Aerial extent, composition, bio-optics and biogeochemistry of a massive under-ice algal bloom in the Arctic

Abstract It has been long thought that coccolithophores are a minor component of the phytoplankton assemblage in Arctic waters, with diatoms typically being more dominant. Little is known about how the phytoplankton communities will change, however, as the Arctic warms. We participated in the 2011 Impacts of Climate on EcoSystems and Chemistry of the Arctic Pacific Environment (ICESCAPE) cruise to the western Arctic, performing a combination of discrete measurements (microscopy, calcification, particulate inorganic carbon (PIC), particulate organic carbon (POC), biogenic silica (BSi)) plus continuous surface bio-optical measurements (absorption, scattering, backscattering and acid-labile backscattering; the latter specific for coccolithophores). Here, we report bio-optical and coccolithophore observations from the massive under-ice algal bloom originally described in Arrigo et al. (2012) . The most intense portions of the bloom were centered in cold Winter Water and there was evidence for nitrate drawdown in the top 10–20 m with strong penetration of silicate rich water into the surface waters. Surface chlorophyll a and particulate absorption at 440 nm approached 30 μg L −1 and 1.0 m −1 , respectively. Particulate absorption of detritus ( a p at 412 nm) was highly correlated to a p at 440 nm associated with chlorophyll a and slopes of the absorption spectrum showed that both dissolved and particulate absorption at 412 nm exceeded that at 440 nm, with slopes, S g , of 0.01 nm –1 . Colored dissolved organic matter fluorescence (FDOM) was high in the bloom but the relative fluorescence yields were low, characteristic of phytoplankton-produced FDOM (as opposed to terrestrially-produced FDOM). Coccolithophore backscattering was elevated in the under-ice bloom, but it only accounted for 10% of the total particle backscattering, relatively low compared to typical subpolar waters further to the south. Total particle scattering was significantly elevated in the under-ice bloom (values of almost 2 m −1 ), likely due to the high abundance of large diatoms. Backscattering probabilities in the bloom were ~1%, again characteristic of diatom-dominated populations with few calcifiers. PIC standing stock in the under-ice bloom was low but measurable while biogenic silica molar concentrations were 150 times greater. POC:PON molar ratios were 6–10, characteristic of healthy, rapidly growing phytoplankton, observations further buttressed by carbon:chlorophyll mass ratios of 50–100. Coccolithophore calcification was low but measurable, reaching 1.75 mg C m −3  d −1 in the under-ice bloom, only 0.4% of the photosynthesis. However, the intrinsic carbon-specific growth rate was 0.4 per day for bulk POC and ~1 per day for bulk PIC, close to maximal growth rates expected at these temperatures. SEM and light microscopy results showed mostly diatoms in the bloom. The coccolithophore, Emiliania huxleyi , was observed, providing unequivocal evidence of the presence of coccolithophores in the under-ice algal bloom.


Polar phytoplankton and coccolithophores
Arctic waters have long been characterized by strong diatom dominance, as evidenced in the first description of diatoms in Arctic sea ice (Ehrenberg, 1841) as well as more recent accounts (Bursa, 1961;Poulin et al., 2011;Saito and Taniguchi, 1978;von Quillfeldt, 2000) that show diatoms to be the significant drivers of Arctic primary production in the upper water column (and under ice). Dinoflagellates are also regularly seen in Arctic waters but at lower biomass than the diatoms (Braarud, 1935;Horner, 1984;Poulin et al., 2011).
Phaeocystis is another common phytoplankter in Arctic waters (Poulin et al., 2011;Sherr et al., 2003) as are nanoflagellates, which can contribute the majority of carbon biomass at specific times (Sherr et al., 2003).
Relative to the other phytoplankton groups, coccolithophores have traditionally been thought to be rare (or absent) in Arctic waters (Poulin et al., 2011) and more abundant in the sub-polar, temperate, sub-tropical and tropical biogeographic zones of the world ocean (McIntyre and Be, 1967;Okada and Honjo, 1973;Winter et al., 1994;Ziveri et al., 2004).
One hypothesized reason for the low abundance of coccolithophores in polar waters has been that they typically show lower growth at temperatures <8°C and in reduced solar radiation (Raitsos et al., 2006), such as in polar waters.
Despite their typically low abundance, coccolithophore blooms have been observed in ice-free polar waters using space-based remote sensing. Evidence from the AVHRR (Advanced Very High Resolution Radiometer) satellite, suggests that the frequency of coccolithophore blooms in sub-polar and non-ice-covered polar Arctic waters has been increasing over twenty years (Smyth et al., 2004). These blooms are probably Emiliania huxleyi but it has been impossible to confirm this due to lack sea-truth data. Polar coccolithophore species besides E. huxleyi were previously described in early taxonomic studies from Resolute Bay (Northwest Passage), West Greenland and South Alaska (genera Pappomonas, Wigmamma, Turrisphaera and Papposphaera) where the water temperature was below 0°C (Manton et al., 1976a;Manton et al., 1976b;Manton et al., 1977). Recent work in the Atlantic Arctic (partially ice-covered/ice edge region north of Svalbard) demonstrated low abundance of coccolithophores (2.5 cells mL -1 ) with species mostly from the family Papposphaeraceae, found in waters <0°C with sub-micromolar nitrate and phosphate (Charalampopoulou et al., 2011). Coccolithophore species observed in this same study included E. huxleyi, Coccolithus pelagicus, Pappomonas sp., Papposphaera arctica and Wigmamma sp.

