Healthy F. vesiculosus plants bearing intermediate amounts of live S. spirorbis were collected for four seasonal experiments in less than 1.5 m water depth in Eckernförde Bay (54∘27′ N, 9∘53′ E; western Baltic Sea, Germany) in March 2013, June 2013, October 2013 and January 2014. The location is described in detail by and . Individual Fucus plants were selected by visual inspection to contain approximately the same volume of blades and similar amounts of S. spirorbis. The typical density of S. spirorbis tubes at the start of the experiments is shown in Fig. 1. The collected plants were transported in a cool box to GEOMAR (Kiel, Germany) for subsequent treatments.

The samples were stained outdoor at the quay at GEOMAR in a closed 10 L transparent plastic box for 3 days in Kiel Fjord seawater with ∼ 50 mg L-1 calcein. The box was continuously bubbled with ambient air. At the start of the staining S. spirorbis were fed with Rhodomonas algae. The staining box was placed in a flow-through water trough with seawater pumped from the Kiel Fjord to keep the temperature close to ambient conditions in the Fjord. The absorption of the dye into newly grown tubes provides a well-defined starting point for growth under the experimental conditions. After 3 days of staining, 12 individual Fucus plants were transplanted into the 12 subunits of the Kiel Outdoor Benthocosms , fixed on a plastic grid at the bottom of the basins under 0.4 m of water. The incubations started immediately after staining. Average S. spirorbis starting populations were on the order of 100–200 specimens per subunit.

The 12 benthocosm subunits were assigned to four treatments and each treatment had three replicates : “control treatment” with ambient pCO2 (380–400 µatm) and water temperature, “+CO2 treatment” with 1100 µatm pCO2 in the headspace of the subunit, “+T treatment” with water temperature elevated by 5 ∘C over ambient conditions and “+CO2 + T treatment” as a combination of both elevated pCO2 and temperature. These conditions are considered as representative for acidification and temperature changes at the end of 21st century . Each benthocosm subunit had a volume of 1500 L and was continuously flushed with ambient fjord water, pumped from 1 m below the surface at a flow rate of about 65 L h-1. Water in the subunits was additionally mixed by artificial waves with a frequency of 30 waves per hour. Four seasonal experiments were carried out: “spring” (4 April–19 June 2013), “summer” (4 July–17 September 2013), “autumn” (10 October–17 December 2013) and “winter” (16 January–1 April 2014). In total, each subunit contained 21 Fucus plants, but only one plant with attached Spirorbis, and a fauna of mollusks, arthropods and echinoderms. Details of the KOB setup and experimental parameters are described in , and . After 10–11 weeks of incubation, the 12 algal plants with S. spirorbis were collected from the benthocosms for freeze drying and further analysis.

S. spirorbis specimens were peeled off from the algal surfaces and photographed under an epifluorescence microscope. The initial and final diameter (in millimetres) of S. spirorbis shells and the length of the newly grown tube segments (mm) were measured after observing the position of the staining line (Fig. 2). The absolute tube length increase (mm) was measured as the length of the newly formed external arc of the tube between the staining front and the terminal tube edge, following .

From the spring, summer and autumn experiments S. spirorbis tubes were collected from some basins for chemical analysis (Ni et al., 2018). From each basin the newly grown tube parts of up to 20 specimens were cut off at the stain line, pooled, bleached, washed, dried and weighed. Bleaching was carried out using sodium hypochlorite with 1 % active chlorine.

The measured length increase and final diameters were normalised by the initial diameter. In our analysis we compared the resulting five growth parameters: (1) initial diameter, Di; (2) final diameter, Df; (3) growth, Gr; (4) growth/initial diameter, Gr/Di; and (5) final diameter/initial diameter, Df/Di. In order to test the robustness of the different parameters we measured Gr and Df of specimens with similar initial diameters in the autumn, winter and spring populations. The results showed that Gr measurements were more sensitive in detecting growth differences than the Df measurements.

Normalisation to the initial diameter was applied because growth of S. spirorbis tubes is strongly size-dependent. However, as the dependence is not strictly linear (see Sect. 3.5) we based all growth rate comparisons on the condition that the initial diameters of the starting populations were in the same range. The clearly bimodal populations in the autumn experiment were treated separately (autumn-big and autumn-small). The summer populations and autumn-small populations, which both were dominated by juveniles, differed significantly from the autumn-big, winter and spring populations, dominated by adults. Therefore no comparisons were carried out between these two sets of populations, because there was very little overlap in the initial sizes (compare Results Section, Fig. 5). Initial diameters of the summer and autumn-small populations overlapped to a high degree, but the medians differed significantly (two-way ANOVA, p=0.02). In order to derive comparable populations with similar Di in the summer and autumn-small data we selected sub-populations that had similar Di ranges and similar median Di values. Tubes outside this Di range were not used in the statistical analysis. For the autumn-big, winter and spring populations the initial diameters were not significantly different, as verified by Tukey's HSD tests.

