The haplo-diplontic life cycle expands niche space of coccolithophores

. Coccolithophores are globally important marine calcifying phytoplankton that utilize a haplo-diplontic life cycle. The haplo-diplontic life cycle allows coccolithophores to divide in both life cycle phases, and has been proposed to allow coccolithophores to expand their niche space. To-date research has however largely overlooked the life cycle of coccolithophores, and has instead focused on the diploid life cycle phase. Through a synthesis of global scanning electron microscopy (SEM) coccolithophore abundance data (n = 2534), we show that the haploid life cycle phase contributes signiﬁcantly to coccolithophore 5 abundance, constituting ≈ 18 % of species abundance for which haploid-diploid pairs are deﬁned. Using hypervolumes to quantify the niche space of coccolithophores, we furthermore show that the haploid and diploid life cycle phases inhabit contrasting niches, and that this allows coccolithophores to expand their niche space by ≈ 17 %. Our results highlight that future coccolithophore research should consider both life cycle stages, as omission of the haploid life cycle phase in current research limits our understanding of coccolithophore ecology. and is contribution number 349 of the AMT programme. The C1-LTER station is part of the national and international Long Term Ecological Research network (LTER-Italy, LTER-Europe, ILTER). Data for the C1-LTER station were obtained in the framework of the EU FP7 MedSeA (Mediterranean Sea Acidiﬁcation in a Changing Climate) project. Environmental and nutrient data were made available through the OGS Italian National Oceanographic Data Center (NODC). This is a scientiﬁc contribution of Project MIUR - Dipartimenti di Eccellenza 2018-2022 to the DISAT of Milano-Bicocca. We would like to thank for the statistical clinics provided by the Institute for Statistical Science 440 at the University of Bristol. Finally we would like to thank everyone who as contributed data to the compilation.

approximately monthly observations between January 1991 to January 1994. For the Mediterranean study, we combine two time-series in the Adriatic Sea by Godrijan et al. (2018) and Cerino et al. (2017), between September 2008to December 2009 and May 2011 to February 2013 at the RV-001 and C1-LTER stations respectively.
For the AMT and Mediterranean case studies, we additionally compiled temperature, salinity, and concentrations of fixed 90 nitrogen (nitrite + nitrate), phosphate, and silicate. For the AMT environmental variables were acquired from the British Oceanographic Data Centre (BODC). For the Mediterranean study, day length was calculated using the MIT Skyfield package in Python.
All data was acquired from supplementary data, online databases, or if neither was available by contacting the authors directly. The data was manually checked for synonyms or misspellings of species names, and where appropriate cell abundances 95 were converted to cells l −1 . All species, or genera if not identified to a species level, were labeled as either heterococcolithophore, holococcolithophore, or 'other', which includes polycrater, nanoliths, and unidentified species. For these categorizations we followed definitions from Cros and Fortuño (2002).
The species and environmental data were compiled in Python, and subsequently analysed in R (R Core Team, 2019). For all analysis we only considered samples within the top 200 m of the water column. On a global scale and regional scale, we 100 calculated the mean and highest observed abundances (the 'maximum abundance') of both hetero-and holococcolithophores.
For the mean abundance calculations the mean was calculated for each sample and then averaged.

Definition of pairs and HOLP-index
Not all heterococcolithophore forming coccolithophore species form holococcospheres. Thus, to better illustrate the proportion of haploid and diploid coccolithophore cells, we reported the ratio between hetero-and holococcospheres of species that form 105 holococcoliths in their haploid phase, which is commonly implemented (Cros and Estrada, 2013;Šupraha et al., 2016).
This ratio is referred to as the 'HOLP-index', and is defined by Cros and Estrada (2013) as: HOLP-index = 100 · paired holococcolithophore abundance paired coccolithophore abundance (1) Species included in the HOLP-index follow the definitions of paired species as defined in Frada et al. (2018) (Table 2) -which is confined to currently understood associations and which is likely to change as our understanding holococcolith 110 species continues to improve. We calculated the HOLP-index on a global and regional level for studies that identified holococcolithophores to a species level, the AMT data set, and the Mediterranean data set. To calculate the mean HOLP-index, the ratios were calculated for each sample and then averaged.

