Constraining CMIP6 model ensemble spread to reduce uncertainty in the representation of the Atlantic water layer temperature in the Arctic Ocean

We aim at reducing the uncertainties in the climate predictions of the Arctic region which is going under rapid changes with global repercussions. We analyse the spread in the Atlantic water core temperature across multi member CMIP6 historical simulations, focusing on different regions of the Arctic Ocean. While the redistribution of heat plays a critical role in the dynamics of the Arctic Ocean basins, it is usually not well represented in climate models, leading to divergent projections of future changes in the Arctic. To address this limitation, we compare CMIP6 model outputs with available reanalysis and observational products, in order to identify the biases within the model simulations and develop new metrics to constrain the model ensemble spread. Such metrics can be used to select the multi model ensemble members and construct a subsample with improved representation of the core temperature evolution over the historical period resulting in a reduced uncertainty in near-term future projections of the Arctic climate.