Three-, two- and one-way ANOVA and Tukey's HSD tests were used for testing statistical significance of differences between the median values from different treatments and seasons. Each treatment had three replicates but, with a total duration of 1 year, seasonal experiments were not replicated. Median values were calculated for each of the treatment replicates based on the measured values, resulting in 12 basin medians for every seasonal experiment. In the three-way ANOVA, the three factors were temperature, pCO2 and season. The temperature and pCO2 factors had two levels, elevated and ambient. It should be kept in mind that the season factor here is a multiple factor which includes a range of parameters/conditions such as fjord temperature, pH, saturation state, nutrients and ontogenetic effects of S. spirorbis. Only differences caused by the temperature and pCO2 offsets between the treatments were tested for statistical significance. The “seasonal” factor had no independent (multi-annual) replicates. Differences between seasonal experiments may consequently arise from any of the above mentioned seasonal factors as well as from other unknown factors.

Assumption of normality of the models' residuals and homogeneity of residual variances were tested with Shapiro–Wilk tests and box plots respectively. Statistical analyses were conducted with R (version 3.2, http://cran.r-project.org), PAST version 3.13; and Microsoft Excel (Data Analysis Tool). A probability value of < 0.05 was considered significant.

For localisation of the calcein stain line S. spirorbis specimens were photographed with an epifluorescence microscope (AxioScope A1, Carl Zeiss, Germany). Polished longitudinal and cross sections were used for electron microscopy. Samples were wet polished with grinding paper followed by polishing solutions of 9, 3 and 1 µm grain size until no more scratches were visible on the polished surface. Backscatter electron images (BEI) and element concentration maps of calcium were taken with a JEOL JXA 8200 “Superprobe” electron microprobe (EMP) at GEOMAR Kiel, Germany. High-resolution (2–3 µm per pixel) maps of calcium were recorded with 50 nA beam intensity at 15 kV, eight accumulations and 100 ms dwell time. Internal structures of stained skeletons were imaged on polished cross sections with a Zeiss Axio Imager.M2 microscope using white field and differential interference contrast. The Cy3 filter set was applied for detection of calcein. Images were acquired with a resolution of 1360 × 1024 pixels. Excitation wavelength was 495 nm. Emission from calcein (517 nm) was recorded.

Temperature and pHNBS in all benthocosm treatments and the fjord water inflow were logged at 2-hour intervals by GHL temperature sensors (PT1000) and pH glass electrodes respectively. Air pCO2 in the head space of the +CO2 treatment subunits was monitored using infrared spectroscopy and kept at a constant level as described by .

Additionally, pHNBS values were measured daily using a Seven Multi1InLab Expert Pro (pH, Mettler Toledo GmbH, Giessen, Germany). The pH electrode was calibrated with NBS pH-buffer solutions (4.001, 6.865) kept at in situ temperature . The pH 9/10 buffer was avoided to prevent impact of possible CO2 contamination under field conditions. The stability of the electrodes' Nernst slope and the applicability of the two-point calibration to higher pH was previously tested by the measurement of a pH 10 calibration solution. Independently calculated pH values (using CO2SYS) based on measured dissolved inorganic carbon (DIC) and total alkalinity (TA) values showed good agreement with the measured pH in the range of 8 to 9 . Discrete water samples were taken as described by and analysed for TA two times a week, as well as for DIC on a monthly base. Water samples for DIC analysis were filled bubble-free into 50 mL Winkler bottles, poisoned by the addition of one drop of saturated mercury chloride (HgCl2) solution and measured via coulometric titration . To remove microbes and particles TA samples were filtered through 0.45 µm Minisart syringe filters (Sartorius SFCA, Sartorius) and measured by potentiometric titration using 0.01 M HCl with a Schott titri plus and an IOline electrode A157. NaCl was added to avoid changes in ionic strength during the analysis. The titration cell was kept at 25 ∘C. Measurements were calibrated using certified seawater standards for DIC and TA .

The speciation in the dissolved carbonate system, including the carbonate ion concentration, was calculated from pHNBS, TA, temperature and salinity using the code of the CO2SYS software package for MATLAB, version 1.1 , with constants recommended for best practice , i.e. K1 and K2 from , KS and KB from , KF from , KW, K1-3P and KSi from and the total boron–salinity relationship from . The K1 and K2 constants from are defined for a salinity range from 19 to 43, while the brackish Kiel Fjord water ranged from about 10 to 20 PSU during the experiments. An alternative set of equations for K1 and K2 is available from for salinities as low as 1. However, as discussed in applications of the latter showed discrepancies on different pH scales. Therefore, we used the constants. Using the latter to calculate carbonate ion concentrations at salinities as low as 10 PSU resulted in offsets of less than 0.5 % compared to the constants, which is negligible for our interpretations.

