CloudSat Observations Show Enhanced Moisture Transport Events Increase Snowfall Rate and Frequency Over Antarctic Ice Sheet Basins

Elevated moisture transport over the Antarctic ice sheet can increase snowfall and ice mass. Previous studies used ground‐based observations, reanalysis products, and atmospheric models to evaluate the relationship between extreme moisture transport and snowfall properties. Here, we build on previous studies by combining reanalysis and CloudSat radar snowfall retrievals to examine impacts of extreme moisture intrusions on changes in snowfall frequency and intensity over glacier basins on the Antarctic ice sheet. We examine the impacts of enhanced moisture transport events on snowfall frequency and intensity over two different regions on the Antarctic ice sheet: the Amery Ice Shelf of East Antarctica and the Thwaites and Pine Island glacier basins of West Antarctica. We determine when the median integrated water vapor transport from reanalysis exceeds the 95th percentile within the glacier basins of interest to define enhanced moisture transport events. We then use CloudSat radar snowfall retrievals to evaluate differences between snowfall frequency and intensity during enhanced water vapor transport events compared to the seasonal means for 2007–2010. We find that enhanced moisture transport events over the Amery Ice Shelf and Thwaites and Pine Island glacier basins coincide with higher snowfall frequency and intensity. These enhanced moisture transport events have the potential to alter surface mass balance within glacier basins, with implications for future rates of sea level rise.


Introduction
The mass balance of the Antarctic ice sheet (AIS) is a key factor controlling the rate of sea level rise (Rignot et al., 2011).Satellite observations over the past three decades show an accelerating negative mass balance across the AIS since 1979 (e.g., Pattyn & Morlighem, 2020;Rignot et al., 2019;Shepherd et al., 2019;Slater et al., 2021;Willen et al., 2021).However, individual regions of the AIS have disparate mass balance trends, with some regions losing ice at accelerating rates, while others are stable or gaining mass (Edwards et al., 2019;Rignot et al., 2019;Willen et al., 2021).For example, the Thwaites and Pine Island glaciers, located in the Amundsen Sea Embayment of the West Antarctic Ice Sheet, show accelerating mass loss throughout the modern record of observations (e.g., Gardner et al., 2018;Joughin et al., 2014;Rignot et al., 2014).On the other hand, short-term events can influence long-term trends in the mass balance of the AIS.For example, the East Antarctic Ice Sheet experienced approximately 300 GT of accumulation associated with extreme atmospheric moisture transport in March 2022 (Wille et al., 2024), comparable to the average yearly mass loss for the region between 2009 and 2017 (Rignot et al., 2019).The increasingly negative mass balances have raised concern that these portions of the ice sheet could transition into an unstable regime through the marine ice shelf or ice cliff instability (e.g., Bassis & Walker, 2012;Bassis et al., 2021;DeConto & Pollard, 2016;Feldmann & Levermann, 2015;Robel et al., 2019).
Mass loss around the fringes of the AIS drives the contribution of the ice sheet to the rate of future sea level rise (Edwards et al., 2021).However, gains in surface mass balance of the AIS at least partially compensate for ice lost to the ocean, with snowfall as the largest positive contributor to surface mass balance (Bromwich, 1988; Lenaerts Supporting Information may be found in the online version of this article.et al., 2016;Medley & Thomas, 2019;Medley et al., 2014;Van Wessem et al., 2014).Future global warming will increase the moisture availability of the atmosphere, potentially leading to higher snowfall frequency and intensity on the AIS (Frieler et al., 2015;Lenaerts et al., 2016;Palerme et al., 2017;Seroussi et al., 2020;Siahaan et al., 2022).However, the magnitude of the increase remains highly variable between models with little consensus on whether increased snowfall accumulation on the ice sheet will be large enough to balance increased dynamic discharge (Edwards et al., 2021;IPCC, 2022).Addressing these discrepancies requires a better understanding of the processes driving current and future precipitation over the Antarctic Ice Sheet.
Precipitation over the AIS occurs most frequently as snow (Behrangi et al., 2016;Bromwich, 1988;Milani et al., 2018;Palerme et al., 2014).Studies utilizing ground-based measurements of snowfall accumulation during extreme precipitation events have linked their occurrence to instances of enhanced integrated water vapor transport (IVT) intruding meridionally from the ocean to the interior of the AIS (e.g., Gorodetskaya et al., 2014;Gorodetskaya et al., 2015;Maclennan & Lenaerts, 2021;Maclennan et al., 2022;Wille et al., 2019Wille et al., , 2021)).These instances of extreme IVT are often identified as atmospheric rivers, defined as narrow transient bands of strong horizontal water vapor transport across midlatitudes (Gimeno et al., 2014).Atmospheric rivers intruding over the AIS are driven by large-scale atmospheric circulation patterns, including atmospheric blocks, that drive both moisture convergence and the meridional transport of water vapor, and redirect moisture poleward (Bozkurt et al., 2022;Massom et al., 2004;Pohl et al., 2021;Terpstra et al., 2021).Extreme precipitation events drive a large fraction of the annual accumulation over many regions of the AIS (e.g., Boening et al., 2012;Servettaz et al., 2020;Turner et al., 2019).For example, snowfall over the Amery Ice Shelf is particularly dependent on extreme precipitation events, where the heaviest 10 days of precipitation per year account for 50% of the total annual precipitation (Maclennan et al., 2022;Turner et al., 2019), and around 60% of total annual snowfall over the Thwaites Glacier basin is associated with extreme precipitation events (Maclennan & Lenaerts, 2021).
Snowfall is a key contributor to the surface mass balance of the grounded portions of the ice sheet.However, measuring snowfall properties, including frequency and rate, is difficult across the remote environment of the AIS.Recent studies have observed snowfall using both ground-based remotely sensed and in situ instruments on the AIS.In situ measurements find high snow accumulation rates on the Thwaites and Pine Island glacial catchments, between 0.2 and 1 m.w.e.y 1 (Johnson et al., 2018), with higher accumulations located closer to the coastline (Medley et al., 2013(Medley et al., , 2014;;Morris et al., 2017).Maclennan and Lenaerts (2021) used ground-based precipitation gauge measurements from automated weather stations (Scambos et al., 2013) to investigate high snowfall events over the Thwaites Glacier and found extreme snowfall corresponds with interactions between blocking over the Antarctic Peninsula and the Amundsen Sea Low.
Although ground-based measurements can yield high temporal resolution, they are limited in spatial coverage and are subject to difficulty from environmental factors such as windblown snow (e.g., Grazioli et al., 2017).Spaceborne observations from the CloudSat satellite (Stephens et al., 2002(Stephens et al., , 2008) ) have been used successfully to evaluate and characterize snowfall over the Greenland and Antarctic ice sheets (e.g., Bennartz et al., 2019;Kulie et al., 2020;Lenaerts et al., 2020;McIlhattan et al., 2020;Milani et al., 2018;Palerme et al., 2014;Palerme et al., 2017;Souverijns et al., 2018).These studies show good agreement with ground-based profiling radar observations of snowfall characteristics over ice sheets (e.g., Lemonnier et al., 2019;Pettersen et al., 2018;Souverijns et al., 2018).Additionally, Milani et al. (2018) evaluated the CloudSat 2C-SNOW-PROFILE snowfall rate intensity and frequency and found agreement between the retrievals and reanalysis products over the interior of the AIS.Previous work has also linked moisture convergence and resulting enhanced IVT to increased snowfall rates over the Greenland and Antarctic ice sheets using ground-based observations and reanalysis (e.g., Gorodetskaya et al., 2014;Maclennan & Lenaerts, 2021;Maclennan et al., 2022Maclennan et al., , 2023;;Pettersen et al., 2022).However, no studies to date have examined satellite observations of snowfall characteristics over regions of the AIS concurrently with periods of enhanced IVT.
Here we assess the influence of enhanced IVT events on snowfall processes over the AIS.We use ERA5 reanalysis to identify events of enhanced IVT over the Thwaites and Pine Island glacial basins and Amery Ice Shelf region and investigate the snowfall processes concurrent with these events using CloudSat radar retrievals of snowfall.We hypothesize that events of enhanced IVT over a given region of the AIS correspond to increased snowfall frequency and intensity.Because definitions of atmospheric rivers can vary by global region, purpose, and even between research groups (Shields et al., 2018), we define an enhanced IVT event as a period where the median IVT on the entire boundary of a designated region exceeds the 95th percentile.The method results in more events than a traditional atmospheric river-selecting algorithm might because our algorithm for detecting enhanced IVT events considers only the median percentile threshold of IVT on a chosen boundary over the ice sheet.Atmospheric river detection schemes usually require a threshold on the water vapor magnitude of transport as well as stipulations on the geometry of shape of the water vapor transport (Gimeno et al., 2014).Our approach includes events that would not be identified by the atmospheric river detection scheme described by Wille et al. (2021) that we use for comparison against the enhanced IVT events.Subsequently, this approach will include enhanced IVT events that may impact snowfall, regardless of whether they are classified as an atmospheric river by a given tracking algorithm.

