, we have organized the discussion into two complementary sections: first, evaluating the plausibility of direct physiological impacts of elevated pH on O. dioica, and second, examining how treatment-driven changes in prey and predator dynamics could have indirectly shaped larvacean population patterns.

Potential direct physiological effects of elevated pH

Key functional traits, house production and prey clearance rates, remained stable across the ΔTA gradient (0–600 µmol kg^-1) and both mineral treatments, demonstrating O. dioica’s capacity to maintain energetically demanding processes under elevated pH (8.1–8.7). This finding extends prior observations of larvacean resilience under acidified conditions [58–60], and now extends this tolerance to more alkaline conditions. Nonetheless, it is important to note that both house production and clearance rates are energetically demanding processes that can be sensitive to environmental stressors, including pH shifts. For example, Bouquet et al. [58] used microcosm life-cycle experiments (~5–6 days from fertilization to spawning) and showed that exposure to elevated pH (≥ 8.4) delayed reproduction by ~2–5 h and significantly reduced egg production, leading to a lower population growth rate (r_max) in O. dioica. Similarly, pH‐driven alterations in biochemical pathways could impair mucus synthesis and structural integrity of the house, although such effects may only emerge under more extreme or prolonged exposures and require further investigation.

Although our incubation experiment revealed no significant impact, subtle or delayed effects at elevated pH remain plausible, particularly under food‐limited or in a multi‐stressor context. Notably, our mesocosm data also did not indicate adverse population-level effects. Some of the highest larvacean densities were observed at the highest ΔTA treatments, particularly in Si-based additions. Although this could suggest tolerance or even a positive association with elevated alkalinity, these patterns likely reflect ecological interactions, such as nutrient-driven prey availability, rather than direct chemical stimulation by pH.

Indirect effects through trophic interactions

Although O. dioica density did not vary with ΔTA, its population dynamics appear to have been shaped by food-web-mediated interactions. Larvacean abundance closely tracked temporal changes in prey availability, picoplankton abundance, which itself differed among Ca- and Si-based treatments. For instance, the sharp increase in O. dioica density observed at ΔTA 0 µmol kg^-1 in the Si-based treatment during Phase II coincided with a prior rise in picoplankton, suggesting a food-mediated response following nutrient enrichment, as previously observed in a mesocosm set-up [85].

We also recorded shifts in larvacean size structure; larger individuals became more prevalent due to ΔTA-mineral interaction. Such trait shifts are also likely food-mediated, as previous studies have shown that O. dioica matures at smaller sizes under low food conditions, with maturation sizes ranging between 600 and 1600 µm, depending on resource levels [86]. Thus, the observed differences in size distribution may reflect indirect responses to changing trophic conditions rather than direct physiological effects of OAE. Specifically, picoplankton was impacted differently due to the mineral type. Such trait-level shifts highlight the importance of considering indirect ecological pathways when assessing the impacts of OAE.

Our study period followed the spring bloom, a period characterized by low macronutrient concentrations and a shift toward smaller phytoplankton [87]. These conditions are favourable for larvaceans, which can efficiently exploit a broad prey size spectrum, including picoplankton [88], which was more prevalent during the early phase of this study. While nutrient addition in Phase II increased overall microbial food availability, the relative contribution of larvaceans to the total zooplankton community also increased. However, this shift likely reflects not only increased larvacean growth but also could be due to a decline in copepod abundance, which may be driven by predation from fish larvae present in the mesocosms [59]. Additionally, it has been documented that larvaceans can contribute relatively more to secondary production under eutrophic conditions [88,89], which is related to their extraordinary growth capacity, following the growth rates of their protozoan prey [53]. Hence, without heavy predation, their ability to utilize prey resources is directly translated into positive population growth rates. The observed reduction of chlorophyll a in treatments with high larvacean abundance further illustrates how prey exploitation by O. dioica can influence phytoplankton dynamics and, by extension, impact phytoplankton standing stocks. Earlier work documented strong top-down control of larvaceans. Tönnesson et al. [90] reported that in Gullmar Fjord, O. dioica removed ~0.06% d^-1 of phytoplankton (0.4% d ⁻ ¹ bacteria), while Sato et al. [91] observed that larvacean peaks coincided with reduced chlorophyll a, and O. dioica removed 0.05–5% d^-1 particulate organic carbon (POC) in eutrophic Tokyo Bay waters. Mechanistically, Scheinberg et al. [92] reported that larvaceans efficiently clear pico- and nanoplankton, and López-Urrutia et al. [93] estimated that larvaceans can account for up to ~10% d^-1 phytoplankton removal and ~40% of total mesozooplankton grazing, consistent with the chlorophyll reduction observed here under high larvacean abundance.

As efficient suspension feeders, O. dioica can exert substantial grazing pressure on microbial prey [88]. Our observed negative correlations between larvacean dynamics and picoplankton, particularly during Phase II, are in line with this mechanism. The potential for OAE to influence not only prey quantity but also prey quality warrants attention. Additionally, the temporal patterns support O. dioica’s known ability to rapidly respond to changes in food availability, due to its short generation times and high reproductive rates [53,58,94].