Arctic primary production and calcification
There is relatively little information on blooms of algae under Arctic ice, primarily due to the high reflectance of sea-ice, and the inability to see such blooms using satellite remote sensing. Observations from ships have provided some evidence that blooms can occur, however. For example, at ice station SHEBA in the western Arctic, chlorophyll concentrations reached as high as 4.3 mg m -3 under the ice during the summer melting of snow overlying the ice (Sherr et al., 2003). Typically, under-ice primary productivity has been assumed to be low due to the strong attenuation of light by ice and snow. Hill et al. (2013) and Matrai et al. (2013), examined historical 14 C primary production and chlorophyll data. Surface productivity rates from regions like the northern Chukchi Sea were typically <10 mg C m -3 d -1 Matrai et al., 2013). Nitrate also is seasonally drawn-down under the ice, the extent of which can be used to estimate annual primary production (assuming a Redfield ratio of C:N in particulate matter and an f ratio of nitrate utilization) (Codispoti et al., 2012;Eppley and Peterson, 1979). Such estimates are within a factor of two of 14 C measurements of net primary production (Codispoti et al., 2012;Hill et al., 2013).
Elevated integrated primary productivity has been documented in waters with >90% ice, with rates as high as 60 mg C m -2 d -1 (but after snow is removed from the ice) (Gosselin et al., 1997).
There is only one previous study of coccolithophore calcification in a partially icecovered region north of Svalbard (Charalampopoulou et al., 2011). In an ice-free fjord and the marginal ice zone, calcification was low, with a subsurface peak of 0.02-0.07 mg PIC m -3 d -1 . In a partially ice-covered region, calcification showed a subsurface peak of 0.6 mg PIC m -3 d -1 (Charalampopoulou et al., 2011). Such rates are extremely low compared to rates measured in more coccolithophore-rich, lower latitude waters (Balch et al., 2007).
The goal of this study was to use a combination of continuous underway and discrete seawater measurements to document under-ice algal features during the ICESCAPE (Impacts of Climate on EcoSystems and Chemistry of the Arctic Pacific Environment) cruise to the western Arctic Ocean, July-August 2011. Moreover, we documented the hydrographic, biological and optical properties of these features and used the data to better understand: bloom magnitude, bloom size, dominant species, pigment-specific absorption, particle scattering and distribution of colored dissolved organic matter (CDOM). Discrete samples provided estimates of the standing stocks of particulate organic carbon (POC), particulate inorganic carbon (PIC), biogenic silica (BSi) plus photosynthesis and calcification rates.
These observations provide a baseline for interpreting future changes in the phytoplankton standing stocks, rates and bio-optical properties of the Western Arctic as the region undergoes climate change (Arrigo et al., 2008).

Cruise details
The ICESCAPE 2011 expedition took place in the western Arctic (Fig. 1A) (Arrigo et al., 2012). Station numbers and their location are shown in Fig. 1B.
Measurements included running a continuous underway system (focused on hydrographic and bio-optical properties) and measuring discrete water samples for a variety of biological and biogeochemical variables. CTD stations typically involved sampling water from eight depths. Of those eight bottles, seven were usually from the euphotic zone and one from deeper in the water column. We also sampled the top Niskin bottle of numerous CTD casts for calibration samples for the continuous underway system.

Underway bio-optical system
The Balch lab bio-optical underway system was run continuously over the course of the trip. It was started on 26 June, 2011 and shut down on 27 July, 2011 with shutdowns for weekly cleaning and calibration (see Section 2.3). This system has been described elsewhere (Balch et al., 2008). Briefly, the seawater source was located at 5m depth on CSCGC Healy.
Water flowed through an ice separator then through insulated stainless steel pipes to the shipboard laboratory. Our flow-through system measured temperature, salinity, chlorophyll a fluorescence, CDOM fluorescence (FDOM) and particle backscattering. Temperature and salinity were first measured with a SeaBird flow-through temperature and conductivity sensor. A WETLabs WETStar CDOM fluorometer was used to measure the fluorescence of colored dissolved organic matter (excitation = 370nm; emission = 460nm). This was plumbed into the flow path just after the temperature/salinity sensors. Next, chlorophyll fluorescence was measured with a WETLabs WETStar chlorophyll fluorometer (excitation = 460nm; emission = 695nm). Particle backscattering at 531 nm (using a WETLabs ECOVSF sensor aimed into a specially-designed container which minimized wall reflectance, hence maximizing the light scattering signal associated with marine particulate matter). First, the system measured particle backscattering of 531 nm light with raw seawater (pH=~8.1) running through the system for one minute. After 60 seconds of data collection (or whatever time period was set in order to have sufficient sample size to achieve standard errors of 0.5x10 -5 m -1 ), the acid controller injected 0.2 µm-filtered, 10% glacial acetic acid into the seawater stream, passing through a mixing coil to thoroughly mix it with the seawater, upstream of the ECOVSF. This reduced the pH to 5.5, below the dissociation point for various mineral forms of calcium carbonate. A pH sensor downstream of the sample chamber measured the pH constantly. Once the pH dropped to 5.5, backscattering was remeasured for an equivalent period of time after which the acid additions stopped and the pH re-equilibrated to raw seawater values and the entire cycle repeated. The difference in backscattering between raw seawater and acidified seawater represented "acid-labile backscattering" (b b '), which can be directly related to the concentration of suspended calcium carbonate (Balch et al., 1996).
The underway bio-optical system had a separate flow loop that passed through a WETLabs ac-9, to measure spectral absorption and attenuation at nine wavelengths: 412, 440, 488, 510, 555, 630, 650, 676 and 715nm. In the flow path to the ac-9 was a solenoid 8 that diverted the seawater stream through a 1µm filter, then a 0.2 µm filter prior to running the water through the ac-9. Every two minutes, the solenoid would alternate between filtered and unfiltered seawater, thus providing absorption and attenuation (at 9 spectral wavelengths across the visible spectrum) for raw and filtered seawater. In turn, this allowed calculation of the absorption and attenuation of total suspended particles and dissolved organic matter. The difference between raw and dissolved ac-9 measurements represented particulate absorption and beam attenuation. Total scattering was calculated as attenuation minus absorption.

