Under‐Ice Mixed Layers and the Regulation of Early Spring Phytoplankton Growth in the Southern Ocean

Under‐ice phytoplankton “blooms” have been observed in the Southern Ocean, although irradiance is extremely low and vertical mixing is assumed to be deep. Most under‐ice data have been collected using Argo floats, as research expeditions during austral fall and winter are limited. Hydrographic measurements under dense ice cover indicate that vertical mixing in weakly stratified systems may be less than previously suggested, and that the accepted determinations of mixed layer depths are inappropriate in regions with extremely weak stratification, such as those under ice. Vertical gradients in density suggest that mixed layers in the Ross Sea in early October are not extremely deep; furthermore, while phytoplankton biomass is low, it has begun to accumulate under ice. Growth rates indicate that phytoplankton growth in the Ross Sea begins in early September. Extending the period of growth may have substantial impacts on carbon biogeochemistry and food web energetics in ice‐covered waters.

• Mixed layer depths under extremely weak vertical stratification (like under ice) are overestimated using conventional criteria • Phytoplankton growth in the Ross Sea is initiated soon after the start of solar day, far earlier than has been previously suggested • Early spring growth can potentially alter our views of Southern Ocean carbon biogeochemistry and food web phenology

Supporting Information:
Supporting Information may be found in the online version of this article.Smith et al., 2013;Smith, Anderson, et al., 2000;Smith, Marra, et al., 2000).A second method is by finding the depth of maximum property gradient, with the property being either density (Lorbacher et al., 2006) or a biological variable (Carvalho et al., 2017), and a third method uses some non-traditional criterion derived from more advanced techniques such as statistical measures or linear fitting (Chu & Fan, 2010;Huang et al., 2018).Underice Argo determinations generally use a density difference of 0.03 kg m −3 , as this has been shown to provide the most robust estimate of mixed layer depths throughout the Southern Ocean.However, in very weakly stratified water columns, it remains unclear if this density criterion provides an accurate estimate of the depth of mixing.
Controls of the onset of phytoplankton growth require estimates of mixed layer depths (Fischer et al., 2014).The classic Sverdrup hypothesis suggests that vertical mixing must be less than a critical depth; at mixing depths less than the critical depth, positive photosynthesis and phytoplankton growth can occur (Sverdrup, 1953).A modification of this approach invokes turbulent mixing depths, rather than the depth of homogeneous distributions of variables (Taylor & Ferrari, 2011).As the depth of turbulent mixing would usually be less than a homogeneous layer induced by vertical mixing, phytoplankton growth could occur earlier in the season and in the presence of apparently deep mixed layers.Estimates of turbulent mixing are absent in polar regions, especially under ice.
The disturbance-recovery hypothesis also requires an accurate estimate of vertical mixing (Behrenfeld, 2010).
Given that a majority of net community production precedes total ice retreat (McClish & Bushinsky, 2023), it is essential to have a clear understanding of mixed layer depths in ice-covered waters to understand the timing and regulation of phytoplankton in much of the Southern Ocean.
We analyzed data collected in the Ross Sea (Figure 1) under conditions of extremely dense ice cover to assess the depth of vertical mixing at the end of austral winter and to understand the impacts of vertical mixing on the timing and magnitude of phytoplankton growth.We then estimated mixed layer depths (or more appropriately homogeneous layers) using a variety of commonly used procedures and compared these results to the vertical profiles of a variety of oceanographic parameters, and conclude that mixed layers, as represented by homogeneous properties, 10.1029/2023GL106796 3 of 10 are shallow under ice-far shallower than estimated using commonly used procedures.Phytoplankton growth had been initiated by the time of these observations, despite the extremely low irradiance conditions, and indicate that growth in the Ross Sea begins earlier than had previously suggested (Smith, Anderson, et al., 2000;Smith, Marra, et al., 2000;Smith & Gordon, 1997).Under-ice productivity throughout the Southern Ocean may occur far earlier than usually assumed and may support substantial impacts on carbon biogeochemistry.

