Antarctic slush-ice algal accumulation not quantified through conventional satellite imagery: Beware the ice of March

J. L. Lieser, M. A. J. Curran, A. R. Bowie, A. T. Davidson, S. J. Doust, A. D. Fraser, B. K. Galton-Fenzi, R. A. Massom, K. M. Meiners, J. Melbourne-Thomas, P. A. Reid, P. G. Strutton, T. R. Vance, M. Vancoppenolle, K. J. Westwood, and S. W. Wright Antarctic Climate & Ecosystems Cooperative Research Centre, University of Tasmania, Private Bag 80, 7001 Hobart, Australia Australian Antarctic Division, Channel Highway, 7050 Kingston, Tasmania, Australia Institute for Marine and Antarctic Studies, University of Tasmania, Private Bag 129, 7001 Hobart, Tasmania, Australia Institute of Low Temperature Science, Hokkaido University, Sapporo, Hokkaido, 060-0819, Japan Bureau of Meteorology (CAWCR), G.P.O. Box 727, 7001 Hobart, Tasmania, Australia Australian Research Council Centre of Excellence for Climate System Science, University of Tasmania, 7001 Hobart, Tasmania, Australia


Introduction
In Antarctica, phytoplankton blooms are typically observed during spring and summer in the vicinity of the sea ice edge as it recedes (Arrigo et al., 2008a), and can also occur within the sea ice itself (Massom et al., 2006).Broad-scale assessment of primary production is derived from satellite estimates of Chl a based on ocean-colour observations in the open ocean.However, ice-covered regions are masked in the processing (Massom et al., 2006;Bélanger et al., 2007).
In autumn, phytoplankton production declines in the Antarctic sea-ice zone due to limitation of growth by seasonal changes in light and temperature.Declining solar elevations reduce incoming photosynthetically active radiation (PAR); reduced stratification allows deeper mixing of phytoplankton to depths with little light; and declining temperatures allow sea ice to reform, further attenuating light entering the ocean (e.g.Mitchell and Holm-Hansen, 1991;Kang and Fryxell, 1993;Smith Jr. et al., 2000;Arrigo et al., 2008b;Arrigo and Lizotte, 2010).
Using cloud-free data from NASA's TERRA satellite MODIS (MODerate resolution Imaging Spectrometer) true-colour bands 1, 2 and 4 (645, 859 and 555 nm), we detected a large-scale region (∼ 30 000 km 2 ) of ice-associated phytoplankton surface accumulation that developed between 21 February and 19 March 2012 (in the early autumn) off Cape Darnley, East Antarctica, centred on 67.5 • S, 70 • E (Fig. 1a, a polynya region during winter).The phytoplankton accumulation was not detected in conventional maps of Chl a from the MODIS ocean-colour bands (443, 488 and 551 nm).This intense early autumn phytoplankton occurrence, which we suggest was caused by a unique combination of physical and biological processes, challenges the assumption that primary production is negligible in this region during early autumn (Smith Jr. et al., 2000).While we were able to perform some opportunistic sampling, this was insufficient to conclusively establish whether this was a conventional phytoplankton bloom.
However, such was the extraordinary concentration of phytoplankton within the seaice matrix, as observed from space and on the surface, that we use the term intra-ice "bloom".
Autumn blooms in consolidated sea ice have been observed around Antarctica previously (Fritsen et al., 1994), including in the interior parts of ice floes in the pack ice zone (Meiners et al., 2012).However, these blooms were relatively small in scale and of low intensity.Increased autumn phytoplankton biomass has been observed in association with coastal polynyas around East Antarctica (Arrigo et al., 2015;Arrigo and van Dijken, 2003) and has been inferred from ice core records (Curran et al., 2002) or hypothesised from models of ice-algal production (see for example Appendix A).Late summer blooms have also been reported from Antarctic waters (e.g.Comiso et al., 1990;Smith Jr. and Nelson, 1990), but only where the recent disappearance of sea ice has belatedly exposed coastal waters to high light levels.In this study, we provide first evidence that enhanced concentration of ice-associated phytoplankton occurs in autumn, and that this is not accounted for in current satellite-based Antarctic primary production estimates.

