Fig. 1: | Nature Communications

Fig. 1

From: Critical Southern Ocean climate model biases traced to atmospheric model cloud errors

Fig. 1

Linear regression of CMIP5 SST biases on AMIP5 net flux biases averaged over 40–60°S. These analyses were undertaken using the 18 models that provided suitable diagnostics for both CMIP5 and consistent AMIP5 experiments (see Methods and Supplementary Table 1). Our observational product uncertainty estimates are ~ 3 Wm−2 for net flux and ~ 0.04 K for SST. Observational product errors do not affect the regression slope, correlation or p values. However, the position of points for individual models and regression intercepts do include a contribution from observational error. The multi-model mean values are plotted with a solid black circle with a cross indicating their estimated observational uncertainties. The small multi-model mean bias estimates compared to the variations in the biases suggest that common structural errors in these parameters are not dominant for CMIP5 SST and AMIP5 net flux biases in this region and set of models