Environmental drivers
We quantified the environmental drivers of hetero-and holococcolithophore abundance and the HOLP-index for the AMT 115 and Mediterranean data sets using Spearman correlations. We calculated Spearman correlations for hetero-and holococcol-ithophores and the HOLP-index relative to temperature, salinity, depth, and concentrations of fixed nitrogen (nitrite + nitrate), phosphate, and silicate for the AMT data set. The same ordinal associations were calculated for the Mediterranean data set, but we considered day length instead of depth, because only the top 30 meters of the water column was sampled, and seasonality is an important driver in this region. To focus on marine systems of coccolithophores, we only considered samples with salinities 120 above 30 ppt. Samples missing any environmental variables were removed. Subsequently the AMT data set included a total of

135
To investigate seasonality we compared monthly hetero-and holococcolithophore abundance data to temporal variations of temperature, salinity, day length, and concentrations of phosphate, fixed nitrogen, and silicate of the BATS and Mediterranean data sets.

Niche overlap and niche expansion
Distribution patterns of phytoplankton are influenced by multiple environmental drivers. These environmental drivers form a n-140 dimensional hyperspace within which hypervolumes can be defined based on where the phytoplankton occur. This hypervolume can be used to quantify niche space (Hutchinson, 1957) and allows comparisons between multiple phytoplankton -in this instance the two life cycle phases of coccolithophores.
Although processing hypervolumes is challenging due to their high dimensionality, methods described by Blonder et al. (2014) allow hypervolume quantification and comparison (for futher discussion see Blonder (2018) and Mammola (2019)).

145
Using this strategy we determine the niche overlap of hetero-and holococcolithophores in hyperspace using the Sørensen-Dice and Jaccard similarity metrics.
We furthermore calculate the 'niche expansion' of the haplo-diplontic life cycle strategy, which we define here as the nonoverlapping region of either phase within hyperspace. In other words: Where NE(A)= niche expansion of A; A = hypervolume A; B = hypervolume B; ∩ = intersection between two hypervolumes; ∪ = union between two hypervolumes We calculated the Jaccard and Sørensen-Dice similarity metrics and niche expansion for both the Atlantic Ocean and Mediterranean Sea data set. For the Atlantic Ocean, nitrogen showed high Pearson correlation to silicate (ρ = 0.95, p < 0.001) and phosphate (ρ = 0.90, p < 0.001). We thus only considered temperature, salinity and the concentration of fixed nitrogen in 155 this region. Although no such correlations were observed for the Mediterreanean data set, to make the niche metrics comparable in both regions, the silicate and phosphate concentration of the Mediterrean data set were also excluded. The environmental data were normalized using z-scores prior to analysis. Niche overlap and niche expansion was calculated only for species for which both life cycle phases were observed.
We used the 'hypervolume' R package (Blonder and David J. Harris, 2018) to conduct our niche overlap and niche expansion 160 analysis. Gaussian kernel density estimation (R function 'hypervolume_gaussian') was used to construct the hypervolume, the overlap metrics were calculated with the 'hypervolume_overlap_statistics' R function, and the volume and intersection of hyper volumes were calculated using the 'get_volume' R function.

Biogeography of coccolithophores 165
Within our compilation heterococcolithophores showed global distribution, while holococcolithophores were noticeably absent at the ALOHA station in Hawaii and (with some exceptions) >50 • S in the Southern Ocean ( Fig. 3 and Table 3).
Highest maximum abundances of heterococcolithophores were observed at high latitudes within the Arctic circle (>66 The regions with the highest mean heterococcolithophore abundance differed from the regions where highest maximum heterococcolithophore abundance were observed. For example, the highest mean abundance was observed in the East Indian Ocean (≈1.96 x 10 5 cells l −1 ); which was higher than the mean abundance in the Southern Ocean (≈8.87 x 10 4 cells l −1 ), North Atlantic (≈9.99 x 10 4 cells l −1 ) and Arctic Circle (≈5.71 x 10 4 cells l −1 ).

