Assessment of sea ice-atmosphere links in CMIP5 models

Abstract

The Arctic is currently undergoing drastic changes in climate, largely thought to be due to so-called ‘Arctic amplification’, whereby local feedbacks enhance global warming. Recently, a number of observational and modelling studies have questioned what the implications of this change in Arctic sea ice extent might be for weather in Northern Hemisphere midlatitudes, and in particular whether recent extremely cold winters such as 2009/10 might be consistent with an influence from observed Arctic sea ice decline. However, the proposed mechanisms for these links have not been consistently demonstrated. In a uniquely comprehensive cross-season and cross-model study, we show that the CMIP5 models provide no support for a relationship between declining Arctic sea ice and a negative NAM, or between declining Barents–Kara sea ice and cold European temperatures. The lack of evidence for the proposed links is consistent with studies that report a low signal-to-noise ratio in these relationships. These results imply that, whilst links may exist between declining sea ice and extreme cold weather events in the Northern Hemisphere, the CMIP5 model experiments do not show this to be a leading order effect in the long-term. We argue that this is likely due to a combination of the limitations of the CMIP5 models and an indication of other important long-term influences on Northern Hemisphere climate.

Appendix: Corrections to surface pressure

Examples of the differences in the pressure fields found in the CMIP5 models can be seen in Fig. 11, where we have plotted the global mean anomaly of surface pressure, sea level pressure, and water vapour pressure (calculated from the water vapour content multiplied by gravitational acceleration) from the Historical and RCP8.5 scenarios for the CMCC-CM, MIROC5 and ACCESS1.3 models.

Fig. 11

Examples of pressure variables, from the historical and RCP8.5 scenarios joined together, from three CMIP5 models as labelled. All values are global anomalies w.r.t. 1960–2000 climatologies, smoothed with a 4 year Hanning window. Surface pressure is shown by the blue line, sea level pressure by the green line and the water vapour pressure shown by the red line. [The surface pressure for the MIROC5 model is not visible but is equal to the water vapour pressure.]

The CMCC-CM model shows no trend in surface pressure over the 250 years of the historical and RCP8.5 simulations (the curves have been smoothed for ease of comparison), suggesting this is a ‘dry’ pressure, i.e. no water vapour is included. The water vapour pressure rises, as would be expected in a warmer atmosphere that can hold more moisture. The sea level pressure shows a drop over the same period. This is likely due to the derivation of sea level pressure over land by extrapolating using the local surface temperature—as the surface temperature rises, the sea level pressure will be lower. 16 of the models in total showed this behaviour—with a flat surface pressure curve but falling sea level pressure.

The MIROC5 model shows an increase in surface pressure exactly equal to that of the water vapour pressure, showing the surface pressure contains a contribution from water vapour. The sea level pressure also shows a rise, but it is lower than that of the surface pressure, due to the competing effect of the extrapolation over land, as described above. 16 of the models in total showed this behaviour—with a surface pressure rise equal to that of the water vapour pressure.

The ACCESS1.3 model shows increasing surface pressure, sea level pressure and water vapour pressure from the year 2000, but the rise in surface pressure cannot be determined from the change in water vapour. There were a total of 9 models which provided one or more pressure variables, but similarly showed no clear relation, or else did not provide both surface pressure and water vapour.

In order to use a consistent pressure for calculating the Northern Annular Mode, we used the surface pressure from only those models which showed a flat surface pressure curve (such as CMCC-CM), and those where we could remove the water vapour pressure to create a new, dry, surface pressure with no trend (such as MIROC5). Those models to which we have applied the correction have a ‘+’ in the ‘PS’ column in Table 2.