Study Regions
We focus on two basin systems with contrasting glaciological and meteorological conditions: the Thwaites and Pine Island glacier basins and the Lambert Glacier-Amery Ice Shelf system.For brevity, we call the combined basins of the Thwaites and Pine Island glaciers the "Amundsen Sea complex," likewise we refer to the Lambert Glacier-Amery Ice Shelf system as the "Amery complex."The Thwaites and Pine Island glaciers show fragmentation and vulnerability (Benn et al., 2022) though they are located in a region of relatively high snowfall for the AIS (Behrangi et al., 2020).The Amery Ice Shelf has shown less mass loss by comparison (King et al., 2009;Li et al., 2020) with a drier surrounding environment compared to the Amundsen Sea Embayment (Behrangi et al., 2020;Palerme et al., 2014).We examine these regions because of this difference in observed mass loss over the last decades, along with their contrasting locations on opposite sides of the continent.After identifying enhanced IVT events using ERA5 reanalysis, we examine the snowfall characteristics for these events with temporally and spatially co-located retrievals of snowfall from CloudSat.
In selecting our regions of focus, we follow previous studies of snowfall over the AIS (e.g., Behrangi et al., 2020;Maclennan et al., 2023) and use the drainage basins outlined in Zwally et al. (2012), called "Zwally basins" hereafter.We consider two pairs of basins on either side of the continent.The two basins on the West, which we call the "Amundsen Sea complex," surround the Amundsen Sea Embayment and drain into the Thwaites and Pine Island glaciers (Zwally et al., 2012, basins 21 and 22; Figure 1).On the eastern side of the AIS, we consider enhanced IVT over two basins flanking the Amery Ice Shelf, the "Amery complex" (Zwally et al., 2012, basins 9 and 11).The combined area of the Amundsen Sea complex is 450,000 km 2 and that of the Amery complex is 439,000 km 2 ; we chose a comparable area since we are measuring the impact of IVT on snowfall characteristics over the entirety of the basins.
The Thwaites and Pine Island glaciers are two of the most rapidly changing glaciers in the world with the potential to contribute 1.2 m of sea level rise (Mouginot et al., 2014).Retreat and disintegration of the Thwaites and Pine The four basins are grouped into two pairs.The boundaries in box A are those we call the Amundsen Sea complex and those in box B are those we call the Amery Ice Shelf complex.Our algorithm is applied to a given set of basins.For a given timestep, the median IVT for the locations on those boundaries matching ERA5 grid points must exceed the 95th percentile with respect to the 1979-2021 climatology.
Island Eastern Ice Shelf glaciers could further contribute to accelerated sea level rise through a loss of buttressing ice in the Amundsen Sea embayment (Bassis & Walker, 2012;Benn et al., 2022;DeConto & Pollard, 2016;Scambos et al., 2017).By contrast, portions of the AIS surrounding the Lambert Glacier-Amery Ice Shelf system in East Antarctica show a slightly positive mass balance seen in observations and modeling (Pittard et al., 2017;Xu et al., 2022;Zhou et al., 2019).The Lambert Glacier-Amery Ice Shelf system covers 16% of the East Antarctic Ice Sheet and the ice flux through the terminus of the Amery Ice Shelf is an important regulator of the mass balance of the ice sheet (Fricker et al., 2000).