Phytoplankton responses to pH elevation remain a critical consideration for trophic interactions. The highest pH observed in our study (8.7) falls within a range where influence on phytoplankton growth has been documented. For example, Oberlander et al. [34] found that approximately half of the coastal phytoplankton taxa experienced reduced growth at pH > 8.5, while other studies reported declines at pH ≥ 9 (ΔTA ≥ 500 µmol kg^-1) [32,33]. These results highlight taxon-specific sensitivities and raise the possibility that prey availability to O. dioica may have been modulated under higher OAE treatments in our study. However, the nutrient-limited nature of our experimental system likely constrained the expression of a pronounced OAE response [69]. In oligotrophic environments, nutrient availability might have exerted greater influence on prey communities than alkalinity perturbations. This emphasizes the importance of accounting for regional nutrient regimes and seasonal dynamics when assessing OAE effects. Although the nutrient additions triggered an overall phytoplankton bloom, and Marín‐Samper et al. [69] reported that higher TA can delay bloom onset, this bloom did not emerge until the experiment’s final phase. As a result, we could not detect any cascading effects of bloom timing on O. dioica within our study window. Moreover, elevated pH can influence prey quality. Paul et al. [95] reported a modest decline in particulate organic nitrogen and a concurrent rise in the C:N ratio of particulate matter. Such changes can reduce the nutritional quality of prey for filter feeders like O. dioica. Additionally, experimental evidence suggests that elevated pH reduces the digestibility of phytoplankton, with consequences for larvacean feeding efficiency and fecundity [58]. Given its reliance on microbial prey, high metabolic demands, and limited energy reserves [96–100].

Finally, we observed positive associations between larvaceans and larger zooplankton (>500 μm), such as copepods, chaetognaths, and hydromedusa. This likely reflects a shared response to nutrient enrichment but could also indicate a shared vulnerability to fish predation, rather than a direct relation to predation-prey interactions among zooplankton. While our data did not indicate strong predation-driven suppression of the larvacean community, we cannot exclude top-down effects, especially from fish predators whose assemblage was not comprehensively characterized throughout the experiment [68]. Combined laboratory and mesocosm experiments demonstrated that early-stage fish (Atlantic herring larvae) were physiologically resilient to OAE conditions, with no significant impacts on either survival or growth [68]. However, alterations in the community composition were observed, particularly under Ca-based OAE treatments. While overall fish biomass increased, these community-level shifts may have modulated predator-prey dynamics and influenced top-down control on the zooplankton population. Previous studies have shown that copepods, medusae, ctenophores, chaetognaths, and other predators can exert substantial pressure on larvacean eggs and juveniles [82,85,89,101], especially under nutrient-depleted conditions when feeding preferences shift. For instance, predation on larvaceans might intensify due to shifts in the feeding preferences of omnivore copepods as well as an expected increase in larger-sized calanoid copepods that are known for their grazing impact on larvacean communities [101]. In summary, although we did not detect consistent trophic cascades, this may be due to the limited sampling resolution and incomplete characterization of predators.

Outlook and implications

The ecological viability of OAE depends not only on its carbon sequestration efficacy but also on its compatibility with marine food webs, biodiversity, and ecosystem functioning. Our findings indicate that O. dioica, due to its physiological resilience and capacity for rapid exploitation of available prey, could continue to play a stabilizing role in the zooplankton community under these tested non-CO₂-equilibrated ΔTA conditions. However, defining operationally realistic ΔTA levels for scalable OAE remains an open question [102], because there is still a gap between small-scale field trials and high-dose experimental studies. For example, Guo et al. [49] tested low ΔTA increases (16−29 µmol kg^-1) using NaOH, olivine, and steel slag, during shipboard incubations in the Equatorial Pacific. They reported limited phytoplankton impacts aside from subtle shifts in cyanobacteria and picoeukaryotes. In contrast, Bednaršek et al. [103] synthesized calcification responses across 84 marine species and identified biological thresholds between 50 and 500 µmol kg^-1, where calcification rates declined by 50%. Their findings highlight that ecological responses can emerge well before extreme pH or TA levels are reached, which highlights the need for more dedicated mechanistic investigations.

Within this context, our mesocosm experiment deliberately extended ΔTA up to 600 µmol kg^-1 to identify potential tipping points. While this magnitude likely exceeds what many operational OAE deployments would target, it falls within proposed regulatory maxima, such as the Isometric protocol’s recommendation to remain below ΔTA ~ 1000 µmol kg^-1 and pH 9, to prevent runaway precipitation and to limit impacts on sensitive marine organisms [62]. The high ΔTA levels in our study are best interpreted as approximating conditions near the centre of an alkalinity discharge plume, where perturbations are most intense, but would be rapidly diluted at larger scales [50]. Such high-exposure tests are valuable for delineating ecological resilience margins, identifying early-warning indicators of ecosystem disruption, and constraining safe-operating spaces for real-world OAE applications.

To refine safe-operating spaces and improve ecological risk assessment for scalable OAE, future experiments should (i) test particulate OAE sources side-by-side with dissolved additions; (ii) add trace-metal characterization and controlled particle loads to isolate chemical vs particulate effects; (iii) study across seasons to capture impacts with changing nutrient supply; (iv) include higher sampling frequency to resolve short-lived predator-prey interactions; and (v) directly quantify larvacean -mediated export using sediment or gel traps and in situ sensors.

Ultimately, it needs to be considered that real-world OAE deployments will occur in a dynamic and multi-stressor-impacted natural environment. Hence, interactions between warming, deoxygenation, and nutrient regimes may lead to synergistic or antagonistic interactions [104]. To support the responsible scaling of OAE, future research should employ factorial, long-term experiments that integrate diverse trophic levels and examine key organismal traits beyond abundance, such as reproduction and metabolic plasticity.