Chronology of 2012 ice-slush phytoplankton event
Daily MODIS true-colour images (pseudo animation in Appendix B) show that initiation of the observed phytoplankton were associated with break-up of landfast sea ice (fast ice) to the east of Cape Darnley between 10 February and 10 March 2012.The initial stages were detected close to this disintegrating fast ice.As the phytoplankton occurrence expanded in scale, it discoloured the sea ice (Fig. 2), which likely consisted of a mixture of pulverised existing ice and newly-formed ice (Massom et al., 2006).This ice was transported by the winds, first to the south (particularly in western Prydz Bay) and then to the north (relevant atmospheric model output can be found in Appendix C) with the change in wind direction occurring around 27 February.We also note evidence of wind-driven strips and patches of sea ice at the southern edge of the "bloom", visible as southeast-northwest aligned streaks west of Cape Darnley around 12 March.This process is presumably driven by katabatic winds.

Water sampling
On 5 March 2012, seven opportunistic seawater samples were taken in the vicinity of the "bloom" by the research and supply vessel Aurora Australis from the uncontaminated seawater line intake system at ∼ 4.8 m depth (see Fig. 1 for the ship track and sampling location).Photos (Fig. 2) suggest a very high biomass of ice-associated algae at the sea surface.Sampling sites were located both within and just outside the bloom region, as identified from MODIS true-colour imagery.Samples were preserved with 1 % vol : vol Lugol's iodine and stored in glass bottles in the dark at 4 • C. Protists were identified using an inverted microscope at 400 and 640 × magnification, and counted in 20 randomly chosen fields of view using a Whipple grid.Cell abundances were converted to biovolumes using the formulae of Hillebrand et al. (1999) and Cornet-Barthaux et al. (2007).Cell carbon values were then calculated from biovolumes using the conversions of Menden-Deuer and Lessard (2000) and Cornet-Barthaux et al. (2007).

Evidence of ice-associated phytoplankton occurrence in autumn for other years
The early autumn, ice-associated phytoplankton phenomenon documented here is not an isolated event in this region.Using MODIS true-colour data, we identified six other years since February 2000 when ice-associated early autumn blooms occurred in the same region : 2000, 2002, 2005, 2006, 2007, 2008 and 2012 (http://lance-modis.eosdis.nasa.gov/cgi-bin/imagery/realtime.cgi).These blooms are generally most pronounced around 1 March ± 5 days, suggesting a regular mechanism, which we explore below.

Biological characteristics of the seawater samples
Seawater samples taken on 5 March 2012 indicated differences in the abundance, composition and size of the phytoplankton at sites inside and outside the bloom (Table D1).Concentrations of Chl a inside the bloom (2.11 mg m −3 average, 3.3 mg m −3 maximum) were three to five times higher than those outside the bloom (0.66 mg m −3 ).
Total protist abundance reached 2.6 × 10 6 cells dm −3 in the midst of the bloom compared to 1.2 to 1.7 × 10 6 cells dm −3 outside.Ice-associated phytoplankton abundance at the surface could have been larger than at 4.8 m depth, but unfortunately could not be measured at the time.Overall, our results show that the phytoplankton biomass was higher in the bloom due to higher densities and the large cell size of the dominant taxa.The phytoplankton community was dominated by members of the diatom genus Fragilariopsis, which comprised 56 to 75 % of all autotrophs sampled.The absolute abundance of Fragilariopsis spp.differed little between sampling sites, but cells sampled within the bloom were larger (most 20 to 50 µm long, with some up to 80 µm long) compared to cells outside (commonly < 10 µm long).The haptophyte Phaeocystis antarctica was the second most abundant taxon, averaging 18 % of all autotrophs by number and reaching 3.1 × 10 5 cells dm −3 at both ice-covered and ice-free locations.Cell concentrations did not vary systematically with ice cover but their small size (∼ 2 to 5 µm in diameter) meant they contributed little to overall protistan biomass.Phytoplankton taxa that inhabit sea ice, namely Entomoneis sp. and Polarella glacialis, were also present at low abundances within the bloom but were absent outside (Table D1).

