Evolution of the Amundsen Sea Polynya , Antarctica , 2016-2021 1

Polynyas are key sites of ice production during the winter and are important sites of biological activity 7 and carbon sequestration during the summer. The Amundsen Sea Polynya (ASP) is the fourth largest Antarctic po8 lynya, has recorded the highest primary productivity and lies in an embayment of key oceanographic significance. 9 However, knowledge of its dynamics, and of sub-annual variations in its area and ice production, is limited. In this 10 study we primarily utilize Sentinel-1 SAR imagery, sea ice concentration products and climate reanalysis data, along 11 with bathymetric data, to analyze the ASP over the period November 2016 March 2021. Specifically, we analyze 12 (i) qualitative changes in the ASP’s characteristics and dynamics, and quantitative changes in (ii) summer polynya 13 area, (iii) winter polynya area and ice production. From our analysis of SAR imagery we find that ice produced by 14 the ASP becomes stuck in the vicinity of the polynya and sometimes flows back into the polynya, contributing to its 15 closure and limiting further ice production. The polynya forms westward off a persistent chain of grounded icebergs 16 that are located at the site of a bathymetric high. Grounded icebergs also influence the outflow of ice and facilitate 17 the formation of a ‘secondary polynya’ at times. Additionally, unlike some polynyas, ice produced by the polynya 18 flows westward after formation, along the coast and into the neighboring sea sector. During the summer and early 19 winter, broader regional sea ice conditions can play an important role in the polynya. The polynya opens in all sum20 mers, but record-low sea ice conditions in 2016/17 cause it to become part of the open ocean. During the winter, an 21 average of 78% of ice production occurs in April-May and September-October, but large polynya events often asso22 ciated with high winds can cause ice production throughout the winter. While passive microwave data or daily sea 23 ice concentration products remain key for analyzing variations in polynya area and ice production, we find that the 24 ability to directly observe and qualitatively analyze the polynya at a high temporal and spatial resolution with Senti25 nel-1 imagery provides important insights about the behavior of the polynya that are not possible with those da26 tasets. 27 28 29 30 31 32 33 34 35 36 https://doi.org/10.5194/tc-2021-250 Preprint. Discussion started: 21 September 2021 c © Author(s) 2021. CC BY 4.0 License.

The location of the Amundsen Sea and our study sites within the context of Antarctica and the Southern

130
Ocean. The background image is from Quantarctica (Matsuoka et al., 2021); (b) The location of the ASP within the tica V2' sea floor topography dataset for our study area to examine alongside our qualitative analysis. This dataset 163 was downloaded from the NSIDC (https://nsidc.org/data/nsidc-0756) and has a grid spacing of 500 x 500 m 164 (Morlighem et al., 2020).

165
We also use our analysis of the imagery to assess the approximate day of summer polynya 'opening' and where ρa is the density of air at standard atmospheric pressure and 0°C, taken as, 1.3 kg·m −3 , cp is the specific heat 253 of air at constant pressure, taken as 1004 J kg −1 K −1 , U (in ms −1 ) is the wind speed at 10 m, taken from the processed 254 ERA5 data and Ta (in K) is the air temperature at 2 m, taken from the processed ERA5 data. Cs and Ce are bulk 255 transfer coefficients for sensible heat and latent heat, respectively and both taken as 0.00144. P0 (in pa) is the surface 256 air pressure and taken from the processed ERA5 data. Lv (in J kg −1 ) is the latent heat of water vaporization, r is the 257 relative humidity. ea (in pa) is the saturation water vapor pressure at the air temperature, rea is the actual water vapor 258 pressure of the air and es (in pa) is the saturated water vapor pressure at the surface temperature and are all calcu-

269
where Td is the dewpoint temperature taken from the processed ERA5 data.

272
The calculated daily heat flux was then cropped, re-aligned, resampled to a 3.125 km 2 grid and reprojected 273 to Antarctic Polar Stereographic using GDAL and QGIS to match the corresponding sea ice concentration data.

