Empirical Run‐Time Bias Correction for Antarctic Regional Climate Projections With a Stretched‐Grid AGCM

This work presents snapshot simulations of the late 20th and late 21st century Antarctic climate under the RCP8.5 scenario carried out with an empirically bias-corrected global atmospheric general circulation model (AGCM), forced with bias-corrected sea-surface temperatures and sea ice and run with about 100-km resolution over Antarctica. The bias correction substantially improves the simulated mean late 20th century climate. The simulated atmospheric circulation of the bias-corrected model compares very favorably to the best available AMIP (Atmospheric Model Intercomparison Project)-type climate models. The simulated interannual circulation variability is improved by the bias correction. Depending on the metric, a slight improvement or degradation is found in the simulated variability on synoptic timescales. The simulated climate change over the 21st century is broadly similar in the corrected and uncorrected versions of the atmospheric model, and atmospheric circulation patterns are not geographically "pinned" by the applied bias correction. These results suggest that the method presented here can be used for bias-corrected climate projections. Finally, the authors discuss different possible choices in terms of the place of bias corrections and other intermediate steps in the modeling chain leading from global coupled climate simulations to impact assessment. Plain Language Summary Climate models are necessary and irreplaceable tools for climate projections, but despite continuous improvement, they still have biases, and their spatial resolution is too low to provide actionable climate change information at relevant small spatial scales. We present a method combining bias corrections and high-resolution climate modeling that allows improving climate projections at regional scales.