Underestimated importance of ice-associated phytoplankton in autumn
The ice-associated phytoplankton in the region off Cape Darnley was not detected in conventional satellite ocean-colour images.Figure 1a shows MODIS level 2 Chl a data in the Cape Darnley region, East Antarctica, acquired on 2 March 2012 10:30 UTC.The highest values are at the fringe of the area that was masked by sea ice and/or cloud cover, and scattered within that region.The MODIS true-colour composite has been overlaid in regions where sea ice and/or cloud cover prohibited Chl a retrievals, and the discoloured sea ice and ice slush is obvious in Fig. 1a.A ship-borne image (taken 5 March 2012) of the sea ice/slush matrix is given in Fig. 2. In this region, the combination of different products from the same MODIS instrument (Fig. 1a) demonstrates that satellite-derived estimates of Chl a are lower than actual values.Also included in Fig. 1a is the ship's track while sampling the bloom three days later.This track shows Chl a calculated from underway fluorometry samples with the highest values of up to 3 mg m −3 inside the bloom area, matching the highest values of the remote sensing Chl a estimates along the edges of the detection.Figure 1b  agreement between remote sensing data and samples in an along-track view where the satellite data are not masked by sea ice and/or clouds, and the high values of Chl a of up to 6 mg m −3 inside the masked area of the bloom patch.Because they have not been detected in the past, large-scale autumn blooms within the sea ice zone have not been factored into calculations of regional-scale primary production for the Southern Ocean (e.g.Arrigo et al., 2008b).These results are evidence that ice-associated phytoplankton production is underestimated in other seasons too (e.g.Massom et al., 2006;Massom and Stammerjohn, 2010).

Proximal drivers of bloom formation in autumn
We hypothesise that the combination of five environmental variables create the conditions that lead to this early autumn bloom formation; sea ice, winds, light, nutrient availability and lack of grazing pressure (see conceptual representation shown in Fig. 3a).In early March, the interaction of these variables allows for an environmental window in which these ice-associated blooms can form (Fig. 3b), as follows.

Sea ice conditions
Based on the MODIS time series and information in Fig. 2, this early-autumn "bloom" appears to be associated with the decay of fast ice and new sea-ice formation.Seaice growth in the Cape Darnley region typically begins in late February to early March (Massom et al. (2013) and Fig. 3b), driven by the air-sea temperature gradient and winds.Frazil ice formation can act to concentrate phytoplankton in the newly-forming sea ice matrix, with frazil crystals "scavenging" algal cells from the water column as they rise (Garrison et al., 1983).One-dimensional biophysical model simulations (detailed in Appendix A) suggest that frazil scavenging (Garrison et al., 1989) is not essential to Chl a accumulation in the ice, but would accelerate it.MODIS ocean-colour imagery indicates that there was chlorophyl in the same region prior to initiation of the ice-associated bloom (see Fig. E1), possibly as a result of decay of fast ice.Fast ice decay occurs through break-up and pulverisation by wind and wave activity.That decay releases both concentrated algal cells and nutrients to the surrounding ice slush (Massom et al., 2006).We suggest that concentrated phytoplankton cells, within this interstitial sea ice matrix between brash ice and floes, comprise the discolouration of sea ice observed in MODIS true-colour imagery (Fig. 1a) and from the ship (Fig. 2).

Wind conditions
Strong winds in the Cape Darnley region in March 2012 are indicated in the highresolution polar realisation ACCESS-P model output (Appendix C).Such conditions (Fig. 3b) favour both the break-up and pulverisation of fast ice, and the formation of new and frazil ice.Wind action may also play a key role delivering nutrients by aeolian transport (Winton et al., 2014) from the continent, as well as increasing the spatial extent by dispersion of the bloom throughout its lifetime.

Light
At the high latitude at which the bloom was observed in March, the availability of incoming PAR is sufficient to initiate bloom conditions if phytoplankton are confined to a shallow ice-slush matrix on the surface.In this way, they are exposed to higher light levels than is the case in ice-free areas where deeper mixing can result in light limitation, or within consolidated sea ice with a snow cover where snow significantly attenuates PAR (Massom et al., 2006).At such a low light angle, sea ice also acts to trap and scatter the light which the phytoplankton can use (Buckley and Trodahl, 1987).

