The absolute abundances (cells per ml) of 22 hard-shelled phytoplankon taxa (comprised of species, genera or higher taxonomic groups) estimated from Scanning Electron Microscope survey of 52 samples collected through 11 austral spring-summers (2002/3 to 2012/13) (part of the L' Astrolabe collection) from the seasonal ice zone of the Southern Ocean (between latitude 62 and 64.4 degrees south, and longitude 135.8 and 150 degrees east)

also included are environmental covariables for each sample: three constructed SAM indices, SST, Salinity, NOx, PO4, SiO4, and the sampling date, time, and location.

Fifty-two surface-water samples were collected from the seasonal ice zone (SIZ) of the Southern Ocean (SO) across 11 consecutive austral spring-summers from 2002/03 to 2012/13. The samples were collected aboard the French re-supply vessel MV L’Astrolabe during resupply voyages between Hobart, Tasmania, and Dumont d’Urville, Antarctica between the 20th October and the 1st March. Most samples were collected from ice-free water, although some were collected south of the receding ice-edge.

The sampled area was in the high latitude SO (Figure 1b) in the south-east corner of the Australian Antarctic Basin, spanning 270 km of latitude between 62°S and 64.5°S, and 625km of longitude between 136°E and 148°E. The area lies greater than 100 km north of the Antarctic continental shelf, in waters greater than 3,000 m depth.

Samples were obtained from the clean seawater line of the re-supply ship from around 3 m depth. Each sample represented 250 ml of seawater filtered through a 25 mm diameter polycarbonate-membrane filter with 0.8 µm pores (Poretics). The filter was then rinsed with two additions of approximately 2 ml of MilliQ water to remove salt, then air dried and stored in a sealed container containing silica gel desiccant. Samples were prepared for scanning electron microscope (SEM) survey by mounting each filter onto metal stubs and sputter coating with 15 nm gold or platinum. Only organisms possessing hard siliceous or calcareous shells were sufficiently well preserved through the sample preparation technique that they could be identified by SEM, and included diatoms, coccolithophores, silicoflagellates, Pterosperma, parmales, radiolarians, and armoured dinoflagellates.

The composition and abundance the phytoplankton community of each sample was determined using a JEOL JSM 840 Field Emission SEM. Cell numbers for each phytoplankton taxon were counted in randomly selected digital images of SEM fields taken at x400 magnification (Figure 2). Each image represented an area of 301 x 227 µm (0.068 mm2) of each sample filter, which was captured at a resolution 8.5 pixels/µm. A minimum of three SEM fields were assessed for each sample, with more fields assessed when cell densities were lower. On average, 387 cells were counted for each sample. Taxa were classified with the aid of Scott and Marchant (2005), Tomas (1997), and expert opinion. Cell counts per image were converted to volume-specific abundances (cells/ml) by dividing by 0.0348 ml of sea-water represented by each image.

A total of 19,943 phytoplankton organisms were identified and counted: 18,872 diatoms, 322 Parmales, 173 coccolithophores, 81 silicoflagellates, and 45 Petasaria. A total of 48 phytoplankton taxa were identified, many to species level. Because the diatoms Fragilariopsis curta (Van Heurck) Hustedt and F. cylindrus (Grunow ex Cleve) Helmcke and Krieger could not be reliably discriminated at the microscope resolution employed, they were pooled into a single taxa-group. Other taxa were also grouped, namely Nitzschia acicularis (Kützing) W.Smith with N. decipiens Hustedt to a single group, and discoid centric diatoms of the genera Thalassiosira, Actinocyclus and Porosira to another. Rare species, with maximum relative abundance less than 2%, were removed from the data prior to analysis as they were not considered to be sufficiently abundant to warrant further analysis (Webb and Bryson 1972, Taylor and Sjunneskog 2002, Swilo et al. 2016). After pooling taxa and deleting rare taxa, twenty-two taxa and taxonomic-groups (species, groups of species and families) remained to describe the composition of the phytoplankton community.

Phytoplankton abundances were related to a range of environmental covariates available at the time of sampling. These included the SAM, sea surface temperature (SST), salinity, time since sea ice cover (DaysSinceSeaIce, defined below), minimum latitude of sea ice in the preceding winter, latitude and longitude of sample collection, the days since 1st October that a sample was collected (DaysAfter1Oct), the year of sampling (year, being the year that each spring-summer sampling season began), the time of day that a sample was collected, and macro-nutrient concentrations: phosphate (PO4), silicate (SiO4) and nitrate + nitrite (hereafter nitrate, NOx).

Water samples for dissolved macro-nutrients were collected, frozen on ship, and later analysed at CSIRO in Hobart using standard spectrophotometric methods (Hydes et al. 2010). Daily estimates of SAM were obtained from the US NWS Climate Prediction Center's website and are the NOAA Antarctic Oscillation Index values based on 700-hPa geopotential height anomalies (NOAA 2017). The variable DaysSinceSeaIce was defined as the time since sea ice had melted to 20% cover (after Wright et al. 2010) as determined from daily Special Sensor Microwave/Imager (SSM/I) sea ice concentration data distributed by the University of Hamburg (Spreen et al. 2008).

To examine the lag in the expression of the SAM on phytoplankton community composition, two response surfaces were constructed relating the variance in phytoplankton community composition explained by the SAM to the temporal positioning of the period over which daily SAM was averaged. These were derived by evaluating separate CAP analyses (described below) based on daily SAM averaged across a range of days {1, 3, 5, … 365} centred on (i) each calendar day individually (1 Jan – 31 Dec) through the year associated with each sample; and (ii) lagged from 1 to 365 days prior to each sample collection date.

Empirical identification of the time between variation in the SAM and the manifestation of this variation in the phytoplankton community structure revealed three modes (maxima) in phytoplankton community composition explained by the SAM. The first was an autumn seasonal SAM mode, which was determined to be the average of 57 daily SAM estimates centred on the preceding 11th March (11th Feb – 8th Apr). This mode explained up to 13.3% of the variance in taxonomic composition (SAM autumn). The second was a spring seasonal mode, which was determined to be the average of 75 daily SAM estimates centred on 25th October (20th Sep – 3rd Dec). This mode explained up to 10.3% of variance in taxonomic composition (SAM spring). Unlike the other modes that were related to the time of year, the third mode was timed relative to the date of sample collection for each sample and comprised the average of the 97 daily SAM estimates centred 102 days prior to each sample collection date. It explained 9.9% of the variance in phytoplankton composition (SAM prior). The mean standard error on estimates of SAM were 0.14 SAM index units for SAM autumn and SAM spring, and 0.13 for SAM prior. Note that SAM prior and SAM spring temporally overlapped to varying extents across the 52 samples and so were not entirely independent covariates: for example, a sample collected in the summer had previous days contributing to both SAM prior and SAM spring.