Variability of Inflowing Current Into the Dotson Ice Shelf and Its Cause in the Amundsen Sea

The inflow of warm and salty Circumpolar Deep Water affects the melting of the ice shelf on the Amundsen Sea, a significant contributor to global sea level rise. Multi‐year mooring data (2014–2016 and 2018–2020) from the front of the Dotson Ice Shelf show the modified Circumpolar Deep Water layer was thicker during 2018–2020 than during 2014–2016. During 2014–2016, Ocean surface stress curl influenced the barotropic process and strengthened southward velocity, while during 2018–2020, it caused lift and downwelling of thermocline depth, increasing the impact of the baroclinic process in ocean circulation. The heat transport to the ice shelf during 2018–2020 (57.42 MW m−1) was half as much as it was during 2014–2016 (111.06 MW m−1) due to a weaker lower layer current. The difference in ocean circulation between two periods, caused by a difference in warm layer thickness, ultimately impacts the heat transport entering the ice shelf cavity.


Introduction
The rapid retreat of the West Antarctic ice sheet (WAIS) plays a critical role in global sea-level rise (Wåhlin et al., 2020).In particular, sea level rose about 6.5 mm  due to volume change in the WAIS, including the Amundsen Sea (AS) (Shepherd et al., 2018).Drained glaciers are buttressed by ice shelves adjoining the sea.Accelerating ice shelf thinning (Depoorter et al., 2013;Pritchard et al., 2012;Rignot et al., 2013) and ground line retreat (Scheuchl et al., 2016) reduces buttressing and makes the edge of the glacier sensitive to ocean-driven conditions (Jenkins et al., 2018).Therefore, to understand the melting process of ice shelves, it is essential to figure out the role of warm water flowing into ice cavities via ocean circulation.
The Dotson Ice Shelf (DIS), located southwest of the AS, buttresses the flow of the Kohler and Smith glaciers (Scheuchl et al., 2016).The DIS has recently experienced significant thinning and retreat.Between 1994 and 2012, the DIS thinned by 2.6 m yr 1 , which was more than 30% faster than the average of the Amundsen Sea sector (Paolo et al., 2015).Basal melting, a significant factor in ice shelf melt, is driven by warm and salty modified Circumpolar Deep Water (mCDW) flowing along the Dotson-Getz Trough (DGT) (Jenkins et al., 2018).Many scholars conducted various investigations to identify the inflow of warm water, its variability, and its mechanism near the DIS, including the DGT.Wahlin et al. (2013) and Dotto et al. (2020) revealed that the intrusion of warm and salty CDW into the DGT is influenced by barotropic undercurrent by the wind.Silvano et al. (2022) showed that the cyclonic wind over the AS is related to the baroclinic undercurrent strength at the shelf break affecting the CDW influx.Inflowed CDW is modified by mixing with surrounding water in the DGT as it reaches the DIS cavity along the eastern bottom layer of the trough (Ha et al., 2014).In front of DIS, Jenkins et al. (2018) argued that an increase in ocean temperature enhances the sensitivity of the melt rate and its relationship with heat transport to ice cavities and ice shelf melting.Yang et al. (2022) revealed how warm seawater capable of melting the ice shelf flows into the ice shelf base, and the melted water is discharged again.These studies present that the heat content and the heat input near the ice shelf from ocean circulation can directly affect ice shelf melting.
Many studies have been conducted to identify the variability of ocean circulation and its causes, which are closely related to melting ice shelves.In Pine Island Bay (PIB), the heat content change could affect the density gradient and the ocean circulation (Webber et al., 2017).Also, ice shelf retreat causes changes in ocean circulation (Yoon et al., 2022).In the DIS, Yang et al. (2022) showed that seasonal variation in ocean circulation under the influence of local atmospheric forcing governs heat transport to the ice shelf.Furthermore, the seasonal variation in heat transport into the ice shelf and meltwater outflow they identified suggests that a summertime survey is insufficient to estimate the melting of the ice shelf.Therefore, it is essential to understand the seasonality of the mCDW circulation in front of the ice shelf to improve our understanding of the oceanic variability responsible for the ice shelf-melting process.The mCDW volume observed in front of the DIS showed clear interannual variability; Kim et al. (2021) argued that this was related to Ekman pumping affected by atmospheric forcing.This suggested that the comprehensive analysis using long-term ocean and atmospheric data is essential to estimate the variability of the mCDW in front of the ice shelf at various time scales.
In this study, we analyzed two sets of multi-year mooring data (2014-2016 and 2018-2020) obtained from the east of the DIS, the main pathway to the mCDW.We identified the seasonal variability of southward flow into the ice shelf caused by ocean and atmospheric conditions differences and investigated the mechanism.

