Abundances and morphotypes of the coccolithophore Emiliania huxleyi in southern Patagonia compared to neighboring oceans and northern-hemisphere fjords

Coccolithophores are potentially affected by ongoing ocean acidification, where rising CO2 lowers seawater pH and calcite saturation state (Ωcal). Southern Patagonian fjords and channels provide natural laboratories for studying these issues 15 due to high variability in physical and chemical conditions. We surveyed coccolithophore assemblages in Patagonian fjords during late-spring 2015 and early-spring 2017. Surface Ωcal exhibited large variations driven mostly by freshwater inputs. High Ωcal conditions (max. 3.6) occurred in the Archipelago Madre de Dios. Ωcal ranged from 2.0-2.6 in the western Strait of Magellan, 1.5-2.2 in the Inner Channel, and was sub-saturating (0.5) in Skyring Sound. Emiliania huxleyi was the only coccolithophore widely distributed in Patagonian fjords (> 96% of total coccolitophores), only disappearing in the Skyring 20 Sound, a semi-closed mesohaline system. Correspondence analysis associated higher E. huxleyi biomasses with lower diatom biomasses. The highest E. huxleyi abundances in Patagonia were in the lower range of those reported in Norwegian fjords. Predominant morphotypes were distinct from those previously documented in nearby oceans but similar to those of Norwegian fjords. Moderate-calcified forms of E. huxleyi A morphotype were uniformly distributed throughout Patagonia fjords. The exceptional R/hyper-calcified coccoliths, associated with low Ωcal values in Chilean and Peruvian coastal upwellings, were a 25 minor component associated with high Ωcal levels in Patagonia. Outlying mean index (OMI) niche analysis suggested that pH/Ωcal conditions explained most variation in the realized niches of E. huxleyi morphotypes. The moderate-calcified A morphotype exhibited the widest niche-breadth (generalist), while the R/hyper-calcified morphotype exhibited a more restricted realized niche (specialist). Nevertheless, when considering an expanded sampling domain, including nearby Southeast Pacific coastal and offshore waters, even the R/hyper-calcified morphotype exhibited a higher niche breadth than 30 other closely phylogenetically-related coccolithophore species. The occurrence of E. huxleyi in naturally low pH/Ωcal environments indicates that its ecological response is plastic and capable of adaptation. https://doi.org/10.5194/bg-2020-449 Preprint. Discussion started: 8 January 2021 c © Author(s) 2021. CC BY 4.0 License.

In contrast to the Norwegian fjord system, E. huxleyi blooms have not been reported in Patagonian fjords but information on 100 coccolithophores in these waters is scarce. A study documenting coccolithophores in the Strait of Magellan found that this group represented a minor fraction of the small-sized phytoplankton (Zingone et al., 2011), but other published studies have not specifically sampled for coccolithophores. The Patagonian shelf on the Atlantic side experiences large E. huxleyi blooms (Poulton et al., 2011;, but satellite observations suggest coccolithophore blooms are of lower intensity in the Pacific waters to the west (Hopkins et al., 2019). These observations raise the question of how coccolithophore communities on the 105 western coast and fjords-channels of Patagonia compare with nearby oceans and to fjord systems in the northern hemisphere.
Here, we evaluated how physical, chemical, and biological features influence the distribution, abundance, and biomass of coccolithophores as well as the proportions of E. huxleyi morphotypes of varying calcification levels throughout southern Patagonia fjords. In particular, three research questions were addressed: i) What coccolithophore assemblages and E. huxleyi morphotypes are present in fjords/channels of southern Patagonia? ii) How do these morphotypes and the co-occurring 110 phytoplankton (mostly diatoms) vary with physical and chemical factors? Focusing on the cosmopolitan E. huxleyi, iii) does the abundance and relative composition of E. huxleyi morphotypes reflect populations in adjacent Pacific, Atlantic, or Southern Ocean waters or instead exhibit similarities to the Norwegian fjord system, suggesting it is shaped by local factors? The 2017 sampling was from the interior WSM, crossing the IC, and ending in the AMD zone (Fig. 1b). In both cruises, surface 120 water (< 5 m) was pumped continuously onboard every 15-20 min for determination of salinity and temperature with a YSI-30 Termosalinometer (Yellow Springs, OH, USA) and pCO2 with a Qubit-S157 CO2-analyzer (Kingston, Ontario, Canada).
The CO2-analyzer was calibrated daily with 0 ppm CO2 (air treated with soda lime) and 403 ppm air-CO2 mixture standard (Indura, Chile). Surface samples for determination of the planktonic assemblages and chemical variables (i.e., concentration of macronutrients, opal, total chlorophyll-a, and the carbonate system parameters) were collected at discrete samplings stations 125 ( Fig. 1a-b). Twenty-one stations were sampled in 2015: five in the AMD , four in the IC (St. 6-9), nine in the WSM (16)(17)(18)(19)(20)(21), two in the OS (St. 13 and 14), and one in the SS (St. 15). Eleven stations were sampled in 2017: three in the AMD , five in the IC , and three in the WSM . CTD vertical profiles were additionally obtained at selected localities on both cruises. In 2015, three casts were performed into the AMD zone and one cast into the SS using a CTD Seabird 19 plus (Sea-Bird Scientific, Bellevue, WA, USA) equipped with 130 photosynthetically available radiation (PAR) and oxygen sensors. Two profiles were performed in 2017 in the AMD zone, by the deployment of a CTD Seabird 25 plus with PAR and oxygen sensors. The depth of 1% of penetration of PAR (euphotic zone) was calculated from maximum surface PAR values. CTD profile binning was 1 m. In both years, samples for the determination of plankton assemblages and chemical variables were obtained at discrete depths using 5-L Niskin bottles to which the CTDs were attached (depths pre-determined from prior studies in the region, aiming to adequately sample the surface 135 mixed layer, pycnocline, and vertical variation in chlorophyll fluorescence). Complete environmental and biological data are provided in supplementary materials (Tables S1-S4).

