Fig. 2: Relationships between SST and surface heat-flux in the midlatitude and subtropical North Atlantic.
Scatter plots of decadal surface heat-flux (Qnet) anomalies (defined positive upwards) versus decadal SST anomalies over the midlatitude North Atlantic region (shown by green boxes in Fig. 1) for (a) the idealised IPSL pacemaker experiments, (b) the idealised UM pacemaker experiments, (c) Historical simulations, and (d) piControl simulations. Each dot shows a value averaged over a different 10-year period and simulation and the ellipses show the two-dimensional Gaussian probability density function calculated across all the dots shown. The black dots show the decadal anomalies from the observational datasets (i.e. HadISST and 20CR). (e, f) As in (a–c) but for anomalies over the subtropical North Atlantic region (shown by green boxes in Fig. 1). Also shown is the regression of decadal surface heat-flux anomalies onto decadal SST anomalies for (g) the midlatitude North Atlantic region and (h) the subtropical North Atlantic region. The regressions were calculated for all unique 140-year periods in the historical and piControl simulations and the box, centre-lines and whiskers show the 5th-25th-50th-75th-95th percentiles of these distributions. For the SST-restoring experiments 10000 random 140-year periods (i.e. 14 decades) were sampled and constructed from the ensemble members to calculate the regression; the box and whiskers show the distribution across these random samples for each experiment. The crosses and dotted lines show the equivalent regression coefficient calculated from the observational datasets. The vertical black lines are an estimate of the 5-95% confidence limits of the regression coefficient calculated from the observations, calculated using a block bootstrap resampling with replacement using a block length of 20-years (repeated 10,000 times).
