Weddell Sea Polynya analysis using SMOS-SMAP Sea Ice Thickness Retrieval

The Weddell Sea Polynya is an anomalous large opening in the Antarctic sea ice above the Maud Rise seamount. After 40 years of absence, it fully opened again on 13 September, 2017, and lasted until melt; staying open for a total of 80 days. 2017, however, actually was not the only year the imprint of the polynya could be identified. By investigating sea ice thickness (SIT) data retrieved from the satellite microwave sensors Soil Moisture Ocean Salinity (SMOS) and Soil Moisture Active Passive 5 (SMAP), we have isolated an anomaly of thin sea ice spanning an area comparable to the polynya of 2017 over Maud Rise occurring in September 2018. In this paper, we look at sea ice above Maud Rise in August and September of 2017 and 2018 as well as all years from 2010 until 2020 in a 11-year time series. Using the ERA5 surface wind reanalysis data, we present the strong impact storm activity has on sea ice and help consolidate the theory that the Weddell Sea Polynya, in addition to oceanographic effects, is subject to direct atmospheric forcing. Based on the results presented we propose that the Weddell 10 Sea Polynya, rather than being a binary system with one principal cause, is a dynamic process caused by various different preconditioning factors that must occur simultaneously for it to occur. Moreover, we show that rather than an abrupt stop to anomalous activity atop Maud Rise in 2017, the very next year shows signs of polynya-favourable activity that, although insufficient, was present in the region. This effect, as will be shown in the 11-year SMOS record, is not unique to 2018 and similar anomalies are identified in 2010, 2013 and 2014. It is demonstrated that L-band microwave radiometry from the SMOS 15 and SMAP satellites can provide additional useful information, which helps to better understand dynamic sea ice processes like polynya events, in comparison to if satellite sea ice concentration products would be used alone.

the given uncertainty of the product, meaning 30%. However, the retrieval does not take into account subtle differences that distinguish the two polar environments. Nevertheless, recently research done on Antarctic phenomena have made use of the SMOS SIT retrieval (e.g., Shi et al., 2021), and more specifically, SMOS SIT retrieval has been used for studying Antarctic polynya (e.g., Heuzé and Aldenhoff, 2018;Mohrmann et al., 2021).
Sea ice concentration (SIC) data (Section 2.2) is necessary to further validate and distinguish SIT data. The SMOS-SMAP 100 retrieval algorithm assumes near-100% SIC when retrieving SIT and since we look at a region prone to polynya and low SIC (Lindsay et al., 2004), it is necessary to consider this factor. The SMOS-SMAP SIT retrieval has no SIC dataset correction implemented because uncertainty of SIC algorithms at high concentration and their covariation at thin thicknesses will cause high errors (Paţilea et al., 2019). Using SIC maps and data in combination with SIT counterparts, we can better infer the location and degree of error in our SIT retrieval. Paţilea et al. (2019) mention specific examples and ratios for the retrieval 105 like sea ice concentration of 90% at 10 cm ice thickness for which the retrieved sea ice thickness is 8.5 cm. Meanwhile, 50 cm ice thickness at 90% sea ice concentration is just 28 cm. Conclusively, all sea ice concentration algorithms show less than 100% SIC for thicknesses below 30 cm (Paţilea et al., 2019). Thus, thin ice thickness data shown in this study should rather be interpreted as a combined ice area and thickness anomaly and not be used to calculate the actual ice volume for the polynya area. However, when the polynya opens, the large heat loss from the ocean often causes thin sea ice to grow, which soon shows 110 up as 100% SIC but will be correctly shown as large-scale thin ice area in the SMOS-SMAP dataset.

ASI Ice Concentration Algorithm
The ARTIST Sea Ice (ASI) algorithm retrieves SIC from the difference between brightness temperatures at 89 GHz at vertical and horizontal polarizations. This polarization difference is then converted into SIC using pre-determined fixed values for 0% and 100% SIC polarization differences known as tie points. It is known from surface measurements that the polarization 115 difference of the emissivity near 90 GHz is similar for all ice types and much smaller than for open water (Spreen et al., 2008).
At such high frequency, atmospheric influence is high also. This effect is dealt with in a bulk correction for atmospheric opacity and by implemented weather filters over open water. Because the Bootstrap (BBA) (Comiso et al., 1997) algorithm uses the 18 and 37 GHz channels, which are less sensitive to atmospheric phenomena, it is also used to essentially filter the produced ASI SIC concentration by setting SIC to zero where the Bootstrap algorithm retrieves less than 5% SIC.