Data Collection
We analyzed hydrographic data obtained from four extensive oceanographic surveys (2014,2016,2018,2020) in front of the DIS (Figure 1a).Temperature, salinity, and dissolved oxygen (DO) profiles were sampled using a Seabird Scientific SBE 911+ conductivity-temperature-depth (CTD).The continuous properties of the water masses and currents were obtained from two multi-year mooring observations at 2-3 km from the DIS.The first mooring was deployed on 8 January 2014, at east of the DIS and was recovered on 19 January 2016 (Figure 1a).The second mooring, located about 1.7 km west of the previous mooring position, was deployed on 26 January 2018, and recovered on 27 January 2020.Current velocities were observed using upward-or downward-looking Teledyne RD Instruments' Acoustic Doppler Current Profiler (ADCP) of various frequencies (150 and 300 kHz).Temperature and salinity were measured using the Sea-Bird Electronics' (SBE) 37-SM and 37-SMP-ODO MicroCAT sensors.

Calculation of Ocean Surface Stress Curl
In order to evaluate the factors influencing the variability of seawater circulation in front of the ice shelf, ocean surface stress curl (OSSC) was calculated from January 2014 to December 2019.The stress on the ocean surface in the Southern Ocean, which is sometimes covered by sea ice, has been calculated from wind stress and sea ice motion (Kim et al., 2017(Kim et al., , 2021;;Yang et al., 2022) OSSC (τ c ) was calculated as follows: The ocean surface stress (τ x o and τ y o ) is given by sea ice and wind stress.We used the sea ice concentration (Advanced Microwave Scanning Radiometer-2; Spreen et al., 2008), sea ice velocity (Polar Pathfinder Daily 25 km EASE-Grid Sea Ice Motion Vector, version 4 data; Tschudi et al., 2010), sea ice thickness (ICESat sea ice thickness; Kurtz & Markus, 2012;Markus et al., 2011), and wind velocity vector 10 m above sea surface (Antarctic Mesoscale Prediction System; Bumbaco et al., 2014).To calculate OSSC, each data was interpolated at intervals of 0.125°latitude and 0.25°longitude and averaged as daily data.