ERA5 Reanalysis
We use the ECMWF Reanalysis v5 (ERA5) reanalysis data products of 850 hPa geopotential height, and the northward and eastward components of water vapor flux (Hersbach et al., 2020) for identification of enhanced IVT events.Previous work showed that ERA5 data products reasonably represent the water vapor transport over the AIS (Naakka et al., 2021).We obtain the magnitude of the total IVT vectors from their northward-and eastward-components at each point on the 0.25°× 0.25°resolution grid using Equation 1: This results in a climatology of IVT at 3-hourly temporal resolution from January 1979 to November 2021.We use 850 hPa geopotential height to calculate monthly averages from 1979 to 2021 that are then used to compute climatological anomalies and calculate 850 hPa geopotential height anomalies by subtracting the climatological monthly means from the 850 hPa geopotential height for identified events (e.g., Pettersen et al., 2018).

CloudSat 2C-SNOW-PROFILE
We leverage the CloudSat snowfall retrievals to examine snowfall characteristics during enhanced IVT events.The CloudSat Cloud Profiling Radar (CPR) aboard CloudSat has a 1.3 km cross-track by 1.7 km along-track footprint, with an orbit extending between 82.5°N and 82.5°S and repeating every 16 days (Stephens et al., 2008).The 94 GHz Cloud Profiling Radar (CPR) on the CloudSat is a near-nadir pointing W-band radar instrument, sensitive to reflectivities to 30 dBZ (Tanelli et al., 2008).CPR is well-suited for observing light snowfall, <1 mm hr 1 LWE (e.g., Kulie et al., 2016;Kulie et al., 2020;Kulie & Milani, 2018;Palerme et al., 2014), as the return signal can be attenuated during high snowfall rates and rainfall (e.g., Chase et al., 2022;Tanelli et al., 2008).The CPR reliably profiles snowfall rates down to approximately 1.2 km above surface due to the radar "blind zone" (Maahn et al., 2014) caused by scattering of the radar signal from the surface.

Gridding the CloudSat Product
We grid CloudSat retrievals of snowfall rates LWE to 1°latitude by 2°longitude spatial resolution (e.g., Mateling et al., 2023;McIllhattan et al., 2020).Each grid box is assigned a unique date and time step, based on the median time of CloudSat's overpass through the grid box.We use only snowfall retrievals where confidence flags equal three or 4, which signifies moderate to high confidence (Wood & L'Ecuyer, 2018).The 2C-SNOW-PROFILE snowfall rates for an individual storm within a gridpoint on the ice sheet can have high uncertainty, as they may be inadequately sampled by the CPR (e.g., Souverijns et al., 2018).However, gridded 2C-SNOW-PROFILE observations have been successfully utilized by compositing observations over long time periods and for collections of events, and over large spatial areas, such as drainage basins on an ice sheet, effectively lowering the uncertainty of the retrieved snowfall rates (e.g., Bennartz et al., 2019;Lemonnier et al., 2019;McIlhattan et al., 2020;Palerme et al., 2014;Souverijns et al., 2018).Furthermore, Ryan et al. (2020) shows that 100 observations from the 2C-SNOW-PROFILE product within 80 km of a given point are sufficient for a root mean square error of less than 30% for accumulation estimates over the Greenland Ice Sheet.Accordingly, we require that grid boxes contain at least 100 observations (i.e., CPR footprints) for the composites of enhanced IVT events, and most grid points have between 250 and 750 observations leading to a further reduction in error (<20%).
We investigate snowfall intensity using the mean absolute gridded snowfall rate, which is the mean snowfall rate for retrievals of which there is ≥0 mm hr 1 LWE snowfall detected for each grid box.Additionally, we define snowfall frequency of occurrence when an observation is ≥0.01 mm hr 1 LWE snowfall rate, which is the minimum threshold due to the sensitivity of the CPR to snowfall (Liu, 2008;Tanelli et al., 2008;Wood et al., 2014).