Underway system calibration
Calibrations of the complete underway system were performed just prior to departure, approximately weekly during the cruise as well as a final calibration after final shut down.
These calibrations were used to estimate biofouling corrections during each operation period.
The protocol was to run 0.2um filtered RO water from the ship's Milli-Q system, under pressure, through the entire flow path prior to cleaning ("a dirty calibration" which provided the endpoint for estimating the optical contribution of biofouling). Then, the system was carefully disassembled and cleaned, reassembled and a "clean calibration" performed (which represented the beginning calibration for the next operational segment, with no biofouling. Post cruise, the biofouling corrections were interpolated between the initial clean calibration and the following "dirty calibration". The backscattering signal associated with the wall of the flow-through container was also estimated by running 0.2um-filtered RO water following cleaning as well as 0.2um-filtered seawater. Daily, biofouling of the wall was estimated by first shunting the inflowing water through a separate 0.2um filter prior to passage through the system and comparing this b bp value to that of pure seawater (Mobley, 1994).

Discrete samples
For the full CTD cast (the "productivity cast"), particulate inorganic carbon (PIC) was measured on 0.2L seawater samples filtered onto 0.4μm pore-size polycarbonate filters, rinsed with potassium tetraborate buffer (Poulton et al., 2006) and biogenic silica (BSi) was measured by filtering 0.2 L seawater onto 45mm 0.4μm polycarbonate filters, stored and measured according to Brzezinski et al. (1989). Particulate organic carbon (POC)/particulate organic nitrogen (PON) was measured using JGOFS protocols (JGOFS, 1996) while coccolithophore counts were processed ashore using polarized light microscopy (Haidar and Thierstein, 2001) (but substituting Norland #74 brand optical adhesive instead of Canada Balsam). Surface and chlorophyll maximum depths were sampled for scanning electron microscope and prepared for analysis ashore according to Goldstein et al. (2003). These same depths were sampled for "live" microscopy using the Filter Freeze Transfer technique (Hewes and Holm-Hansen, 1983), with samples filtered on 0.4um polycarbonate filters prior to transfer and then samples examined using an AO-Spencer Model 10 microscope equipped with epifluorescence and polarization optics. Nutrient samples were run on an AA3 autoanalyzer for nitrate, nitrite, ammonium, phosphate and silicate (but only nitrate and silicate results will be discussed here).
At the daily productivity cast, samples were taken for measuring primary production and calcification from the 30L Niskin samples (with Silicone O-rings). Water was sampled from 6 light depths: 38.6%, 21.1%, 11.7%, 3.5%, 1.9% and 0.3%. Estimation of those light depths was performed based on the percent light as measured by the scalar PAR sensor aboard the CTD, scaled to the above-water downwelling PAR irradiance measured from the superstructure of USCGC Healy. Given that standard depths were typically sampled (surface, 10m, 25m, 50m, 100m plus the chlorophyll fluorescence maximum), the percent of surface PAR was estimated at each standard depth, then the closest Niskin bottle to each target light depth was chosen for productivity incubation. Often, the water column was only 30-40m and the euphotic depth was shallower still. It was common that water from a single Niskin bottle would be used for more than one simulated in situ incubation sample since the depth range sampled by the Niskin bottle encompassed several standard light depths. Water samples for incubation were transferred from Niskin bottles to incubation bottles inside the ship's enclosed hanger. Water samples for 14 C carbon fixation measurements were prefiltered through 200 μm nitex mesh to remove large grazers. Incubations were performed in 70 mL polystyrene tissue culture bottles that were previously thoroughly cleaned with 10% HCl, then ethanol, 4 rinses with ship's distilled water and finally 3 rinses of polished reverseosmosis water, then rinsed three times with each sea water sample prior to filling.
Incubations were performed in simulated in situ conditions on-deck, corrected for both light quantity (using bags made of neutral-density shade cloth) and quality (spectral narrowing using layers of blue acetate as bag inserts). Bottle transfers between the CTD hanger and radioisotope van were always done in a darkened themal cooler to reduce light and temperature shock to the phytoplankton. Deck incubators consisted of a white plastic tub open to ambient sky light, chilled using surface seawater from the ship's flowing sea water system. The daily PAR was measured using the ship's PAR sensor set on top of the ship's meteorological mast. All filtrations were performed using 0.4 μm pore-size polycarbonate filters. Following the microdiffusion step, filters and sample "boats" were placed in scintillation vials with 7mL of Ecolume scintillation cocktail. Samples were counted using a Beckman-Coulter LS6500 scintillation counter with channel windows set for 14 C counting with calibration checked with a sealed 14 C standard. Counts were performed for sufficient time to reach 2% precision or 20 minutes for samples with lower counts. Blank 14 C counts were always run for scintillation cocktail as well as the phenethylamine CO 2 absorbent. 14 C counts with a 5% isotope discrimination factor assumed for the physiological fixation of 14 C-HCO 3 (as opposed to 12 C-HCO 3 ). Aerial integrations of carbon fixation to the base of the euphotic zone were based on the PAR attenuation measured during the CTD cast and depth integrations were performed using trapezoidal integration. Photosynthesis and calcification measurements were normalized to fluorometer-derived chlorophyll concentration. Samples for chlorophyll analysis were filtered on 25mm, GF/F filters (Whatman) then submerged in 5mL of 90% acetone, extracted for ~ 24h at 3°C. Following centrifugation, the fluorescence of the supernatant was analyzed using a Turner 10-AU fluorometer (Turner Designs, Inc.), previously calibrated with chlorophyll standard (Sigma) (Holm-Hansen et al., 1965).
Intrinsic, carbon-specific growth rates for POC (μ POC ) and PIC (μ PIC ) (units d -1 ) were estimated by dividing the rates of photosynthesis or calcification (in units of moles m -3 d -1 ) by POC or PIC concentrations (moles m -3 ), respectively.

