CLSW was retrieved from a ship equipped with a closed-loop scrubber system in September 2020. The CLSW was stored in a plastic container and kept in dark conditions at approximately 5°C until the end of the third experiment, nine months after receiving the CLSW.

The alkalinity and pH of the undiluted CLSW were determined two months prior to the first experiment. The initial (0 h) alkalinity and pH of each treatment were determined at each experiment from separately prepared bottles with the same concentration as the experimental bottles, but without algae culture and zooplankton (n=5). Finally, alkalinity and pH were determined in all replicates of all treatments at the end of the scrubber exposure phase (72 h). pH measurements were corrected to the total scale (pH_T) using standard curves of TRIS (Tris/HCl) and AMP (2-aminopyridine/HCl) buffer solutions with a salinity of 25 PSU. Alkalinity was measured by titration (TA05 plus/TW alpha plus, SI Analytics, Mainz, Germany).

Samples of undiluted CLSW were sent for analysis of metals and organic chemicals (that have previously been detected in CLSW, e.g. Thor et al., 2021) to an accredited laboratory (ALS Scandinavia AB, Danderyd). The CLSW was analyzed prior to each experiment to ensure that the chemical components of the CLSW were stable throughout the storage. Validation samples were collected from at least one replicate of each treatment in each experiment at the start (0 h) and end (72 h) of the respective exposures.

The water was analyzed for metals according to the SS-EN ISO 17294-2:2016 and US EPA Method 200.8:1994 standards (ALS Scandinavia AB, package V-5-bas). The water samples were acidified with 1 ml HNO₃ per 100 ml prior to analysis.

The water was also analyzed for organic chemicals with GC-MS (ALS Scandinavia AB, package OV-21H in November and OV-21B in March and June) including aliphatic, aromatic and polyaromatic hydrocarbons. Water from the March and June experiments were also analyzed for additional monoaromatic hydrocarbons that could have been present in the CLSW.

Toxicity data for zooplankton was retrieved from the ECOTOXicology Knowledgebase (accessed on [2022-05-04]) (Olker et al., 2022). Hazard data for freshwater and marine meso and microzooplankton with exposure lengths of<3 days were included, and only concentrations that caused 50% effect on the organisms were included (EC50, ED50, IC50, ID50, LC50, LD50). The mean of each effect concentration, together with the mean detected concentrations of each substance in the undiluted CLSW, was used to calculate a cumulative toxic unit (TU) of the scrubber water. The cumulative (TU) of the scrubber water was defined using the principles of CA by calculating the sum of each TU, i.e. the ratio of concentration ( c ) to effect concentration ( E C x ), for each individual substance ( i ) in a mixture. This cumulative ratio indicates whether a mixture is close to a concentration that causes toxicity at the level of chosen effect concentration ( E C x ). The cumulative TU was defined as:

Experiments were conducted using the same CLSW in November 2020, March 2021, and June 2021 ( Table 1 ) to capture effects related to seasonal differences in composition of zooplankton. The ambient conditions during each of the experiments were selected to reflect the water temperature and the light regime during the respective season ( Table 1 ).

Mesozooplankton communities (0.2-20 mm) were collected from the Gullmar fjord near the monitoring station Släggö (N 58° 15.5’, E 11° 26.0’). The communities were sampled with vertical tows from 0-30 m using a 200 µm plankton net fitted with a plastic bag at the end, and resuspended in filtered seawater (0.2 µm; FSW). The communities were acclimated to the experimental conditions in a thermoconstant room with natural ambient conditions for 24 h prior to the start of each experiment ( Table 1 ). The water was aerated lightly, and the mesozooplankton were fed at excess with a mixture of the cryptophyte Rhodomonas salina (GUMACC, strain no 126) and the diatom Conticribra weissflogii (GUMACC, strain no 162).