Enhanced IVT Event Identification
We define an enhanced IVT detection on a glacier basin as a time step in which the median of the IVT percentiles for each point on the basin boundary is greater than or equal to the 95th percentile.An enhanced IVT event is a sequence of time steps that start and end with an enhanced IVT detection with the condition that each detection is separated by no more than 24 hr from the next.Independent enhanced IVT events are therefore separated by more than 24 hr (e.g., Pettersen et al., 2022).We calculate the enhanced IVT detections for each glacier basin individually, and then combine the lists of detections for each basin complex (i.e., we combine the list of detections for Zwally basins 21 and 22 which comprise the Amundsen Sea complex, and likewise for basins 9 and 11 of the Amery complex).
The IVT percentile rankings are calculated using ERA5 reanalysis data spanning 1979 through 2021 and grouped by season-December-January-February (DJF), March-April-May (MAM), June-July-August (JJA), and September-October-November (SON).We then match gridded 2C-SNOW-PROFILE data for identified enhanced IVT events by selecting the CloudSat overpasses with median timesteps that fall between the start and end times of each event.
Unlike many atmospheric river detection schemes, the enhanced IVT selection method of this study does not consider the geometry of IVT percentiles to determine events.However, given that previous work connects atmospheric rivers to extreme precipitation over Antarctica (e.g., Gorodetskaya et al., 2014;Maclennan et al., 2022), we also evaluate our event list against the atmospheric river algorithm defined in Wille et al. (2021).
If an atmospheric river is detected anywhere on a basin complex boundary using the Wille et al. (2019Wille et al. ( , 2021) ) algorithm, commonly used for atmospheric river research over Antarctica (e.g., Maclennan et al., 2022Maclennan et al., , 2023;;Turner et al., 2022), during an enhanced IVT event, then that event is flagged as atmospheric river-concurrent.We focus our analysis on the period between the years 2007 and 2010.The 4-year period between 2007 and 2010 encompasses the observational record prior to the CloudSat transition to daylight-only operations after a battery anomaly, which the impacted the record of southern high-latitude snowfall (Milani & Wood, 2021;Stephens et al., 2018).The number of events which coincide with atmospheric rivers between 2007 and 2010 is provided in Table 1 of the results, along with the total number of enhanced IVT events and the means and standard deviations of the event durations by season.

Enhanced IVT Event Snowfall Differences
We compute the snowfall rate and frequency differences by subtracting seasonal means from the enhanced IVT events means at each grid point.We assess statistical significance with a bootstrap random sampling method (Wilks, 2011).We have a set of enhanced IVT events for a given season and basin complex.For each event in that set of enhanced IVT events, we randomly select an observational period for a given grid point in the 2C-SNOW-PROFILE product such that the enhanced IVT event and the observational period have the same duration.We composite the randomly selected observation periods in order to generate a mean snowfall rate for the set.We repeat this process 1000 times to generate 1000 mean snowfall rates for the given grid point grid point.If the magnitude of snowfall rate in the enhanced IVT event composite falls within the top 5% of mean snowfall rates given by the 1000 composites of random observation periods, then that grid point is said to have a statistically significant snowfall rate.We repeat the process for each grid point in the domains shown in respective Figures 5  and 6.We perform the same bootstrapping technique to determine grid points in Figures 5 and 6 with statistically significant differences of snowfall during enhanced IVT events.

Enhanced IVT Event Characteristics
We first examine the number of enhanced IVT events and the event duration over the Amundsen Sea complex and over Amery complex as a function of season, between the years 2007 and 2010, shown in Table 1.In the Amundsen Sea complex, enhanced IVT events appear most often in JJA for the 2007-2010 period, with 23 total events.SON and DJF during the same period have 13 enhanced IVT events each, and MAM has 14.The mean durations of enhanced IVT events varies with season.The Amundsen Sea complex events in DJF, MAM, JJA and SON have average durations of 31, 27, 29, and 16 hr, respectively.The standard deviations of the events per season fall between 14 and 30 hr.Atmospheric rivers are concurrent for roughly half of the IVT events in the Amundsen Sea complex for all seasons (see Section 3.2.2 for a description of how we determine atmospheric river concurrence).
The Amery complex in SON has 26 events, the highest of each season.The seasons DJF, MAM, and JJA have 22, 20, and 18 events, respectively.The average duration of Amery complex events is between 23 and 30 hr for each season, with standard deviations of event duration between 17 and 28 hr.We note that the magnitude of the standard deviations is similar to that of the means of event durations; this shows the high variability of the time span of enhanced IVT events.Nearly half of the enhanced IVT events also had a concurrent atmospheric river detected over the Amery complex boundaries, similar to the Amundsen Sea complex.Tables of event start and end times and durations of each detected enhanced IVT event, including if there was a concurrent atmospheric river identified, are provided in the supplementary data (see Table S1 and S2).

Amundsen Sea Complex Enhanced IVT Event Synoptic Characteristics
Figure 2 shows the seasonal composites of IVT percentiles (Figures 2a-2d) and 850 hPa geopotential height climatological anomalies (Figures 2e-2h) during the identified events detected over the Amundsen Sea complex.
Events composed over the four seasons show enhanced IVT across the West Antarctic Ice Sheet up to the Transantarctic Mountains, which are a barrier to further atmospheric transport.The 850 hPa geopotential height anomaly composites (Figures 2e-2h) generally show positive anomalies to the east of the Amundsen Sea complex over the Antarctic Peninsula between 30°W and 120°W, and negative anomalies to the West, between 120°W and 180°W, over the Ross Sea.Both the magnitude and location of the positive and negative pressure anomaly pattern varies with season.During DJF (Figure 2e), the positive 850 hPa geopotential height anomalies are less than 60 m, although the negative anomaly is less than 90 m between 60°S and 70°S.MAM shows positive 850 hPa geopotential height anomalies, ≥110 m, over the Antarctic Peninsula and Weddell Sea (Figure 2f).The negative anomaly between 120°W and 180°W in MAM is weak compared to that of all other seasons, never exceeding 50 m.JJA and SON show large 850 hPa geopotential height anomaly magnitudes, displaying high anomaly values >120 m and negative anomaly values < 120 m.The 850 hPa geopotential height anomaly composite for SON displays a zonal wave 3 pattern, where three deep negative anomalies are separated by longitude and interspersed with less-distinct high anomalies (i.e., Raphael, 2004).For all four seasons, the 0 m anomaly contour is located over on the Amundsen Sea Embayment and the Amundsen Sea complex intersecting at the coastline.The pattern of composite 850 hPa geopotential height anomaly for JJA (Figure 2g) extends westward compared to the other seasons, with the negative 850 hPa geopotential height anomaly centered over the Ross Ice Shelf and the high 850 hPa geopotential height anomaly toward the Amundsen Sea, and the 0 m contour intersecting the coast at an oblique angle.