Cruise details and hydrographic observations
The general study area of Healy cruise 1101 was the western Arctic (Fig. 1A). The period that the Healy 1101 cruise was in the vicinity of the under-ice algal bloom was between calendar days 183-205. During this period, the southern extent of the ice edge receded north ~100km (Fig. 1B). Water temperatures over the top 5m showed the presence of coldest waters (<-1°C), indicative of Winter Water (Rudels et al., 1990) in the far western portion of the study area, near stations 54-57 ( Fig. 1B,C). The next coldest waters were observed in the northern extent of the study area, over the Canadian Basin (Station 100; Fig.   1C). Highest salinities were observed along the southern end of the cruise track, extending (in patches) to station 67 (Fig. 1D), usually associated with waters of 2 to 5 o C. Lowest salinities were found in the 0 to -1 o C water of the Canada Basin (Fig. 1D).

Chlorophyll, absorption and fluorescence observations
Chlorophyll concentrations (derived from the continuous underway fluorescence measurements calibrated to discrete chlorophylls) reached greatest values of ~30ug L -1 in the western portion of the study area, where Winter Water reached the top 5m (see white contour line in Fig. 2A). This was the site of the under-ice bloom described earlier (Arrigo et al., 2012). Lowest chlorophyll a values were seen in the Canadian Basin, (~300X lower at 0.1 ug L -1 ( Fig. 2A)). Using a chlorophyll concentration of >2μg L -1 as the criterion for the bloom the largest horizontal dimension measured in the under-ice algal bloom, using the continuous underway system, was ~140km ( Fig. 2A).
Particle absorption was also highest in the under-ice algal bloom, with elevated values of ~1m -1 , reaching 100km from the ice edge and lowest values in the Canadian Basin (north of station 95; Fig. 2B). Absorption of colored dissolved organic matter (CDOM; a g412 ) was elevated within the under-ice bloom and lowest in the Canada Basin (Fig. 2B). Absorption of both CDOM plus detrital matter (a gp412 ) was elevated in the under-ice bloom, twice the magnitude of a g412 (Fig. 2C). Values of a gp412 were also elevated near shore (Fig. 2D). The proportion of total absorption at 412 nm contributed by the dissolved (<0.2um) fraction was generally 70-90% over the study area except in the under-ice bloom where only 30-50% of the total absorption was contributed by dissolved materials (Fig. 2F).
Chlorophyll-specific absorption (Fig. 2E) were calculated by first subtracting dissolved absorption from the total particulate and dissolved absorption at all wavelengths, in order to estimate particulate absorption. The particulate absorption was then calculated at each wavelength, subtracting the residual absorption at 715nm to correct for scattering effects (Bricaud et al., 1988).The absorption cross section of chlorophyll at 440nm (a* p440 ) was calculated by dividing the particulate absorption (m -1 ) by the chlorophyll concentration (units mg m -3 ). Average values of a* p440 were 0.025 m 2 (mg Chl) -1 in the western Winter Water as 13 well as in the cold waters of the Canadian Basin (Fig. 2E). Highest absorption crosssections were seen in the warmest, high salinity waters near the Alaskan coast.
Both a p440 and a p412 were well correlated to chlorophyll biomass. The plot of particulate absorption at 440nm (a p440 ) versus chlorophyll concentration (Fig. 3A) had a Y intercept of 0.002 m -1 , barely above zero (Table 1), indicating that particulate absorption of phytoplankton was virtually all associated with viable, chlorophyll-containing phytoplankton, not detritus. Further, a p412 (which would normally be expected to be representative of particulate detritus) was highly correlated to chlorophyll with a slope of 0.025 m 2 (mg Chl) -1 and Y intercept of 0.005 m -1 ( Fig. 3B; Table 1). The high correlation between a p412 and a p440 can be seen in Fig. 3C, with an r 2 = 0.975 and slope of 0.926 (Table 1). Thus, a p412 was as good proxy of chlorophyll a as a p440 , not detritus. Values of dissolved absorption at 412nm (a g412 ) had a positive but far reduced correlation with chlorophyll a, however, with only a factor of two increase in a g412 observed over >2 orders of magnitude of chlorophyll (Fig. 3D).
The relation was still statistically-significant (Table 1).
The slope of the absorption spectrum of dissolved material between 412 and 440 nm ( Fig. 4A), S g (nm -1 ) , was calculated according to Stedmon and Markager (2001)as: S g =((Ln(a g412 / a g440 )/(440-412)) A comparable slope for the detrital and particulate absorption, S pg (nm -1 ), was also calculated by substituting a pg412 and a pg440 in place of a g412 and a g440 , respectively, in the above equation  (Fig. 4C). The relative fluorescent yield of the combined dissolved/detrital material was calculated as the FDOM (from raw, unfiltered seawater) divided by the a g412 . The term "relative" is used here because FDOM excitation wavelength (370nm) did not match the absorption wavelength measured by the ac-9 (412nm).
The lowest relative FDOM fluorescent yield was observed in the under-ice algal bloom while highest values were observed in the Canada Basin region. Relatively low values were also seen near the coastline of Alaska (Fig. 4D). Highest concentrations of FDOM were in the under-ice bloom, (likely produced by the intense phytoplankton growth) but this FDOM had low relative fluorescent yields (Fig. 4D). The nature of this FDOM can be evaluated through its relation to other bio-optical variables. For example, FDOM was significantly correlated with CDOM (as a g412 ) but the dynamic range in FDOM was less than a factor of 2 over a 10X variation in a g412 (and the squared coefficient of correlation was only ~0.3; Fig. 5A). FDOM was better correlated to the chlorophyll a concentration than a g412 (Fig. 5B). The best-fit power function to those results accounted for almost 60% of the variance ( Table 1). The relative FDOM fluorescence yield also was inversely correlated with the chlorophyll concentration ( Fig. 5C) such that the under-ice bloom showed the lowest fluorescent yields, accounting for about 25% of the variance. However, relative FDOM fluorescence yield was strongly inversely correlated to S g (Fig. 5D) suggesting that the most weakly-colored CDOM and detritus (low S g ) had the highest relative fluorescent yield. Note, negative S g values as shown in Fig. 5D indicate that a gp412 <a gp440 (which only occurred in the clearest, most oligotrophic waters with extremely low chlorophyll and low suspended particulate matter, such as in the Canada Basin).

