Supplementary Text

Basin-scale trends
Over regions down to basin scale (basin outlines are shown in Figure 1f), estimating the SAMΣ and/or Niño3.4Σterms has a substantial impact on the estimated long-term linear trend (see Table S1). For example, most of the linear trend in Droning Maud Land (Fig. 2d-f; basins 305-7 Fig. S4, S8) is accounted for by SAM (Table S1) in addition to major accumulation events(57 ). The long-term linear signal in the Denman Glacier region (basin 311) appears largely driven by SAM-related SMB (Fig. S10d-e, Fig. S11) and switches 19 Gt/yr to mass gain after the removal of the SAMΣterm. Long-term purely linear mass loss in the Totten Glacier region (basin 312) is reduced by the same amount, to -9±9 Gt/yr, suggesting much of the mass loss there is near-instantaneously related to SAM. By contrast, purely linear mass loss becomes more evident in Victoria Land (basin 313-4) after removing the effects of climate variability.
In West Antarctica, purely linear mass loss is reduced by about 30% from Marie Byrd Land to the Bellingshausen Sea (basins 321-323 Fig. S9), again suggesting substantial near-instantaneous response to changes in SAM, particularly through SMB (Fig. S10d-e). Linear mass loss in the southern Antarctic Peninsula (basin 324) reduces to negligible levels once the SAM component is removed, while the purely linear trend in the northern Peninsula (basin 325) increases from ‑21±2 to ‑36±3 Gt/yr keeping the same trend for the entire Peninsula (Fig. 4).
We note that these purely linear trends are subject to GIA model error, although the change in linear rate ascribed to climate variability is not. We also note that the overall mass change remains the same over the GRACE period and the distinction here is the attribution of some of the multi-decadal change to SAM and ENSO.