Generation of the Amundsen Sea Low by Antarctic Orography

The Amundsen Sea Low (ASL) is a distinctive feature of the Southern Hemisphere high latitude atmospheric circulation, regulating regional Antarctic climate, meridional heat transport, ocean circulation, and sea‐ice in the Amundsen‐Bellingshausen Seas. Most previous research on the ASL has focused on its variability with only a few studies attempting to understand why the climatological ASL exists. These studies have proposed different hypotheses to explain the presence of the ASL, however, a clear understanding of the mechanisms responsible for the generation of the ASL remains uncertain. Here we use an atmospheric general circulation model to show that the ASL is a consequence of the interaction between Antarctic topography and the westerly wind jet, with negligible influence from low‐latitude teleconnections. A nonrotating fluid flow simulation further suggests that the ASL can be explained by flow separation resulting from the interaction of westerly winds with the topography of Antarctica.

(and therefore ASL) are projected to be dominated by an increase in greenhouse gases (Raphael et al., 2016), with recovering ozone playing a lesser role (Goyal et al., 2021a;Thompson et al., 2011).
Given the importance of the ASL for changes in the West Antarctic region, including its impacts on the West Antarctic Ice Sheet, most previous work has focused on the variability of the ASL pattern and on its past and projected changes as well as their implications for the West Antarctic climate. However, only a small number of studies have focused on what gives rise to the climatological ASL in the first place. Baines and Fraedrich (1989) were the first to consider this problem, conducting rotating tank experiments with a physical scaled-down model of Antarctica, including realistic orographic features placed in homogeneous (fresh water) and stably stratified fluid. Their results suggested that the ASL might be generated by flow separation because of an interaction between the westerly jet in the southern high latitudes and Antarctic orography (refer to Figure 1a here showing Antarctic orography). However, a quantitative assessment was not possible with their experimental set-up, with the model results based on visual inspection of the flow. Subsequently, Walsh et al. (2000), using a simple 2-layer baroclinic model of the atmosphere, found a weak ASL pattern in simulations with a flat Antarctic land-mass (i.e., no orography over Antarctica). They suggested that even though the baroclinicity created because of the presence of Antarctic orography strengthens the ASL pattern, the climatological low pressure might be generated either by the shape of the Antarctic continent or because of synoptic activity from the SH mid-latitudes in the absence of Antarctic topography. Both these studies were carried out using a simplified representation of the Antarctic climate system, omitting important physics such as katabatic winds and with limited representation of Antarctic orography and stationary waves in the atmosphere. More recently, Fogt et al. (2012) used multiple reanalysis products to suggest that the ASL might be generated and maintained by a net influx of cyclonic storms reaching the ASL region. They proposed that the ASL is present in the climatological mean because of the high density of cyclonic storms from both lower latitudes and from the baroclinic zone around Antarctica ending up in the Amundsen-Bellingshausen Seas. However, Raphael et al. (2016) found that the seasonal location of the ASL does not always coincide with the center of maximum storm activity. Several other recent studies (e.g., Yiu & Maycock, 2019, 2020 have linked the existence of the ASL to zonal asymmetries in tropical sea surface temperatures (SSTs) via planetary scale Rossby wave activity from the tropics to the southern high latitudes.
Given these competing hypotheses, a clear understanding of the mechanisms responsible for the generation of the ASL remains uncertain. In this study, we use an atmospheric general circulation model supplemented with a nonrotating fluid dynamics simulation, to examine the mechanisms responsible for the generation of the ASL.