Methods
Temperature and salinity data were collected from the RVIB N.B.Palmer using a dual-sensor SeaBird 911+ CTD system, a Chelsea fluorometer, and a SeaTech transmissometer at two locations in the southern Ross Sea (Figure 1).A full description of the cruise sampling is provided in Supporting Information S1.Fluorescence was converted to chlorophyll values using a regression of discretely determined values and fluorescence values at the depth of sample collection.Transmission data were converted to particulate organic carbon (POC) concentrations using the calibration of Gardner et al. (2000) which included discrete POC measurements from these casts.
The temperature sensors had a resolution of 0.0002°C and an accuracy of ±0.001°C; conductivity sensors had a resolution and accuracy of 0.00004 and ±0.0003 S m −1 .Photosynthetically active radiation was measured on a trace metal-clean CTD system using a BioSpherical quantum sensor (Model QSP-2100), and surface irradiance measured using a Model QSL-2100 quantum sensor.The NOAA/NSIDC Climate Data Record of Passive Microwave Sea Ice Concentration (version 4) were downloaded from the National Snow and Ice Data Center (https:// doi.org/10.7265/efmz-2t65).
Mixed layers were estimated using multiple methods.The first was using seawater density difference criteria of 0.005, 0.01, 0.02, and 0.03 kg m −3 from the stable surface value (Dong et al., 2008).The second was by determining the maximum property gradient using the Brunt-Väisälä frequency (N 2 ; Lorbacher et al., 2006), and the third by the relative variance method (RV) (Huang et al., 2018, details in Supporting Information S1).
Net growth rates (µ in day −1 ) were calculated assuming exponential growth using the equation where N 0 is the initial biomass and N t the biomass at time t.Estimates of the date of phyto-plankton growth onset were made with the same equation using the growth rate calculated from the data and calculating the number of days required to reach the observed biomass on 17 October.

Results
Ice conditions vary spatially in the Ross Sea, as the Ross Sea Polynya forms near the ice shelf at 175°W due to off ice-shelf winds and alters ice concentrations both spatially and temporally (Figure S1 in Supporting Information S1).Ice concentrations near the Victoria Land coast were >95% during both stations, whereas those at 178°W were ∼85% during our first sampling and decreased to 75% by the time of the second occupation.Ice thickness was ca. 1 m; snow cover varied but averaged ca. 10 cm.Sampling in both locations required removal of ice to allow for instruments to be lowered through the ice; as such, the top 10 m of the water column were likely disturbed, and the data excluded from our analyses.Surface irradiance ranged from 57 to 100 mol photons m −2 day −1 ; irradiance just below the ice surface averaged 182 μmol photons m −2 s −1 .Vertical profiles of in situ irradiance at the earliest occupations of stations gave euphotic zones (the depth of the 1% isolume) of 88 m, and an average attenuation coefficient of 0.052 m −1 (Table S1 in Supporting Information S1).
Estimates of mixed layer depths (Z mix ) at all locations were dependent on the criterion used to define that depth (Table 1).Z mix estimates based on density differences of 0.005, 0.01, 0.02, and 0.03 kg m −3 gave mean mixed layers of 50, 73, 153, and 192 m, more than a three-fold difference among estimates.Average maximum gradient estimates (using Brunt-Väisälä frequencies) averaged 101 m, and estimates using the RV gave average Z mix of 55 m.These large differences greatly impact our assessment of the environment under sea ice and the regulation of phytoplankton growth.
Vertical profiles of density at the eastern station (178°W) demonstrated that a homogeneous layer had been created prior to the first occupation and maintained throughout the study (Figure 2); mean vertical mixed layers (Z mix ) using the most conservative density difference criterion (0.005) were 67 m and decreased only slightly at the time of the last occupation (mean Z mix = 63 m).As stratification increased, all estimates of mean Z mix tended to converge.At the western station, mean Z mix at St. 10 was 121 m, but decreased to 45 m at St. 15.As salinity controls density at these temperatures, this indicates that ice melting had begun in October under complete ice cover.Advection of waters from the south under ice would result in the same reduction in mixed layer depths.
Vertical profiles of density show the stratification at the eastern station, and the extremely weak stratification at the western station (Figure 2).Expanding the density profiles for Stations 10 and 15 reveal weak but identifiable stratification (Figure 3) and indicate that mixed layers were less than 50 m at Station 10, and ca.20 m at Station 15.Z mix were substantially less than the depth of the euphotic zone.Profiles of chlorophyll (Figure 4) clearly indicate that phytoplankton growth had begun.At St. 3 chlorophyll levels in the surface mixed layer had increased by an order of magnitude relative to waters below 150 m, and chlorophyll had doubled when the station was re-occupied 12 days later.A non-uniform chlorophyll distribution was observed at these stations; while the surface layer was seemingly homogeneous in hydrographic properties, it was not with respect to vertical phytoplankton distributions.Chlorophyll at the western station (St.10) was 2.5 times that of deep-water concentrations at the first occupation and increased further (4-fold greater than deep water values) 8 days later.Vertical variations were even greater at the western station, with chlorophyll concentrations showing a distinct near-surface maximum at St. 15.Surface reductions in chlorophyll may have resulted from photoinhibition of fluorescence to the low-irradiance acclimated phytoplankton, but also could reflect effects of horizontal advection or additions of low chlorophyll ice melt.POC concentrations were similar to those of chlorophyll except for the absence of surface reductions (Figure 5).At St. 3 POC was distributed evenly in the upper 80 m, consistent with the homogeneous distribution of density.Greater vertical variability was observed at St. 13, and the homogeneous distribution of POC was confined to the upper 60 m.Both stations showed a marked increase from deep-water concentrations, increasing by 150% and Estimates of net growth rates were made from the increases in integrated (0-100 m) chlorophyll, mean chlorophyll and mean POC levels (upper 50 m) between the two occupations (Table S2 in Supporting Information S1).Growth rates from POC increases at the western station were 0.034 day −1 and for the eastern station and 0.046 day −1 for the western station, reflective of the greater irradiance closer to open water due to reduced ice concentrations.Growth rates derived from chlorophyll were ca.60% greater at the eastern location, but those based on integrated chlorophyll at the western location were less than those of the POC-derived estimates due to the fluorescence reduction in the upper 40 m.Growth rates from mean chlorophyll concentrations were 69% greater than those derived from POC.The difference between the two estimates may reflect the sensitivity of the POC estimates, but also may reflect a physiological shift to optimize energy capture in low irradiance environments.Growth rates, regardless of the method of estimation, are very low-far below that estimated from temperature (0.65 day −1 ; Eppley, 1972) and from direct isotopic incorporation estimates in ice-free systems (Smith et al., 1999;Smith & Gordon, 1997).
Using the observed biomass in October and the estimated growth rates, it is possible to estimate the approximate onset of phytoplankton growth.For Station 10 (occupied on 17 October), growth at the observed growth rate must have begun in early September.Furthermore, if growth rates declined linearly as a function of irradiance, growth would have been initiated in late August, soon after the return of positive surface irradiance (17 August).
Early growth can provide a stimulus to the local food web and potentially influence the temporal patterns of the vertical flux of carbon.