Salinity and concentrations of Ca, Si and P were measured in all benthocosm treatments 2 times a week. Dissolved Si, P and Ca were analysed by inductively coupled plasma optical emission spectrometry (iCAP 6300 DUO, Thermo Fisher Scientific) after appropriate dilution. The accuracy and precision was routinely checked with the certified seawater standard CASS-5 as previously described . PO4 was also measured by spectrophotometry using a QuAAtro nutrient analyser (SEAL Analytical; ). Accuracy and precision checked by replicate analyses of a solution from powdered phosphate salts were better than 8 % RSD. Si and P concentrations were usually too low to have a substantial impact on alkalinity. Alkalinity and salinity behaved conservatively in our experiments and showed no significant systematic variability on diurnal timescales (). Calcium concentrations ranged from about 3.5 to 6 mM and were closely coupled to salinity (R2 > 0.9).

The saturation state in the benthocosm treatments with respect to the calcium carbonate of S. spirorbis tubes was calculated considering the shell composition. It has been shown that the thermodynamic stability of biogenic Mg-calcites differs from pure calcite and varies with the Mg content. The MgCO3 content of S. spirorbis tubes is about 10 ± 1 mole% (Ni et al., 2018; ). Unfortunately, the solubility of S. spirorbis has not yet been explicitly determined. According to different experimental studies biogenic Mg-calcite with about 10 mol% MgCO3 has a solubility which is thermodynamically equivalent to aragonite . Therefore, the saturation state with respect to aragonite (Ω) was taken as an estimate for the Mg-calcite forming the S. spirorbis shell. It should, however, be kept in mind that the solubility of biogenic Mg-calcites may not only differ with shell composition, but may also depend on crystal ordering, trace element impurities and other mineralogical factors . Most of these factors increase the solubility of Mg-calcite.

Saturation states in the benthocosms at the measured in situ temperatures and salinities were calculated from carbonate ion concentrations, calcium ion concentrations and the apparent solubility constant (Ksp*) of aragonite : Ω=[Ca2+]⋅[CO32-]/Ksp*. Only pH and temperature were measured with 2-hourly resolution . All other parameters ([Ca2+], TA, salinity, Si, P) were interpolated to calculate diurnal variations of Ω (Fig. S1 in the Supplement). Linear interpolation is justified by the conservative behaviour of these properties. The resulting 2-hourly resolved time series of Ω were used to estimate the mean saturation state and the percentage of time when treatments were undersaturated with respect to S. spirorbis tube Mg-calcite.

Average diurnal amplitudes of saturation state, pH and temperature were calculated as follows: the pH and temperature time series from and the resulting saturation values have an interpolated resolution of 10 min. We averaged all values of the period of interest into 24 bins of 1 h length, resulting in a mean value for each hour of the day. The minimum and maximum values of the resulting mean diurnal cycle define the mean diurnal amplitude. The resulting values were averaged for each of the four different treatments. Each experimental period was subdivided into four sub-periods with durations of 17–19 days and mean diurnal amplitudes of each sub-period were calculated as explained above.

We use the daily insolation sum measured at the GEOMAR meteorological observatory

www.geomar.de/service/wetter

Variations of physical and chemical parameters (TA, pH, temperature, salinity, etc.) in the control treatments of the four seasonal experiments are shown in Fig. S2. The calcium carbonate saturation state of the seawater (Ω) in all basins was dominantly controlled by the pH. Average diurnal cycles showed a minimum in pH and Ω around sunrise followed by a late afternoon maximum (Fig. S1). The pH values showed strong diurnal fluctuations in all treatments. Average day/night pH differences were smallest (< 0.05) in December 2013 and largest (up to 0.6) in June, July and August 2013 and February and March 2014 (Fig. 3). The pH amplitudes clearly follow insolation showing saturation behaviour at high insolation values, most pronounced in the ambient CO2 treatments (Fig. 4). Generally, pH values declined from the spring to the autumn experiment and reached a minimum in November and December (Fig. 3).

Saturation states closely followed the pH dynamics. Average saturation was highest during the spring experiment when all treatments were generally oversaturated with respect to aragonite and Mg-calcite (Ω > 1). Basin waters were undersaturated (Ω < 1) only 6 to 51 % of the time during the spring experiment (Table 1). The lowest saturation states occurred during the autumn experiment with Ω < 1 during 81 to 100 % of the experiment. Average autumn saturation ranged from 0.6 to 0.8 (Table 1). It was only slightly elevated during daytime (Ωmax⁡ of 0.6 to 1.1, Fig. 3). Average day–night differences in Ω largely tracked the diurnal pH amplitudes with smallest differences during the autumn experiment (< 0.1 in December 2013) and large fluctuations in February and March, June and July (up to 2.2, Fig. 3).