Climate Model
We use the National Center for Atmospheric Research (NCAR) Community Earth System Model (CESM v1.2.2) which was part of the Coupled Model Intercomparison Project 5 (CMIP5). The atmospheric component of the model is the Community Atmosphere Model Version 4 (CAM4, Neale et al., 2010) and is coupled to the Community Land Model Version 4 (CLM4, Oleson et al., 2010). CAM4 is run with a 1.875 × 2.5° finite volume grid with 26 hybrid sigma levels. CAM4 is forced with monthly averaged interannually varying SSTs and sea-ice. SST and sea-ice fields are obtained from a fully coupled run of NCAR CESM integrated in a preindustrial control configuration.
A Computational Fluid Dynamics (CFD) analysis in a nonrotating frame of reference is also carried out using the ANSYS Fluent Academic Research v2019 R3 fluid dynamics simulation software. A smoothed three-dimensional (3-D) model of Antarctica on a Mercator projection is constructed using the ANSYS SpaceClaim software over a coarse tetrahedron mesh for numerical discretization. The height of Antarctica is kept constant at all longitudes. This idealized 3-D model of Antarctica is placed in a westerly wind flow of 10 m/s, matching the typical climatological westerly wind velocity around Antarctica. In the real world, easterlies driven by the effects of gravity (katabatic winds) and the pressure gradient force are present near the edge of the Antarctic orography. These polar easterlies are not present in this simulation, but the model configuration is reasonable because the effect of these katabatic winds on the localized circulation features around Antarctica is minimal compared to the westerlies (Baines & Fraedrich, 1989). Moreover, in the region west of the Ross Sea and near the Antarctic Peninsula, these easterlies are only present for a few hundred meters above the surface, with the climatological winds aloft being the prevailing westerlies.

Model Simulations
Model simulations with the NCAR CESM were carried out to understand the generation of the ASL pattern. For all simulations, the model uses a preindustrial control configuration (i.e., preindustrial levels of greenhouse gases, aerosols, and other forcing). A control simulation (CTRL, Figure 1b) is carried out with interannual varying monthly SSTs and sea-ice. The control simulation has a realistic land-sea configuration as well as realistic orography over all landmasses including Antarctica. A second simulation (NO TOPO , Figure 1c) is carried out in which the topography is removed everywhere (i.e., flat land at sea level). Another simulation (ANT TOPO , Figure 1d) is then integrated in which only the topography over Antarctica is added to the NO TOPO simulation (refer to Figure 1a showing Antarctic orography). In a third experiment (EAST-ANT TOPO , Figure 1e), topography is only present over East Antarctica (i.e., topography over the West GOYAL ET AL. 10.1029/2020GL091487 3 of 9 Antarctic including the Antarctic Peninsula is flattened to sea level). Finally, in the ROTATED ANT (Figure 1f) simulation, Antarctica is rotated by 100 degrees toward the west, while all other land-ocean features north of Antarctica remain at the same location as in the CTRL simulation. SST and sea-ice fields south of 62°S are also rotated by 100 degrees west in the ROTATED ANT simulation. To remove any sudden discontinuities in SST and sea-ice fields in the ROTATED ANT simulation, both SST and sea-ice fields are smoothed using linear interpolation in the buffer region between 57° and 62°S. In all simulations where topography has been removed, surface properties remain unchanged; in particular, all Antarctic surface parameters including land-ice properties and vegetation are retained from the CTRL simulation in all experiments considered. All simulations are integrated for 120 years. The first 20 years are discarded as a spin-up period, and the remaining 100 years are used for the analyses presented here.

Data and Methods
Monthly MSLP and surface (at 60-m elevation) zonal winds from the model simulations are analyzed in this study. Monthly averaged MSLP from the European Center for Medium Range Weather Forecasts (ECMWF) Reanalysis (ERA-Interim; Dee et al., 2011) from 1979 to 2016 is also used for model evaluation. The ASL index is defined as the minimum MSLP in the South Pacific sector of the Southern Ocean (between 60°-75°S and 180°-310°E; after Turner et al., 2013). For the ROTATED ANT simulation, ASL is calculated as minimum MSLP between 60°-75°S and 80°-210°E. As the zonal MSLP in simulations with differing topography and/ or land-sea configurations is found to be different, a relative ASL index is used. The relative ASL index is defined as the minimum MSLP between 60°-75°S and 180°-310°E after the zonal mean MSLP has been removed.