Amery Complex Enhanced IVT Event Synoptic Characteristics
Figure 3 shows the seasonal composites of IVT percentiles and 850 hPa geopotential height monthly anomalies for events detected over the Amery complex (analogous to Figure 2).The composites show high percentiles of IVT across the East Antarctic for all 4 seasons, particularly between 60°E and 120°E.The IVT vectors indicate weaker flow onto the Amery complex compared to the corresponding composites for the Amundsen Sea complex.In DJF (Figure 3a), the flow of enhanced IVT runs largely parallel to the 60°E meridian and intersects the coastline perpendicularly nearest the Amery Ice Shelf.In the other three seasons, the vectors of IVT are directed southwest toward the Amery complex, with the strongest IVT magnitude impinging the coast around 90°E.The event composites indicate high IVT percentile values beyond the boundaries of the Amery complex for each season, which extend far inland.In all seasons, we see that anomalously high 850 hPa geopotential height values are located east of the Amery complex (Figures 3e-3h).DJF (Figure 3e) and SON (Figure 3h) show weak negative anomalies west of the Amery complex during events.In the other seasonal composites, positive 850 hPa geopotential height anomalies persist across the entire AIS, although the highest 850 hPa geopotential height anomalies are consistently located east of the Amery complex.An area of anomalously negative 850 hPa geopotential height is shown in the MAM composite over the Southern Ocean north of 60°S, further from the AIS.Similar to the composites for the Amundsen Sea complex, a 0 m anomaly contour follows from lower latitudes in the Southern Ocean to the Amery complex coastline boundary for each season, with the exception of MAM.

CloudSat Seasonal Snowfall Observation Composites
Composited seasonal snowfall frequency of occurrence and mean absolute snowfall rate from CloudSat retrievals are shown in Figure 4. Per our methods, snowfall occurrence is defined as any CloudSat retrievals that show ≥0.01 mm hr 1 LWE snowfall rate.The snowfall frequency is the fraction of snowfall occurrence as a percentage of all available CloudSat observations per season between 2007 and 2010, for a specific grid point (Figures 4a-4d).The mean, gridded snowfall rates and frequencies shown in Figure 4 agree with previous work using CloudSat snowfall retrievals in the Antarctic region (e.g., Kulie & Milani, 2018;Milani et al., 2018;Palerme et al., 2014).The frequency of snowfall is generally higher over the Amundsen Sea complex compared to the Amery complex for all seasons.Snowfall frequency over the Amundsen Sea complex is lowest in DJF (Figure 4a), with snowfall frequency of 15%-30% of all gridded CloudSat observations.MAM, JJA, and SON show a snowfall frequency of 20%-35%.Over the Amery complex, the snowfall frequency varies little by season, varying between 10% and 20% between 2007 and 2010.However, JJA displays somewhat lower snowfall frequencies over the Amery complex, between 5% and 15%.The gridded mean absolute rate is shown seasonally in Figures 4(e-4h).The absolute mean rates over the Amundsen Sea complex are lowest during DJF, between 0.02 and 0.04 mm hr 1 .MAM and SON both show average rates between 0.03 and 0.08 mm hr 1 , while JJA shows the highest rates of up to 0.15 mm hr 1 .For the Amery complex, DJF has the highest rates in the seasonal composites, between 0.01 and 0.06 mm hr 1 , while each of the other seasons show average rates between 0.01 and 0.04 mm hr 1 .

Amundsen Sea Complex Event Snowfall Characteristics
Figures 5a-5d shows the difference in snowfall frequency between the identified enhanced IVT events and the seasonal mean CloudSat observations from 2007 to 2010 (Figures 4a-4d).Positive values (blue shading) indicate more frequent observations of snowfall during enhanced IVT events, while negative values (red shading) indicate less frequent.Enhanced IVT events over the Amundsen Sea complex show increased snowfall occurrence over the basins for all four seasons.DJF and JJA have similar frequency differences over the basin, showing ∼22% increase on average over the basin complex.MAM showed a 30% increase and SON indicates a 36% increase in snowfall occurrence, however, this season has the fewest events and the shortest average event duration.
Figures 5e-5h shows the difference in CloudSat gridded mean absolute snowfall rate for Amundsen Sea complex enhanced IVT events compared to CloudSat seasonal gridded mean absolute rates.The area within the Amundsen Sea complex boundary shows an increase in snowfall rates with seasonal variation.DJF shows the lowest increase in mean rates with a basin average of +0.07 mm hr 1 .MAM and JJA show average mean rate increases of approximately +0.13 mm hr 1 each within the basin complex.SON shows the highest mean increase with an average of +0.20 mm hr 1 in the basin complex.The maximum mean snowfall rate increase for any one grid point in the composite of enhanced IVT events is 0.57 mm hr 1 , shown in MAM.Importantly, both snowfall frequency and rate increases are seen extending beyond the bounds of the Amundsen Sea complex to the west, particularly in JJA.