The sizes that the S. spirorbis shells reached before the experiments in their natural environment are indicated by the initial diameters. They reflect the size distributions under natural conditions. In contrast, the final diameters of our specimens reflect changes from the initial sizes under experimental conditions. Note that only stained specimens were included in the analysis. Therefore, juveniles that settled during the experiments and specimens that did not calcify during the staining were not included.

The final and initial diameters of 2782 stained and photographed S. spirorbis tubes from all four seasonal experiments were in a range of 0.2 to 4.0 mm (Fig. 5). The tube with the biggest final diameter (∼ 4.0 mm) was found in the winter experiment. The smallest measured shell diameters (0.2 mm) occurred in summer and autumn (Fig. 5b, c). The size distributions of the shells indicate distinct populations that, in summer and autumn, were separated by a minimum in shell counts at a diameter of about 1.3 mm (Fig. 5b, c). Accordingly we classified S. spirorbis specimens into two general populations: “small” (diameter < 1.3 mm) and “large” (diameter > 1.3 mm). Seed et al. (1981) observed reproduction of S. spirorbis at shell diameters > 1.9 mm. Therefore, our small populations consist of juveniles, while the large populations are mostly adults but may include immature specimens.

Large specimens were observed in the starting populations of all seasons (Fig. 5b–e), including spring and summer, which is in accordance with the maximum life span of S. spirorbis of about 1.5 years . Most small specimens grew to large sizes during the ∼ 10 weeks of the experiments (Fig. 5f–i). The majority of S. spirorbis in the maximum size range (> 3 mm) were seen at the end of the winter and spring experiments (March, June; Fig. 5a, h, i). The initial shell diameters of the S. spirorbis autumn population showed a clear bimodal distribution (Fig. 5c). A juvenile population with a modal diameter of 0.6 mm (autumn-small) was clearly separated from an “adolescent/adult” population with a modal diameter of 1.8 mm (autumn-big). Initial diameters in the intermediate range of 1.4–1.5 mm were scarce. A similar size distribution was found in the summer experiment. However, the large population had very few specimens in summer (Fig. 5b).

Juveniles occurred in all four seasons but were rarely observed in winter and spring. The proportion of juveniles in the initial populations decreased from July (Fig. 5b) to April (Fig. 5e). Accordingly, the majority of the S. spirorbis specimens at the start of the summer and autumn experiments were in the juvenile stage (< 1.3 mm), while the winter and spring experiments were dominated by large specimens (Fig. 5). The modal initial diameter increased systematically with the sequence of the seasons from July (∼ 0.7 mm, Fig. 5a, b) until April (∼ 2.4 mm, Fig. 5a, e). The spring, winter and autumn-big populations started with similar initial diameters (modes of 1.8 to 2.5 mm, Fig. 5c–e) and all grew into a typical final diameter range (modes of 2.5 to 2.8 mm, Fig. 5g–i) representing the most common size of adult S. spirorbis.

As visible in Fig. 5, the diameter increase of S. spirorbis tubes during the experiments strongly depended on the season and the initial size distribution of the populations. Diameter increases ranged from 4 µm day-1 for the adult-dominated population in spring to 20 µm day-1 for the juvenile population in autumn. Modal diameter increases of the summer, autumn-big and winter populations were similar (Fig. 5) with values of about 10 µm day-1. This ontogenetic influence has to be taken into account when interpreting growth rates in terms of temperature and saturation state.

SEM pictures of S. spirorbis sections (Fig. 6) show a relatively rough and irregular outer tube wall surface whereas the inner surface is smooth. The internal wall structures consist of convex-forward lamellae or chevrons (Fig. 6b). New lamellae were laid down by the worm on the anterior tube surface, forming curved convex-forward layers, wrapping the end of the tube wall to completely cover the end of the anterior tube wall with a new layer. Thin crescent pores exist in the wall interior between the chevrons (Fig. 6b). These pores taper towards the inner and outer rims of the tube wall where the chevron lamellae fuse into a dense, calcium-rich wall (Fig. 6). The high calcium concentrations indicate that strongly calcified dense layers, not organic layers, armour the inner and outer tube wall surfaces.

A comparison of a cross section (Fig. 6a) and a longitudinal section (Fig. 6d) through S. spirorbis shells reveals the complex shape of the growth lamellae. The convex-forward layers are additionally curved upward, forming convex-upward lamellae in longitudinal sections. The convex-upward lamellae were built upward successively from the bottom on both sides of the tube and then converge at the tube top. The growth direction is indicated by the convex layering.