Climatological Mean ASL
For the purposes of model validation, the simulated ASL from CTRL is compared with the ERA-Interim reanalysis (Figure 2a and 2b). The model is able to capture the magnitude, location, and spatial extent of the ASL with reasonable fidelity. The simulated climatological mean ASL is at approximately the same location as the ERA-Interim reanalysis, however, the simulated ASL in CTRL is deeper when compared to ERA-Interim ( Figure 2). The simulated zonal mean MSLP around Antarctica in CTRL is also lower compared to the reanalysis. In fact, after removing the zonal mean, the relative ASL is actually deeper in the ERA-Interim reanalysis as compared to CTRL (Figure 2c and 2d). Because the modeled absolute ASL is deeper relative to the reanalysis, this can lead to model biases in Antarctic sea-ice and winds in the Amundsen-Bellingshausen and Ross sea sectors (Coggins & McDonald, 2015). The model captures the seasonal timing in the ASL depth, with a deeper ASL during spring and a weaker ASL during summer, although the simulated seasonality is slightly too strong ( Figure 2e). The model is also able to capture the seasonal variation in the meridional and zonal location of the ASL, with the ASL being poleward and eastward in winter and equatorward and westward (i.e., closer to the Antarctic Peninsula) during summer (Figure 2f and 2g).
To examine the role of orography in generating the ASL pattern, the NO TOPO simulation (Figure 1c) is carried out in which all landmasses are set to a fixed flat sea level elevation in the model (i.e., the topography is removed everywhere). In this simulation, no climatological low-pressure center is found in the Amundsen Sea region (Figure 1c), suggesting a critical role played by topography in generating the ASL pattern. The zonal wave 1 pattern present in the SH mid-latitudes found in the CTRL simulation (Figure 1b) shifts poleward in the NO TOPO simulation because of the absence of any topographic barrier provided by Antarctica, and therefore a clear zonal wave 1 pattern is now present around Antarctica with a high pressure in the Amundsen Sea region (Figure 1c). This result contrasts with the findings of Walsh et al. (2000) who found an ASL to be present, albeit much weaker, in the absence of Antarctic orography using a simple 2-layer baroclinic model.
Next, we examine the role played by orography over Antarctica alone in generating the ASL pattern. To do this, another simulation (ANT TOPO ) is integrated in which only topography over Antarctica is added to the NO TOPO simulation (Figure 1d). In this experiment a clear low-pressure center is now present in the Amundsen Sea region (Figure 1d) with an ASL magnitude almost as deep as in CTRL (10.8 hPa in ANT TOPO compared to 11.2 hPa in CTRL). The location of the ASL is also similar to CTRL (76.7°S, 201°E in ANT TOPO and 76.7°S, 202°E in CTRL). The climatological pressure field around Antarctica is marginally weaker in the ANT TOPO simulation as compared to CTRL, however the strength and location of the ASL relative to the ambient large-scale pressure field are comparable (e.g., compare Figures 1b and 1d). Clearly the presence of Antarctic topography can largely explain the presence of the ASL pattern.
Next, to examine the factors controlling the climatological location of the ASL, another simulation (EAST-ANT TOPO ) is integrated in which topography over western Antarctica (including over the West Antarctic Ice Sheet and the Antarctic Peninsula) is removed (Figure 1e (b) shows the climatological mean MSLP in ERA-interim and the CTRL simulation, respectively. Panels (c) and (d) represent the climatological mean MSLP respectively in ERA-Interim and the CTRL simulation when the zonal mean has been subtracted. Panels (e), (f), and (g) represent the monthly climatology in ASL depth, meridional location and zonal location respectively in ERA-interim and the CTRL simulation. Shading in panels (e), (f), and (g) represents interquartile range based on the year-to-year variability in the ASL depth, meridional location, and zonal location, respectively. ASL, Amundsen Sea Low; MSLP, mean sea level pressure; CTRL, control. remained unchanged. The climatological location of the ASL in this simulation shifts eastward toward the Antarctic Peninsula and the ASL is no longer trapped in the Amundsen-Bellingshausen Seas region as found in CTRL (Figure 1e). This clearly shows that the mountains in the western part of Antarctica constrain the low-pressure center in the Amundsen Sea region.
Zonal variability in the tropical heating can create stationary waves in the SH high latitudes (Inatsu & Hoskins, 2004;Trenberth et al., 1998). Several past studies have demonstrated a strong influence of the tropics in generating ASL variability (Clem et al., 2017;Pope et al., 2017;Raphael et al., 2016). Zonal variability in tropical heating has strong impacts on the ASL via upper tropospheric planetary Rossby waves which move poleward and eastward from their tropical source region (Inatsu & Hoskins, 2004). These waves are present in the climatological mean and also show strong variability at monthly to interannual timescales. The climatological mean ASL is located southward of the crest of the climatological Zonal Wave 1 in the SH extratropics, which is known to be strongly influenced by deep convection over the Indo-Pacific warm pool in the tropics (Peña-Ortiz et al., 2019). As such, zonal asymmetries related to tropical heating may be expected to play a role in setting the characteristics of the climatological mean ASL. To test this possibility, another simulation (ROTATED ANT ) is carried out in which Antarctica is rotated westward by 100 degrees while all other landmasses are kept at the same location as in CTRL (refer to Figure 1f and methods for details). Even though the Amundsen-Bellingshausen Seas are now away from the climatological zonal wave 1 crest, a clear ASL pattern is still found in this simulation (Figure 1f). Furthermore, there is no change in the depth of the ASL in the ROTATED ANT simulation compared to CTRL (11.2 hPa in both ROTATED ANT and CTRL). There is also almost no change in the zonal location relative to the Antarctic continent; that is, it is now shifted westward by 100 degrees similar to the 100 degrees westward shift of the Antarctic landmass in this simulation (101.5°E in ROTATED ANT and 202°E in CTRL). Even though there is a northward shift in the meridional location of the ASL in the ROTATED ANT simulation as compared to CTRL, its magnitude is rather small (74.9°S in ROTATED ANT and 76.7°S in CTRL). These results suggest that tropical teleconnections have little role in setting the climatological mean ASL.