Seawater with microzooplankton (20-200 µm) was collected with a 1.7 L Ruttner water sampler (Hydro-X, Swedaq) near the Släggö station from approximately 2 m depth. Subsamples were collected from the water during each of the seasons and checked using a microscope (Leica Leitz Labourlux S) to confirm ciliate presence. The water was gently siphoned through a submerged 200 µm filter to remove the larger zooplankton, and was then stored at the respective experimental conditions for approximately 6 h before the start of the second phase of each experiment.

The mesozooplankton communities were exposed to two levels of CLSW and a negative control in 2 L glass bottles with light aeration for 72 h (Control, 1.5%, 3%, n=5) ( Figure 1 ). The mesozooplankton were carefully concentrated on a 200 µm mesh submerged in FSW, and transferred to a glass bottle from which subsamples were collected using a 10 ml pipette with a wide-opening tip. The subsamples were fixed in ethanol and counted on a stereo microscope in order to determine the approximate copepod concentration in the community, however, this data is not shown and was not used in any analyses in this study. Aliquots of the concentrated community were then distributed to each experimental bottle to ensure a community containing approximately 60-80 copepodites and adults per liter ( Table 1 ).

C. weissflogii culture was retrieved from Gothenburg University marine algae culture collection (GUMACC, strain no 162), and the communities were fed at excess with 600 µg carbon/l based on equivalent spherical diameter (11.9 µm; average between the three experiments) at the experimental start. The algae concentration was determined every 24 h from measurements on a particle counter (Beckman Coulter Z2 Particle Cell Counter & Size Analyzer), and new culture was added to maintain a 600 µg carbon/l concentration in each bottle.

The mesozooplankton were carefully filtered out on a 200 µm mesh submerged in FSW after 72 h of exposure. Copepod eggs that were produced by the broadcast spawning copepods during these 72 h were filtered out on a 60 µm mesh from the remaining water, and transferred to plastic petri dishes with FSW. The eggs were left to hatch for another 72 h under the same ambient conditions, and then fixed in 15 ml L^-1 acidified Lugol’s solution to estimate the total egg production and hatching success of the broadcast spawning copepods in the community.

After the exposure phase, the mesozooplankton were transferred to 2 L polycarbonate bottles with seawater containing microzooplankton ( Figure 1 ). 1 ml L^-1 EDTA was added to each bottle to detoxify metal ions naturally occurring in the seawater to ensure normal microzooplankton growth during the incubation. The bottles were incubated for 23 h on a plankton wheel together with 250 ml control bottles containing microzooplankton only (ciliate controls). At the end of the incubation, the mesozooplankton were stained in neutral red and fixed in a 2:8 glycerol:FSW mix for live/dead assessment and frozen at -18°C (November), or fixed in a 5:95 glycerol:ethanol mix and kept at 5°C (March and June). To ensure a quick fixation of the microzooplankton, 200 ml subsamples were collected in dark bottles pre-filled with 3 ml acidified Lugol’s solution at a final concentration of 15 ml L^-1 for later identification of the ciliates. Start samples were collected and fixed from separately prepared bottles of both meso- and microzooplankton communities in order to characterize the communities prior to the exposure.

The neutral red fixed mesozooplankton samples were thawed to room temperature and counted using a Leica Wild M8 stereo microscope on an illuminated stand. Dead and alive individuals were separated based on neutral red uptake and copepods were categorized into male, female, copepodite or nauplius. The ethanol fixed samples were transferred to filtered seawater and counted using the same approach, with the exception that any intact individual was considered alive. Only live individuals were included in the further analyses. Copepod eggs and nauplii were counted using a Leica Wild M8 stereo microscope on an illuminated stand, and the ciliates in the microzooplankton samples were counted in 50 ml Utermöhl chambers with an inverted microscope (Leica Leitz DM IL). All mesozooplankton, copepod eggs, and copeopod nauplii were counted in each replicate, and all ciliates in a 50 ml subsample from each replicate were counted. Community copepod egg production was estimated by adding the number of unhatched eggs and nauplii in each sample, and hatching success was estimated by dividing the number of nauplii by the total number of unhatched eggs and nauplii. Taxonomy of microzooplankton was determined based on reference literature (Enckell, 1980; Gissel Nielsen and Hansen, 1999; Larink and Westheide, 2011) relevant to the Baltic and North Sea, and in dialogue with personnel working with identification of zooplankton species for the Swedish monitoring program. Species identification was made to species level, when possible, or to genus followed by sp. or spp. accordingly.