ERA5 Climate Reanalysis
ERA5 Climate Reanalysis data is used to study direct atmospheric forcing on the opening of the polynya as well as on anomalous regional sea ice thinning to conclusively answer whether the Weddell Sea Polynya is purely ocean-driven or maintained by a combination of both processes. from 1950 onwards. It replaces the ERA-Interim reanalysis (spanning 1979 onwards) and is based on the Integrated Forecasting System (IFS) Cy41r2. ERA5 benefits from a decade of developments in model physics, core dynamics and data assimilation (Hersbach et al., 2020). In addition to a significantly enhanced horizontal resolution of 31 km, compared to 80 km for ERA-Interim, ERA5 has hourly output throughout, and an uncertainty estimate from a 10-member ensemble of data assimilations with 3-hourly output. mean absolute deviation = 2.2 hPa; mean bias = 0.8 hPa). On these grounds, ERA-I was deemed accurate for gathering signs of storm activity as it skillfully represented MSLP variability near Maud Rise. ERA5 is a reanalysis with a higher temporal and spatial resolution than ERA-I. It improves upon its predecessor in terms of information on variation in quality over space 135 and time as well as an improved troposphere modelling. As a result, for the purposes of this study, it should offer a better, or at least identical, assessment of the wind speeds near Maud Rise that are going to be cross-referenced with the presented SIT retrievals in this study.  (Fig. 1). Here the advantage of the SMOS-SMAP ice thickness retrieval shows its strength compared to the traditional sea ice concentration datasets. This section presents findings that suggest a previously unrecognized similarity between the two September anomalies. year has been shown via satellite imagery; most commonly via SIC retrieval (e.g., Campbell et al., 2019). Here the advantages of SMOS-SMAP SIT retrieval are limited by the high open water fraction but nevertheless help to demonstrate the full extent of the anomaly that SIC maps of the region can only partially depict (see Fig. 3).     prevailed that year. Atmospheric data (Fig 4a) in the form of wind speed derived from u and v components of wind velocity 160 vectors are presented as daily average (in blue) and maximum (in red) magnitude in the region of interest. Notably we can see the highest mean (9-10 Sep) at the start of the SIT anomaly and the highest maximum at the same time as the peak of the anomaly (17-18 Sep). Thus wind could have contributed to the formation of the 2018 "polynya thin ice" event. Interpreting the wind speed results shown in Fig. 5a as compared to the lower polynya area and thickness plots, we see that the 165 highest maximum (in red) and mean (in blue) wind speed magnitude both coincide with the 13 September polynya opening date. From the ASI SIC record (Fig. 5b), we can see both the similarities it shares with the SIT record (Fig. 5c) as well as clear differences that will be further discussed below. Important to note is that the blue line in both SIC and SIT records represents the area that is classified as open water. These lines are also present in the 2018 Fig. 4b and 4c but are consistently at 0 km 2 and therefore hidden because of the overlap with low SIC and low SIT lines.

175
The polynya maps in Fig. 2 and Fig. 3 are useful for accessing fine details of low SIT distributions as well as comparing SIT retrieval with ASI SIC (Fig. 3). By capturing the low sea ice thickness anomaly in 2018 and at the beginning of the 2017 polynya event in the SIT record we can infer that there were residual polynya-favourable effects that produced a forcing that was insufficient to open the polynya but sufficient to still impact the overlying sea ice. This is similar to the 1970s polynya cases, where the 1973 polynya resulted in a much larger iteration of the Weddell Sea Polynya visible from 1974 to 1976. Cheon

185
In order to analyse the time periods during which the polynya of 2017 (Fig. 5) and the sea ice anomaly of 2018 (Fig. 4) occurred, we view the respective time series. 2017 in Fig. 5c shows a progression of events in terms of SIT of how the polynya came to be. First and foremost we have a major regional ice thinning early August that peaks on the fourth of August much like the minor Weddell Sea Polynya of 2016 that also peaked on 4-5 August of that year (see Fig. A1 in the appendix). Looking at 5b we can see how much smaller the area affected by SIC variations is and how it is different in behaviour to the SIT time series.