Nutrient availability
Primary production in the Southern Ocean is strongly limited by the availability of nutrients (Blain et al., 2007).Following spring blooms, the exhaustion of nutrients, particularly iron, together with increased grazing pressure and decreased stratification leads to a decline in production in the open ocean (de Baar et al., 1995).In the region of the Cape Darnley bloom, the presence of large diatoms is consistent with iron enrichment (Coale et al., 2004).Sea ice and fast ice has been shown to contain high concentrations of iron relative to adjacent areas of open water (van der Merwe et al., 2011), so its subsequent pulverisation by waves likely released iron in the bloom region, to initiate or maintain algal growth.
Other potential sources of iron to support this early autumn bloom include: iceberg melt, wind-driven snow deposition, perhaps containing iron-rich dust, from the continent and wind-driven upwelling (see Fig. F1).Of these, iceberg sources of nutrients are difficult to quantify for the Cape Darnley bloom, but fast ice break-up was observed and upwelling can be inferred from wind strength and direction.Fast ice break-up in western Prydz Bay was observed in MODIS true-colour imagery ten days before the detection of the ice-associated bloom (Appendix B).

Grazing
Within the bloom region, the habitat provided by sea ice provides a protective matrix against zooplankton grazing (Krembs et al., 2000).Microbial grazers, such as heterotrophic flagellates, dinoflagellates and ciliates, were more abundant in the bloom, but their grazing was insufficient to limit phytoplankton growth.In the absence of samples to indicate metazoan grazer densities, it is not possible to quantify total grazing pressure for the ice-associated bloom.

Climate indices
The discovery of this bloom and the satellite-based evidence of other such blooms in different years and regions, and implications for primary production estimates, raises the question of inter-annual variability.Our initial investigation suggests that the positive phase of El Niño-Southern Oscillation (ENSO) and the Southern Annular Mode (SAM) may play a role (see discussion in Appendix G).Understanding these drivers of inter-annual variability, and their association with blooms of this type, is important for estimating the scale of this additional source of primary production.

Conclusions
This intra-ice phytoplankton "bloom" appears to be the consequence of a combination of sea ice conditions resulting from breakup of fast ice and initial growth stages of new sea ice.We suggest that the combination of sufficient light, nutrients released from disintegrating ice, subsequent wind redistribution and a reduction in grazing pressure allows for this phytoplankton bloom.Blooms of this type are a potentially important source of primary production in the Southern Ocean, which is not currently quantified with ocean-colour remote sensing, and therefore have implications for Southern Ocean food webs and biogeochemical cycling (Murphy et al., 2012).

Appendix A: One-dimensional bio-physical model results
The model couples a simple N-P module with the halo-thermodynamic model (Vancoppenolle et al., 2010).This model captures ice thermodynamics, halo-dynamics, a radiation scheme, and a simple nutrient-phytoplankton ice algae model, characterised by one species of ice algae (diatoms) and two limiting nutrients (nitrates and silicates).Provided that iron is plentiful, the algal biomass observed in MODIS true-colour imagery and in-situ (from the research vessel Aurora Australis) originates either from (i) in-ice growth of algae, or (ii) scavenging of algal cells from the water column by rising frazil ice crystals.These alternative hypotheses were tested using simple calculations and model simulations.

A1.1 In-ice growth of ice algae
We first compute the amount of Chl a that is achievable from the in-situ growth of ice algae by consumption of the the locally available nutrients.Assuming that 30 % of the oceanic concentration of silicates of C Si i = 40 mmol m −3 is stored in the sea ice during formation (the rest being rejected by brine gravity drainage) and fully exhausted by ice algae, then we get that chlorophyll a concentration in sea ice is: where r Chl a C = 0.15 mg mmol −1 is a standard chlorophyll : carbon ratio in sea ice diatoms and r C Si = 9.09 is the C : Si molar ratio (Sarthou et al., 2005).

A1.2 Scavenging of frazil crystals
The Cape Darnley post-polynya is highly productive in late summer (see Arrigo and van Dijken, 2003), with surface Chl a concentrations reaching up to 5 mg Chl a m −3 .Assuming that a fraction α = 10 % of seawater Chl a with a concentration C

A2 Model simulations
The model is designed to represent un-deformed level (columnar) ice, we apply it here to the sea ice mixture that characterised the Cape Darnley early autumn bloom.This is based on the assumption that nutrient pathways in level ice (e.g.brine channels) that are captured in the one-dimensional model, behave similarly to the frazil mixture found between pancakes (although this mixture likely occupies more volume than brine channels, because of the constant swell forcing that tends to prevent full ice consolidation).
As Chl a is a particulate tracer, we assume 100 % of incorporation (C ).The model therefore arguably provides a rough, minimal, representation of the Cape Darnley bloom conditions.