175
Although holococcolithophores showed low abundances at at high latitude regions in the Southern Hemisphere, highest maximum holococcolithophore abundances were observed in the Arctic circle (>66 • N) (≈2.23 x 10 5 cells l −1 ). High maximum abundances were additionally observed in the Mediterranean Sea (≈1.38 x 10 5 cells l −1 ). Generally low maximum abundances were observed at tropical and subtropical basins such as the Arabian Sea (≈5.63 x 10 3 cells l −1 ); with exception of the East Indian Ocean (≈3.10 x 10 4 cells l −1 ). Medium maximum abundances were observed in the North Atlantic Ocean (≈3.00 x 180 10 4 cells l −1 ). On average the Mediterranean Sea had the highest mean holococcolithophore abundance (≈6.00 x 10 3 cells l −1 ), followed by the East Indian Ocean (≈4.40 x 10 3 cells l −1 ), and the Arctic Circle (≈1.70 x 10 3 cells l −1 ). The lowest mean abundances were observed in the Southern Ocean (≈4.00 cells x 10 2 l −1 ), and Arabian Sea (≈4.60 x 10 2 cells l −1 ).
Overall holococcolithophores contributed 7.3 % (±16 %) to total coccolithophore abundance globally, with their highest contribution observed in the Mediterranean Sea (16.5 % ±22.7 %) (Table 3). However, on a regional basis (Table 4) holo-185 coccolithophores generally contributed less than 6 % to total coccolithophore abundances. The contribution of holococcolithophores to paired species was much higher than when all hetero-and holococcolithophores are considered ( observed.

Vertical distribution
In the global data set heterococcolithophore abundance is evenly distributed with depth, while holococcolithophore abundance is highest in the top 50 m of the water column ( Fig. 4).
For holococcolithophores the vertical distribution pattern is mainly driven by paired holococcolithophore species which 195 constituted ≈62.2 % to total coccolithophore abundance. Two currently unpaired holococcolithophores also contribute to the depth distribution trend with Helladosphaera cornifera (for which the association has to be further confirmed) constituting ≈8.1 % of total holococcolithophore abundance, and Corisphaera gracilis (for which no pair has been described) constituting ≈3.6 % of total holococcolithophore abundance. Subsequently paired holococcolithophore abundances broadly followed the same patterns observed when all holococcolithophores were considered.

200
In comparison to holococcolithophores, depth distribution of heterococcolithophores was driven by unpaired species -in particular E. huxleyi which constituted ≈59.2 % of total heterococcolithophore abundance, but also by the presence of unpaired deep water species such as Ophiaster formosus, Florisphaera profunda, Calciopappus caudatus, and Oolithotus antillarum.
However, although paired heterococcolithophores only contributed ≈5.7 % to total heterococcolithophore abundance, the depth distribution trends of paired and total heterococcolithophores species were similar.

Environmental drivers of niche partitioning
To further understand the distribution patterns observed on a global basis and within the water column we investigated the environmental drivers of hetero-and holococcolithophore abundance in the Atlantic Ocean (with the AMT data set) and the Mediterranean Sea. For the Atlantic Ocean data set, the environmental drivers were considered in the context of their distribution within the water column, whereas for the Mediterranean the environmental drivers were considered within PCA 'niche 210 space'. These observed patterns were then further corroborated through Spearman analysis.

Atlantic Ocean
In the Atlantic Ocean both hetero-and holococcolithophores have highest abundances in the top 50 m of the water column ( Fig.   5). However, a noticeable difference between hetero-and holococcolithophore distribution ( Fig. 5a and Fig. 5d respectively) is the absence of holococcolithophores below the deep chlorophyll maximum (DCM) (Fig. 5l). The DCM tends to occur at 215 1-10 % irradiance levels, and is closely linked to the nutricline and thermocline (Poulton et al., 2006). The difference in depth distribution between hetero-and holococcolithophores, and the absence of holococcolithophores below the DCM may therefore be influenced by a combination of light limitation, high nutrient concentrations, and cold water temperatures at depth. This suggests that heterococcolithophores might be better adapted to exploit such conditions. Although differences in sinking rates -which are conceivably higher in the more heavily calcified heterococcolithophores could also factor into the difference in

