Unprecedented mass gain over the Antarctic ice sheet between 2021 and 2022 caused by large precipitation anomalies

The Antarctic ice sheet (AIS) is susceptible to global climate change, and its mass loss has been 92 ± 18 Gt yr −1 between 1992 and 2020. Given the current intensive global warming, we investigate the AIS mass changes from January 2003 to December 2022, using the newly released satellite gravimetry and atmospheric datasets. The results show that the continuous mass loss in the AIS between 2003 and 2020 was 141.8 ± 55.6 Gt yr−1 . However, the AIS showed a record-breaking mass gain of 129.7 ± 69.6 Gt yr −1 between 2021 and 2022. During this period, the mass gain over the East AIS and Antarctic Peninsula was unprecedented within the past two decades, and it outpaced the mass loss in the Amundsen sector of the West AIS from 2003 to 2022. Basin-scale analysis shows that the mass gain mainly occurred over Wilhelm II Land, Queen Mary Land, Wilkes Land, and the Antarctic Peninsula due to anomalously enhanced precipitation. Further investigation reveals that during 2021–2022, a pair of symmetrically distributed high-low pressure systems, located at approximately 120°W and 60°E in the Southern Ocean, drove the observed abnormal precipitation and mass accumulation.


Introduction
The Antarctic ice sheet (AIS) melting has the potential to raise the global mean sea level (GMSL) by 58 m, and even a slight sea-level rise has direct societal and economic implications for coastal areas (Bars et al 2017, Oppenheimer et al 2019).Therefore, estimating its mass changes and understanding the driving factors is crucial.The AIS mass changes include SMB (surface mass balance) and ice dynamic processes.In the SMB, most ablation is from sublimation, and the important dynamic process is discharge across the grounding line.If the ice mass over the AIS is balanced, snowfall accumulation will balance surface ablation and ice discharge (Rignot et al 2019).However, numerous reports consistently indicate that the ice mass has been in imbalance, and they concur that ice mass loss over the AIS has outpaced gains from the mid-1990s through 2020 (The IMBIE team 2018).Specifically, the mass loss over the AIS contributed 7.4 ± 1.5 mm to GMSL from 1992 to 2020 (Otosaka et al 2022).
Mass changes over the AIS occur from decadal to interannual timescales (Li et al 2022).The interannual variations affect the estimation of long-term trends (Zhang et al 2020) and alter the shortterm contribution to sea-level changes (Bodart and Bingham 2019).For instance, 350 Gt net ice gain over dronning Maud land (DML) from 2009 to 2011 is assessed by Boening et al (2012), which is equivalent to a decrease in GMSL rise at a rate of 0.32 mm yr −1 over the three years.Moreover, relative to the average change from 2002 to 2017, the extreme El Niño event resulted in a mass increase of 277 ± 91 Gt over the AIS between June 2015 and March 2017 (Bodart and Bingham 2019).The Amundsen Sea sector in the West AIS, the primary source of Antarctic mass loss, experienced a reduction in the total mass loss rate between 2019 and 2020 (Davison et al 2023).The elevation change records revealed a rapid mass loss in the Amundsen Sea sector from 2003 to 2021, while the mass loss slowed down between 2019 and 2021 (Yue et al 2023).Nonetheless, the mass balance over the entire AIS in the recent two years (2021-2022) remains unclear.Given the intensive global warming and the impact of Antarctica on GMSL rise, it is imperative to extend the record of mass balance over the AIS and analyze the driving factors behind its short-term mass changes.
During the past decade, the mass balance record over the AIS has been significantly consolidated, benefiting from the data collected by the GRACE (Gravity Recovery and Climate Experiment) mission (Bamber et al 2018).As a successive mission, GRACE-FO (GRACE Follow-On) tracks changes in gravity since the end of the GRACE mission, allowing quantification of the cumulative mass change between the two missions (Sasgen et al 2020).With the GRACE and GRACE-FO observations, it has become possible to map the mass changes of the entire Antarctic over longer periods (Velicogna et al 2020).Therefore, in this study, we analyze the cumulative mass change characteristics over the AIS from January 2003 to December 2022 (20 years) and calculate the biennial mass change rate to demonstrate the short-term mass balance status.Then, we illustrate the geographical patterns of mass change and analyze the direct cause of the mass change.Finally, we reveal the basin-scale interannual mass variation and the primary driving factors.