Forcing and initial conditions
The sea ice model is forced at boundaries with the following numbers: sea surface salinity = 34 g kg −1 ; nitrate concentration in seawater = 30 mmol m −3 (typical value from that region, see Sarmiento and Gruber, 2006); (from the formula of Kovacs (1996) giving bulk ice salin-ity for young ice as a function of its thickness); (no biological activity in the ocean).In experiment 2, we tested the role of the maximum specific growth rate µ, which is rather uncertain.It would be bigger if the diatoms are adapted to high-light levels (ocean diatoms) and lower if they are adapted to low-light levels (ice diatoms).In ex-periment 3, the seawater that sea ice is forming from is typical of bloom conditions (≈ 1 mg Chl a m −3 ), but there is no harvesting of organic matter by frazil crystals.In experiment 4, the seawater is typical of bloom conditions, and we assume that initial concentration in forming ice is enhanced by frazil scavenging (initial concentration of Chl a in sea ice ≈ 10 mg Chl a m −3 ).

A3 Results
The model grows 50 cm of ice thermodynamically within the 30 day simulation period (Fig. A1).The cold air temperatures imply low temperatures and high brine salinities in the upper ice, which is improper to ice algae growth, as can be seen from the limitation factors in the lower panels of the Fig. A1 (0 = improper to ice algal growth, 1 = ice algal growth is possible).The four sensitivity experiments give quite different blooms, particularly in terms of Chl a concentration and timing of maximum Chl a values (Fig. A2).Integrating Chl a vertically (Fig. A3a) and dividing by ice thickness to get mean concentrations in sea ice (Fig. A3b) indicates the following differences between experiments: -Experiment 1 gives low concentrations that are probably unrealistic; -Experiment 4 (frazil scavenging experiment) gives the fastest ice algal growth rate and the most abundant ice algae; -Experiments 2 and 3 are in between these extremes.Experiment 2 indicates that light-acclimated diatoms with higher µ might outcompete ice algae in the bloom.
The concentrations in sea ice are high and the timing of maximum concentrations (around day-of-year 60) is consistent with observations.Experiment 3 indicates that even without frazil scavenging, accumulation of organic matter from the upper ocean creates a sufficient basis for ice algal growth.

A4 Conclusion
Model simulations suggest that frazil scavenging would lead to rapid and intense ice algal development (Experiment 4).Sensitivity experiments also suggest that substantial activity is possible in the ice (i) with no frazil scavenging but accumulation of surface material in forming ice (Experiment 3); (ii) with no activity in the water, but with lightacclimated diatoms outcompeting ice diatoms in the slush (Experiment 2).On the basis of these results, it is not possible to distinguish between alternative hypotheses for the origin of algal cells in the ice-associated bloom.climate variability that influence the Southern Ocean, specifically the El Niño-Southern Oscillation (ENSO) and the Southern Annular Mode (SAM).The Southern Oscillation Index (SOI) is an index describing the atmospheric oscillation between El Niño (warm) and La Niña (cool) temperature states in the tropical eastern Pacific Ocean.Vance et al. (2013) showed that there is a circumpolar ENSO signal in the westerly winds around Antarctica.This strengthening was further enhanced by interaction of ENSO with the SAM.By austral summer in El Niño (La Niña) years, tropospheric anoma-lies develop into a circumpolar ridge-like (trough-like) anomaly across Antarctica that weakens (strengthens) circumpolar westerlies (Fig. G1) and interacts with the SAM (L' Heureux and Thompson, 2006).

Discussion
In the Cape Darnley region, these ice-associated early autumn blooms occurred in years when La Niña and positive SAM conditions generated high winds (Fig. G1), which is one of the important drivers identified for this bloom (see Fig. 3b of the original paper).The short satellite dataset (2000)(2001)(2002)(2003)(2004)(2005)(2006)(2007)(2008)(2009)(2010)(2011)(2012) limits our ability to confirm this ENSO-SAM-bloom relationship in a statistically significant way, but it occurs more often than by chance (Fogt et al., 2011) and is likely linked to greenhouse gas forcing (IPCC, 2007).Therefore, these previously unquantified blooms may increase in frequency, further contributing to primary production in the Southern Ocean.

