Understanding inter-model spread in Arctic amplification: insights from a simple model

The Arctic is warming up to four times faster than the global mean, a phenomenon known as Arctic amplification (AA). AA is a robust feature of CMIP6 models, but projections show large inter-model spread, leaving its mechanisms uncertain. This accelerated warming results from local and remote processes contributing to projection differences. This thesis investigates the processes controlling Arctic warming and the source of the inter-model spread in AA. Diagnostics from climate models are combined with a reduced complexity conceptual model. Three CMIP6 general circulation models (GCM) with different degrees of AA are analyzed. AA is quantified from near surface temperature anomalies and feedback contributions are estimated with a radiative kernel method. The analysis shows that models with stronger AA reproduce observed Arctic warming more realistically, while models with weaker amplification underestimate Arctic and global temperature changes. Model differences are linked to Arctic feedback processes. Surface albedo and lapse rate feedbacks emerge as dominant contributors to Arctic warming and to the inter-model spread. A vertically resolved single column conceptual model of the Arctic climate system is used to further investigate model differences. The model simulates the coupled evolution of the ocean mixed layer, sea ice and atmosphere and is calibrated with parameters from the analyzed GCMs. It reproduces the relative differences in Arctic warming and sea ice loss across the three GCMs. Mechanism denial experiments isolate the role of individual processes. Results show that within the conceptual framework lapse rate and water vapor feedbacks exert the strongest control on Arctic temperature changes. Reduced complexity models can capture key features of climate system evolution, reproducing the inter-model spread encoded in the calibration parameters. A simplified framework helps interpret Arctic warming mechanisms and sources of uncertainty in AA projections.