225
Within the upper water column, heterococcolithophores showed highest abundance at higher latitudes (>35 • N and >30 • S), which is associated with a shallow mixed layer, lower salinity, and lower temperature, as well as increasing silicate concentrations in the southern hemisphere. Holococcolithophores meanwhile showed highest abundances at both high latitudes and in the Atlantic subtropical gyres. The HOLP-index ( Fig. 5f) was highest within the Atlantic subtropical gyres, with a higher proportion of holococcolithophores in the Northern subtropical Gyre, which is associated with a shallower DCM relative to 230 the Southern subtropical Gyre. This shallowing of the DCM on the AMT is however likely a seasonal signal as described by Poulton et al. (2006) and Poulton et al. (2017).
Spearman correlations (Table 6) suggests holococcolithophores are significantly (p<0.05) negatively correlated to phosphate, fixed nitrogen, silicate and depth and significantly positively correlated to temperature and salinity. Paired holoccolithophore and the HOLP-index showed the same correlation trends as holococcolithophores.

235
On the contrary, heterococcolithophores are only significantly and negatively correlated with depth and phosphate. While for paired heterococcolithophores significant negative correlations were observed with depth and silicate.
Thus hetero-and holococcolithophore abundance in the Atlantic Ocean seems primarily driven by the depth of the DCM both in terms of vertical and latitudinal distribution. Highest abundances of both hetero-and holocococlithophores are observed above the DCM, and heterococcolithophores are present below the DCM while holococcolithophores are not. In terms of 240 latitude highest abundances of heterococcolithophores correspond to shallow DCM depth which occurs in higher latitude regions, and highest abundances of holococcolithophores occur in subtropical regions with deep DCM depths.

Mediterranean Sea
For the Mediterranean Sea long term time series, niche separation of hetero-and holococcolithophores within the PCA niche space (Fig 6), is primarily driven by Principal Component 1 (PC1) which is positively associated with temperature and day 245 length and negatively associated with salinity, fixed nitrogen, silicate and phosphate (see Table 7). Heterococcolithophores are most abundant at low PC1 values (i.e. the left quadrants of Fig. 6a) which corresponds to low temperatures and short day lengths, and high salinity and concentrations of fixed nitrogen, silicate and phosphate. Holococcolithophores are most abundant at high PC1 values (i.e. the right quadrants of Fig. 6b), which corresponds to high temperatures and long day lengths, and low salinity and concentrations of fixed nitrogen, silicate and phosphate.

250
The pattern observed in the PCA niche space is also apparent in the Spearman correlations ( Table 6) which indicate that heterococcolithophores are significantly negatively correlated to temperature, and day-length, and significantly positively correlated to phosphate, fixed nitrogen, silicate and salinity. For paired heterococcolithophore species the only significant correlation observed was a positive correlation with silicate.
Holococcolithophores showed an opposite pattern to heterococcolithophores, and are significantly positively correlated to 255 day-length and temperature, and significantly negatively correlated to salinity, fixed nitrogen, silicate and phosphate. Paired holoccolithophores and the HOLP-index showed significant positive correlation to temperature and day length, but no significant correlations with the other environmental variables were observed.

General environmental trends
Our statistical analysis shows that in both the Mediterranean Sea and Atlantic Ocean holococcolithophores are generally found 260 in low nutrient and warm environments and high light availability. However, an opposite trend was observed between the Atlantic Ocean and Mediterranean Sea in terms of correlation to salinity, with holococcolithophores positively correlated to salinity in the Atlantic Ocean and negatively correlated to salinity in the Mediterranean Sea. This difference in correlation to salinity may be explained by the different drivers of salinity in both regions. In the Atlantic Ocean, low salinity occurs at high latitudes, while high salinity corresponds to mid-ocean gyres due to higher evaporation in tropical and sub-tropical regions. In 265 contrast, at the coastal site in the Mediterranean Sea, low salinity is strictly related to direct freshwater input and associated nutrients. As such salinity may be simply correlated to other environmental drivers, rather than be a driver itself.
Statistically significant correlations were the same when all holococcolithophores, paired holococcolithophores or the HOLPindex was considered at both locations -however fewer significant correlations were observed for paired holococcolithophores and the HOLP-index.