Mechanism
We next examine whether the climatological ASL is a consequence of the large-scale surface westerlies, for example, as a result of flow separation due to orographic features over Antarctica. To help understand this, we examine how variability in the strength of the large-scale mid-latitude westerly winds affects the ASL. Care must be taken while interpreting these results, as these variables are not independent of each other. Moreover, the relationship might also be influenced by the large-scale variability related to the SAM, which manifests as a seesaw in the pressure field between SH mid-latitudes and Antarctica. In this analysis we find that there is a large-scale relationship between the strength of the surface westerlies and the ASL beyond the local effect of the ASL (Figure 3a). In particular, a significant correlation of −0.59 is found between the strength of the large-scale mid-latitude westerly wind anomalies and the ASL index anomaly, with a deeper ASL associated with stronger surface westerlies (Figure 3a) suggesting a persistent relationship between the interannual variability in the mid-latitude westerly winds and the ASL. Moreover, the seasonal variability in the depth of the ASL is also strongly related to the seasonal variability in the strength of the mid-latitude westerlies (Figure 3b). In particular, the depth of the climatological ASL shows a maximum during austral spring consistent with the stronger mid-latitude surface westerlies during this season, and it exhibits a minimum during austral fall consistent with weaker westerly winds during fall (Figure 3b).
The previous analysis suggests that stronger westerlies may generate a stronger ASL. Following on from the previous analysis, we next test whether the ASL-like clockwise circulation can be created by flow separation due to the interaction of westerly winds with the topography over Antarctica. To do this, we undertake an idealized experiment in a nonrotating frame of reference where a westerly wind (with realistic strength) is introduced at the western boundary, to the west of the Ross Sea, and allowed to interact with the Antarctic coastline. The 3-D model of Antarctica has a smoothed yet realistic shape of the Antarctic continent. In this model, there is no variation in wind strength or topography with height (i.e., equivalent to a 2-D barotropic model). The model configuration is reasonable given that the circulation at polar latitudes is equivalent barotropic . The model is placed in an idealized westerly flow of 10 m/sec (equivalent to the climatological zonal mean surface westerly wind strength near the Antarctic margin) and 50 ensembles are carried out with each ensemble comprising of 200 simulation time-steps. The simulation clearly shows that a clockwise circulation is generated in the lee of Victoria Land due to the interaction between the westerly winds and Antarctic topography (Figure 4). The westerly flow adjacent to the continent separates from the coast as it reaches the Ross Sea embayment. This gives rise to a cyclonic circulation that extends to the Antarctic Peninsula. This flow separation can be understood as similar to a flow around a solid body; for example, a sphere or an aircraft wing, which forms lee side eddies because of separation of the flow from the surface. The clockwise circulation is contained in the Amundsen-Bellingshausen Seas region by the topography in the West Antarctic region including the Antarctic Peninsula.
GOYAL ET AL.