Plankton assemblages
Samples for the determination of planktonic organisms through the Utermöhl (1958) method were collected only in 2015. For that, duplicate 100-mL water samples were pre-filtered through 200-µm Nitex mesh, fixed with a formaldehyde-glutaraldehyde 140 solution (1% formaldehyde, 0.05% glutaraldehyde, 10 mM borate pH 8.5) and stored at 4º C. In the laboratory, water samples were brought to room temperature, gently homogenized and sedimented into 100-mL chambers for 24-48 h before counting and identification. The absolute abundances of the microplankton (20-200 μm in size) and coccolithophores (~6 µm in diameter) were estimated with an inverted microscope (Olympus CKX41) connected to a digital camera (Motic 5.0). For counts of diatoms, dinoflagellates, and other planktonic cells greater than about 40 μm, the whole chamber was examined at 145 200× magnification. When large chain-forming diatoms were in high density, between 5-60 randomly selected fields of view were examined at 200× magnification until reaching 100 chains. For counts of small diatoms, naked flagellates (including small flagellates and athecated dinoflagellates), and coccolithophores, between 1-4 transects (to reach ≥ 100 cells in total) were analyzed at 400× magnification. Counts of total coccolithophores were performed with a 40× objective with cross-polarized light (Edmund Optics polarizers 54926 and 53347). 150 In both cruises, samples for the identification and quantification of coccolithophores through scanning electron microscopy (SEM) analysis were obtained by filtering 200-300 mL of surface water, immediately after sampling, onto 0.8-µm polycarbonate filters that were subsequently dried at room temperature. For the identification of coccolithophores and E.
huxleyi morphotypes, a portion of each dried filter was cut, sputter-coated with gold and examined either in a Quanta FEI 250 or Quanta FEG 250 SEM (both FEI, Hillsboro, Oregon, USA). As water samples for light microscopy counts were not available 155 for two samples from 2015 (St. 3 and St. 5 at 8 m) and all samples from 2017, total coccolithophores abundances were obtained from SEM counts for those samples. On average, 70 images per filter were captured at 1,500× magnification (276×184 µm per frame), covering 3.5 mm 2 of the filter area corresponding to 1.8-3.4 ml of water analyzed. The coccolithophores abundances were calculated using the following equation: To check for differences in coccolithophore counts obtained through sedimentation + inverted light-microscopy versus 160 filtration + SEM examination (hereinafter SEM and Utermöhl counts, respectively), polycarbonate filters from three selected Utermöhl samples (showing higher, medium and lower coccolithophores abundances) were analyzed with SEM as outlined above. Coccolithophores SEM counts were consistently about twice as high compared to Utermöhl counts (average 1.84), agreeing with the correction factor suggested by Bollmann et al. (2002). Thus, all total coccolithophore counts obtained by the Utermöhl method were multiplied by 1.84 to be comparable to SEM counts. To estimate the absolute abundances at species-165 and morphotype-level, the relative abundance of each coccolithophore species or E. huxleyi morphotype determined from SEM counts was multiplied by the absolute abundance of total coccolithophore cells. Saturation curves obtained for each sample confirmed that the number of analyzed coccospheres (minimum 40 coccospheres per sample) was enough to capture the specific/morphotype diversity. SEM images taken at 20,000-25,000× magnification were used to categorize E. huxleyi cells in the different morphotypes 170 according to the morphology of distal shield and the central plate of the coccoliths (following Young and Westbroek, 1991;Young et al., 2003;Hagino et al., 2011;von Dassow et al., 2018). Given high morphological similarities in the A morphotype coccoliths with those found by Young (1994) in Norwegian-fjords, they were here classified as lightly-, moderate-, and robustcalcified, based on the morphology of distal shields and central plates ( Fig. 2; Table 1). Moreover, two extremely heavilycalcified A-morphotypes were observed: the A-CC (with closed central area but distal shield elements mostly not fused) and 175 the R/hyper-calcified (Table 1; Fig. 2). These two morphotypes are sometimes grouped as "over-calcified" (Cubillos et al., 2007;Saavedra-Pellitero et al., 2019). However, we have observed in culture that they remain distinct under different physiological stresses (e.g., Mella-Flores et al., 2018;von Dassow et al., 2018). Due to frequent overlap in coccolith distal shield lengths and coccosphere diameters observed in moderate-and robust-calcified A-forms (Table 1), we consolidate them into one group (hereafter jointly referred to as "moderate-calcified A-morphotype") for statistical analyses. Moreover, we 180 classified the malformed (teratological), incomplete, weakly-dissolved and collapsed coccoliths, following the terminology and definitions of Young and Westbroek (1991) and Young (1994). Although these malformed and collapsed coccoliths were observed in < 9% of the morphotype-A cells, it was almost always possible to classify those abnormal coccospheres into one of the above-mentioned morphotypes (Fig. S2). SEM images were also used to measure the orthogonal coccosphere diameters and, when available, coccolith distal shield length (ImageJ software version 1.48 for Mac OS). 185   (Hillebrand et al., 1999; see Table S4). For E. huxleyi, a spherical geometric shape was assumed and the maximum diameter used for biovolume calculations. Biovolume calculations were then converted to carbon biomass by using  Young (1994) and the present study. Similarly, the "Type A overcalcified" in Fig. 3c of Cubillos et al. (2007) corresponds to the A-CC morphotype here (as distal shield elements are not fused or only partly fused in most coccoliths) while Fig. 3d of the same reference, identified also as "Type A overcalcified" appears to show both nearly complete fusion of distal shield element as well as nearly complete over-calcification covering the central area, so corresponds to the R/hyper-calcified morphotype in the present study. b B and C morphotypes are distinguished by distal shield diameters > 4.5 or < 3.5 µm, respectively, with B/C being intermediate. protists (pg C cell -1 = 0.216 × volume 0.939 ). We assumed a constant cytoplasm diameter to be 60% of the mean E. huxleyi coccosphere diameter (O'Brien et al., 2013), whereas cytoplasm volumes of 50% and 78% were used for diatoms and dinoflagellates, respectively (i.e., total cellular volume minus frustule or theca and vacuole volumes; Sicko-Goad et al., 1984).
Absolute abundance data were standardized to cells L -1 and multiplied by specific carbon contents per cell (pg cell -1 ) to derive total carbon biomass (Total C; µg C L -1 ). We used the biogenic silica concentration (µmol opal L -1 ) as a proxy of diatom 195 biomass along the 2017 track, as samples for microscopy counts were not available. For this, the bSi concentration was converted into carbon units using the average net silicate to carbon ratio of 0.52 (mol/mol) found by Brzezinski et al. (2003) in the Southern Ocean. There was a significant linear relationship between diatom carbon biomass estimated with microscopy/allometry and those calculated from bSi concentration in 2015 samples (R 2 = 0.60, p < 0.05, slope = 0.8; N = 11), with an offset (16 µg C L -1 ) likely from other contributors to bSi (e.g., silicoflagellates, radiolarians) as well as the contribution 200 of Minidiscus spp. (data not shown) not included in microscopy/allometric carbon estimation. The presence/absence of diatoms was confirmed qualitatively in 2017 by SEM images at 1,500× magnification, and a semi-quantitative evaluation was made as follows: low (few cells), intermediate (at least one species with several cells or chains) and high (many species with several cells or chains). It should be kept in mind that there can be substantial variation in diatom carbon-biomass estimated by microscopy vs. bSi, due to variability in diatom C:Si ratios (Leblanc et al., 2018). 205

