Antarctic ice mass variations from 1979 to 2017 driven by anomalous precipitation accumulation

Global mean sea level (GMSL) has been rising at an accelerated rate, increasing from ~1.9 mm/yr over the 20th century to ~3.1 mm/yr in recent decades. While mountain glaciers and the Greenland Ice Sheet have been primary contributors, the Antarctic Ice Sheet (AIS) contributed ~0.3 mm/yr over the past three decades. IMBIE2 reported that AIS mass loss rates increased from about −47 Gt/yr (1992–2006) to −194 Gt/yr (2007–2017), suggesting an apparent recent acceleration. AIS mass balance reflects the difference between ice discharge (D) and surface mass balance (SMB), the latter dominated by precipitation for Antarctica. Conventional projections often assume slow variability in D and negligible long-term trends in precipitation. However, prior comparisons of GRACE data with SMB reanalyses indicate that precipitation anomalies can account for a substantial fraction of regional AIS mass loss accelerations. This study aims to reassess AIS mass changes since 1979, quantify the contribution of precipitation accumulation to SMB and total mass change, and diagnose the climate modes (e.g., SAM, ENSO) modulating these variations to better interpret apparent accelerations in AIS mass loss and their implications for sea level rise.

The study builds on: (1) century to recent-decade GMSL trends and budgets attributing increased rates to cryospheric mass loss; (2) IMBIE2’s pan-Antarctic mass balance assessment showing a step-like increase in AIS mass loss around 2007; (3) reanalysis and regional climate model work (ERA-Interim/ERA5, RACMO, MAR) showing Antarctic precipitation variability with limited linear trends; (4) findings that SMB variability can obscure detection of accelerations in ice sheet mass loss; (5) evidence that SAM and ENSO modulate Antarctic precipitation patterns; and (6) satellite geodesy (GRACE, altimetry) and remote sensing showing regional accelerations in ice discharge. Prior work also identified spurious barometric pressure signals contaminating GRACE-based Antarctic mass trends and accelerations, motivating improved corrections.

• Data and period: Precipitation fields from the ERA5 reanalysis (1979–2017) over Antarctic drainage basins (Zwally et al. definition). Comparable model series (ERA-Interim, RACMO2.3p2, MARv3.6.4) examined for consistency. GRACE CSR RL06 mascon solutions used for 2003–2017 surface mass change. IMBIE2 AIS mass balance series (1992–2017) used for total mass evaluation.
• SMB construction: Surface mass balance approximated by accumulated (time-integrated) precipitation. Linear trends (mean precipitation rate) removed to focus on variability and acceleration in the accumulated series (ΔSMB). Acceleration rates estimated by quadratic fits a0 + a1 t + a2 t^2 to ΔSMB at grid points; only statistically significant accelerations (95% CI) mapped.
• Climate mode diagnostics: Performed Empirical Orthogonal Function (EOF) analysis on area-weighted ΔSMB fields (matrix decomposition ΔSMB = U S V^T). Interpreted spatial modes (U columns) and principal components (V columns), with explained variance from singular values. Applied Rotated EOF (REOF) with VARIMAX rotation to uncouple modes and better isolate patterns associated with SAM/ENSO. Compared time-integrated, detrended SAM (ΔSAM) and ENSO (SOI; ΔENSO) indices with REOF principal components; computed correlations. Additional linear regression at each grid point projected normalized ΔSAM onto ΔSMB to predict SAM-driven SMB patterns and amplitudes.
• GRACE–SMB comparison and pressure correction: Compared detrended GRACE mass change (ΔM) to ERA5 ΔSMB (2003–2017). Recognizing barometric pressure modeling errors in GRACE over Antarctica, applied EOF to the difference ΔM − ΔSMB (after 600 km Gaussian smoothing) to identify the continent-scale synoptic first mode interpreted as pressure error, which was then removed. Estimated GRACE-based ΔSMB′ by subtracting an independent ice discharge acceleration term (ΔD) and correcting for pressure error.
• Ice discharge estimation: For 1992–2017, used IMBIE2 total AIS mass (M). Computed detrended variability ΔM and subtracted ERA5 ΔSMB to estimate ΔD (assuming negligible inland meltwater). Fitted a quadratic to ΔD to quantify acceleration. Regional analyses separated West Antarctica from the whole AIS where noted.
• Statistical inference: Trends and accelerations reported with 95% confidence intervals. Spatial filtering and area weighting applied as appropriate. Units: accelerations generally in Gt/yr^2; correlations reported for REOF PC–index comparisons.