270
The trend for heterococcolithophore is less clear when comparing the two sites: an opposite trend to holococcolithophorese.g. high nutrients and low temperatures -is observed in the Mediterranean Sea, but not in the Atlantic Ocean where many of the correlations were not significant and heterococolithophore were negatively correlated to phosphate. This negative correlation to phosphate is potentially due to deeper sampling in the Atlantic Ocean combined with high phosphate concentrations in deep and light limited waters skewing correlations, which highlights the need to consider sampling and DCM depth when com-275 paring environmental correlation between studies. It may furthermore be due to the presence of mixotrophic or heterothrophic coccolithophores at depth in the Atlantic Ocean, which are not found in the shallow coastal waters of the Mediterranean Sea.

Niche overlap and niche expansion
We conducted niche similarity and niche expansion calculations on both the Atlantic Ocean and Mediterranean data sets to quantify niche space in these regions. For niche overlap we considered the Jaccard overlap and Sørensen-Dice overlap metrics 280 which range from 0 to 1, with 1 signifying complete overlap. For niche expansion we considered the relative amount each life cycle contributed to the total niche space. In the Atlantic Ocean the niche overlap of paired species was high for both the Jaccard overlap and Sørensen-Dice overlap metrics (0.84 and 0.91 respectively, Table 8). However, for individual species the overlap metrics were highly variable ranging from 0.11 -0.74 and from 0.20 -0.81 for the Jaccard overlap and Sørensen-Dice overlap metrics respectively. The niche expansion was higher for heterococcolithophores than holococcolithophores when all 285 paired species were considered (see Table 8), but was again highly variable for individual species. The holococcolithophore phase of C. mediterranea, S. bannockii, H. wallichii, and C. leptoporus for instance all contributed more to the total niche space than their heterococcolithophore life cycle phase.
In the Mediterranean Sea niche overlap values were smaller, and niche expansion values were larger than in the Atlantic Ocean (Table 9). Niche expansion of heterococcolithophores was also higher than holococcolithophores when all paired species 290 were considered, but like in the Atlantic Ocean species specific exceptions were observed. The holococcolithophore phase of C. mediterranea, S. histrica, S. strigilis and C. leptoporus all contributed more to the total niche space than their heterococcolithophore life cycle phase in this region. In the Mediterranean Sea the niche space of S. molischii is of particular note, as no overlap between the two life cycle phases was observed, and the two unique components were of similar size (0.51 and 0.49 for hetero-and holococcolithophores respectively).

295
Although quantitative interpretation of niche space is difficult since niche space will vary depending on the number of environmental axes included (Blonder et al., 2014), these results highlight that holococcolithophores contribute significantly to the niche space of coccolithophores, in some instances contributing more to total niche space than the heterocococlithophore phase. In this context C. pelagicus is particularly relevant as this species contributes significantly to the global carbonate flux  These results additionally suggest that the niche expansion and difference in niche preference between the two life cycle phases is higher in the Mediterranean Sea than the Atlantic Ocean. It is however not clear if this is because the haplo-diplontic life cycle is better suited to more variable coastal environments as suggested by Godrijan et al. (2018) or due to higher temporal sampling resolution in the Mediterranean Sea data set compared to the Atlantic Ocean data set.

Seasonality of coccolithophores 305
Hetero-and holococcolithophore abundance highly varies with season at both the BATS station in the Atlantic Ocean and the long-term stations in the Mediterranean sea (Fig. 7). Both locations experience a peak of heterococcolithophores in the winter, followed by a peak of holococcolithophores at the end of spring, and in early summer. In the Atlantic Ocean, the heterococcol-https://doi.org/10.5194/bg-2020-194 Preprint. Discussion started: 24 June 2020 c Author(s) 2020. CC BY 4.0 License.
ithophore are present in high abundance for a longer period of time -overlapping with the spring peak in holococcolithophore abundance (Fig. 7a).

310
At both locations the peak of the holococcolithophore bloom occurs in the spring and summer when water temperatures rise and the day length is longest, while heterococcolithophore abundance is highest in the winter when temperature is lowest and day length shortest. The seasonality of peak hetero-and holococolithophore abundance may furthermore correspond to seasonal changes in mixed layer depth (MLD), as both the Atlantic Ocean and Mediterranean Sea experience increased mixing in the winter and higher stratification in the summer.