Discussion
Using Argo technology, the early spring accumulation of Southern Ocean phytoplankton has been documented (Hague & Vicchi, 2021;Horvat et al., 2022), but the mechanisms supporting growth have not been elucidated.One suggestion was that irradiance attenuation by the ice-snow interface has been overestimated, but attenuation rates have been exhaustively investigated in a variety of habitats and quantified (Arndt et al., 2017).Snow is the major attenuator, and because it is not necessarily uniform in space, can drive spatial variations in growth.It also has been suggested that phytoplankton acclimate to extremely low irradiance levels (Hague & Vicchi, 2021), but there is no evidence from other deep-living assemblages that extreme acclimation occurs (Platt et al., 1983).Estimates of under-ice mixed layer depths (and hence irradiance availability) have used the accepted Argo criterion (a density difference of 0.03 kg m −3 from a stable surface value).Our results suggest that the use of this criterion It also has been observed using Argo data that there can be vertical variations in biological properties within a seemingly homogeneous surface layer (Carranza et al., 2018).These variations were suggested to arise due to difference in scales; that is, phytoplankton growth and accumulation can occur rapidly relative to the energy

Figure 1 .
Figure 1.Map of ice concentrations found on (a) 17 October and (b) 30 October 1996 in the southern Ross Sea.Locations of stations analyzed are indicated in panel (a); inset shows the location of the Ross Sea.

Figure 2 .
Figure 2. Vertical density distributions at the four stations in the Southern Ross Sea.All were plotted using the same scale for comparison.Stations 3 and 13 were in the east, and 10 and 15 were in the west closer to the coast.

Figure 3 .
Figure 3. Same data as in Figures 2c and 2d but plotted on an expanded density scale.

Figure 4 .
Figure 4. Vertical distributions of chlorophyll at the four stations analyzed.

Figure 5 .
Figure 5. Vertical distributions of particulate organic carbon at the four stations analyzed.
Note.When the mixed layers exceeded the depth of the cast, values are indicated as being greater than the depth listed.Density or stratification values used in the threshold method are indicated in parentheses.*,unable to calculate; nd, not detected.Mean and standard deviations (St Dev) calculated using only the absolute values detected.Mixed Layer Depths (Z mix ) Estimated by the Threshold Method, the Maximum Buoyancy Frequency Method, and the Relative Variance Method(RV) Criterion/station and cast Z MIX (Δ 0.005) Z MIX (Δ 0.01) Z MIX (Δ 0.02) Z MIX (Δ 0.03) Z MIX (Δ N 2 ) Z MIX (RV)