Research brief: Mean state biases reduce natural variability in climate models

How trade winds changed: Schematic representation of the AGCM interbasin links and modelled surface flux changes. a, Schematic representation of the changes in atmospheric circulation induced by the Atlantic SST trend with the observed Atlantic climatology. b, Schematic representation of changes in atmospheric circulation induced by the Atlantic SST trend in the experiment with the biased CMIP5 climatology. The underlying surface radiation (contours) and latent heat flux (LHF, shading) changes in both panels are modelled changes from the corresponding CAM4 simulations where SSTs are set to climatology in the Pacific and Indian Oceans.


From 1992-2012, Pacific trade winds strengthened to levels never seen in the observational record. Although climate models could reproduce the timing of the change in the winds, they are not able to reproduce the magnitude of the strengthening as most models produce trends that are less than half the magnitude of those observed.

This trade wind strengthening had a significant impact on the Earth’s climate system. It led to sea level rise in the Western Pacific that was 3-times larger than the global mean and was associated with the slowdown in global average surface temperatures during the early 21st Century, referred to as the pause or hiatus.

A new paper in Nature Climate Change reveals why climate models struggled to reproduce the strength of the trade winds.

The key is the mean state bias of Atlantic Ocean basin in climate models. The mean state is another way of saying the normal conditions found in ocean basins when large influences, like the El Niño Southern Oscillation, are at not at play.

Climate researchers have long recognized that climate models don’t precisely capture this mean state, with clear temperature biases across ocean basins being one example of how this appears. 

Separately from this, natural variability in this system is not perfectly captured either, with the underrepresentation of decadal variability in the Pacific Ocean being a particularly thorny issue.

These two issues have often been treated separately. However, after investigating why climate models could not produce the acceleration in trade wind strengthening, it became apparent that solving mean state biases could also significantly improve the representation of Pacific decadal variability in climate models.

They found that because there is an atmospheric connection between ocean basins, such as the Pacific and Atlantic, the mean state bias in each ocean played a prominent role in how atmospheric connections played out. This had a direct impact on how natural Pacific decadal variability was represented in climate models.

However, the results of this paper also explain how variability in climate models can be better represented. This means the underestimate of natural Pacific decadal variability in climate models can potentially be corrected before a solution has been found for mean state bias issues.

UNSW logo ANU logo Monash logo UMelb logo UTAS logo