315
No clear seasonal patterns were observed for fixed nitrogen or silicate concentrations at either location. Which suggest that although hetero-and holococcolithophores abundance is correlated to nutrient concentrations on spatial scales (see Sect. 3.3), on a seasonal scale other drivers such as temperature, light availability, and mixed layer depth predominate.
It is important to note that on a species level, individual species do not exclusively follow the seasonal hetero-holococcolithophore trends described above, as illustrated in detail by the original publications (Cerino et al., 2017;Godrijan et al., 2018). For in-320 stance for Syracosphaera molischii and Syracosphaera pulchra the holococcolith rather than heterococcolith phase is the dominant life cycle phase in these time series. Furthermore the holococcolithophore phases of S. molischii, Syracosphaera histrica, Algirosphaera robusta and Acanthoica quattrospina are observed in the winter -a period when total holococcolithophore abundance is lowest. Finally on a individual level, succession does not immediately follow the previous life cycle phase with several months of absence observed between peak abundance for some species (Cerino et al., 2017;Godrijan et al., 2018).

325
This highlights that grouped hetero-holococcolithophore abundances represents a generalization that might not always represent patterns observed for individual species. These differences from generally observed patterns could be due to variations in life strategy -such as mixotrophy, motility and grazing susceptibility -independent of life cycle phase. Suggesting that functional traits different from the life cycle phase may determine the niche space these species inhabit.

330
Our meta-analysis shows that holococcolithophores are important contributors to coccolithophore abundance and ecology, contributing ≈7.3 % to total coccolithophore abundance. Our analysis furthermore shows that haploid cells play an important role in coccolithophore species that calcify in their haploid phase, accounting on average for ≈18.3 % of their total abundance.
Although holococcolithophore contribution to calcium carbonate production is likely small due to their lower cellular CaCO 3 content -which is an order of magnitude lower than heterococcolithophores (Daniels et al., 2016;Fiorini et al., 2011a, b) -their 335 role in the carbonate cycle in present, past and future oceans is not to be underestimated. A shift towards a higher proportion of holococcolithophore cells, would result in lower global calcium carbonate production which would subsequently result in lower CO 2 outgassing on short time scales. Furthermore the ballasting effect of coccolithophores would be reduced if a shift towards more lightly calcified haploid cells occurred (Hoffmann et al., 2015).
In terms of the ecological niche space -which is the environmental range a species inhabits -hetero-and holococcol- However, some exceptions occur. For instance in the AMT data set, although heterococcolithophores are more evenly distributed with depth, maximum abundance of heterococcolithophores is in surface waters, and subsequently heterococcolithophores are negatively correlated to nutrients. Nonetheless the relation to turbulence holds: heterococcolithophore abundance is highest in well mixed high latitude waters and holococcolithophore abundance is highest in stratified sub-tropical regions.

360
Finally, many species specific exceptions occur. We highlight examples on a seasonal scale in our Mediterranean data set discussion (see Sect. 3.6), but exceptions were also noted along the AMT (see discussion in Poulton et al. (2017), and in other Mediterranean studies (Šupraha et al., 2016;?;Skejić et al., 2018). Which means that caution should be used when considering the niche space model for individual species.

Niche overlap and expansion 365
Our study showed that the niche volume of coccolithophores is larger when holococcolithophores are included in coccolithophore niche space. This tells us two things: first, studies focused solely on heterococolithophore are underestimating coccolithophore habitat and thus inaccurately represent the coccolithophore functional group in modelling and physiological studies, which means that we might be underestimating their ability to compete with other phytoplankton, as well as the range of environmental conditions they can tolerate. Secondly, we are underestimating the importance of coccolithophore primary 370 productivity and carbonate production by not including accurate assessments of their abundance or activity.
This might be of particular relevance for E. huxleyi, the diploid phase of which has been of particular research focus due to high abundances (approx 59.2% in our compilation). Although our meta-analysis does not include haploid abundance data of this species, we suspect, following upon our findings on the haploid/diploid paired species, that the haploid phase of E.
huxleyi is also ecologically relevant. Previous studies suggest that the haploid life cycle phase of E. huxleyi can increase its https://doi.org/10.5194/bg-2020-194 Preprint. Discussion started: 24 June 2020 c Author(s) 2020. CC BY 4.0 License.
niche space due to streamlined metabolism (Rokitta et al., 2011), and variations in response to bacterial (Mayers et al., 2016;Bramucci et al., 2018), and viral pressures (Frada et al., 2008). Although it should be noted that in some instance, morphology rather than ploidy level seems to be the primary driver for observed differences in E. huxleyi (Frada et al., 2017). Overall, observations in the haploid stage of E. huxleyi are extremely limited due to difficulty of identifying the haploid phase with regular light microscopy, highlighting the need for developing new techniques to account for this potentially important life 380 cycle stage. Further development of the COD-FISH method as described by Frada et al. (2012) in particular would be relevant in this context.

