An analysis of present and future seasonal Northern Hemisphere land snow cover simulated by CMIP5 coupled climate models

The 20th century seasonal Northern Hemisphere (NH) land snow cover as simulated by available CMIP5 model output is compared to observations. On average, the models reproduce the observed snow cover extent very well, but the significant trend towards a reduced spring snow cover extent over the 1979-2005 period is underestimated (observed: (-3.4 +/- 1.1)% per decade; simulated: (-1.0 +/- 0.3)% per decade). We show that this is linked to the simulated Northern Hemisphere extratropical spring land warming trend over the same period, which is also underestimated, although the models, on average, correctly capture the observed global warming trend. There is a good linear correlation between the extent of hemispheric seasonal spring snow cover and boreal large-scale spring surface air temperature in the models, supported by available observations. This relationship also persists in the future and is independent of the particular anthropogenic climate forcing scenario. Similarly, the simulated linear relationship between the hemispheric seasonal spring snow cover extent and global mean annual mean surface air temperature is stable in time. However, the slope of this relationship is underestimated at present (observed: (-11.8 +/- 2.7)%degrees C-1; simulated: (-5.1 +/- 3.0)%degrees C-1) because the trend towards lower snow cover extent is underestimated, while the recent global warming trend is correctly represented.