Fig. 3: Increased sensitivity of an extratropical response with a northward-broadened anomaly pattern.

a, Comparison of the extratropical response to tropical forcing over the 1900–1999 (blue bars) and 2000–2099 (red bars) periods. Models that simulate a decrease are greyed out. Error bars on the multi-model mean are calculated as 1.0 s.d. of 10,000 inter-realizations of a bootstrap method. The horizontal dashed line indicates observation. b, Inter-model relationship between changes (2000–2099 minus 1900–1999) in multi-year La Niña numbers and in the extratropical response to tropical forcing, both scaled by the increase in global mean SST of each model. Linear fit (solid black line) is shown together with correlation coefficient r and corresponding P value. c,d, Multi-model mean pattern of the extratropical response to tropical forcing in MAM(1) during 1900–1999 (c) and its change during 2000–2099 (d). The response pattern is obtained in each model by first regressing grid-point MAM(1) SST (colouring) and 10-m wind (vectors) anomalies onto the normalized time series of SST expansion coefficient from SVD and then multiplying the regression coefficients by the s.d. of the SST expansion coefficient, separately for 1900–1999 and 2000–2099 (see text for details). Extratropical anomalies are more sensitive to tropical SST forcing and the first La Niña of a multi-year La Niña features further northward-broadened anomalies in the North Pacific during 2000–2099 than during 1900–1999.