Chemical analyses
Macronutrients, opal, total chlorophyll-a (chlo-a), pH and total alkalinity (AT) were determined as described in Torres et al. (2020). Full carbonate system parameters (including Ωcal) were estimated from pH, AT, salinity, temperature (25 °C as input and in situ temperature as output conditions), pressure (0 dbar as input and depth as output conditions) using CO2Sys Excel macro spreadsheet version 2.1 (Pierrot et al., 2006) with Mehrbach set of solubility constants (Mehrbach et al., 1973) refitted 210 by Dickson and Millero (Dickson and Millero, 1987). To extrapolate full carbonate parameters from pCO2 (onboard sensor) and salinity measurements where alkalinity samples were not directly available (due to mismatch in chemical and biological sampling along the IC-WSM 2015 track), the regression curve for the salinity-AT relationship (µmol kg -1 ) = 63.4 × salinity + 101 (R 2 = 0.99, N = 186; Torres et al., 2020) was used to derive AT estimated from salinity. It is important to note that this relationship has been stable for over a decade in Patagonia (Torres et al., 2011;Torres et al., 2020). pCO2 values delivered by 215 the onboard sensor (underway sampling) correlated with pCO2 calculated from AT-pH pairs (discrete sampling) in the same 2015 samples (R 2 = 0.56, p < 0.001; N = 17), with an overestimation of 6 µatm (2%). The differences between measured and calculated pCO2 values are small compared to the high ranges in the variability of pCO2, salinity, AT, and pH, and should not affect the objectives of the present study. Exceptionally, the calculated pCO2 values for SS were overestimated by up to 36% concerning pCO2 measurements (comparing 15 readings from the sensor with three calculated values). This disagreement 220 could be due to various local factors that increase the sensitivity of calculated pCO2 to AT (Abril et al., 2015) or pH uncertainties due to differences in salinity between buffers and samples. Therefore, it should be kept in mind that in the case of SS, the uncertainties in the carbonate system could be more substantial. Finally, in order to calculate the bicarbonate [HCO3 -] to proton [H + ] ratio (in mol/µmol), the [HCO3 -] was divided by the antilog10 of -pH (total H + scale).

Statistical analysis 225
All statistical analyses were performed in R software using packages freely available on the CRAN repository (R Core Team, 2019). As measurements for nitrate, phosphate, DSi, bSi, and chlo-a in 2015 were scattered and uncoupled from plankton sampling, these variables were used only descriptively and not included in the statistical analyses.