All data were analyzed using R Version 1.4.1717 (^© 2009-2021 RStudio, PBC) with the exception of the SIMPER analyses and Analysis of Similarities (ANOSIMs) that were performed using PRIMER version 7.0.21 (Clarke and Gorley, 2015). The differences between meso- and microzooplankton abundances, copepod egg production and hatching success in control and treatments were analyzed with one-way ANOVAs and Dunnett’s post-hoc tests. The assumptions of residual normality and homogeneity of variances were checked visually from residual and Q-Q plots. The heat plots and clusterings (hierarchical complete-linkage) were performed using the package pheatmap (version 1.0.12). The effects of CLSW exposure on mesozooplankton diversity, and the indirect on microzooplankton diversity were tested using ANOSIMs based on Bray-Curtis dissimilarity. The beta diversity (compositional differences between treatments) was visualized through an NMDS. The SIMPER analysis and the Bray-Curtis distances on which the ANOSIM run were both based on untransformed taxa abundances of live individuals of all life stages.

The undiluted closed-loop scrubber washwater (CLSW) contained ten individual metals and 15 individual organic substances ( Table S1 ) and had a pH_T of 6.9 and an alkalinity of 11 030 mmol/l ( Table S2 ). Several aromatic and aliphatic substances could not be individually identified in the CLSW, and were instead grouped based on number of carbon atoms. None of the analyzed organic substances were detected in the control water used in the experiments, whereas all metals except for cadmium and lead were detected, nevertheless at substantially lower levels than in the CLSW ( Table S1 ).

Ecotoxicity data was found for 19 out of the 25 detected individual metals and organic substances in the ECOTOX database, and the cumulative toxic unit (TU) for the combination of these in the undiluted CLSW was estimated to 17 ( Figure 2 ). This corresponds to TUs of 0.26 and 0.5 for the 1.5% and 3% treatments respectively. Vanadium was the main toxicity-driver and contributed with 57% of the total toxicity followed by copper (26%), benzo[ghi]perylene (BghiP) (11%) and nickel (5%). These four substances were together responsible for >99% of the total toxicity of the CLSW based on the detected individual chemicals.

The analysis of scrubber water contaminants at the start and end of each experiment generally confirmed the nominal dilutions of 1.5% and 3%, with the exeption of the aromatic substances ( Figure S1 ). Many of these substances were not detected in the diluted treatments since their concentrations in the undiluted CLSW were close to their respective detection limits. The pH and alkalinity in the controls varied between 8.1-8.4 depending on season and was reduced by 0.1-0.2 in the 1.5% treatment and by 0.3-0.5 in the 3% treatment at the start of the exposures ( Table S2 ).

The total copepod community egg production varied depending on season but the relative differences between control and CLSW treatments were similar, except for in March ( Figure 3A ). The egg production was increased to 127% and 113% in the 1.5% CLSW treatment during November and June and reduced to 78% in March, but none of them were significantly different from the control (November p=0.17, March p=0.41, June p=0.78, Table S3 ). However, the 3% CLSW treatment had a substantial effect on the community egg production which was reduced to 35-42% (p<0.05, Table S3 ).

Hatching success of the eggs was between 76-80% in the control during all seasons ( Figure 3B ). The CLSW reduced the viability of the eggs independent of season starting at 1.5% with a hatching success between 36-57% (p<0.05, Table S3 ). The viability was even more reduced by 3% CLSW where the hatching success was as low as 3-27% (p<0.05, Table S3 ).