274
Next, where SIC was < 0.7 (i.e. pixels considered as part of open polynya) daily ice production volume, V, was esti-

288
Caution should be used when interpreting the absolute numbers produced by the ice production model, par-289 ticularly because the input data is modeled climate data not necessarily always representative of reality, and the 290 model itself is a simulation sensitive to uncertain parameter settings (Cheng et al., 2017;2019). Nevertheless, we 291 opted for this method due to the difficulty of directly measuring and tracking thin ice thickness in the polynya (e.g.

292
Tian et al., 2020) to estimate ice production, and the potential to compare our daily ice production results to results

296
In order to assess changes in the ASP in the context of changes in SIC at a broader spatial scale, SIC was 297 analyzed for the larger area defined in Fig. 1a. The same SIC dataset described in section 2.2 to obtain polynya area 298 was cropped to the broader region. The daily data was plotted spatially for all available days 1 November 2016 -31 tially. Additionally, the total SIC for each day was calculated by calculating the sum of all percentage SIC values in 301 the study region. These total SIC values should only be considered useful for analyzing relative changes in SIC in 302 our study period.

305
In order to analyze how polynya behavior relates to changes in wind conditions, mean wind speed and di-306 rection, and daily wind speed was calculated from ERA5 wind data. To obtain mean wind speed and direction, 307 ERA5's monthly wind speed and direction product was downloaded and cropped to the ASP study area (Fig. 1b) for 308 the period 1 November 2016 to 31 December 2020 and the mean calculated for the whole period, and plotted spa-     (Table 1). Through

507
Analysis of the spatial distribution of ice production across all years reveals that the mean daily ice produc-    Daily mean wind speed at the polynya site and polynya area during the winter period has a weak but signif-521 icant positive correlation (0.33, P <0.05) (Fig. 8). Many day-to-day variations in polynya area are not correlated with 522 wind speed, however it is clear that notable spikes in polynya area do often occur on days with high wind speed 523 (Fig. S3). For example, the three highest polynya areas recorded in 2020 after April all occur alongside 524 the three highest spikes in wind speed post-April. By viewing the mean spatial distribution of wind it is clear that 525 the ASP forms in an area of relatively high winds (Fig. 9). A band of high winds with a mean speed of around 8 -9 526 ms -1 exists along the coast from Thwaites Glacier, over the Thwaites Iceberg Tongue and into the eastern area of the 527 ASP study area, where the main polynya originates.

528
The mean wind direction throughout the ASP study area is approximately southerly. While this direction 529 corresponds to the direction in which the polynya sometimes forms northward off the Dotson Ice Shelf, it does not 530 correspond to the more typical westward formation off the iceberg chain. open area on the seaward side of icepack and at a distance from the area used to calculate wind speed (Fig. 11b)).

534
The Pearson Product Correlation Coefficient is 0.33 (P < 0.05).  11. SIC for the broader ASP region on two days in 2017, during and following a summer of record-low SIC.

575
The area corresponds to that shown by red box in Fig. 1a.

578
Analysis of the SIC over a broader area also shows daily changes in the polynya area and how it relates to 579 changes in the icepack. The mean monthly cycle of the polynya can be seen in Fig. 10 and presents a similar picture 580 of the polynya as described in sections 4.1-4.3. The broader icepack has a minimum total SIC of 1 975 921 in Janu-581 ary and remains similar in February. From March the broader icepack can be seen to expand in area as the polynya 582 begins to close and continues to increase until a peak of 6 588 615 cumulative SIC in September. From October the 583 icepack begins a marked decline into summer. Interannual variation can be seen in Fig. S4 and Video S2, with maxi-584 mum icepack area occurring in August or September each year, and minimum icepack in January or February.

594
tive limit on the ASP's ice production is the previously discussed tendency for polynya-produced ice to inhibit fur-666 ther opening of the polynya due to blockages and reversals in ice drift. This process could also partly explain why sen Sea to the west, rather than traveling away from the coast after formation. This westward flow of the ice away microwave or sea ice concentration datasets for analyzing daily changes in polynya area or ice production. However, we find that the ability to directly observe and qualitatively analyze the polynya at a high spatial and temporal reso-

733
lution, year-round, with Sentinel-1 imagery provides important insights that are not possible with those other da-