Fig. S12. Results of multi-variate regression using time-differenced GRACE WAIS basin data . Panel a shows the time-differenced GRACE data and the SAM and ENSO components of the model fit to it (Equation 3). Panel b shows the GRACE data with the cumulative sum of the SAM and ENSO model components, and the residual after removing the SAM and ENSO components of the model. Panelc compares the time-integrated SAM+ENSO model with that estimated using unmodified (undifferenced) GRACE data (as in Fig. 3). The shaded regions are the 1-sigma uncertainty of the respective models.
Table S1 . Regression parameter estimates with their 2-sigma uncertainties. Also shown are the contributions to the trends from Marshall SAMΣ and Niño3.4Σ and the contribution to the trends from the ICE6G_D and Caron et al. 2018 GIA models. The SAMΣ and Niño3.4Σcoefficients are d and e , respectively. The percentage of the trend bunivariate due to SAM and ENSO, reported in the text, is calculated by 100*(bunivariate -bmultivariate )/bunivariate.The sum of the basin regression estimates for each of the ice sheets is also shown.