Survival in the control ranged between 55% (November) and 76% (June), and copepod survival after CLSW exposure was significantly lower than the control at all instances (p<0.05), except after the 1.5% exposure in November where it was lower but not significantly (p=0.12, Table S4 ). The magnitude of the reduced survival depended on season with 8-43% lower survival than control in the 1.5% treatment, and 19-32% lower than control l in the 3% treatment ( Figure 4 ). The survival was generally lower from 3% exposure than from 1.5% exposure, apart from in March.

Ciliate densities (as % of the mean density prior to exposure) followed an inverse pattern to the abundance of surviving copepods during all three seasons, with the most prominent pattern in November and March ( Figure 4 ). The ciliate abundance was 8-26% higher after pre-exposure to 1.5% CLSW, and 15-61% higher after pre-exposure to 3% CLSW ( Figure 4 ). The increase of ciliates relative to the control was significant during November and March (p<0.05, Table S4 ), but not during June (1.5% p=0.44, 3% p=0.09, Table S4 ), although the pattern was similar.

The mesozooplankton diversity was affected from 1.5% CLSW exposure in March and June, and from both 1.5% and 3% CLSW in November ( Figures 5A–C , Table S5 , Figure S2 ). The copepod species Paracalanus parvus and Temora longicornis were less sensitive to the CLSW and were present in larger relative proportions after the exposure, whereas Centropages hamatus, Centropages typicus and Calanus spp. were more sensitive and present in smaller proportions ( Tables S6–S7 ). The sensitivity of A.clausi depended on season and was less sensitive in November and more sensitive in June. The sensitivity of each copepod taxa appeared to be closely linked to the developmental stage, as C. typicus, Calanus spp. and A. clausi each had the highest abundance of nauplii or copepodites during the seasons when they were most affected by the CLSW ( Figure 6 ).

Changes in mesozooplankton diversity had a moderate effect on microzooplankton diversity. Significant effects were observed for the 1.5% CLSW in November and for the 3% CLSW in March ( Figures 5D–F , Table S5, Figure S2 ). The ciliate Strombidium spp. generally increased in numbers relative to other taxa when mesozooplankton had been pre-exposed to CLSW ( Tables S6–S7 ). The average abundance and variation of each mesozooplankton and microzooplankton taxa prior to and after the incubations can be found in Tables S8–10.

Our findings demonstrate that closed-loop scrubber washwater (CLSW) is toxic to marine mesozooplankton. The cumulative toxic unit (TU) for the undiluted CLSW indicates that the combination of the individually identified substances in the CLSW is 17 times the concentration that affects 50% of a population. However, ecotoxicity data could not be included for all substances in the CLSW, so it cannot be excluded that more substances contribute to the overall toxicity, in which case the cumulative TU would increase.

Vanadium, Ni and Cu are all metals with TUs larger than 1 in the undiluted CLSW, indicating that their respective individual concentrations are high enough for them to be toxic to zooplankton. The concentrations of these metals in the CLSW vary across different studies and are sometimes in the same order of magnitude as what was detected in this study, but sometimes substantially lower (Hermansson et al., 2021; Thor et al., 2021). V, Ni and Cu are all found at high concentrations in particulate matter (PM_2.5) in ship exhaust from ships running on residual fuel oils such as intermediate fuel oil (IFO) and heavy fuel oil (HFO) (Celo et al., 2015; Corbin et al., 2018), and can be transferred from the exhaust to the water recirculated in the closed-loop scrubber. Total emission factors for metals such as Cu and Zn have been shown to be higher when using an open-loop scrubber together with HFO, than when running on HFO alone (Hermansson et al., 2021), which may be attributed to use of sacrificial anodes for marine growth protection, lubrication oils and corrosion in the piping system caused by the acidic open-loop scrubber water (Kim and Jeong, 2019). The CLSW is less acidic than the open-loop scrubber water, so the contribution from corrosion should be smaller in comparison to use of open-loop scrubbers.