GRACE/GRACE-FO data
Using the GRACE/GRACE-FO observations, the mascon (mass concentration) approach can estimate monthly cumulative mass anomalies relative to a given time at specific grid locations on the Earth's surface.Here, we used newly released 0.25 • × 0.25 • mascon solutions (RL06, version 02) with an extended timespan provided by the CSR (Center for Space Research) at the University of Texas at Austin (Save 2020) to estimate mass changes over the AIS from January 2003 to December 2022.The CSR derives its mascon solutions with spherical harmonic coefficients up to degree and order of 120 (Save et al 2016) after the C 20 (degree 2 order 0) coefficients and C 30 (degree 3 order 0) coefficients for GRACE-FO period are replaced with those from SLR (Satellite Laser Ranging) (Loomis et al 2019(Loomis et al , 2020) ) and the absent degree-1 coefficients are added with those from the Technical Note-13 (Save 2020).The glacial-isostatic adjustment (GIA) correction is completed with the ICE6G-D model (Peltier et al 2018).

Meteorological fields
The Regional Atmospheric Climate Model 2 (RACMO2), Version 2.3p2, with a spatial resolution of 27 km × 27 km, is used to estimate the SMB over the AIS from January 1979 to December 2022 (van Wessem et al 2018).Since there are potential errors and biases in particular precipitation datasets (González-Herrero et al 2023), the average of the precipitation fields from the RACMO2, Modern-Era Retrospective analysis for Research and Applications version 2 (MERRA-2; at 0.5 • × 0.625 • spatial resolution, Gelaro et al 2017), and the European Centre for Medium-Range Weather Forecasts Reanalysis version 5 (ERA5; at 0.25 • × 0.25 • spatial resolution, Hersbach and Dee 2016) reanalysis products is used to explore the relationship between the precipitation and the interannual mass variations over the AIS.The spatial resolutions of the RACMO2 and the precipitation fields are adjusted to match the resolution of CSR mascon solutions through spatial averaging.

Processing strategies
Since the study period is from January 2003 to December 2022, the cumulative SMB/mass/dynamic anomalies are relative to those in January 2003.Given that RACMO2 provides simulations of absolute SMB, the SMB anomalies are derived by removing their 1979-2008 mean (Velicogna et al 2014, Rignot et al 2019).Then, the cumulative SMB anomalies are derived by time-integrating the SMB anomalies from January 2003.Since there is no systematic bias between the GRACE and GRACE-FO time series (Velicogna et al 2020), the 11 month gap between them is bridged using cumulative SMB anomalies, and the remaining missing data are filled with cubic spline interpolation.Then, the gridded cumulative dynamic mass anomalies are calculated by removing the cumulative SMB anomalies from the cumulative mass anomalies (Diener et al 2021).The monthly time series of cumulative anomalies for an individual basin or region is calculated by integrating the values of the grids within the basin or region.The corresponding contribution to GMSL rise is determined by using the relationship that 362.5 Gt mass variation corresponds to 1 mm GMSL change (Cogley 2012).Monthly precipitation anomalies and cumulative precipitation anomalies are derived following the same approach used for RACMO2 SMB.The interannual mass (or precipitation) variation is computed using a 13 month moving average filter on the detrended time series of cumulative anomalies.Consequently, positive interannual mass (or precipitation) variation indicates higher cumulative mass (or precipitation) relative to the mean rate over the study period.In contrast, negative values indicate a reduction in mass (or precipitation).
The long-term mass change trend, or biennial mass change rate (denoted as a in equation ( 1)), is determined by least squares fitting to the time series of cumulative anomalies M (t), using the following equation: where t i is the ith time tag in years; a 0 and a 0 are the constant and biennial mass change trend (or rate); the last four terms on the right-hand represent the annual and semi-annual signals.The biennial mass change rate calculation employs time series that commence in January of the initial year and conclude in December of the second year.It should be noted that the rates for SMB and dynamic mass change are relative to the average SMB from the reference years 1979-2008.

Error estimates
The error for the cumulative mass/SMB/dynamic anomaly at a 95% confidence level (or twosigma) is derived from the law of error propagation based on the fitting error in equation ( 1), where σ a0 , σ a1 , σ a2 , σ a3 , σ a4 , σ a5 are the onesigma uncertainties for corresponding parameters in equation ( 1).The error for the mass change trend (or rate) at a 95% confidence level is derived by where σ GIA denotes the standard deviation of the GIA model, which is empirically calculated by a multimodel comparison as

Cumulative mass anomalies and biennial mass change rates between 2003 and 2022
The time series of cumulative mass anomalies and their contribution to GMSL for the AIS, EAIS, WAIS and APIS are depicted in figures 2(a)-(d).Results