Amery Complex Event Snowfall Characteristics
The differences between the snowfall frequency of occurrence during the identified enhanced IVT events for the Amery complex and seasonal CloudSat observations are shown in Figures 6a-6d.The CloudSat observations over the Amery complex generally show increases in snowfall frequency over the AIS between 60°E and 120°E during enhanced IVT events.The average increase of snowfall frequency in the basin in MAM is 4.6% during enhanced IVT events.DJF shows an average increase of 12.5% over the Amery complex during enhanced IVT events.JJA and SON show larger increases over the Amery complex during IVT events, with basin averages of 23% and 27%, respectively.However, snowfall frequency increases extend beyond the bounds of the basin complex.The boundaries of Amery complex and the region inland, containing the Lambert glacier which feeds into the Amery Ice Shelf, display the largest increases of snowfall frequency during enhanced IVT events compared to the increases to the east of 40°E, except for MAM.The composite for MAM (Figure 6b) shows snowfall frequency increases over the basin complex and the AIS more broadly, albeit with limited coherence.The spatial pattern of the snowfall frequency increases in JJA and SON are similar in presentation, where positive frequency increases are shown on the ice sheet in the composites between 50°E and 100°E.
The differences between mean absolute snowfall rate during enhanced IVT events and seasonal CloudSat observations are shown in Figures 6e-6h.DJF and JJA show an average increase in mean rates for grid points within the bounds of the basin complex of approximately +0.06 mm hr 1 .SON shows an average increase in basin complex of +0.04 mm hr 1 .MAM shows no spatially coherent increase in the composite (Figure 6f) and has a correspondingly small mean, only 0.01 mm hr 1 , of the snowfall rate differences in the basin.DJF, MAM, and SON also show snowfall rate increases on grid points located between the coastline and the basin boundary on the flanks of the basin complex.The western flank is defined as the grid points between the basin boundary, the coastline and the 92°E meridian.The eastern flank is defined as the grid points between the basin boundary, the coastline and the 59°E meridian.The snowfall rate increases are strongest in DJF, with 0.21 and 0.17 mm hr 1 for the grid points within the western and eastern flanks of the Amery complex, respectively.

Discussion
Using CloudSat snowfall retrievals, we find that enhanced IVT events correspond to higher snowfall rates and frequency over both the Amery and the Amundsen Sea complexes.This reinforces previous studies using groundbased observations and reanalysis that highlight the effect of extreme moisture transport on snowfall (e.g., Maclennan & Lenaerts, 2021;Maclennan et al., 2023;Servettaz et al., 2020;Swetha Chitella et al., 2022;Terpestra et al., 2021).Unlike previous studies that specifically examined the impacts from atmospheric rivers, our analysis shows that increased snowfall is present during enhanced IVT events, for both the Amundsen Sea and Amery complexes.It is worth noting that atmospheric river detection depends on the prescribed conditions of the detecting algorithm.Moreover, the patterns of 850 hPa geopotential height anomalies that we find during enhanced IVT events over the Amundsen Sea complex show similarities to synoptic conditions found in prior studies of snowfall extremes related to atmospheric rivers over portions of the West Antarctic Ice Sheet (e.g., Djoumna & Holland, 2021;Maclennan et al., 2023;Maclennan & Lenaerts, 2021;Scott et al., 2019;Swetha Chitella et al., 2022;Wille et al., 2021).Synoptic conditions associated with enhanced IVT events over the Amery complex show ridging in locations of the Indian and Southern Oceans known for generating atmospheric blocking patterns associated with moisture transport poleward (Hirasawa et al., 2013;Massom et al., 2004;Servettaz et al., 2020;Terpestra et al., 2021).Enhanced IVT events over glacier basins on the Antarctic Ice Sheet also coincide with elevated rate and frequency shown by the CloudSat CPR over the same region, including deep into the interior, for all seasons.We note that both atmospheric river-and non-atmospheric river-concurrent enhanced IVT events show snowfall increases over the ice sheet, shown in Figures S1 and S2 in Supporting Information S1.

Synoptic Drivers of Enhanced IVT Events
Our results suggest that an anomalous Amundsen Sea Low and high pressure anomalies over the Antarctic Peninsula are associated with enhanced IVT events (Figure 2).The central pressure of the Amundsen Sea Low varies seasonally, the central pressure is at a maximum in December or January and at a minimum in August or September (Hoskins et al., 2013).The Amundsen Sea Low follows an annual seasonal migration, from a summer (DJF) mean longitude at 100°W westward to a winter (JJA) mean longitude of 160°W (Hoskins et al., 2013;Raphael et al., 2016).This seasonal variation is apparent in Figure 2, which show areas of negative 850 hPa geopotential height anomalies further west compared to other seasons and close to the continent in JJA.
The 850 hPa geopotential height anomalies shown in Figure 2 imply that a deeper low-pressure system, and/or a westward shift in the location of the low relative to monthly climatology, are associated with enhanced moisture transport toward the Amundsen Sea embayment.Positive 850 hPa geopotential height anomalies are positioned over the Antarctic Peninsula and surrounding Southern Ocean during enhanced IVT To the East of the Amundsen Sea embayment (Figure 2), and the development of ridging high pressures over the Antarctic Peninsula can lead to blocking conditions, which are associated with marine air intrusions over the West Antarctic Ice Sheet (e.g., Djoumna & Holland, 2021;Scott et al., 2019;Wille et al., 2021).Previous work has found that coupling between an atmospheric block over the Antarctic Peninsula and the Amundsen Sea low can drive extreme snowfall events over the West Antarctic Ice Sheet, including Thwaites Glacier (Maclennan & Lenaerts, 2021;Swetha Chitella et al., 2022).A similar 850 hPa geopotential height pattern consisting of ridging over the Antarctic Peninsula and a low to its east is shown during enhanced IVT events by this study, suggesting that an anticyclone coupled with anomalously low 850 hPa geopotential height to its west drive elevated moisture transport over the Amundsen Sea complex.The moisture transport during enhanced IVT events then increases snowfall processes over the grounded portions of the ice sheet in the Amundsen Sea complex.
Similarly, our results over the Amery complex show that anomalous ridging is a driver of enhanced IVT events.The 850 hPa geopotential height anomalies in Figure 3 show positive 850 hPa geopotential height anomalies between 70°E and 160°E, with amplitudes weakest in DJF and strongest in JJA.Blocking anticyclones in a similar location over the Southern Ocean are known to drive warm air intrusions deep into the interior of the East Antarctic Ice Sheet (e.g., Hirasawa et al., 2013;Massom et al., 2004;Scarchelli et al., 2011).These blocks can be associated with intense snowfall events resulting from the transport of moist marine air over the ice sheet (e.g., Massom et al., 2004;Servettaz et al., 2020;Terpstra et al., 2021).Additionally, similar atmospheric blocking mechanisms have been shown to drive moisture transport toward and onto the Greenland Ice Sheet (e.g., Pettersen et al., 2022;Ward et al., 2020).