In addition, the inner and outer sides of each convex-forward layer of the tube walls are asymmetric (Fig. 7). The fluorescent, stained skeleton outlines the pattern of lamellae which were accreted during the 3-day staining period. The newly grown lamellae cover a large area along the inner tube wall surface, while little new material is attached to the outer tube wall surface.

The bottom of the tube, which was attached to the substrate, is relatively thin and characterised by parallel planar lamellae. An idealised sketch of the S. spirorbis tube structures is shown in Fig. 8. Where the wall of a new whorl attached to an older whorl it formed a thickened wedge-like structure partly filling the gap between the old and new whorl (Figs. 6d, 8). These wedges are usually calcium-rich, densely calcified (Fig. 6e), increasing the stability of the shell. The tube diameter of the whorls and the tube wall thickness generally increased as the S. spirorbis shell grew (Fig. 6d). The wall thickness ranges from about 30 to 180 µm. It is thicker in the fully developed shell parts and tapers towards the tube opening (Fig. 6a, b).

Shell corrosion (Fig. 9) occurred in all treatments during all seasons, but was most commonly observed in the high pCO2 treatments of the autumn and winter experiments (Fig. 10, Table 2). In the basins of these treatments up to 75 % of the specimens showed corroded shells. On average, the proportion of corroded samples (Pcorr) was highest in the autumn +CO2 treatment and in the winter +CO2 + T treatments, with treatment averages of 58 and 62 % respectively. In contrast, corrosion was nearly absent in the control treatments, where in all four seasons Pcorr values were lower than 1.5 %. Additionally, corroded specimens were nearly absent in all spring treatments, except for the +CO2 + T treatment.

The percentage of corroded samples was clearly related to the saturation state (Fig. 10). Except for one basin from the spring +CO2 + T experiment Pcorr was below 10 % when average saturation (Ω) was above 1. For average saturation Ω > 2 corroded shells were completely absent. On the other hand, although Pcorr = 0 % was observed in basins with an average saturation as low as 0.8 (basin D2, autumn control), corrosion frequencies generally increased in undersaturated basins. For Ω < 1 we observed a significant inverse correlation between Pcorr and saturation state: Pcorr(%)=-143±72⋅Ω+131±51;R2=0.54;n=17;p<0.0008. Notably, shells grew significantly even in undersaturated waters. Thus corrosion selectively affected the previously grown parts of the shell (Fig. 9b).

Pcorr was independent of temperature in autumn, winter and spring (R2=0.03, n=42, p=0.28), but temperature may have fostered corrosion and bioerosion in the summer experiments. Ambient temperature treatments of the summer experiments showed very low Pcorr values (Fig. 10). However, the few recovered samples from the elevated temperature experiments were highly corroded and showed very little net growth. Unfortunately, because very few specimens were recovered from these treatments of the summer experiment, Pcorr values could not be determined.

Strong bioerosion by microborers was observed in a cross section of a summer control specimen. Numerous microborings of about 5 to 45 µm diameter affected the outer tube wall (Fig. 11). The microborings penetrated the whole tube wall. This is in contrast to the shell corrosion of the other seasons, which mostly affected the outermost layer of the tube wall (Fig. 9b).

The length of new tube segments that grew during an experiment (Fig. 2: “growth”, Gr) varied considerably between populations and seasons, ranging from less than 0.1 up to 7.3 mm. This corresponds to a range in growth rates of 1 to 100 µm day-1. The longest newly grown tube in all experiments (7.3 mm) occurred in the autumn-small population. Growth was found to be inversely correlated with the initial diameter of the shells (Di), i.e. smaller tubeworms generally grew faster than bigger ones (Fig. 12). The correlation is highly significant: Gr(mm)=-1.1±0.05⋅Di(mm)+5.17±0.09;R2=0.41;n=2783;p=0; for Di ranging from 0.2 to 3.5 mm.

Growth of the winter populations showed the highest variability of all treatments, ranging from 0.4 to 6.3 mm (Fig. 12). Growth rates and initial sizes in winter were similar to those of the autumn-big populations. This indicates that the tubeworms from these two experiments were in the same developing stage, although they represented different generations of S. spirorbis populations (Sect. 3.2, Fig. 5).

In a subset of specimens from the spring and autumn (control, +CO2 + T) and the summer (control, +CO2) experiments average weights of newly grown tube segments, Wt, were determined (Table S1 in the Supplement). The results show similar weight increases in spring and summer of 0.1–0.9 and 0.2–0.6 mg shell-1 respectively. In contrast, Wt values in autumn were significantly larger, ranging from 1.2 to 2.1 mg shell-1. As visible in Fig. 12, Gr varied seasonally (Fig. 12). For the weighed specimens mean Gr ranged from 2.2 to 3.8 mm and 3.8 to 5.4 mm in summer and autumn respectively. It was only 1.0 to 2.3 mm in spring. We accordingly normalised Wt by Gr. This resulted in overlapping Wt/Gr ranges for spring and autumn of 0.1–0.4 and 0.3–0.4 mg mm-1 of tube respectively (Fig. 13). The summer shells increased their weights by only 0.1–0.2 mg mm-1 of tube.