Optical scattering measurements
Optical scattering properties were elevated in the under-ice algal bloom. For example, the acid-labile backscattering--that backscattering associated with suspended calcium carbonate--while generally low, had the most elevated values in the under-ice algal bloom (Fig. 6A). Total particulate backscattering (Fig. 6B) was also elevated within the under-ice algal bloom, such that b b ' only represented, at most, 10% of the total particulate backscattering (Fig. 6C). Total scattering in the under-ice algal bloom reached values as high as 2m -1 with an order of magnitude decrease in the Canada Basin (Fig. 6D).
Backscattering probability (b~b = b bp /b p ; indicative of all minerogenic scattering, but not just for calcium carbonate) had values of 1% in the under-ice bloom and values up to 3-4% in the Canada Basin and in the open, warm waters south of the ice margin. Low values were seen in the southeastern portion of the study area (Fig. 6E). Waters with highest particle scattering ( Fig. 6D) also had highest particle beam attenuation (Fig. 6F). Indeed, particle backscattering, particle scattering and particulate attenuation all showed similar patterns (compare relative patterns in Figs. 6B, D and F).

Chemical and biogeochemical observations
Vertical sections of PIC, POC and BSi through the under-ice algal bloom all were elevated in regions where the Winter Water reached closest to the surface (Fig. 7). PIC showed elevated values just above the sediments at about 50m, near the shelf break at the most northwesterly position of the cruise, as well as in the region close to the coast of Alaska.
Ice-free waters away from the ice edge had low PIC concentrations and elevated POC and BSi. The most elevated PIC in surface waters was seen in the shallowest part of the sections, in ice-free waters, for both legs shown in the section. (Fig. 7A). POC and BSi were highly elevated under the ice, and had a subsurface peak which extended southeast of the ice edge, in the same area where PIC was low (Fig. 7A-C). Deepest waters along the section had lowest values of POC and BSi.
Ratios of PIC:POC were extremely low (~0.25%) in surface waters at the ice edge and within the under-ice algal bloom whereas the ice-free waters over Hannah Shoals (with elevated PIC; Fig. 7A) had PIC:POC ratios of 1.5-2% (Fig. 8A). Highest PIC:POC ratios were found at 100-150m depth at the shelf break. POC:PON molar ratios of the particulate material in the under-ice algal bloom were elevated above   (Fig. 8B).
POC:Chl a ratios in the under-ice algal bloom were generally low (50-100 except at the northwest corner of the survey area where there was a region with clearly elevated POC:Chl a ratios (Fig. 8C). Highest POC:Chla ratios in surface waters were found off the NW coast of Alaska.
The nitrate section through the under-ice bloom showed clear evidence of drawdown in the top10-20m as well as evidence of elevated nitrate at the shelf break which was associated with cold Winter Water (Fig. 9A). Silicate drawdown in surface waters also occurred in the under-ice bloom but concentrations of 30μM silicate were observed at the surface at station 54 (Fig. 9B). Residual nitrate (defined as the nitrate concentration minus the silicate concentration) (Townsend et al., 2010) showed negative values of -20 to -40μM under the ice, emphasizing the strong reduction of nitrate relative to silicate (Fig. 9C).
Primary production and calcification showed highest values within the under-ice bloom. While the primary production rates were high on any standard (~400 mg C m -3 d -1 ; Fig. 9A; Table 2), the calcification rates were only 0.4% of the primary production values (Fig. 10B). Carbon fixation dropped off rapidly in the ice free waters, as well. Primary production and calcification both attenuated with depth. Integrated primary productivity rates in the bloom approached 3g m -2 d -1 whereas integrated calcification was ~10 mg m -2 d -1 (Table 2). The C:P ratio in the bloom averaged 0.33% over the water column. Chlorophyllnormalized primary production was 5 gC (g Chl) -1 d -1 (Table 2). Integrated calcification normalized by integrated chlorophyll was also low, 0.02 gC (g Chl) -1 d -1 (Table 2). Intrinsic, carbon-specific growth rates for POC (μ POC ) approached 0.4 d -1 (Fig. 10C) while μ PIC approached 1d -1 (Fig. 10D). Integrated chlorophyll biomass in the center of the under-ice bloom was 490 mg m -2 (Table 2).

Microscopy
Scanning electron microscopy results from the under-ice algal bloom showed strong dominance by diatoms, with Chaetocerous sp, Fragilariopsis sp. and Thallasiosira sp. (Fig.   11A-F). Coccoliths of the coccolithophore, Emiliania huxleyi were also observed. While the coccoliths were >4um in diameter (which typically is a trait more characteristic of the type B morphotype) , there were traits that align with Type A morphotypes--the distal shield was larger than the proximal shield, the radial elements were robust, and the elements in the central area were curved (Fig. 11G, H)  .

Size of bloom based on continuous underway measurements
The hydrographic measurements made by our surface underway system clearly showed the coolest waters (-1.6°C under the ice with salinities of 30-31), characteristic of Arctic Winter Water (Coachman and Aagaard, 1974;Coachman and Barnes, 1961;Rudels et al., 1990;Rudels et al., 2004). Based on the continuous surface hydrographic data, the maximum horizontal length-scale of the Winter Water mass was about 150km (Fig. 1), close to the length of the elevated chlorophyll concentration for the bloom (~140km; Fig. 2A).

