The nature of the Arctic lapse-rate feedback: Spatial distribution, seasonality and trends in ERA5 and CMIP6 data

The Arctic amplification is driven by several intertwined causes including the interplay of locally positive radiative feedbacks. The lapse-rate feedback (LRF) is a dominant driver of Arctic amplification and arises from the vertically non-uniform warming in the troposphere. In the Arctic, the LRF enforces a positive radiative feedback as the warming is most pronounced at the surface, but becomes smaller at higher altitudes which feedbacks positively on the initial greenhouse effect. This stands in contrast to the processes in the tropics, where a stronger warming of the upper troposphere dampens the greenhouse effect.

We investigate the nature of the Arctic LRF by using ERA5 Reanalyses and CMIP6 models to compute the feedback via simplified radiative transfer calculations (radiative kernels).

The Arctic LRF is unique in terms of its geographic distribution, seasonality and time evolution. From a global perspective, the LRF is most positive in Arctic winter, but shows the strongest seasonality as it becomes negative in summer over the sea ice covered ocean. Our trend analysis shows that the positive winter LRF increased strongly during the past 30 to 40 years. This increase during boreal winter mediates the annual response and accounts for all Arctic surface types which we define as sea ice, sea ice retreat, open ocean and land. A special focus lies on regions of retreating sea ice, where the positive LRF is strongest throughout the year.

Our results are embedded in previous studies on the changing Arctic atmospheric energy budget through CO2-driven climate change. They show strongly increasing surface heat fluxes over areas of retreating sea ice which is mostly compensated by a decrease in atmospheric transport convergence, both of which can shape the maximum of the high-latitude positive LRF.

We finally carry out an inter-model comparison of linear trends of the Arctic LRF during the past 30 years of historical CMIP6 simulations. This includes more than 50 models to determine the performance of each model by relating to reanalyses data.