Observed Warm Layer Thickness and Current
The vertical structure of temperature investigated from four times CTD observations in front of the DIS was distinctly different throughout the observation period (Figure 1b).The thermocline depth separating the mCDW (below 600 m) and WW (100-400 m) showed a significant difference between the summers of 2014 and 2020.The depth of the 0.5°C temperature became shallower from ∼690 m in 2014 to 430 m in 2020.This interannual variation in the thermocline depth caused a dramatic change in seawater temperature in the middle layer (400-600 m).In 2014 and 2016, the average temperatures in the middle layer were 0.96 and 0.98°C, respectively.However, they increased to 0.56 and 0.30°C in 2018 and 2020, respectively.The increase in the volume of warm and salty water in the lower layer in 2018 and 2020 was also observed in time-series variation in the vertical salinity structure (e.g., The average depth of the 34.35 isohaline during 2018-2020:574 m, during 2014-2016:605 m) (Figure 1c).It suggested that the mCDW layer became thicker during 2018-2020 compared to 2014-2016, not only during the summer-time but throughout the entire observation period.
The strength and seasonal cycle of the southward flow also showed notable dissimilarity during the two mooring periods (Figures 2a and 2b).The average southward flows during 2018-2020 were 1.64 cm s 1 and 6.67 cm s 1 in the upper (400 m) and lower (680 m) layers, respectively, which decreased compared to the average southward flow during 2014-2016 (upper layer 5.11 cm s 1 , lower layer 17.21 cm s 1 ).During 2014-2016, a relatively distinct seasonal cycle appeared at all layers, with southward flow strengthening in January and weakening in May.During 2018-2020, the weakest southward flow occurred in May, similar to earlier observations.However, the strongest southward flow appeared in October and November, and the peak was not distinct compared to the previous observations.The time-series meridional current velocity was separated into depth-averaged velocity and its residual current to examine the interannual variability of the southward velocity and the seasonal cycle (Figures 2c and 2d).During 2014-2016, the depth-averaged southward flow was strongest at 11.25 cm s 1 in December 2015 and weakest at 0.24 cm s 1 in May 2015.On the contrary, minimum and maximum southward currents of 1.83 and 8.01 cm s 1 were observed in May and October 2018, respectively, during the 2018-2020 mooring period.The standard deviation of depth-averaged southward flow was ±4.25 cm s 1 in this period, which was weaker than that during 2014-2016 (±5.16 cm s 1 ).The time-series variation in the residual current showed a significant difference between the two mooring periods (Figures 2c and 2d).During 2014-2016, the residual current in southward flow between depths of 400-600 m was a relatively gentle temporal variation compared to the depth-averaged southward flow.However, during 2018-2020, the temporal fluctuation ranges of the southward flow's vertical gradient (400-600 m) extended, and its standard deviation increased by about 1.7 times compared to 2014-2016.Based on these results, it can be inferred that the influence of the baroclinic component increased during 2018-2020, indicating that the baroclinic process plays an essential role in the temporal variation of southward flow, along with the barotropic process.

Variation of Density and Residual Current
Changes in seawater density structure induce a baroclinic pressure gradient transferred to the lower layer, which modifies the vertical profile of the current velocity, along with the barotropic pressure gradient caused by the difference in sea surface elevation (Stewart, 2008).In this study, the locations of the eastern and western mooring stations are far from each other, there is a limit to the evaluate the effect of the baroclinic component between two mooring stations.Therefore, we intended to evaluate the baroclinic effect from the temporal variability of density data during both mooring periods.
The densities at each layer showed seasonal variation in the 2014-2016 and 2018-2020 mooring data (Figures 2e  and 2f), but seasonality decreased with depth.In the summers of 2018 and 2020, the thermocline depth was lifted, and the mCDW layer was thicker than that in the summers of 2014 and 2016 (Figure 1b).The signal of seasonal variation was amplified during 2018-2020 compared to that during 2014-2016 period in the upper layer (Figure 2f).These results suggest that the effect of the baroclinic process on meridional current velocity in the mid-depth layer (400-600 m) was possibly more significant during 2018-2020.During 2014-2016, the range of the meridional current variation at 400 m depth was ∼12.6 cm s 1 (May 2015:3.2cm s 1 , December: 9.4 cm s 1 ) and the range of the depth-averaged southward flow variation was 11.5 cm s 1 (Figures 2a and 2c).Thus, the barotropic process might have led to variability in the southward flow during 2014-2016.On the contrary, the range of the meridional current variation at 400 m depth during 2018-2020 (May 2018:5.7 cm s 1 , October: 6.8 cm s 1 ) was the same as that during 2014-2016, but the range of the depth-averaged southward flow variation decreased to 9.8 cm s 1 during 2018-2020 (Figures 2b and 2d).At a depth of 550 m, the weak northward flow was 1.7 cm s 1 in April 2019, when density decreased sharply at the upper layer (Figure 2f), exhibiting a difference with the variability of the meridional flow at the upper layer.
In Figure 3a, although the distinct decreasing trend of depth-averaged density in April 2019 was not reflected in the fluctuation of residual current at 400 m, except for this period, the intra-seasonal variation of residual current demonstrated sound agreement with average density.The 31-day moving average density and residual current had a significant correlation of r = 0.26 (at zero lag) in a 99% confidence interval (Figure S1 in Supporting Information S1).Assuming that there was no temporal variation in density at the DIS center, an increase in density on the eastern side in winter and spring promoted a southward flow.In contrast, a sharp decrease in density in the upper layer during autumn can increase the northward flow in the mid-depth layer and countervail the southward flow in the lower layer.