190
Only the "below 80%" SIC shows some variability, however, not very correlated to the SIT time series. This is especially true during the brief period (6-12 Sep) leading up to the polynya, which is promising because it suggests a lack of low SIC-induced SIT values due to the SMOS-SMAP SIT retrieval full ice cover assumption. In total, compared to the 50 · 10 3 km 2 of below 100% SIC area, less than 50 cm thick ice spans over 300 · 10 3 km 2 of the region of interest. Following the period mentioned (6-12 Sep), we have the sudden peak (12-13 Sep) in both lower sea ice concentrations and thin sea ice. Based on Fig. 3, we see 195 that this, at first minor opening in sea ice, paved way to the Weddell Sea Polynya. From an oceanographic point of view this would imply heat exchange with the atmosphere which would cool the surface water layer and destabilize the water column.
This destabilization, further facilitated by the effect of the Taylor column, isolates the water mass above Maud Rise (Muench et al., 2001). The lack of stratification in the waters surrounding the Antarctic continent, would trigger convection cells able to bring up warm Deep Water from below. In Fig. 3 we can see the much larger scale effect this is having on SIT rather than 200 SIC and how peaks of low SIC and low SIT do not coincide in Fig. 5b and Fig. 5c, respectively. Instead, we see the low SIT area peak occurring 5-6 September following the 4 September peak in low SIC area which is what we expect considering how the convection cell would not simply cease immediately after the the smaller openings in ice freeze up; but rather its effects would be "felt" in the general location for days to come. With the ocean destabilized, coupled with heavy storm activity as can be seen in Fig. 5a Fig. 4 shows that 2018 is less anomalous than 2017 for the first one and half months until the sea ice anomaly begins to form on the 6 of September 2018. There is an initial thinning and occasional sporadic "below 80%" SIC events distributed throughout the period. Notably, the event on 24 August and 31 August, seen in Fig. 4b seem to suggest lead openings in thick pack ice as there is no thinning recorded in the SIT retrieval for those days. The sea ice anomaly itself, as can be seen in Fig. 4c, is very well defined in the SIT record and has a clear beginning and an end. Notable is that the two consecutive low SIT area peaks are 210 characterised by more extreme case of thinning during the first smaller peak reaching a prolonged period (7-13 September) of ice thinner than 20 cm followed by a much larger area of ice thinner than 50 cm (15)(16)(17)(18)(19)(20)(21). This anomaly follows a period of relatively strong mean and maximum wind speed from 3 August to 13 Sep towards the East and Southeast directions ( Fig. 4) that could imply that wind-driven ice advection influences the sea ice anomaly as any attempt at refreezing ice that has been broken apart by wind would require newly formed thin ice. Similarly, low SIC and strong winds would enhance heat loss 215 from the ocean and cause upwelling warm water, which would melt the ice from below. Fig. 7 depicts hourly wind conditions during the start of the sea ice anomaly on 7 September 2018: the strong westerly winds (blowing towards the East) common for this region occasionally show a more northerly component roughly where the sea ice anomaly began to form at the same time.
Through the comparison of our SIC data with ERA5 atmospheric data we can infer when wind can force the Weddell Sea forcing is a strong contributing factor especially towards the start of the polynya. Thereafter also oceanic upwelling of warm water due to the reduced stratification plays an increasing role.
Lastly, we use the SMOS SIT retrieval instead of the combined SMOS-SMAP to analyze years before 2015 (the year when SMAP was put into orbit) to make a consistent 11 year SIT time series over the months of July, August, September and October ( Fig. 1) to fully include the freezing periods of the relevant region over the years. Notably, the sea ice thinning of 2018 is by 230 are periods of near-100% SIC and low SIT as during pre-polynya periods. When the polynya is open, the SIT signal from the retrieval is unlikely to provide accurate ice thickness data due to large areas of open water influencing the signal. As mentioned before, low SIC affects the SIT record. Other sources of uncertainty can be flooded ice and slush caused by snow pushing 275 down the sea ice such that water floods from the sides or from below through cracks in the sea ice. However, we do not have indication that this happened here. Due to the potential uncertainties in this study the SIT record serves mainly as an indicator of anomalous sea ice activity rather than a means by which to quantify the exact degree of thinning in the region.
In 2018, a polynya-free year, SIT retrieval has shown that the beginning and end of a sea ice anomaly that, at its peak (18 Sep: <50 cm sea ice region with an area of 300 · 10 3 km 2 ), reached an estimated area larger than the United Kingdom. It is apparent 280 that the low SIT anomaly covered a much wider area than where low SIC (most likely minor lead openings) is recorded. As such, SMOS-SMAP SIT analysis is a method by which the Maud Rise region can be better monitored on a more frequent basis. This type of analysis, able to detect anomalous activity above Maud Rise with high temporal resolution, paves way to a better understanding of the underlying processes that not only drive the polynya but are in fact affecting the sea ice more often than previously thought possible. An extension of the 11-years SMOS time series is needed to better quantify the regularity 285 and how often such polynya-type ice anomaly events occur. As both SMOS and SMAP are science missions with no planned follow ups there is a chance that we will have a gap in the current L-band radiometry capability in space. However, with the future, operational Copernicus CIMR mission (planned launch 2028; https://cimr.eu/) some continuation of the SIT time series will be possible.
In conclusion, the classification of the Weddell Sea Polynya as a purely-open ocean polynya has been challenged and clear 290 links between wind speed magnitude and polynya conditioning have been found. As for SIT retrieval from L-band microwave radiometers like SMOS ad SMAP: it is an effective tool at monitoring sea ice conditions above Maud Rise and capable of collecting more substantial information than its SIC counterpart. Rather than substitute SIC retrieval though, the two should be used in conjunction with one another to aid the scientific understanding of the processes taking place and it should be added as yet another tool at trying to understand the unique and complex processes present in the Maud Rise region.

Appendix A A1 The 2016 Polynya Event
In Fig. A1 we show the 2016 polynya in the same format as the 2017 polynya and 2018 ice thinning anomaly.