Generally, this is in agreement with smaller final shell sizes in summer (Fig. 5) and consequently smaller tube widths (Fig. 2). Assuming a cylindrical tube geometry and a constant wall thickness of 0.1 mm the measured tube width values (Table 3) allow for the estimation of average shell densities: 1.1 ± 0.3 and 1.8 ± 0.3 g cm-3 (±1 SD) for the summer and autumn specimens respectively. This indicates that density and/or tube wall thickness of the summer tubes was 38 ± 13 % lower compared to the autumn tubes. The difference is significant (t test, p=0.005).

The aim of the study was to detect influences of elevated pCO2 and temperature on growth and destruction of calcareous tubeworm shells under near-natural conditions in different seasons. The temperature manipulations produced consistent offsets of 4–5 ∘C between the respective treatments (Fig. 3). However, the basin water acidification (pH, saturation state Ω) induced by elevated pCO2 was more complex. The average pH and Ω values were highest in the control treatments and lowest in the +CO2 + T treatments. Intermediate values occurred in the +CO2 and +T treatments. At the same pCO2 level, pH was lower in the elevated temperature treatments Table 1;. This was probably caused by biological activity or nutrient cycling. It cannot be explained by the carbonate chemistry, which would result in higher pH at elevated temperatures under otherwise constant conditions . The mean pH difference between the +CO2 + T and the control treatments was 0.2 units in summer and autumn and 0.4 units in spring and winter (Table 1). This simulated pH change is in good agreement with the predicted pH decrease at the end of this century .

The seasonal fluctuations of pH (0.4 to 0.6) and Ω (0.9 to 1.9) exceeded the respective differences between treatments (pH: 0.2 to 0.4, Ω: 0.2 to 1.2). This has to be considered when comparing data from different seasons (e.g. Fig. 13). In addition, the strong diurnal cycles of pH (≤ 0.6) and saturation (≤ 2.2) complicate interpretations of carbonate chemistry impacts on tube growth and corrosion (e.g. Fig. 10).

Such interpretations are further hampered by potential impacts from the diffusive boundary layer (DBL) forming at the surface of Fucus, the substrate of S. spirorbis tubes . Photosynthetic activity during the day can elevate pH and saturation state in the algal DBL compared to the bulk fluid. Average saturation of the autumn experiment was as low as Ω = 0.6 in the bulk fluids of some treatments. To elevate saturation from this value to slight oversaturation (Ω=1.1) pH has to be increased by log⁡(1.1/0.6), i.e by about 0.3 pH units. A pH elevation of this magnitude was reported by at a DBL thickness corresponding to the height of S. spirorbis tubes. However, these observations were made in stagnant water while conditions in the benthocosms were quite turbulent due to artificial waves generated every 2 min . Additionally, considering that insolation was reduced during the autumn experiment (Fig. 3), it appears unlikely that photosynthesis-driven daytime pH elevation (Fig. 4) was sufficient to overcome undersaturated water conditions in the DBL. This means that S. spirorbis was able to build tubes with above-average rates (Gr of ∼ 5 mm, Table S2; Fig. 12) in spite of constant undersaturation (tΩ<1=100 %) in the autumn treatments A1, B1 and E2 (Table 1). S. spirorbis tube growth in undersaturated water was previously observed by .

S. spirorbis reproduces and releases larvae predominantly during the warm seasons . Larvae settle in episodic pulses that may be coupled to fortnightly lunar or tidal cycles . The episodic larval settlement provides an explanation for the presence of distinct populations in our experiments (Fig. 5b–e). In line with previous studies, we found living (actively calcifying) juveniles at the beginning of all four seasonal experiments, i.e. in January, April, July and October. This indicates that the Eckernförde Bay S. spirorbis population reproduces throughout the year although juveniles were very rare in January and April. At the end of the summer and autumn experiments (September and December respectively) we found numerous unstained living S. spirorbis specimens on the Fucus thalli which had shell diameters < 1.3 mm. These juveniles obviously had settled during the experiments, indicating continuous reproduction in the benthocosms from July to December.