Conclusions
Multiple hypotheses have been suggested in the past to explain the generation of the climatological mean ASL in the southern high latitudes. In this study, we put these hypotheses to test and then further examined the mechanisms responsible for creating the climatological ASL using an atmospheric general circulation model as well as a nonrotating fluid dynamics simulator. We show that the ASL is generated because of the presence of Antarctic topography interacting with the westerly wind jet and that it is trapped in the Amundsen-Bellingshausen Seas because of the presence of topography in the west Antarctic region. We also show that teleconnections from the tropics play little role in generating the climatological ASL pattern. A simulation carried out using a fluid dynamics simulator further suggests that an ASL-like clockwise circulation can be generated by flow separation using an idealized Antarctic landmass in the presence of westerly airflow, without the effects of rotation.
Our results show that the ASL is a consequence of the interaction between westerly airflow with the topography of Antarctica, however, easterlies are present close to the Antarctic margin, which are generated from downslope katabatic winds. These katabatic winds may also have a role in controlling the depth and location of the climatological mean ASL. The katabatic winds are stronger in the Ross Sea sector of Antarctica (Parish & Cassano, 2003) and even though they may play a role in modulating the climatological mean ASL, it is difficult to disentangle the effects of katabatic winds from the effects of flow separation by Antarctic topography in a climate model. An attempt to do this was made by Parish and Cassano (2003), who tried to disentangle the effects of katabatic winds from the effects of Antarctic topography by running climate model simulations without explicit longwave radiation. Although they were able to suppress katabatic winds in their model simulations, they however found that the low-level winds resulting from the pressure gradient force resembled katabatic winds, with directions being tied to the underlying terrain. Even with this limitation of not being able to disentangle these two effects, our results still strongly suggest that the ASL is a consequence of the Antarctic topography alone, as the katabatic winds themselves are present only because of the existence of elevated Antarctic topography. Despite the generation of the ASL being related to an interaction between westerly flow and Antarctic topography, the variability of the ASL is influenced by remote factors. For example, tropical variability related to the El-Niño Southern Oscillation (ENSO) can create significant changes in the depth of the ASL during the austral winter, with depth of the ASL being significantly lower during the La-Niña phase of ENSO as compared to the El-Niño phase (Raphael et al., 2016;Turner et al., 2013). The SAM, which manifests as a seesaw in atmospheric pressure between southern mid-and high-latitudes, is also known to control variability in the ASL ; with the ASL being deeper during the positive phase of the SAM and shallower during the negative phase. These variations are effectively superimposed on the climatological mean ASL pattern, which we have shown here to be due to an interaction between the westerly wind jet and Antarctic orography.

Data Availability Statements
All data analyzed in this study are freely available to download. ERA-Interim data can be downloaded from the ECMWF website (https://www.ecmwf.int/en/forecasts/datasets/reanalysis-datasets/era-interim). Data for the atmospheric model simulations required to reproduce the figures can be downloaded from Goyal et al., 2021b.