Effects of OSSC on Density Variability
In polynya, wind and seasonal changes in the contraction/expansion of sea ice cause spatial imbalances of stress on the ocean surface, leading to complex influences on ocean currents.The positive (negative) OSSC at the eastern margin of the polynya can generate convergence (divergence) of surface water and accelerate southward (northward) flows by barotropic processes (Yang et al., 2022).In addition, positive (negative) OSSC is accompanied by Ekman downwelling (upwelling), causing a decrease (increase) in seawater density at the pycnocline (Kim et al., 2017).
In a previous study using 2014-2016 mooring data, a strong positive OSSC in the summer accelerated the southward flow throughout the water column (Yang et al., 2022).During 2018-2020, the maximum OSSC appeared in autumn (April 2018) (Figure 3c) and the OSSC was greater than that during 2014-2016 (Figures 3b  and 3c) but, the depth-averaged southward flow was weakened compared to 2014-2016 (Figures 2c and 2d).It suggests that the OSSC-induced barotropic processes may not dominate the southward flow during 2018-2020.The variability of seawater density in the mid-depth layer (300-500 m) was affected by the OSSC.Downwelling (upwelling) accompanying positive (negative) OSSC decreased (increases) the density, particularly in the middepth layer.From autumn 2018, density at a depth of 400 m gradually increased as the positive OSSC decreased.In spring (October-November), the OSSC approached zero and the density was maintained at 27.52 kg m 3 .Subsequently, with the increase in OSSC, the density decreased accordingly, and in the autumn of 2019 (March-May), it was at least 27.47 kg m 3 .The spatial distribution of OSSC (Figure S2 in Supporting Information S1) showed that in the spring of 2018, the OSSC on the eastern flank of the DIS was weakly negative and the zonal gradient was minimal.By contrast, in the spring of 2019, the OSSC increased significantly at the eastern flank, and the zonal gradient was steep.

10.1029/2023GL105404
The seasonality of the OSSC for the 2018-2020 period differed from that of the 2014-2016 period.On the eastern slope of the DIS, the OSSC reached a maximum during summer during 2014-2016; however, during 2018-2020, a maximum OSSC appeared during autumn (Figures 3b and 3c).In addition, during this period (autumn 2019), the southward flow slowed over the entire water column, and the residual current above a depth of 600 m was northward.In the spring of 2018 (October-November), when the OSSC was almost spatially homogeneous, the southward flow accelerated, and the residual current was southward below 500 m depth.During 2014-2016, the positive OSSC in spring and summer caused a southward flow by barotropic processes and led to a density increase due to a positive density gradient in the meridional direction.In contrast to the earlier mooring period, during 2018-2020, the positive OSSC accompanied by downwelling caused a density decrease in the mid-depth layer and led to northward flows via the baroclinic process.