In addition to temperature, fecundity of S. spirorbis is affected by salinity and food supply and increases with individual size and age . Salinity fluctuated strongly during the experiments (from ∼ 10 PSU in June to ∼ 20 PSU in November–January, Fig. S2), but our data do not allow one to draw conclusions about salinity impacts on reproduction. Food supply for the filter-feeding Baltic tubeworms is generally lower in winter and increases when increased light availability promotes phytoplankton growth in spring, summer and autumn. Juveniles were rare in the initial populations in April, when the water temperature was still < 10 ∘C and in January when temperatures had decreased to < 5 ∘C (Fig. 3). In April, however, phytoplankton biomass is already high in the Kiel Bay area . Therefore, temperature probably dominates over food availability in controlling S. spirorbis reproduction in Eckernförde Bay.

The S. spirorbis investigated in this study displays the typical chevron lamellae microstructure (Fig. 6) that has been reported for Spirorbis spirorbis and for many other serpulid species e.g.. We observed a complex three-dimensional shape of the S. spirorbis chevron lamellae with convex-forward curving layers that show convex-upward curving substructures (Fig. 8).

S. spirorbis tube walls are purely calcitic and two-layered with an irregularly oriented prismatic (IOP) ultrastructure in the chevrons of the wall's core and a spherulitic prismatic ultrastructure (SPHP) of the thin outer wall region . The IOP chevrons and the SPHP structure are also common in a range of other serpulid genera Crucigera, Floriprotis, Pyrgopolon, Spiraserpula;. pointed out that this complex internal tube architecture is difficult to explain with simple pasting models for serpulid calcification, i.e. secretion of calcium carbonate granules or of a mucus paste with small calcite crystals that are molded into the calcitic tube. It is likely that extracellular organic matrices and scaffolds play a role in tubeworm biocalcification . Thin layers of organic matrix could be secreted onto the surface of the growing shell, as indicated by the chevron-like pores between growth lamellae (Fig. 6b).

Chevron-like accretion of new tube lamellae is indicated by the shape of the shell's stain line (Fig. 7). The figure additionally shows that synchronously with the accretion of new chevron lamellae new material was added in a thin layer to the inner tube wall. This wall thickening is in agreement with the observed tapering of the tube walls near the tube mouth (Fig. 6a, b). The asymmetric chevron lamellae structure of the S. spirorbis shells reported here (Fig. 7) has not been recorded previously in serpulid tubes. It shows that the inner and outer tube wall linings are differently constructed. S. spirorbis prefers to consolidate the inner surface of the tube while constructing new chevron layers.

In many S. spirorbis specimens the chevron lamellae of the central tube wall became visible as ring structures when the outer tube wall layer broke off or dissolved (Fig. 9). The outer tube wall layer appears to be susceptible to corrosion in spite of its massive densely calcified nature (Fig. 6).

In a recent study incubated S. spirorbis in a laboratory experiment for 30 days at three different pCO2 levels (450 µatm, Ω=1.8; 1200 µatm, Ω=0.8; 3150 µatm, Ω=0.4). They used specimens from the same site as in the current study, i.e. Eckernförde Bay. The tubes exhibited substantial dissolution at the highest pCO2 conditions (3150 µatm, Ω=0.4), but not in experiments with lower pCO2, even though waters were slightly undersaturated with respect to S. spirorbis calcite (1200 µatm, Ω=0.8). In contrast, in the current study corrosion of S. spirorbis shell surfaces was common (> 10 % of the shells) when average seawater saturation state was below ∼ 0.9, indicating corrosion starting even under mildly undersaturated conditions. Shell corrosion increased with decreasing saturation when the seawater was undersaturated (Ω < 1; Fig. 10), but occurred in only a few experiments for Ω > 1. It was completely absent at Ω > 2.

The S. spirorbis tubes in our experiments may have been more susceptible to shell corrosion compared to those of for several reasons. First, the duration of the benthocosm experiments was much longer (> 70 days) than the laboratory experiments (30 days). Second, during our experiments saturation state in the benthocosms fluctuated strongly between day and night (Fig. 3). Even with a mean saturation of 1 the shells may have been exposed to strongly undersaturated water during nighttime. did not record pH or Ω on diurnal timescales, but the low biomass (1 g per 0.6 L flask) and constant vigorous gas bubbling most likely prohibited strong diurnal fluctuations of saturation states in their experiments. Third, with the more natural conditions in the benthocosms (unfiltered seawater, presence of natural fauna and flora) bioerosion by microbes may have fostered corrosion of the shells. Therefore, our experiments indicate that under natural conditions S. spirorbis can be significantly affected by shell corrosion at acidification levels expected for the end of the century.