The only organic substance with a TU above 1 in the undiluted CLSW was BghiP, which is one of many PAHs that are found in the exhaust when ships run on HFO (Sippula et al., 2014). The concentration of BghiP in our study was 1.1 µg/l, which is higher than the 0.031 µg/l and 0.07 µg/l that have been detected in other CLSW (Hermansson et al., 2021; Thor et al., 2021). The differences in both metal and PAH content of the CLSW in our study, relative to other studies, indicate that the composition of the water is highly variable, which could be attributed to engine load (to which capacity the engine is run at) and fuel type (Sippula et al., 2014).

The PAH content in the scrubber water is regulated by the PAH_phe threshold, which is based on optical online sensor measurements of UV or fluorescence detection of phenantrene (MEPC, 2015). However, our analysis of the CLSW content shows that this is neither the most abundant PAH nor the most toxic one, which suggests that this type of threshold is unsuitable for ensuring that the scrubber water is safe to discharge. The PAH_phe has previously been criticized by GESAMP (Group of Experts on the Scientific Aspects of Marine Environmental Protection) because of its vague definitions and methodology for determination as well as questionable level of protection (Linders et al., 2019).

The pH of undiluted CLSW has been recorded in the range of 3.5-8.2 (Hermansson et al., 2021; Thor et al., 2021), and hence the CLSW used in this study was at the higher end with a pH at 6.9. Similar levels of CO₂ induced reductions in pH have caused contrasting effects on mesozooplankton depending on organism type and measured endpoint. Growth and reproduction in Acartia spp. were not affected at pH 7.02-7.4 (Kurihara et al., 2004), whereas hatching success in Calanus finmarchicus exposed to pH 6.9 was severely impaired (Mayor et al., 2007). As for sea urchin larvae, both growth and development have been reduced by pH in the range of CLSW (Kurihara and Shirayama, 2004; Dorey et al., 2013).

While the undiluted CLSW had a pH low enough to potentially cause adverse effects, the diluted CLSW treatments used in this study had pHs that were similar to the control water. The 3% treatment with pH 7.6-7.9 is in fact within the natural range in the Gullmar fjord where we collected the zooplankton used in the experiments (Dorey et al., 2013), and thus we assume that the main driver of effects in this study should be the contaminants rather than acidification.

Copepod reproduction and survival, as well as mesozooplankton predation on ciliates, were affected by the CLSW starting at the lowest tested concentration of 1.5%, all independent of seasonal community structure. Mesozooplankton diversity was affected from 1.5% CLSW, but there was no indirect effect on microzooplankton diversity, which indicates that the mesozooplankton species in the exposed community feed on the same ciliate species as those in the control community, only to a lesser extent.

The community egg production was slightly increased during two out of three seasons in response to 1.5% CLSW exposure, despite the reduced copepod survival, which indicates that copepods experience stress through hormesis (Calabrese, 2008) at this level of exposure. Increased copepod egg production has been observed both at community level, in response to water accommodated fractions of ultra-low sulphur fuel oil (ULSFO) (Jönander and Dahllöf, 2020), and at population level (Acartia bifilosa) in response to a CO₂ induced 0.3 reduction in pH (Engström-Öst et al., 2014). When the exposure level increased to 3% CLSW, the remaining surviving copepods could no longer compensate by producing more eggs, and instead the egg production was reduced substantially. Given that both total egg production and hatching rate was impacted at both exposure levels, the effective reproductive output of the community becomes severely reduced. Effects on copepod hatching success from individual PAH exposure has been shown to occur at levels 2-1000 times higher than their respective concentrations in the undiluted CLSW (Bellas and Thor, 2007; Jensen et al., 2008). This suggests a mixture effect of the substances in the CLSW that together causes additive or synergistic toxicity.

Individual exposure to toxicity-driving metals in the CLSW have been shown to cause contrasting toxicities to copepods. Ni reduces hatching success in Acartia pacifica, and nauplii production in Tigriopus japonicus and Apocyclops borneoensis at 10 µg/l, corresponding to 0.3% CLSW (Mohammed et al., 2010). Cu, on the other hand, reduces egg production rate in A. tonsa at 30 µg/l, corresponding to 4% CLSW (Lauer and Bianchini, 2010). Although these studies only cover a fraction of all the substances present in the CLSW, they indicate that metals may have a larger role than PAHs in driving the effects on copepod reproduction.