Concluding remarks
Our compilation provides insight into the distribution of hetero-and holococcolithophores, but also highlights many gaps in the data distribution and our knowledge on coccolithophore ecology. There is for instance a lack of SEM observations in the 385 Pacific Ocean (2 studies in this compilation), and there are a limited number of time series available, which are particularly valuable due to the seasonal nature of these organisms. Patchiness of data combined with the patchiness of coccolithophore blooms is a challenge in fully assessing marine ecosystem functioning, and in providing global abundance estimations. Beside limitations of in situ measurements, size, POC and PIC measurements of paired hetero-and holococcolithophore species are sparse, in particularly for holococcolithophores.

390
Such measurements are needed for global organic carbon and carbonate production estimates, which are critical for biogeochemical estimates, including model studies. Models which could then be used to contextualize in situ observations in biogeochemical context, and which could test response to environmental pressures presented by anthropogenic CO 2 emissions. Modelling approaches could furthermore be used to investigate drivers of distribution trends difficult to acquire with in situ measurements such as the role of competition with other phytoplankton and the influence of top down control on distri-395 bution trends, both of which have shown to be important drivers of coccolithophore distribution in previous studies (Monteiro et al., 2016;Nissen et al., 2018).
A pertinent environmental driver not covered in our meta-analysis due to limited data, is the influence of carbonate chemistry within the haploid-diploid niche space. As the haploid and diploid phases of coccolithophores vary in their calcification status, they may thus show different responses to carbonate chemistry. A study by Triantaphyllou et al. (2018) for instance found that 400 holococcolithophores increased abundance in low pH waters. If this holds true on a global level, and holococcolithophores inhabit lower pH waters in terms of their niche space, this would have important implication in the context of ocean acidification. In particular because meta-analysis (Ridgwell et al., 2009;Krumhardt et al., 2017) and modelling (Ridgwell et al., 2007;Krumhardt et al., 2019) suggest a shift towards lower global calcification rates in response to ocean acidification and warming.
It should however be noted that the response of heterococcolithophores to ocean acidification is both strain and species depen-405 dent (Langer et al., 2006(Langer et al., , 2009Meyer and Riebesell, 2015), and global calcification rates might be more impacted by shifts in species compositions rather than individual response (Ridgwell et al., 2009). will furthermore be highly valuable for future modelling approaches. In this context a better understanding of the triggers of 410 phase transition would additionally be highly desirable, as the lack of haploid-diploid pairs of the same strain limits genomic approaches.

Conclusions
Our analysis shows that holococcolithophores constitute a large proportion of total coccolithophore abundance (≈18 % for paired species). Our study furthermore shows that hetero-and holococcolithophores have contrasting environmental preference, 415 and that therefore the haplo-diplontic life cycle expands the niche space coccolithophores can inhabit by ≈17 %. Although our findings are limited to holococcolith forming species, lab studies suggest similar patterns are likely to be observed for other coccolithophore species such as E. huxleyi, and raises the question how much the haploid phase of this species contributes to global coccolithophore abundance.
These results highlight the need to include haploid cells into coccolithophore studies, both in the context of environmental 420 studies, modelling approaches, and physiological studies. We limit our understanding of these organisms by just focusing on one life cycle phase, particularly in the context of coccolithophore response to climate change, as increased stratification in a warming climate may favour the haploid life cycle of coccolithophores.   34 https://doi.org/10.5194/bg-2020-194 Preprint. Discussion started: 24 June 2020 c Author(s) 2020. CC BY 4.0 License.