Geographical patterns for mass change trends
To investigate the mass balance before and after the observed mass gain, figure 4 depicts the geographical patterns of trends from January 2003 to December 2022 and the biennial rates of total mass changes between January 2021 and December 2022.The estimated trends and biennial rates of individual basins for the two periods are shown in supplementary table S1, in which Basin 24, 25, 26, and 27 are integrated due to the relatively small area coverage.Additionally, corresponding trends and rates for cumulative SMB/dynamic anomalies are presented to describe the direct cause of total mass change.
From 2003 to 2022, the Amundsen Sea sector and the APIS exhibited clear mass loss patterns.They are primarily driven by ice dynamic imbalances from the fast-flow glaciers of Pine Island, Thwaites, Getz (Rignot et al 2014), and several Bellingshausen Sea glaciers, caused by submarine melting (Mouginot et al 2014, Khazendar et al 2016) and iceberg calving (Berthier et al 2012).Besides, the surface mass reduction in the Getz glacier also contributed to the total mass loss though to a lesser extent.Moreover, a prominent mass gain trend is observed in the Kamb ice streams, primarily driven by ice dynamic changes.For the EAIS, the geographical pattern of the mass change trend is more complex.The mass loss trends are evident in Queen Mary Land, George V Land, and Coats Land.In Queen Mary Land, the sustained reduction in cumulative SMB is primarily responsible for these signals.Conversely, in Coats Land and George V Land, the mass losses are predominantly driven by ice dynamic changes from the Totten, Moscow, Cook, Mertz, and Ninnis glaciers.In DML, Enderby Land, and Kemp Land, a widespread pattern of modest ice mass gain extended along much of the coastline and reached several hundred kilometers inland, closely associated with SMB accumulation.However, ice dynamic losses may offset a small portion of the mass accumulation.
The short-term mass change rates in figures 4(d)-(f) provide additional insights for the mass changes in the recent two years.Compared to the past two decades, from 2003 to 2020, the amplitude and even the sign of mass change (gain or loss) between 2021 and 2022 changed significantly over some regions.Most strikingly, there were clear mass loss signals over the APIS from 2003 to 2020.However, during 2021-2022, the mass gain was evident across most regions of the APIS, excluding the narrow peripheral areas in Basin 27.Moreover, there were significant mass gain signals over Basin 12 (Wilhelm II Land and Queen Mary Land) and Basin 13 (Wilkes Land).Basin-scale mass change rates in table S1 show that Basin 12, 13, and APIS are the top three contributors to the observed 2021-2022 mass gain over the AIS, with the biennial rates of 74.2 ± 8.0 Gt yr −1 , 63.2 ± 11.6 Gt yr −1 and 34.1 ± 3.8 Gt yr −1 respectively.Besides, the mass gain rate over Basin 5-6 (DML) increased twofold from 24.0 Gt yr −1 to 48.1 Gt yr −1 .These mass gains are spatially matched with the cumulative SMB change rates, indicating that enhanced SMB accumulation is the direct cause of the observed total mass gain.

Causes for the unprecedented mass gain in the recent two years
To explore the underlying factors of the pronounced interannual variation of cumulative anomalies in total mass over the AIS, especially the observed mass gain in the recent two years, we calculate the interannual variation of cumulative anomalies in total mass and precipitation and derive corresponding crosscorrelation coefficients for individual basins.The results for the focused regions of Basin 12, 13, and APIS are shown in figures 5(a)-(c).
The cross-correlation coefficients for individual basins over the AIS range from 0.47 to 0.92, with a mean value of 0.75, indicating that the interannual changes in cumulative mass anomalies are primarily driven by cumulative precipitation anomalies (Kim et al 2020).For Basin 12, 13, and APIS, the interannual variations between cumulative anomalies in precipitation and total mass have a high correlation (figures 5(a)-(c)).For Basin 12 (Wilhelm II Land, Queen Mary Land) and Basin 13 (Wilkes Land), the total mass anomalies rapidly accumulated with 361 ± 23 Gt from January 2020 to December 2022, and the contemporary accumulation of precipitation anomalies was 389 ± 77 Gt.For the APIS, the total mass anomaly increase was 95 ± 47 Gt from May 2015 to December 2017, while the total mass anomaly from July 2020 to December 2022 was 111 ± 54 Gt.The corresponding increases in cumulative precipitation anomalies were 66 ± 31 Gt and 109 ± 83 Gt, respectively.Specifically, for Basin 13, two notably extreme precipitation events in March 2022 and October 2021 (supplementary figure S2) accounted for approximately 38% of the total precipitation anomalies over the three years (2020-2022).These two striking precipitation events have been linked to atmospheric rivers (Clem and Raphael 2022, Blanchard-Wrigglesworth et al 2023, Wille and the East Antarctica heatwave project 2023).For both Basin 12 and APIS, 26 out of 36 months (2020-2022) recorded precipitation levels exceeding the two-decade average (supplementary figure S2), indicating that the observed mass anomalies are related to the increased frequency of aboveaverage precipitation.
As the normalized 500 hPa geopotential height serves as a crucial indicator of atmospheric pressure and motion, we present, in figures 5(d) and (e),