• Accumulated precipitation controls inter-annual AIS mass anomalies during the GRACE era: Time-integrated precipitation (ΔSMB) explains most inter-annual variability in Antarctic mass change from 2003–2017.
• Bi-polar SMB acceleration pattern since 1979: Significant accelerations in accumulated precipitation produce negative acceleration (mass loss) in the Pacific sector and positive acceleration (mass gain) in the Atlantic and Indian sectors. Spatial average acceleration over AIS is near −0.3 Gt/yr², with regional values of about +2 Gt/yr² (Atlantic–Indian) and −2 Gt/yr² (Pacific).
• SAM is the dominant modulator of multi-decadal SMB variability: The first REOF mode explains 30% of variance and its PC correlates strongly with ΔSAM (r = 0.89) and moderately with ΔENSO (r = 0.62). SAM-driven regression reproduces the observed bi-polar spatial pattern and amplitude. In two opposing regions, SAM explains ~72% (1.5 ± 0.1 of 2.1 ± 0.1 Gt/yr) and ~67% (−1.6 ± 0.1 of −2.4 ± 0.2 Gt/yr) of the multi-decadal acceleration in ΔSMB.
• Validation with GRACE after pressure correction: Raw GRACE ΔM shows higher frequency variability and larger apparent acceleration (−8.6 ± 2.1 Gt/yr²) than ERA5 ΔSMB (−5.1 ± 1.9 Gt/yr²), due to inclusion of ΔD and barometric pressure errors. After removing ΔD (~−7 Gt/yr²) and correcting the continent-scale pressure mode, GRACE-derived ΔSMB′ agrees well with ERA5 ΔSMB; GRACE-based ΔSMB′ acceleration is ~1.6 Gt/yr² over 2003–2017.
• Ice discharge has accelerated steadily since 1992: From IMBIE2 M minus ERA5 ΔSMB, ΔD exhibits a steady acceleration of 8.7 ± 0.3 Gt/yr² (sense: increasing discharge each year), not an abrupt change.
• Apparent 2007 step in AIS mass loss is SMB-driven: The abrupt increase in IMBIE2 mass loss rate after 2007 results from the combination of steadily accelerating discharge and a contemporaneous SMB shift. SMB mitigated discharge-driven acceleration before ~2007 and enhanced it afterward. ERA5 annual precipitation trends increased (1994–2005) then decreased (2006–2014), consistent with the integrated ΔSMB transition around 2007.
• Contribution of SMB to the 1992–2006 vs 2007–2017 rate difference: Of the ~147 Gt/yr increase in AIS mass loss rate reported by IMBIE2, −39 Gt/yr (~27%; ~0.11 mm/yr of GMSL) is attributed to SMB changes, with a larger fraction (~41%) in West Antarctica.

The study set out to disentangle the roles of surface mass balance (precipitation accumulation) and ice discharge in recent AIS mass changes. Findings show that while precipitation rates have negligible long-term linear trends, their time integration yields substantial inter-annual to multi-decadal SMB variability that strongly influences observed mass change. The dominant multi-decadal SMB pattern is bi-polar across Antarctica and closely tied to SAM variability, indicating large-scale atmospheric circulation governs precipitation accumulation anomalies. Correcting GRACE for barometric pressure errors and removing discharge contributions confirms that reanalysis-based SMB captures observed variability. Consequently, the perceived abrupt increase in AIS mass loss around 2007 largely reflects a phase change in SMB superimposed on a steady acceleration in ice discharge, rather than a sudden shift in ice dynamics. These insights emphasize that SMB variability must be accounted for to accurately diagnose ice dynamic changes and to improve projections of AIS contributions to sea level rise.

We examined long-term AIS SMB (accumulated precipitation) variations for 1979–2017 using precipitation fields from the ERA5 reanalysis. Even though AIS precipitation rates do not exhibit a significant trend, SMB shows strong inter-annual and multi-decadal variations. We found that multi-decadal SMB variations are related to SAM. AIS SMB modulated by SAM shows a distinct bi-polar pattern with negative acceleration in the Pacific Sector and positive acceleration in Atlantic and Indian Sectors. Model predictions of SMB variations are observed by satellite geodetic observation like GRACE. After correcting for SMB, we found a steady acceleration of ice discharge of −8.7 ± 0.3 Gton/year² for the period 1992–2017. The apparent abrupt change in 2007 is not associated with a change in ice dynamics but instead with SMB variations mostly in the West AIS.

• Reanalysis dependence: SMB results rely on ERA5 precipitation fields; while cross-checked with other models, uncertainties remain due to data assimilation and forcing (e.g., ECMWF).
• GRACE corrections: GRACE-based SMB required empirical removal of a continent-scale barometric pressure error using EOF; residual errors and noise may persist, particularly at high frequencies.
• Assumptions: Inland meltwater contribution to mass balance was assumed negligible; sublimation and runoff were considered minor for AIS, which may not hold locally.
• Mode separation: Ordinary EOF modes mixed multiple signals; although REOF/VARIMAX improves interpretability, some ambiguity in mode partitioning remains.
• Spatial/temporal smoothing: A 600 km Gaussian filter was applied for GRACE–SMB differencing, potentially spreading signals beyond glacier outlets and affecting localization.
• Period and generalizability: Analyses span 1979–2017 (or 1992–2017 for IMBIE2), which may not capture longer-term climate variability beyond these windows; confidence intervals indicate non-negligible uncertainties (e.g., ΔSMB accelerations).