Fig. 2. Results of multi-variate regression . Climate index regression coefficients (a, b ) and the percentage of the partial-variance explained (R2) by the combination of the SAMΣ and lagged- Niño3.4Σ terms are shown (c ). The purely linear trends from a univariate (no climate indices) regression (d ) and multivariate regression (e ) and their difference (f ) are shown. In a, b, d, and e the hachured regions are not significantly different to zero at 95% confidence level. Basin outlines are shown in (f ) with basin numbers given in Fig. S3.