Impacts of Enhanced IVT Events on Snowfall
Enhanced IVT events correspond to increased snowfall frequency and intensity over both the Amery and Amundsen Sea complexes, consistent with previous studies.For example, Adusumilli et al. (2021) found increases in surface height in West Antarctica as observed by ICESAT-2 during extreme snowfall events in 2019, and atmospheric rivers made up 63% of these events.Maclennan et al. (2023) investigated a case study of three atmospheric river landfalls over the Thwaites Glacier, and found these events corresponded to observed increases in surface height on the Thwaites Eastern Ice Shelf due to snowfall deposition.Our results illustrate that increased snowfall frequency and rates from enhanced IVT events (which include atmospheric rivers over the Amundsen Sea complex, see Table 1) extend beyond the coasts and penetrate deep into the interior of the West Antarctic Ice Sheet (Figure 5).Increased snowfall rates and frequency are present over the entire Amundsen Sea complex for all four seasons, suggesting that enhanced IVT events lead to increased snowfall contributions over the interior of the ice sheet.
Orographic lift of moist air over steep portions of the ice sheet enhances the development of snowfall (Bromwich, 1988), including the ice sheet around the Amundsen Sea (Lenaerts et al., 2018).The strongest increases we observe in snowfall rate and frequency for the Amundsen Sea complex are located along the coast, indicating that the snowfall related to the enhanced IVT events originates from moisture flow in the atmosphere traveling perpendicular to the coastline up the slope of the ice sheet around the Amundsen Sea Embayment.
The association we find between enhanced IVT events and increased snowfall processes over the Amery complex aligns with previous work, which used reanalysis data and model results to show that regions of persistent high pressure to the east of a corridor of poleward moisture transport can lead to instances of extreme snowfall over the East Antarctic (Baiman et al., 2023;Servettaz et al., 2020;Terpstra et al., 2021).Additionally, Gorodetskaya et al. (2014) found that the presence of atmospheric rivers resulted in anomalously high snowfall and accounted for 46%-80% of the annual accumulation over Dronning Maud Land on the East Antarctic between 2009 and 2012.Similarly, Boening et al. (2012) used Gravity Recovery and Climate Experiment (GRACE) estimates of mass balance alongside CloudSat retrievals of snowfall and found a consistent relationship between anomalous snowfall and mass gains over Dronning Maud Land, from 2009 to 2011, particularly during 2 months (May 2009 and June 2011).Snowfall rates and frequency are increased during enhanced IVT events over the Amery complex in DJF, JJA, and SON, suggesting that these events can lead to elevated snowfall conditions over the region.
Elevated IVT conditions over ice surfaces have been shown to negatively impact ice sheet mass balances.Atmospheric river landfalls have been shown drive melt over the ice shelves around the Antarctic Peninsula (Wille et al., 2019(Wille et al., , 2021) ) and the Greenland Ice Sheet (Mattingly et al., 2020(Mattingly et al., , 2023)).This melt results from adiabatic heating as a result of foehn wind generation (Bozkurt et al., 2018;Mattingly et al., 2023;Wille et al., 2019Wille et al., , 2021) ) and radiation as a consequence of cloud formation over the ice surface (e.g., Mattingly et al., 2018Mattingly et al., , 2020;;Wille et al., 2019).At the same time, snowfall resulting from elevated IVT may replace lost mass from the Antarctic Ice Sheet (e.g., Boening et al., 2012).Atmospheric river intensity and frequency of occurrence is expected to increase as the climate warms (Mattingly et al., 2023;Wang et al., 2023).The balance of these potential impacts are important to consider when examining elevated IVT as a factor in surface mass balance over the AIS.

Conclusions
Mass loss to the ocean around the fringes of the AIS contributes to increased rates of sea level rise (Edwards et al., 2021).However, this loss is partially compensated by surface mass balance gains, and snowfall over the grounded portions of the ice sheet is the main factor driving positive mass balances (Bromwich, 1988;Lenaerts et al., 2016;Medley et al., 2014;Medley & Thomas, 2019;Van Wessem et al., 2014).Previous work has found that extreme snowfall events, with high potential for depositing mass on the AIS, are linked to strong intrusions of water vapor from lower latitudes into the AIS (Gorodetskaya et al., 2014(Gorodetskaya et al., , 2015;;Maclennan et al., 2022;Maclennan & Lenaerts, 2021;Wille et al., 2019Wille et al., , 2021)).These results used ground-based observations, reanalysis, and modeling, each of which have strengths and limitations with respect to assessing the link between extreme moisture transport and snowfall processes.Our study complements these previous results through our examination of enhanced IVT events and collocated observations by CloudSat radar retrievals of snowfall rate and frequency.Our work shows that the snowfall frequency and intensity increase during enhanced IVT events over the examined glacier basins, including deep into the interior of the ice sheet.The enhanced IVT events show similarities to atmospheric drivers of extreme snowfall events found by previous work over regions surrounding the Amery Ice Shelf and Thwaites Glacier, respectively.These previous results relied on ground-based observations with limited geographic scope, and reanalysis data and model results with potential bias in reproducing precipitation processes.Our results complement these previous studies with spaceborne radar observations with broad geographic coverage.Additionally, our results indicate that ridging over the Antarctic Peninsula may play a role in generating and maintaining enhanced IVT events over the Amundsen Sea and Amery complexes, suggested by the presence of broad areas of ridging in the composites of 850 hPa geopotential height anomalies for each region.This ridging may be the result of atmospheric blocking observed over the Antarctic Peninsula in previous study of extreme snowfall over Thwaites Glacier (Maclennan & Lenarets, 2021).
Our algorithm for detecting enhanced IVT events considers only the median percentile threshold of IVT on a chosen boundary over the ice sheet.The method captures more events than traditional atmospheric river-selecting algorithms, which usually require a threshold on the water vapor magnitude of transport as well as stipulations on the geometry of shape of the water vapor transport.Enhanced IVT events detected by our method include every atmospheric river detected on the glacial basin boundaries by the Wille et al. (2019) method and a number of unclassified events of strong moisture transport.
Evidence links enhanced IVT to processes negatively impacting AIS surface mass balance.Our study examines the impacts of enhanced IVT intrusions on snowfall properties alone and does not consider the radiative effects of extreme moisture transport nor the hydrologic processes affecting surface mass balance on the ice sheet, such as surface melt and refreezing.The combined influence of winds and orography complicate the impacts of enhanced IVT on snowfall and ultimately surface mass balance.These complicating factors necessitate further study of the impacts of anomalous moisture transport events on the regional surface mass balance of the AIS.