A circumpolar pressure signal of ENSO
The interaction between ENSO and SAM during DJF is primarily the result of coinci-dent La Niña-SAM positive, El Niño-SAM negative or ENSO-weak SAM events occur-ring more often than by chance (Fogt et al., 2011).Other iterations (i.e.La Niña-SAM negative) resulted in anomalous transient eddies in the mid-latitudes that opposed rather than reinforced the SAM-ENSO interaction.During La Niña-SAM positive events (Fig. G1, upper panel), an anomalous circumpolar trough over the Antarctic continent and coastal regions is paired with a decidedly un-SAM-like zonal wavenumber 3 (ZW3) pattern at mid-latitudes.ZW3 is associated with transitions from strongly meridional to strongly zonal flow (Raphael, 2007).ZW3 is thought to be associated with ENSO, and an Australian sector wave 3 index (AZ3 -van Ommen and Morgan, 2010) is strongly correlated with SOI during summer (Table G1).
The NCEP/NCAR reanalysis (Kistler et al., 2001) was used to investigate GPH anomalies.All reanalysis datasets have high latitude data problems related to data scarcity and trend effects in the Southern Hemisphere, particularly prior to 1979 (Hines et al., 2000;Kistler et al., 2001).Data coverage improves greatly after 1979, thus we used composites from 1979-2009.
The Supplement related to this article is available online at doi:10.5194/tcd-9-6187-2015-supplement.

6189
Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper |

6191
Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper |

Discussion
Paper | Discussion Paper | Discussion Paper | Discussion Paper | demonstrates the close 6193 Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper |

6195
Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper |

6197
Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | frazil crystals over an oceanic depth of h o = 50 m, the resulting average Chl a concentration in 30 cm thick sea ice should be: Chl a m −3 .(A2) These two scenarios give equivalent (high) Chl a concentrations and do not distinguish between our two hypotheses for the bloom source.Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | silicate concentration in seawater = 40 mmol m −3 (typical value from that region, see Sarmiento and Gruber, 2006); atmospheric heat fluxes from empirical formula for sensible, latent, shortwave and long wave fluxes, fed by daily NCEP temperatures, winds and specific humidity (multiplied by 0.83 to correct a known systematic bias), together with a climatology of cloud fraction; a prescribed very small snowfall rate.Model runs start at day-of-year 50 (19 February) when the surface energy budget becomes negatives, which (approximately) corresponds with the onset of ice formation.6199 Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Runs last 30 days, until day-of-year 80 (20 March), and are initialised with standard thin ice characteristics: ice thickness = 5 cm; snow depth = 2 cm; ice salinity = 22.93 g kg −1

-
nitrate concentration = 20.23 mmol N m −3 (same ice-ocean ratio as for salt); silicate concentration = 26.97mmol Si m −3 (same ice-ocean ratio as for salt).The concentration of carbon stored in ice algae C DA i is also prescribed at the oceanic boundary (C DA o ).As this concentration in the ocean is used as an initial concentration in forming new ice, accreted at the ice base, we made several sensitivity experiments.Exp.3: C DA o = 5 mmol C m −3 ; µ = 1.3 × 10 −5 s −1 ; -Exp.4: C DA o = 50 mmol C m −3 ; µ = 1.3 × 10 −5 s −1 , where µ is the maximum specific photosynthesis rate.Experiment 1 corresponds to < 0.1 mg Chl a m −3

6201
Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Figure F1.MODIS true-colour image, illustrating potential ice seeding mechanisms, including (1) wind-driven formation of strips and patches, (2) disintegrating fast ice, and (3) wind-blown snow off the face of the Amery Ice Shelf.

Figure G1 .
Figure G1.Composite images of 500 hPa geopotential height (GPH500) anomalies from December to March.Upper panel: composite of years that are coincident in both the upper quartile of Southern Oscillation Index (November to February) and SAM (December to March) averaged summer values (1979-2012) are composited.Lower panel: composite of bloom years identified by satellite images from 2000-2012 (December to March).Cape Darnley region is identified with a red dot.Contours are 10 m.

Table D1 .
Biological characteristics of opportunistic samples.