Interpreting the absorption properties of the under-ice algal bloom
The particulate absorption at 440nm showed similar trends to the chlorophyll concentration, as expected (Fig. 2B), however, the chlorophyll specific absorption at 440nm (the absorption cross section, a* p440 ) averaged 0.027(SE = ±9.6x10 -5 ) m 2 (mg Chl) -1 over the study region (Figs. 2E; 10A), well within the range observed for phytoplankton (Bricaud et al., 1983), in particular diatoms (Bricaud et al., 1988;Sathyendranath et al., 1987). Such variability is known to be a function of pigment composition, cell size and internal chlorophyll concentration. The predominance of low values of the absorption cross section (a p * 440 ; Fig. 2E) suggest that the pigments were highly packaged, characteristic of large diatoms. However, the a p * 440 values observed near the coast (0.10-0.23 m 2 (mg Chl) -1 ) were far higher than expected for phytoplankton and these may have resulted from other sources of absorbing particulate matter, or the presence of photoprotective pigments. It should be noted that the two cruise legs with such high a p * 440 , southeast of Hanna Shoals were performed at the end of the cruise (calendar day 204-205; July 23-24), almost one month after the earlier section through the under-ice bloom, and water temperatures had warmed 3.5-4°C and light levels would have been higher, making phytoplankton cells more high-light-adapted.
The shape of the particulate absorption spectrum contains information on phytoplankton size. Ciotti et al. (2002) normalized the spectral absorption at a given wavelength, λ (a ph(λ) ; m -1 ), by the mean absorption across the visible spectrum (<a ph >) and demonstrated that, for 440nm light, the closer the value of a ph(440) /<a ph > to 1.5, the greater the proportion of microplankton in the sample and alternatively, the closer the value to 3, the larger proportion of smaller phytoplankton. Using this technique, they were able to discriminate between picophytoplankton (<0.2μm), ultraphytoplankton (2-5μm), nanophytoplankton (5-20μm) and microphytoplankton (>20μm). Moreover, they could model the normalized phytoplankton absorption of any assemblage using combinations of just the micro-and pico-phytoplankton spectra. Ciotti et al. (2002) used the methanol extraction technique (Kishino et al., 1984) to unequivocally measure the spectral absorption of particulate detritus which they then subtracted from the total particulate absorption spectrum to calculate phytoplankton absorption (a ph(λ) ).
Unfortunately, we had no methanol extraction data so we had to use other means to ascertain if a p (λ) approximated a ph (λ). In over half of the study area, >95% of absorption at 412nm was from dissolved material, hence absorption by particulate detritus was minimal (thus, at 440 nm, a p was probably close to a ph ). Particulate absorption at 412nm was only significant in the under-ice algal bloom (see Fig. 2F where a g412 /a pg412 was 20-50%) as well as close to the Alaskan coast. However, the carbon:chlorophyll ratio in the bloom was ~50 (Fig.   8C), more representative of actively growing phytoplankton than assemblages dominated by particulate detritus (Geider, 1987). Further, the plots of a p440 and a p412 versus chlorophyll showed that particulate absorption of phytoplankton was virtually all associated with viable, chlorophyll-containing phytoplankton, not detritus ( Fig. 3; Table 1). The reduced correlation between chlorophyll a and a g412 (Fig. 3D) is consistent with other sources of a g412 than just phytoplankton, such as terrestrial sources.
In short, while the a p412 was elevated in the bloom, it strongly covaried with chlorophyll, indicative of minimum amounts of particulate detritus. We conclude that detrital absorption at 440nm was negligible in the bloom such that a p (440) would have approximated a ph (440). This allowed calculation of a ph(440) /<a ph > (Ciotti et al., 2002) along the cruise track ( Fig. 12A) as well as the resultant fraction of picoplankton that would have been expected in the assemblages (S f ; Fig. 12B). Values of a ph(440) /<a ph > varied from 1.5-2, suggesting that the entire study area was strongly dominated by microplankton (Ciotti and Bricaud, 2006;Ciotti et al., 2002). This conclusion was entirely consistent with the scanning electron microscopy results, as well (Fig. 11).

CDOM, FDOM and fluorescence yield
These results suggest that under-ice phytoplankton were an important source of FDOM and that FDOM fluorescence accounted for only ~30% of the variance in CDOM.
The change in dissolved absorption between 412 and 440nm, normalized by the change in wavelength, S g (Roesler and Perry, 1989) has been suggested to vary as a function of the source of CDOM. Steeper slopes typically are more representative of lignin-rich, terrestrially-derived materials (Stedmon and Markager, 2001). In this study, S g values of 0.01 in the bloom were more representative of low-colored, autochthonous, marine CDOM (Carder et al., 1989), as opposed to highly-colored, terrestrially-derived CDOM (Stedmon and Markager, 2001) (Fig. 4A). In the Canada Basin, absorption at 440nm was greater than at 412nm, likely due the extremely low CDOM concentrations there.
The CDOM fluorometer used here had excitation/emission peaks of 370 and 460nm, respectively. These correspond roughly to the red-shifted, "Peak C", humic-like, CDOM fluorophore originally described by Coble (1996). FDOM can be produced by a variety of different compounds, and the fluorescence yields can be affected by a multitude of physical and chemical factors (including pH, temperature, hydrogen bonding, metal binding, etc.).
Biology is also involved since bacteria (Rochelle-Newall and Fisher, 2002) and phytoplankton (Romera-Castillo et al., 2010) are both sources of FDOM. Our results showed no relation of relative fluorescence yield to water temperature (results not show).
Moreover, the data support the conclusions of Romera-Castillo et al. (2010) that phytoplankton are producers of FDOM (Fig. 5B). The strong linear relation between relative fluorescent yield and S g was not expected but shows a highly predictable continuum of relative fluorescence yield across these under-ice waters. With such a strong relationship, these results would suggest an alternative way to predict S g using FDOM fluorescence yield.