Conclusions
We quantified the mass change over the AIS from January 2003 to December 2022 using the GRACE/GRACE-FO and atmospheric datasets.We analyzed the biennial mass change rates to demonstrate short-term mass change.The continuous mass loss in the AIS between 2003 and 2020 was 141.8 ± 55.6 Gt yr −1 .Subsequently, the AIS showed a mass gain of 129.7 ± 69.6 Gt yr −1 between 2021 and 2022, which was record-breaking within the past two decades of GRACE/GRACE-FO records.During this period, the mass gain over the East AIS and Antarctic Peninsula reached unprecedented levels within the past two decades, surpassing the mass loss observed in the Amundsen sector of the West AIS.Basin-scale analysis shows that the mass gain trends mainly occurred over Basin 12 (Wilhelm II Land and Queen Mary Land), Basin 13 (Wilkes Land), and the Antarctic Peninsula, at rates of 74.2 ± 8.0 Gt yr −1 , 63.2 ± 11.6 Gt yr −1 and 34.1 ± 3.8 Gt yr −1 respectively.The correlation analysis between the interannual variations of cumulative anomalies in precipitation and total mass suggests that the enhanced precipitation over the coastal EAIS and APIS primarily causes the observed mass gain.Further exploration indicates that the enhanced precipitation is driven by a pair of symmetrically distributed high-pressure systems over the southern ocean surrounding the Antarctic continent, which altered the direction of the prevailing westerly winds and transported more moisture toward the AIS.
Overall, our study continuously monitors the mass change of the AIS for the recent two decades and produces biennial updates of Antarctica's mass change rates in 2021-2022.The assessed biennial mass gain is unusual and offsets the previous contribution to GMSL of 1.3 mm.Our findings emphasize the impact of changing atmospheric circulation on the AIS and hold significant importance for projections of future sea-level rise.

Figure 1 .
Figure 1.Regional distribution of basins (1-27), with the EAIS colored in light orange, the WAIS in green, and the APIS in pink.Basin boundaries are obtained from Zwally et al (2012); the glacier and land information are referred to Mouginot et al (2017) and Cox et al (2023).

Figure 2 .
Figure2.Time series of the cumulative SMB, dynamic, and total mass anomaly relative to January 2003.A 3-month average is applied to the observed total mass anomaly, while a 13-month average is applied to the dynamic mass anomaly to minimize the possible high-frequency noise.

Figure 3 .
Figure 3. Biennial mass change rates of the total, dynamic mass changes, and cumulative SMB for the AIS, EAIS, WAIS, and APIS between January 2003 and December 2022 (a)-(d).
(165.9 ± 14.8 Gt yr −1 ).Such a reduced amplitude of the mass loss for the WAIS in 2019-2020 has been rare since the late 2010s.The deceleration of mass loss in the WAIS during 2019-2020 is generally consistent with the previously observed slowdown in the Amundsen Sea sector from 2019 to 2021, as indicated by ice sheet elevation changes from multi-altimeter observations (Yue et al 2023).In the APIS, there was a positive mass change rate in 2015-2016, as Bodart and Bingham (2019) reported.During 2021-2022, the mass change rate again turned positive, and exceeded the previous record of mass gain observed in 2015-2016, reaching an unprecedented large magnitude within the past two decades.

Figure 4 .
Figure 4. Trends of the total mass change from GRACE/GRACE-FO, cumulative SMB, and dynamic mass anomaly from January 2003 to December 2020 (a)-(c) and corresponding biennial rates from January 2021 to December 2022 (d)-(f).The corresponding errors are shown in supplementary figure S1.

Figure 5 .
Figure 5. Interannual variations of cumulative anomalies in total mass (red) and precipitation (blue) for basins 12, 13, and APIS (a)-(c); biennual averaged precipitation anomaly and normalized 500 hPa geopotential height anomaly for 2021-2022 (d) and (e), with arrows representing the corresponding wind field anomaly.