Discussion and Conclusion
From 2014 to 2020, two multi-year moorings and four hydrographic observations (Figures 1b and 1c) on the eastern side of the DIS indicated a distinct interannual variation in warm layer thickness and mCDW volume.In 2014, the mCDW volume decreased and shifted remarkably to the eastern slope, constructing a strong zonal density gradient (Figure 3d) and a robust southward flow by baroclinic processes with the barotropic process in the lower layer.Additionally, the thin dense mCDW layer and the smoothness of the vertical gradient of the pycnocline in the mid-depth layer reduced the effect of the baroclinic process on the temporal variation of meridional flow induced by up/downwelling of the isopycnal derived by OSSC.Thus, the temporal fluctuation range of the residual current was relatively narrow compared with that of the depth-averaged flow.In 2018, the mCDW was extensively distributed below 400 m in front of the DIS, and the isopycnal became shallower toward the west in contrast to 2014 (Figures 3d and 3e).The increased volume of dense mCDW and intense vertical inclination of the pycnocline augmented the effect of the baroclinic process on meridional flows.As such, changes in the density structure may lead to different responses from the ocean conditions to the OSSC.
The average depths of the pycnocline for 2014-2016 and 2018-2020 periods from the mooring data were 500 and 400 m, respectively (Figures 3b and 3c).In the cold spell (2014-2016), a positive OSSC in the summer generated a strong southward flow by the barotropic process.Additionally, the barotropic southward flow to the DIS front, with a relatively low density owing to downwelling, caused an increase in density.In winter, the sea surface was covered by sea ice, the OSSC was weakened, and the southward flow and density decreased at the averaged pycnocline depth (Yang et al., 2022).In contrast, during the warm spell (2018-2020), OSSC was close to zero and density at the averaged pycnocline depth (400 m) was maximum in spring, the southward flow was dominant due to the positive pressure gradient in the zonal direction (Figure 3c).Moreover, dominant southward flow increased with depth owing to the intrusion of the mCDW along the eastern flank.Increased the OSSC in autumn caused downwelling, the pycnocline deepened, and density decreased at the averaged pycnocline depth.As a result, the baroclinic pressure gradient in the zonal direction reversed and changed to a northward flow at the pycnocline depth.However, below the pycnocline, where the vertical gradient of density was relatively weak, the downwelling effect did not significantly decrease the density and maintain the southward flow.
In polynya, changes in surface density due to the melting and forming of sea ice and the cooling and heating of the ocean surface layer by heat exchange with the atmosphere affect thermocline depth.Therefore, buoyancy flux causes the thermocline to rise/fall and increase/decrease its density (Webber et al., 2017), which is one of the critical factors contributing to ice shelf melt (Moorman et al., 2023;St-Laurent, et al., 2015).The buoyancy flux, estimated from air-sea heat flux and freshwater flux obtained from SOSE (Southern Ocean State Estimate; Mazloff et al., 2010) data, was positive in autumn to winter (i.e., increases buoyancy flux from ocean to air), which coincided with the density decrease in April (Figure S3 in Supporting Information S1).This can be interpreted as the thermocline deepening and density decreasing as it the less buoyant at the sea surface due to the formation of sea ice and cooling of the ocean surface.However, in winter 2018-2020, when the buoyancy flux is positive, the density shows an increase (July-October).Although seasonal fluctuations in buoyancy affect changes in the vertical density structure, this study focused on the ocean density changes caused by OSSC.
The seasonal cycle of OSSC showed a difference during the two mooring periods (Figures 3b and 3c), which may affect ocean circulation with changes in ocean conditions not independent of OSSC-induced interannual variability of Ekman pumping (Kim et al., 2021) the seasonal cycle of OSSC may change owing to the combined action of sea ice concentration and wind speed.The strong positive OSSC in April 2018 and March 2019 was due to decreased sea ice concentration and stronger winds (Figure S4 in Supporting Information S1).In addition, the increase in wind speed in February 2014 and January 2015 caused a strong positive OSSC.These results indicate that the local atmospheric circulation represented by OSSC, which determines the ocean conditions in the DIS, is associated with larger-scale atmospheric variability.