Significance of bloom magnitude
The levels of productivity found in the under-ice bloom represent some of the highest levels found in nature (Balch et al., 1992;Morel and Maritorena, 2001). Integrated primary production rates in the bloom center (2.86g m -2 d -1 ) were well above under-ice rates observed previously (Gosselin et al., 1997), yet the water column assimilation efficiency of 5.8 gC (g chl) -1 d -1 was still well below the theoretical maximum for phytoplankton (Falkowski, 1981).
In an integrated sense, calcification rates represented only 0.33% of the integrated productivity (Table 2). Similarly, in a pure coccolithophore culture (or dense coccolithophore blooms in nature), one would expect a chlorophyll-normalized calcification of ~1.19 gPIC fixed (g chl) -1 h -1 (or 28.6 gPIC fixed (g chl) -1 d -1 ) (Balch et al., 2007). The chlorophyll normalized values observed in this study (0.02 gPIC fixed (g chl) -1 d -1 ; Table 2) were three orders of magnitude less than this, simply due to the dominance of diatoms (Table   2).

Coccolithophores were present in the under-ice algal bloom
Optical scattering, PIC, SEM and calcification results demonstrated the presence of coccolithophores in the under-ice algal bloom but not as a dominant part of the community.
Values of acid-labile backscattering, compared to subpolar or subtropical waters, were low, approaching the sensitivity of the technique and certainly indicative of a non-bloom, background population of coccolithophores (Balch et al., 2004;Balch et al., 2005;Balch et al., 1996). As a percentage of the total backscattering ( Fig. 6C; 4-6%) this is lower than the typical percentage of backscattering typically attributed to coccolithophores, even in oligotrophic gyres . PIC:POC ratios were characterized by low values (<0.2%) in the under-ice bloom, too, again suggestive of low biogeochemical impact of coccolithophores in this under-ice bloom.
In general, the distribution and fixation of PIC mirrored that of POC, and it appeared that coccolithophores were responding in the same manner to increased light penetration through the ice as were the diatoms and other algal groups (Figs. 10). Given the elevated nutrients found the Arctic Winter Water, all groups were released from light limitation together. Barber and Hiscock (2006) described algal communities in which all the phytoplankton groups responded with increased growth rate to enhanced iron, they also observed that the picoplankton response was more muted than that of the diatoms because picoplankton were selectively grazed down by the fast-responding microzooplankton (Landry, 2002). In the case of the under-ice algal bloom, it is possible that enhanced growth of coccolithophores (by release from light limitation) was also muted by grazing by fastresponding protistan predators. Indeed, standing stocks of POC and BSi showed highcovariance under the ice (Fig. 7); both showed a subsurface "tongue" that extended out from under the ice on the northern part of the under-ice algal bloom while PIC, on the other hand, was actually reduced in that feature and greatest PIC concentrations were observed almost 100km away from the ice edge, in ice-free water of 3-4 o C (Figs. 1 and 5). The appearance of E. huxleyi in waters of the Southern Ocean also occurs at such temperatures Cubillos et al., 2007;Gravalosa et al., 2008;Holligan et al., 2010;Mohan et al., 2008).
Elevated concentrations of PIC and BSi also were observed just above the sediments near the shelf break, at 40-60m depth, suggesting that resuspension also may have been important source of these biogenic mineral particles. This could be seen in the PIC:POC ratios which were extremely low in the under-ice algal bloom (0-0.2%) but elevated with depth, with highest values observed in the 100-150m-deep, northwest portion of the under-ice algal bloom (at 3% ; Fig. 8A). As noted above, resuspension may have influenced this ratio at the shelf break. Alternatively, preferential remineralization of POC over PIC could also have produced this pattern (Honjo et al., 2008). The saturation states for calcite (Ω calcite ) and aragonite (Ω aragonite ) showed that waters were saturated for calcite and aragonite in this region (Bates personal communication), thus dissolution of calcite and aragonite would have been unlikely.

Nutrient limitation and a mismatch in carbon standing stocks versus rates of carbon fixation
Molar ratios of C:N were, for the most part, greater than the 6.6 C:N Redfield ratio (Redfield et al., 1963) in the top 20m of the under-ice feature (Fig. 8B) but given that nitrate levels under the ice had been depleted to micromolar levels (Fig. 9A), then it would be expected that the populations would have shown signs of nitrogen limitation. This interpretation was buttressed by the low POC:Chlorophyll ratios in under-ice algal bloom waters where nitrate depletion had only occurred in the upper 10m of the water column.
Indeed, regions of elevated POC:PON corresponded to high POC:Chlorophyll, for example at the northwest corner of the study area. Deeper Winter Water under the ice was rich in nitrate and silicate as evidenced by the covariance of the -1.6 isotherm and isolines of silicate and nitrate ( Fig. 9A and B). Moreover, Winter Water was elevated in silicate relative to nitrate (by 40μm at 60m depth; Fig. 9C). In regions where chlorophyll and POC were highest, the residual nitrate was closer to zero, suggesting that the phytoplankton community uniformly drew down nitrate and silicate towards zero. This could have resulted from either adjustments of diatom cell quotas as they consumed the silicate (Baines et al., 2011) or depletion of nitrate by the majority diatom assemblage and minority, non-diatom phytoplankton plus silicate depletion by just the diatoms, in such a way that both were depleted together. What seems clear is that where physical processes brought Winter Water upwards under the ice, the release of the phytoplankton from light limitation by melt ponds then allowed nutrient drawdown to occur such that all algae began to show signs of both nitrogen limitation (increased C:N) and silicate limitation (reduced silicate with increasing BSi).
Overall, the history of the bloom formation caused a mismatch in the standing stocks and rates of fixation of particulate organic and inorganic carbon. Highest carbon fixation (for photosynthesis and calcification) was observed further into the ice, in regions where the standing stocks of POC and chlorophyll were not the highest. Elevated standing stocks of PIC and POC were found in waters where nutrients had already been depleted whereas highest productivity rates were seen where nutrients had not yet been depleted.