Environmental gradients
Physical-chemical data obtained from the surface at the discrete sample stations both in 2015 and 2017 were combined in a 230 unique matrix and standardized by subtracting the mean and dividing by the standard deviation (Legendre and Legendre, 2012). We used the varclus function in the 'Hmisc' package based on Spearman's correlation to detect redundant environmental variables (N = 32). Temperature, salinity, in situ pH, and Ωcal were selected as they are non-redundant based on Spearman's correlation < 0.75 (Fig. S3) and they are easiest to interpret from a biological or cell physiological point of view.
To these four, we also included CO2. It was moderately correlated with pH (Spearman correlation = 0.8), but represents the 235 substrate for photosynthesis and is typically incorporated as a driving variable in ocean acidification studies. The selected standardized variables were then used in two separate cluster analyses to recognize groups of sampling stations with similar characteristics in 2015 (N = 21) and 2017 (N = 11). For that, Euclidean distance matrixes were first calculated based on selected standardized variables (vegdist function in the 'vegan' package) and then included in hierarchical cluster analyses based on the Ward method using the hclust function in the basic 'stats' package. 240

Testing for a relationship of Emiliania huxleyi vs. diatoms
To characterize the diatom species associated with the different E. huxleyi morphotypes and other coccolithophore species, we performed a non-metric multidimensional scaling (nMDS) using the metaMDS function in the 'vegan' package, based on the log-transformed [ln (x+1)] coccolithophore and diatom abundances in Patagonia (only 2015) and the other coastal and oceanic locations (von Dassow et al., 2018) (N = 52). The function heatmap of the basic 'stats' package was then used to plot the 245 abundance of the coccolithophore and diatom species related to the clusters based on the nMDS scores of species/morphotypes and samples. As both the nMDS and OMI (see below) analyses suggested a clear separation between Patagonia fjords and the other coastal/oceanic areas, we used the IndVal analysis (Dufrene and Legendre, 1997) to identify indicator species for both areas, based on log-transformed abundances (indval function in 'labdsv' package).
We aimed to assess how E. huxleyi and diatom biomasses were related to each other and with the environmental conditions 250 throughout Patagonia fjords. However, the different methods used to estimate diatom biomass in both years precluded the use of E. huxleyi:diatom ratios. Moreover, the use of regression-based analyses was not recommended due to the absence of a linear relationship between E. huxleyi and the different physical-chemical variables. To overcome these limitations, we created three categorical variables for both E. huxleyi and diatom biomasses (low, intermediate, and high) based on their < 25, 25-75 and > 75 percentiles, respectively. We then performed a correspondence analysis (CA) using the function cca in the 'vegan' 255 package, based on the presence or absence of these new categoric variables in each sample (N = 32), followed by fitting the standardize physical-chemical variables to the CA plot using the envfit function (10,000 permutations).

Niche analysis
We used the outlying mean index (OMI) analysis (Dolédec et al., 2000) to assess the main environmental conditions associated with the realized niche of the different E. huxleyi morphotypes. The OMI index represents the marginality (i.e., niche position) 260 and measures the distance between the average habitat conditions used by a given population and the average environmental conditions across the study area (represented by the point where the two multivariate axes intersect at zero). The tolerance (Tol) accounts for the dispersion of samples containing organisms of the population from the average environmental condition (i.e., niche breadth), whereas the residual tolerance (RTol) accounts for the proportion of the variability unexplained by the variables included in the analysis (Dolédec et al., 2000). Thereby, a species having a low OMI (species score close to zero, 265 located in the center of the multivariate space) and high Tol is one that utilizes a wider array of resources and maintains populations within a wider variety of conditions (i.e., generalist), when compared with the specialized and less resilient species with more restricted realized-niche associated to high OMI and low Tol (Dolédec et al., 2000).
The OMI analysis was performed using the niche function in the 'ade4' package (Dray and Dufour, 2007), considering simultaneously the data obtained for 2015 and 2017 (N = 32). To compare the patterns observed in Patagonia to other localities 270 in the south eastern Pacific, a complementary OMI analysis was performed, including records of coccolithophore assemblages and E. huxleyi morphotypes from nearby coastal and oceanic waters (published by von Dassow et al., 2018) in addition to the data used in the first analysis (N = 64). A 1.84× correction factor was applied to these data, as coccolithophore counts from von Dassow et al. (2018) were obtained by the Utermöhl method. In both cases, data were arranged in two matrices, one containing the coccolithophore abundances and a second matrix with the standardized physical-chemical variables. 275 Coccolithophore abundances were previously Hellinger-transformed (Legendre and Legendre, 2012). Since Hellinger transformation is obtained by the squared root of relative abundances, the potential biases from comparing data from both SEM and Utermöhl counts was minimized. The statistical significance of the morphotypes/species marginality was tested using the Monte Carlo method included in the 'ade4' package (10,000 permutations). The envfit function in the 'vegan' package (Oksanen et al., 2007) was then used to fit the five environmental variables to the OMI scores (10,000 permutations). 280

The late-spring southern Patagonia 2015
The hierarchical clustering based on the surface values of the selected physical-chemical variables in the austral late-spring 2015 (i.e., salinity, temperature, Ωcal, pH and pCO2) showed a clear separation between the sampling station at the Skyring https://doi.org/10.5194/bg-2020-449 Preprint. Discussion started: 8 January 2021 c Author(s) 2021. CC BY 4.0 License. Sound (SS; st. 15) and the other localities (Fig. 3a). The other stations were grouped in two main clusters: one cluster composed 285 of stations in the Archipelago Madre de Dios (AMD) and the Inner Channel (IC) and a second one composed mainly of stations in the western part of the Strait of Magellan (WSM). Samples from the Otway Sound (OS; sts. 13-14) were distributed between the two clusters. The cluster separation seemed to be mainly related to temperature and salinity dissimilarities, while stations 4 and 15 differed from others by their relatively low pCO2/high Ωcal and high pCO2/low Ωcal conditions, respectively (Fig. 3b).
Surface salinity ranged from > 29 in the AMD and southernmost WSM stations to as low as 17 in the SS (st. 15), with 290 intermediate values throughout the IC and in the OS (range: 26-29; Fig. 3a-b). A north-south gradient of decreasing surface temperature was recorded from 9.0-10.0 °C around the AMD zone to 7.1 °C near Helado Sound (sts. 17-18; Fig. 3a-b, S4a).