Heat transfer to the ice shelf, determined by the ocean's current velocity and heat content, shows significant differences between the two periods (Figure S5 in Supporting Information S1).The average depth-integrated heat transport in 2014-2016 is estimated to be approximately 111.06 MW m 1 , approximately twice that of 2018-2020 (57.42 MW m 1 ).Although the average heat content in 2018-2020 was 5.59 MJ m 3 , higher than that in 2014-2016 (4.12 MJ m 3 ), the average southward flow was 3.34 cm s 1 , only half of the 2014-2016 (7.46 cm s 1 ).The seasonal mean ocean heat content in 2018-2020 was highest in winter at 6.32 MJ m 3 , when the southward current was 4.20 cm s 1 , and the heat transfer amount was 82.15 MWm 1 .Winter heat transport was higher than the overall average in 2018-2020 but was about 1.2 times less than the winter (101.72 MWm 1 ) in 2014-2016.It can be inferred that more heat entered the ice shelf cavities in 2014-2016 when OSSC affected the barotropic component southward velocity variation and heat content (Yang et al., 2022) and evident that the current significantly contributes to the heat inflow to the ice shelf.According to a recent study (Wåhlin et al., 2020), the barotropic current may be partially blocked by the ice wall and thus have a small impact on heat transfer processes under the ice shelf.However, the meridional density gradient formed by the strong downwelling in front of the ice shelf induces a change in the zonal gradient of isohaline and isotherm and variability of the baroclinic southward flow caused by the barotropic southward current.As evidence, heat transfer to the ice shelf driven by the barotropic current in 2014-2016 matched well the variability of meltwater discharge in the western flank of the ice shelf (Yang et al., 2022).Meanwhile, although the heat content of the ocean at the ice shelf front increased significantly in 2018 compared to 2014 (Kim et al., 2021), the basal melt rate from altimetry on the DIS showed that the melt rate did not significantly increase in 2018 compared to 2014 (Adusumilli et al., 2020).
In this study, we found a significant change in mCDW volume between the two mooring periods on the eastern flank of the DIS.During the cold spell (2014-2016), the southward flow was strengthened by the barotropic process induced by the OSSC (Yang et al., 2022).By contrast, when the volume of the mCDW increased during the warm spell (2018-2020), the pycnocline became shallow, and OSSC-induced up/downwelling caused significant density changes in the middle layer, increasing the effect of the baroclinic process on local seawater circulation.We found that, despite the higher heat content, less heat flowed into the ice shelf during the warm spell when the southward current was weak.Consequently, changes in ocean conditions cause changes in the warm water intrusion into the ice shelf in response to the atmosphere.However, this study did not suggest the effect of ice shelf melting by the intrusion of warm water.Nevertheless, understanding ocean circulation through ocean conditions and its mechanisms is essential for predicting the future variability of mCDW flowing into the ice shelf and understanding long-term ice shelf melting trends.

•
The modified circumpolar deep water layer was thicker in 2018-2020 than that during 2014-2016 • The baroclinic effect plays a more important role in the variability of the current entering the ice shelf during 2018-2020 • Differences in the seasonal cycle of the ocean surface stress curl can affect ocean circulation by changing ocean conditions Supporting Information: Supporting Information may be found in the online version of this article.

Figure 1 .
Figure 1.Study area, station locations and time series of salinity and potential temperature at eastern flank of the Dotson Ice Shelf (DIS) mooring station.(a) Coastal line (Nitsche et al., 2007), bathymetry, and observation stations near the DIS.The yellow and green stars show mooring in 2014 and 2018, respectively.The color-coded symbols represent CTD stations.(b) Vertical profile of potential temperature of the CTD observed at 2-year intervals since the summer of 2014 at the mooring position (station in the red box of Figure a).(c) Vertical and temporal distribution of salinity and potential temperature in 2014-2016 (left) and 2018-2020 (right).