Conclusion-Under-ice coccolithophores and global change
The aforementioned observations demonstrate that coccolithophores were present in the under-ice algal bloom but that their relative contribution to carbon cycling was minor compared to the carbon fixation by the diatom-dominated assemblages. Coccolithophore presence was observed analytically (ICP-OES), optically and microscopically. The low numbers of coccolithophores in the under-ice bloom is consistent with previous observations of algal communities in ice-covered waters.
A comparison of the calcification rates in the under-ice algal bloom of ICESCAPE with those rates measured on polar coccolithophores by Charalampopoulou et al. is informative. Highest total calcification rates at their ice edge station was ~0.6 mg C m -3 d -1 at 20m depth, some 40% of what we observed for the under-ice algal bloom (1.5 mg C m -3 d -1 ). Charalampopoulou et al. (2011) also measured calcification in the marginal ice zone and a Svalbard fjord and found calcification rates 20-75 times lower than the calcification rates in Chukchi Sea under-ice algal bloom (0.02-0.07 mg C m -3 d -1 ).
This data set also provides the opportunity to compare the growth rates of the phytoplankton (μ POC in Fig. 10C, clearly dominated by diatoms) to rates predicted in the classic treatise by Epply (1972) on the effects of temperature and phytoplankton growth.
Using his equation 1 (or 1a), the predicted maximal growth rate of phytoplankton in Winter Water of -1.6°C would have been 0.77 doublings d -1 (= 1.11 d -1 specific growth rate). The highest POC-specific growth rates that we observed (~0.35 d -1 ) were ~30% of the maximum growth rate predicted by the Eppley (1972) equation, possibly reflecting the effect of the previous nitrate draw-down in the surface waters. Moreover, if the intrinsic rates of increase of PIC were coupled to coccolithophore growth rates, as shown previously (Fritz and Balch, 1996), then the observed PIC-specific growth rates (0.9 d -1 ) would have been much closer to the maximal growth rates predicted by Eppley (1972). This would have been expected anyhow given the well-known observations that coccolithophores such as E. huxleyi can maintain higher growth rates at lower nitrate concentrations than, for example, diatoms, due to their significantly lower half saturation constants for nitrate uptake (Eppley et al., 1969;Margalef, 1978).
The hypothesis central to the formation of the under-ice algal bloom is that melt ponds on top of the ice allowed light to penetrate the meter-thick ice, thus releasing under-ice algae from severe light limitation (Arrigo et al., 2012). One can then ask how long it would have taken for phytoplankton with chlorophyll concentrations below the limit of detection to grow to the levels seen in the under-ice bloom. This also provides insights whether the bloom could have formed in place or was somehow advected and concentrated there.
Assuming a background concentration of chlorophyll of 0.01 mg m -3 prior to the bloom (0.04 mg m -3 is typically used as the limit of detection for the fluorometric chlorophyll technique using standard practices and volumes (Parsons et al., 1984)), then using the logistic growth equation, and the above μ POC of 0.35 d -1 , then it would have taken ~23d to reach a chlorophyll concentration of 30mg m -3 (i.e. time (days) = ln[30/0.01]/0.35), assuming no grazing or other loss terms. Thus, for the above hypothesis to be consistent with our observations, than the melt ponds would have had to be present for at least three weeks prior to our measurements for there to be sufficient time for the high chlorophyll levels to form.
Moreover, it is assumed that once nitrogen became limiting, then the diatom growth would have slowed (alternatively, grazing might have also reduced the net growth below maximal growth rates).
This same calculation can be done for PIC, however, in this case, the estimated intrinsic calcification rate, μ PIC , was greater (0.9 d -1 ; Fig. 10). Even beginning with just 10 E.
huxleyi cells L -1 (each containing15 coccoliths of 0.2pgPIC (Balch, 1991;Balch et al., 1991) 26 (or 0.25pmoles PIC/cell) for a total of 2.5pmoles PIC L -1 ), then after 23d, the water would have contained several mmoles PIC L -1 . Such was not the case, however, with PIC levels only 0.15 μmols L -1 in surface waters of the under-ice algal bloom (Fig. 7), thus suggesting other forces were acting on the coccoliths (sinking, grazing and/or dissolution) to keep concentrations low, or simply that such PIC-specific growth rates of coccolithophores were not sustained for 23d (most likely).
The under-ice bloom observations first described by Arrigo et al. (2012)  Given intrinsic calcification rates, it is not known why there are not more coccoliths present in these waters. It has been shown that pH has a strong influence on coccolithophores in polar waters (Charalampopoulou et al., 2011). Ocean acidification will cause the largest decline in carbonate saturation states in high latitude, polar waters (Feely et al., 2009), especially after the polar ice cap melts, allowing more efficient air-sea gas equilibration. A key point, however, will be the balance that warming/release from light limitation will play in encouraging coccolithophore growth in Arctic waters versus the inhibitory role that increasing ocean acidification will have on coccolithophore production and growth in polar waters, in the face of climate change.     Table 1.        Fraction of phytoplankton that are picoplankton (Sf) calculated according to equation 3 of Ciotti et al. (2002) for measurements at 440nm. See text for details. In all panels, isopleths of temperature and the ice edge on day 182 are shown for reference as in Fig. 1C.