E. huxleyi populations were mostly composed of the moderate A morphotype (which also included cells of the robust-calcified
A-morphotype; see Methods) ( Fig. 3e, Table 3). Few lightly-calcified A cells were observed. Among all samples, only eight 310 total O and C and no B or B/C morphotype cells were detected. The heavily A-CC and R/hyper-calcified morphotypes were restricted to the AMD zone.
Three vertical profiles were performed in the AMD estuarine zone (Fig. 4): one at the "limestone" western AMD basin (st. 3), one between the western and eastern AMD basins (st. 4), and one at the easternmost basin (st. 5, Fig. 4, S1). All samples were taken within the euphotic zone (1% PAR), which extended down to 36 and 30 m in sts. 3 and 4, respectively (st. 5 was 315 conducted at night, Fig. S7a-c). All three sites were sharply stratified in the upper 10 m, with pycnoclines around 5 m.
Temperature decreased while salinity increased with depth (Fig. 4a-c). In station 3, Ωcal varied little with depth, remaining in the range 2.5-2.7 (Fig. S7g) even across the shallow pycnocline. In station 4, Ωcal showed the highest values and highest range; AT, pCO2, and HCO3increased sharply with depth in the upper 10 m, while Ωcal decreased from over 3.6 to less than 2.9 at 9 m, and 2.7 at 29 m (Fig. S7h). Finally, at station 5, Ωcal decreased steadily from 3 at the surface to 2.4 at 25 m (Fig. S7i). Based 320 on total chlo-a levels and fluorescence signals, most photosynthetic biomass occurred in the upper 15 m of the water column in sts. 3-4, peaking at the surface in st. 4 (5.4 µg L -1 ). However, in station 5, total chlo-a levels and fluorescence signals were more constant with depth, dropping proportionally less by 25 m from maximal values, when compared to the other stations. The "Silicate" estuarine zone (st. 4-5) showed higher photosynthetic biomass and bSi than the "limestone" site ( Fig. 4a-c, S7d- E. huxleyi mostly occurred in the well-illuminated upper layer, most notably in the "limestone" western AMD waters, where 325 diatom abundance and biomass were low compared to the communities recorded in "silicate" sites ( Fig. 4d-f). Although the moderate A morphotype was predominant at all depths in the three stations ( Fig. 4g-i), a higher proportion of the A-CC morphotype (up to 31% relative abundance) was observed in the upper 15 m of the "limestone" western AMD when compared to the other two stations. The lightly-calcified and R/hyper-calcified morphotypes were present in a lower proportion (<10%).   Table S1). However, the values from that sample are shown in parenthesis for comparison. n.a = no available data.

Survey
Late     Fig. 3. See Fig. S7 for additional variables.

The early-spring southern Patagonia 2017 330
The hierarchical clustering based on the physical-chemical conditions in austral early-spring 2017 indicated a separation between the WSM and the IC, whereas the three stations in the AMD were distributed between the two clusters (Fig. 5a).
The dominant coccolithophore during early-spring 2017 was again E. huxleyi (> 96%). Abundances ranged from 1.69×10 4 to 9.06×10 4 cells L -1 (Fig. 5e). The E. huxleyi carbon biomass averaged 0.5 ± 0.3 µg C L -1 (in 11 samples), reaching both maximal 345 and minimal values (0.2 and 1 µg C L -1 , respectively) in the southern IC (sts. 25-27). In contrast, the opal-derived diatom carbon biomass averaged 40±17 µg C L -1 , with lower values (< 18 µg C L -1 ) in the AMD st. 30 and IC st. 25 (Fig. 5d). While fixed samples for standard microplankton analysis were not available, large chains of Skeletonema spp., Thalassiosira spp., and Chaetoceros spp. were noted as frequent in samples observed by SEM (Fig. S8), and were likely significant contributors to opal. Similar to the 2015 survey, the moderate A morphotype dominated the E. huxleyi assemblages along the 2017 track 350 (Fig. 5e). Cells of the lightly-calcified A morphotype were sporadically observed, whereas the highly-calcified A-CC and R/hyper-calcified morphotypes were again restricted to the AMD zone (Fig. 6, Table S1; note the low abundances of the R/hyper-calcified morphotype were only detected at other depths, so do not appear in Fig. 5e).
Two CTD profiles were performed in the AMD zone: one in the "limestone" western AMD basin (st. 30) and another profile southwest of Escribano Island at the "silicate" eastern AMD basin (st. 32) (Fig. 6a,b). The profiles covered the euphotic zone 355 (down to 27 m in st. 32; st. 30 was conducted at night) as well as sub-surface layers (25-75 m depth). In both stations, temperature and salinity increased with depth, with maximum density stratification between 5-10 m. In the western AMD profile, Ωcal was low at the surface (2.1) due to the lowest salinity, but rose to a maximum at 5 m due to a minimum in pCO2 and increasing salinity, whereas below 20 m, pCO2 rose and Ωcal dropped (Fig. S9e). At the eastern AMD site, in contrast, Ωcal increased with depth despite increasing pCO2 (Fig. S9f). At both stations, photosynthetic biomass was mainly confined to the 360 upper 25 m of the water column, with chlo-a peaks at 5 and 10 m (0.7 and 3.1 µg L −1 , respectively), and dropping close to zero below 40 m in the western AMD, while remaining near 1 µg L −1 even at depths ≥50 m in the eastern AMD ( Fig. 6a-b, S9a-b). The eastern AMD zone exhibited higher chlo-a and [bSi] when compared to the western AMD, despite no depletion of phosphate, nitrate and DSi were observed at either site ( Fig. 6a-b, S9a-d).
At both sites, E. huxleyi dominated the coccolithophore assemblages across all depths, with another six coccolithophore species 365 observed in sub-surface AMD waters (Table S3). E. huxleyi and diatom abundances were highest in the surface at both sites ( Fig. 6c-d). However, reflecting the chlo-a profile, estimated diatom biomass remained relatively high at depth compared to surface values (dropping by only about 40%), and E. huxleyi abundance and biomass also dropped less with depth in the eastern AMD compared to the western AMD (Fig. 6c-d). In both stations, the composition of E. huxleyi morphotypes was similar at all depths, characterized by the predominance of the moderately-calcified morphotype followed by highly-calcified A-CC (Fig.  370 6e-f). The lightly-calcified and R/hyper-calcified morphotypes were either undetected or represented a minor fraction of coccolithophore assemblages.