Copepod survival in the controls relative to start abundances ranged between 55-76%, but part of the mortality can be attributed to the sampling process. Even when sampling with a non-filtering cod end net, there can be 13% mortality due to damage (Elliott and Tang, 2009). Furthermore, field-sampled communities with varying demography means that some individuals will die due to age during the experiments.

The copepod survival was affected starting at 1.5% CLSW across all seasons, similar to what has previously been observed in juvenile C. helgolandicus where mortality rate was increased starting at 0.1% (Thor et al., 2021). Our higher effect concentration could be explained by the fact that we exposed a community of several species with potentially different sensitivities, as well as of different life stages. Furthermore, in both studies the effect was observed at the lowest tested concentration so it cannot be ruled out that survival is affected at even lower concentrations of CLSW.

Mesozooplankton diversity was affected starting at 1.5% CLSW during all three seasons, which indicates that the community always consists of species with varying sensitivity to the CLSW. We also observed that the relative sensitivity of different taxa changed with season, and that those with the most juveniles at the time of each experiment were the most sensitive. Like many other organisms, copepod sensitivity to chemicals is higher among juveniles than adults (Saiz et al., 2009; Heuschele et al., 2022), although the specific mechanisms responsible are unknown for this present study. Copepods molt between each of their 12 life stages, and abnormal molting as well as complete inability to molt has been found in C. helgolandicus copepodites exposed to 0.1% CLSW (Thor et al., 2021), which could play a role in their higher sensitivity.

The reduced survival of the copepods increased the ciliate abundance, which is a clear sign of a reduced predation pressure. Although there were no significant effects on ciliate diversity, Strombidium spp. generally increased more than other taxa, which could indicate a higher growth rate or a better predator avoidance behavior in this group. In accordance with this, Strombiidid ciliates relieved of predation responded quicker than other taxa to increasing food densities in mesocosm studies (Löder et al., 2011). It has been shown that individual mesozooplankton-ciliate interactions is taxa specific (Wiackowski et al., 1994), which confirms that the loss of certain mesozooplankton taxa will likely play an important role in structuring the microzooplankton community. The direct toxicity of CLSW to microzooplankton was not tested in this study, and so it is possible that the interactions between the trophic levels would be different if survival was reduced among the microzooplankton taxa due to direct exposure. Indeed, survival in freshwater ciliates has been reduced at concentrations of Cu and Ni corresponding to 1.5% CLSW (Madoni and Romeo, 2006). Since ciliates are important grazers on small phytoplankton (Löder et al., 2011; Haraguchi et al, 2018), and are known to be sensitive to metals, investigating direct toxicity of CLSW on both copepods and ciliate communities could reveal different indirect effects and interactions between these two trophic levels.

Closed-loop scrubber washwater (CLSW) is toxic to marine mesozooplankton at levels<1.5%, independent of the seasonal species composition. Short-term exposure impairs both copepod reproduction and the ability of the entire mesozooplankton community to predate on microzooplankton and alters community biodiversity. The cumulative toxic unit of 17 for undiluted CLSW, based only on detected substances, suggests that the CLSW can be toxic at levels below what was tested in this study. Given that each single ship equipped with a closed-loop scrubber can discharge 126-150 m³ of CLSW per day (Hermansson et al., 2021), it is likely that substantial adverse effects on the zooplankton will occur near such discharges. Some of the substances in the CLSW are listed as priority substances in the EU Water Framework Directive (Directive 2013/39/EU) and therefore have the potential to cause harm to aquatic systems, as well as reduce the ability to achieve good chemical status in coastal areas. Furthermore, some have half-lives of 100-1000 days in shallow water, and even up to several 1000 days once they reach the sediment (Tansel et al., 2011). This means that the continued use of closed-loop scrubbers without appropriate regulation of discharge limits for more compound groups, will contribute both to adverse effects, and an accumulation of toxic contaminants in areas near shipping routes and ports.