Corrosion in the mostly undersaturated waters of all autumn and the high-CO2 winter experiments (tΩ<1>70 %; Table 1, Fig. 10) was likely induced by mineral dissolution. In contrast, during the summer experiment when waters were mostly oversaturated (tΩ<1<50 %) the tubes were affected by bioerosion. Boring organisms play an important role in the ecology of many marine habitats . Microborings were observed in a S. spirorbis shell from the summer control experiment. They probably affected the stability of the worm tube (Fig. 11). The few tubes recovered from residual Fucus thalli in the summer experiments with elevated temperatures (+T and +CO2 + T) were mostly broken and strongly corroded (Fig. 16). These observations hint at a detrimental influence of elevated summer temperatures on S. spirorbis shells, either directly by affecting the worm's metabolism or indirectly through the reduction of grazing organisms and increased anti-fouling activities of the Fucus host plants . Additionally, irreversible damage of Fucus algae at high summer temperatures (> 27 ∘C; ) leads to substratum loss for S. spirorbis, which preferentially settle on Fucus .

S. spirorbis tubes that grow during the warm season might be especially susceptible to mechanical stress and bioerosion. As shown in Sect. 3.5, summer tubes were significantly lighter than expected for their size. This indicates thinner and/or less dense tube walls compared to autumn and spring specimens. With the increasing frequency and duration of summer heatwaves in central Europe predicted for the 21st century , increased bioerosion and loss of substratum could severely affect future S. spirorbis populations in the Baltic Sea.

Tube growth rates of S. spirorbis in our experiments were strongly controlled by the ontogenetic development. Growth rates were highest for juveniles and decreased when the worms got older and the tubes reached the maximum diameter range (Fig. 12). Similar growth–age relationships were found previously for S. spirorbis and other serpulid worms .

However, if only large specimens of similar initial sizes are considered, S. spirorbis tubes grew more rapidly in autumn and winter (October–March) than in spring (April–June), with the slowest growth occurring in the large summer population (July–September, Fig. 12). We know of no comparable published data. Reports of more rapid tube growth in summer compared to winter for several temperate serpulid species usually refer to the ontogenetic effect described above, i.e. enhanced growth of juvenile serpulids. The enhanced cold season growth of large S. spirorbis in our experiments was quite unexpected. Water was frequently undersaturated with respect to Mg-calcite during autumn and winter (48 to 98 % of experimental time, compared to 9 to 43 % in spring and summer, Table 1, Fig. 3). Other factors may play a role, like food availability and salinity. Food supply is generally lower in winter than during the warm seasons when increased light availability promotes phytoplankton growth. It consequently provides no explanation for the observed enhanced cold season growth of large S. spirorbis. Salinity was high during November–January and lowest in June. Spring salinities between 10 and 15 PSU contrasted with autumn and winter values between 16 and 21 PSU (Fig. S2). Enhanced calcification at higher salinities was previously observed in Baltic bivalves and may potentially provide an explanation for enhanced growth of large S. spirorbis during the autumn and winter experiments. No significant treatment effects on growth were detected (see below). Consequently, the cold season growth enhancement of large S. spirorbis is not an artefact of increased temperatures and pCO2 levels in the benthocosms. We suggest that, in addition to possible salinity effects, the reduced growth of adult specimens during the warm seasons reflects enhanced reproductive activity during this time , re-allocating energy resources from calcification to reproduction and thus reducing tube growth capacities.

The influence of increased pCO2 on the growth of S. spirorbis worm tubes was previously studied in the experiments of . A significant growth rate reduction was only observed for large specimens at the highest pCO2 (3150 µatm, Ω=0.4). No significant growth rate reduction was found at the intermediate pCO2 level of 1200 µatm. In agreement with these results we found no significant change in tube growth parameters (Gr, Gr/Di, Df/Di) when elevating pCO2 from ambient levels to 1100 µatm (Fig. 14), corresponding to average saturation values as low as Ω=0.6 (Table 1). We detected no significant impact of saturation state on growth (tube length or diameter) in any season. However, the average weights of newly grown tubes correlated with saturation states in the spring and autumn experiments (Fig. 13).

Apparently, the Baltic S. spirorbis worms are able to build their tubes with little changes in extension rates at pCO2 levels as high as 1100–1200 µatm. Notably, these pCO2 values are in the range of their natural habitats 385 to 2500 µatm;. However, the tubes that are formed at lower CaCO3 saturation may be more fragile. This is in line with results from cultured juvenile worm tubes of the tropical serpulid species Hydroides elegans, which showed reductions in shell hardness and wall thickness at lowered pH and CaCO3 saturation states .

As discussed in Sect. 4.4 high temperatures in the +T and +CO2 + T treatments of the summer experiment (average T of 24 ∘C, Table 1) led to high mortality and strongly reduced growth of S. spirorbis tubes (Fig. 16). The only other significant temperature influence on growth was found in the juvenile populations of the autumn experiment. The higher temperature in the +T treatment induced higher growth rates of the juvenile populations compared to the control treatment (Fig. 14). There was, however, no significant temperature influence on the growth parameters at elevated pCO2 (+CO2 + T treatment), possibly indicating interactions between the effects of temperature and pCO2 on growth.