Emiliania huxleyi abundance vs. diatoms
The nMDS depicted a clear separation between the Patagonia fjords and the oceanic/coastal areas regarding the composition of coccolithophorid and diatom assemblages (Fig. S10). The IndVal analysis (Table S5) identified only the E. huxleyi 375 moderate-calcified morphotype as an indicator of the fjord locations, along with the diatoms Thalassiosira spp., Stephanopyxis turris, Leptocylindrus spp. and Chaetoceros spp. The coastal/oceanic locations were more characterized by the lighty-calcified and A-CC morphotypes and the other coccolithopore species (i.e., G. ericsonii, G. muellerae, G. parvula), as well as the diatoms cf. Lioloma spp., Pennate diatoms (< 50 μm length), Nitzschia spp., cf. Pseudo-nitzschia cuspidata, and cf.

Asteromphalus sarcophagus. 380
The two first axes of the CA accounted for 60% of the total explained variability and indicated that the highest E. huxleyi and low diatom biomasses were associated with increasing temperatures (Fig. S11). Intermediate E. huxleyi biomasses were associated with high diatom biomasses and increasing gradients of salinity, pH and Wcal, whereas low E. huxleyi biomasses were associated with intermediate diatom biomasses and increasing pCO2. However, none of the considered environmental variables had a significant fit in the envfit test. 385

Niche analysis of Emiliania huxleyi morphotype responses to environmental conditions
The OMI analysis depicted differences in the realized niches of the E. huxleyi morphotypes throughout Patagonia fjords in 2015 and 2017 (Fig. 7a, Table S6). The OMI plot showed station 15 from 2015 as an outlier, characterized by extremely low salinity and high pCO2. The OMI axis 1 (91.02% of explained variability) was negatively related to Wcal, whereas the OMI axis 2 (8.42 % of explained variability) was positively related to salinity and pH and negatively related to temperature and 390 pCO2. The envfit test indicated that all variables had a significant fit (R 2 > 0.88, p < 0.01; Table S7). Only the moderate A and R/hyper-calcified morphotypes showed significant OMIs (p < 0.05, Table S6). The moderate A morphotype was the most generalist (OMI = 0.07, Tol = 1.23), observed in all samples (except st. 15 in 2015). The R/hyper-calcified morphotype, observed exclusively in the AMD zone, was the most specialized morphotype (OMI = 4.77, Tol = 0.75). The A-CC morphotype (OMI = 1.43, Tol = 1.68), observed in the AMD and northern IC, showed intermediate habitat preferences (Fig. 7a), but the 395 OMI for this morphotype did not meet the threshold for significance (p = 0.060). The RTol for the R/hyper-calcified morphotype was 12% (Table S6), indicating that most variability in its realized niche was accounted for by the environmental variables included in the analysis.
The complementary OMI analysis (Fig. 7b, Table S8) indicated a clear separation between the Patagonian fjords and coastal and oceanic waters off central and northern Chile and Peru. The OMI axis 1 (74.57% of explained variability) was negatively 400 related to temperature, salinity and Wcal, whereas the OMI axis 2 (25.32 % of explained variability) was positively related to pH and negatively related to pCO2. The envfit test indicated that all variables had a significant fit (R 2 > 0.88, p < 0.01; Table   S9). All coccolithophore species and E. huxleyi morphotypes showed significant OMIs (p < 0.05, Table S8). The lightlycalcified, moderate-calcified and A-CC morphotypes, characterized as the most generalists, showed similar realized niches (OMI = 0.25-0.75, Tol = 2.85-2.96), whereas the R/hyper-calcified form was again the most specific of the E. huxleyi 405 morphotypes (OMI = 5.25, Tol = 1.97) and restricted to the coastal upwelling and AMD zones (Fig. 7b, Table S8). Regarding the other coccolithophore species, G. muellerae was common to both coastal and oceanic areas but still showed a higher degree of specialization (Tol = 1.03) than the E. huxleyi morphotypes, whereas C. leptoporus, G. ericsonii and G. parvula showed preference for oceanic conditions with low Tol values (0.15-0.38; Table S8). The R/hyper-calcified morphotype, G. ericsonii and G. parvula showed very low RTol (< 4.6%), indicating that most of their realized-niche variation was accounted for by 410 environmental variables in the analysis (Table S8).

Patagonian coccolithophore communities dominated by E. huxleyi
Emiliania huxleyi was the only coccolithophore widely distributed along the fjords and inner channels of southern Patagonia and always represented > 96% of total coccolithophore abundance and biomass, during both early/late spring. The low diversity 415 of coccolithophores assemblages, dominated by E. huxleyi, is common to both the Patagonian and Norwegian fjord systems.

