Assessment of dynamical downscaling using the Regional Arctic System Model for sub-seasonal to decadal Arctic climate prediction

• Main Authors: Maslowski, W., Craig, A., Seefeldt, M., Comeau, D., Lee, Y., Osinski, R., Veneziani, M., Cassano, J., Clement Kinney, J.
• Format: Conference Object
• Language: English
• Published: 2023
• Subjects: Arctic; Pacific; Climate change; North Atlantic; Sea ice
• Online Access: https://gfzpublic.gfz-potsdam.de/pubman/item/item_5021000

The Regional Arctic System Model (RASM) is a fully coupled limited-domain climate system model that has been used for modeling and prediction of Arctic climate change at time scales from weeks to decades. Its pan-Arctic domain extends south to ~30 o N in the North Pacific and to ~40 o N in the North Atlantic oceans. The RASM ocean and sea ice model components are configured on rotated sphere meshes with the default configuration of 1/12 o (~9.3km), while the atmosphere and land components are configured on a 50-km grid. Its boundary conditions along the lateral boundaries and in the upper atmosphere are derived either from global atmospheric reanalyses or from Earth System Model (ESM) simulations. RASM hindcasts allow comparisons of results with observations in place and time, which is a unique capability not available in ESMs. Within this framework, RASM has been used to downscale output from three global models: the National Center for Environmental Predictions Coupled Forecast System version 2 (CFSv2), the National Center for Atmospheric Research Community Earth System Model initialized Decadal Predictability Large Ensemble (CESM-DPLE), and the Department of Energy Energy Exascale Earth System Model (E3SM). Forced with CFSv2, RASM has produced probabilistic intra-annual forecasts each month for the past 4+ years. The CESM-DPLE and E3SM output was used for downscaling multi decadal simulations. Here, we present a review of some of these results, including evaluation of RASM sea ice predictive skill in comparison with observations and relative to the original global output using the common metrics to quantify model skill.