Figure 1 .
Figure1.The outlines of the basins we use in this study.The four basins are grouped into two pairs.The boundaries in box A are those we call the Amundsen Sea complex and those in box B are those we call the Amery Ice Shelf complex.Our algorithm is applied to a given set of basins.For a given timestep, the median IVT for the locations on those boundaries matching ERA5 grid points must exceed the 95th percentile with respect to the 1979-2021 climatology.

Figure 2 .
Figure 2. Composite (a-d) IVT seasonal percentiles and (e-h) 850 hPa geopotential height monthly anomalies averaged per season during enhanced IVT events over the Amundsen Sea complex.IVT seasonal percentiles (shading) averaged over all events detected by the algorithm on the Thwaites and Pine Island glacier basin for (a) DJF, (b) MAM, (c) JJA, (d) SON.Percentiles are calculated relative to seasons in the 1979-2021 climatology.Contouring indicates IVT magnitude in kg m 1 s 1 .The three contours on each IVT plot are spaced every 50 kg m 1 s 1 .The contour furthest south is 50 kg m 1 s 1 .Arrows indicate composite IVT magnitude and direction during events.Composite 850 hPa geopotential height monthly anomalies during events averaged seasonally are shown for (e) DJF, (f) MAM, (g) JJA, (h) SON.Monthly anomalies are calculated by subtracting the 1979-2021 monthly mean 850 hPa geopotential height from the average 850 hPa geopotential height during enhanced IVT events.At each point.Contours indicate anomaly values every 10 m.

Figure 3 .
Figure 3. Composite IVT percentiles and 850 hPa geopotential height anomalies during enhanced IVT events over the Amery complex.IVT percentiles (shading) averaged over all events detected by the algorithm on the Amery glacier basin for (a) DJF, (b) MAM, (c) JJA, (d) SON.Percentiles are calculated relative to the 1979-2021 climatology.Contouring indicates IVT magnitude in kg m 1 s 1 .The three contours on each IVT plot are spaced every 50 kg m 1 s 1 .The contour furthest south is 50 kg m 1 s 1 .Arrows indicate composite IVT magnitude and direction during events.Composite 850 hPa geopotential height anomalies during events are shown for (e) DJF, (f) MAM, (g) JJA, (h) SON.Anomalies are calculated by averaging the 850 hPa geopotential height during events and subtracting the 1979-2021 mean at each point.Contours indicate anomaly values spaced every 10 m.

Figure 4 .
Figure 4. Gridded seasonal ≥0.01 mm hr 1 LWE snowfall observation frequency (%) from the 2C-SNOW-PROFILE CloudSat product for the 2007-2010 are shown at the top.The grid box color scale represents the magnitude of the snowfall frequency (the fraction of observations in the period).Gridded seasonal mean absolute snowfall rates (mm h 1 ) from the 2C-SNOW-PROFILE CloudSat product for the 2007-2010 are shown on the bottom, plotted in log 10 scale.Grid box color scale indicates the magnitude of the retrieved mean absolute snowfall rate (the rate found for instances where snowfall rate is ≥ 0 mm hr 1 ).The hole in the center of each image shows the extent of CloudSat's orbit (82°S).

Figure 5 .
Figure 5. Differences in snowfall frequency using a ≥0.01 mm hr 1 LWE threshold (%, top) and mean absolute snowfall rate (mm h 1 , plotted in log 10 scale, bottom) between the observations during enhanced IVT events and all observations for a given season between 2007 and 2010 over the Amundsen Sea complex for (a) DJF, (b) MAM, (c) JJA, and (d) SON.Positive values (blue shading) indicate a greater frequency and intensity of observed snowfall during an event.The basin bounds of Zwally basins 21 and 22 are outlined in black.Grid points with statistically significant differences at the 95th level are marked with black crosses.

Figure 6 .
Figure 6.Differences in snowfall frequency using a ≥0.01 mm hr 1 LWE threshold (%, top) and mean absolute snowfall rate (mm h 1 , plotted in log 10 scale, bottom) between the observations during enhanced IVT events and all observations for a given season between 2007 and 2010 over the Amery complex for (a) DJF, (b) MAM, (c) JJA, and (d) SON.Positive values (blues) indicate a greater frequency and intensity of observed snowfall during an event.The white grid boxes are where there are fewer than 100 observations from CloudSat for all events during a season.The basin bounds of Zwally basins 9 and 11 are outlined in black.Grid points with statistically significant differences at the 95th level are marked with black crosses.

Table 1 A
Wille et al. (2021)nced IVT Event Characteristics by SeasonThe number of events which contain at least one detection of an atmospheric river by the algorithm presented inWille et al. (2021)on the glacier complex bounds are given in the second column.