Abundance of E. huxleyi in Patagonia compared to nearby oceans
During the early/late spring, standing stocks of E. huxleyi in the Patagonian fjords and inner seas were moderate compared with those documented in nearby coastal and oceanic regions and within the range of background stocks reported in the Norwegian fjords and North Sea (Table 4 and references therein). The high E. huxleyi abundances typical of spring blooms in the Norwegian fjords were not observed in either early or late spring in the present study despite the similar temperature, 425 salinity, and Ωcal conditions in both fjord systems. No E. huxleyi blooms have been reported in Patagonia fjords, although this might be due to limited observations and methodological issues. For example, many phytoplankton studies in the area (e.g., -de-Souza et al., 2008;González et al., 2013) as well as standard phytoplankton monitoring in the zone (Vivanco and Seguel, 2009) often rely on samples fixed with acid-Lugol's, which would not preserve coccoliths, or have only focused on larger phytoplankton size classes (e.g., Paredes et al., 2014). 430

Variation in E. huxleyi with environmental factors
It has been previously proposed that the realized niche of E. huxleyi is partly defined by physical and chemical conditions unfavorable to large diatoms (Tyrrell and Merico, 2004;Hopkins et al., 2015). During late-spring Patagonia fjords, E. huxleyi reached higher abundances in the southern IC when the temperature was above 8 °C and macronutrients, and larger diatoms were the lowest, consistent with the pattern previously reported more generally for nano-phytoplankton based on size-435 fractionated chlo-a for this geographic area (e.g., Cuevas et al., 2019). However, the CA analysis showed that the lowest levels of E. huxleyi were associated with intermediate levels of larger diatoms, and intermediate levels of E. huxleyi were associated with highest levels of larger diatoms, suggesting a unimodal relationship between these two planktonic groups, possibly affected by environmental and biotic factors not assessed in this study.
The Ωcal -a parameter assumed to constraint calcification in coccolithophores -was subject to large spatial variations in 440 surface waters, from relatively high Ωcal levels in the AMD zone (range: 2.1-3.6), moderate Ωcal in the interior WSM, low Ωcal in the southern IC (range: 1.5-2.2) and sub-saturating in the SS (0.5). The range of surface Ωcal recorded along southern Patagonia was comparable to those reported for the Norwegian seas (Jones et al., 2019). Whereas the highest Ωcal values observed at the AMD were not as high as those observed in the global ocean (Takahashi et al., 2014), the lower values at the southern IC were comparable to values reported previously (range: 1.8-2.8) from high CO2 upwelling conditions in central and 445 northern Chile (Beaufort et al., 2008(Beaufort et al., , 2011von Dassow et al., 2018).

Comparison of E. huxleyi morphotypes in Patagonia to nearby oceans vs. Norwegian fjords
There was some variability in the vertical distribution of the E. huxleyi morphotypes in the water column. The lightly-calcified coccoliths appeared associated with subsurface waters in both seasons sampled at the locations, so they might be associated with intrusion of these waters. However, the samples within the euphotic zone were generally similar to each other within a given sample station. Thus, for the purposes of the questions in this study, the use of surface samples to describe morphotype 455 distributions is expected to be reasonable, but we caution that more subtle patterns could have been revealed if more vertical profiles have been obtained.
The E. huxleyi populations in the Patagonian fjords were completely distinct from surrounding coastal or open ocean populations in the eastern South Pacific, the Southern Ocean, and the Atlantic. The Atlantic Patagonian Shelf E. huxleyi populations are reported to be dominated by B/C morphotypes (Poulton et al., 2011. Southern open ocean populations 460 of E. huxleyi are dominated by B morphotypes (including the B, B/C, C, and O types; Saavedra-Pellitero et al., 2014;Saavedra-Pellitero et al., 2019), and A morphotypes were reported to represent only a small fraction. However, C and O morphotypes were very rare in Patagonian inland waters, and B and B/C morphotypes were undetected. Although the moderate-calcified and robust-calcified A morphotypes have also been shown to be present in eastern South Pacific coastal and open ocean waters (von Dassow et al., 2018), the dominance of these A morphotypes was particular to Patagonian interior waters, as revealed by 465 the IndVal analysis (these A moderate-calcified and robust-calcified A morphotypes were consolidated for final statistical analyses as they are not easily distinguished by objective morphological characters and were present in all samples, and preliminary analysis revealed completely overlapping realized niches). The moderate-calcified and robust-calcified A morphotypes are also observed as dominant in the Norwegian fjords (Table 4) (Young, 1994). The lightly-calcified A morphotype was rare, and did not show any clear pattern in its distribution. The A-CC morphotype has been associated with 470 coastal upwelling zones in the Atlantic (Giraudeau et al., 1993;Smith et al., 2012;Henderiks et al., 2012) but not reported from the Norwegian fjords or the Southern Ocean. In both early/late spring, the R/hyper-calcified and A-CC E. huxleyi appeared only at the Pacific border of southern Patagonia (AMD zone). Thus, E. huxleyi populations of both Patagonian fjords and Norwegian fjords share a similar morphotype composition.  Birkenes and Braarud, 1952Berge, 1962Paasche and Kristiansen, 1982Erga, 1989Kristiansen et al., 1994Young, 1994Fernández et al., 1996Samtleben et al., 1995Baumann et al., 2000Egge et al., 2015Gran-Stadniczeñko et al., 2017van Bleijswijk et al., 1991Charalampopoulou et al., 2011Zingone et al., 2011 This study This study Okada, 2006 Beaufort et al., 2008;2011Menschel et al., 2016Alvites, 2016von Dassow et al., 2018McIntyre et al., 1970Saavedra-Pellitero et al., 2014Charalampopoulou et al., 2016 Saavedra

Niche analysis E. huxleyi morphotypes related to carbonate chemistry conditions 475
The broader niche-breadth by the moderate-calcified A morphotype contrasted with the marginal niche of the R/hyper-calcified forms in Patagonia (Fig. 7a). In order to extend the realized-niches derived in Patagonia, we complemented the OMI analysis with a sample set of nearby oceanic and coastal sites (data from von Dassow et al., 2018), in some of which the moderatecalcified A morphotype, unlike in Patagonia, was less abundant than other E. huxleyi morphotypes and coccolithophore species. According to OMI analysis, the niche-differentiation along Patagonia is mostly driven by the pH/Ωcal conditions, but 480 temperature and salinity conditions also become important. In this extended domain, both the moderate-calcified A morphotype and the A-CC morphotype appeared to be generalists, with high Tol values (Fig. 7b). The lightly-calcified morphotype also appeared to be a generalist. However, we caution that while the lightly calcified E. huxleyi were almost exclusively lightly-calcified A morphotype in Patagonia, there was a continuum of lightly-calcified A, B, and B/C morphotypes (and some lightly calcified cells were difficult to classify among these types) in the coastal and oceanic sites. In contrast, the 485 very distinct R/hyper-calcified morphotype exhibited restricted preferences in terms of Ωcal, temperature, and salinity, but a broad niche in terms of CO2 and pH (Fig. 7b).
The R/hyper-calcified morphotype, in which there is both fusion of distal shield elements and closure or partial closure by over-calcification of the central area, has so far only been reported as prevalent in high CO2/low pH upwelling zone of the eastern South Pacific (Beaufort et al., 2011;Alvites, 2016;von Dassow et al., 2018), although it has seen (and reported as rare) 490 in both Australian waters (Cubillos et al., 2007)  the Southeastern Pacific that correlates with the Ωcal, temperature, and salinity of its realized niche.
A striking result from the OMI analysis was that all the E. huxleyi morphotypes, even the more specialized R/hyper-calcified type, exhibited much greater niche breadth (higher Tol values) than the other coccolithophore species. The three Gephyrocapsa species are very close relatives of E. huxleyi and phylogenetically should be considered as congenerics (Bendif et al., 2016;Bendif et al., 2019), but all showed lower niche breadth than the E. huxleyi morphotypes. The small G. parvula and G. ericsonii 500 showed Tol values that were more than 10-fold lower than the most specialist E. huxleyi morphotype. Despite the evidence for a genetic underpinning of E. huxleyi morphotype (Krueger-Hadfield et al., 2014), as well as evidence of a high level of genomic content variability in E. huxleyi (von Dassow et al., 2015), phylogenetic and phylogenomic evidences do not clearly support for it to be split into different species (Bendif et al., 2016;Filatov, 2019). If the ubiquitous taxon is less susceptible to environmental change compared to marginal taxa (i.e., marginality or richness vs. tolerance are inversely correlated; Dolédec 505 et al., 2000;Hernández et al., 2015), the exceptional generalist behavior exhibited by E. huxleyi compared to other coccolithophores suggests it may be more plastic and more adaptable in the face of environmental change.

Conclusions
Our study of how E. huxleyi populations and morphotypes respond to the highly dynamic physical and chemical environments of southern Patagonia yielded seven principal findings: 510 1. The only coccolithophore that was a regular and ubiquitous component of the phytoplanktonic assemblages throughout the surface waters of the southern Patagonian fjords/channels was E. huxleyi. It occurred under a wide range of carbonate chemistry conditions and was only absent in the SS zone where Ωcal < 1.
2. Although E. huxleyi never reached more than a small fraction of total phytoplankton biomass, it reached moderate abundances comparable to adjacent coastal and oceanic areas, and within the lower range of stocks reported from 515 Norwegian fjords.
3. E. huxleyi abundance was highest when assemblages of large diatoms were lowest, in waters with lower macronutrients, consistent with it being most important in the absence of large diatoms.
4. In terms of morphotypes, the E. huxleyi populations in the southern Patagonian fjords/channels were similar to Norwegian fjords and very distinct from populations previously documented in the eastern South Pacific, Southern 520 Ocean/Drake Passage, and the Patagonian Shelf of the Atlantic.
5. Niche analysis shows that the moderate A morphotype and A-CC morphotypes are generalists, whereas the R/hyper-calcified morphotype has a more marginal (specialized) realized niche.
6. The association of the R/hyper-calcified morphotype to high Ωcal in southern Patagonia, where Ωcal is driven principally by freshwater input, contrasts with its dominance of the upwelling system of central Chile to Peru,525 where low Ωcal is due to high CO2. This morphotype occupies a narrow range of Ωcal values compared to the ACC and moderate A-morphotypes. 7. The moderate A, A-CC, and R/hyper-calcified E. huxleyi morphotypes all display higher niche breadth (more generalist behavior) than closely related coccolithophores, suggesting that E. huxleyi may be ecologically more plastic and have more capacity for adaptation in the face of environmental change than other coccolithophores. 530

Data availability
All data resulting from this study are available from the corresponding author upon request. The scanning electron micrograph image datasets can be found at https://doi.org/10.5281/zenodo.4292020

Sample availability
Material for SEM characterization (filter sections) are in Dr. von Dassow's laboratory and can be requested.

Competing interests 555
The authors declare no competing interests

Acknowledgments
This study was supported by the Comisión Nacional de Investigación Científica y Tecnológica (now Agencia Nacional de