We analyze the response of Arctic sea ice to a warming climate using earth system models (ESMs), including those from the Coupled Model Intercomparison Project Phase 6 (CMIP6). Our comprehensive evaluation compares historical sea ice representation against satellite observations, the Pan-Arctic Ice Ocean Modeling and Assimilation System, and the Regional Arctic System Model (RASM). We found that while the CMIP6 multi-model mean captures the mean annual cycle and the 1979–2014 sea ice trends, individual models vary significantly in spatial distribution and sea ice decline rates. Notably, only 40% of the CMIP6 models and 13% of the ensemble members accurately depict the observed trends and acceleration in sea ice area (SIA) decline. Sea ice volume simulations show even greater spread and uncertainty, indicating a need for improved observational constraints. Our findings highlight pronounced regional model biases and errors in ice edge and thickness, particularly in marginal and shelf seas, underscoring the limitations of current models in capturing key physical processes potentially linked to oceanic forcing. The sea ice trend analysis suggests that models with higher ocean heat transport better simulate sea ice declines, pointing to an emergent constraint related to ice-ocean interactions and the need for enhanced modeling of processes like frazil ice growth.

While global ESMs project a continuous sea ice decline over decadal time scales, achieving reliable seasonal forecasts remains challenging. To address this, we utilize a regional climate model, RASM, which allows us to forecast Arctic sea ice on timescales from weeks to six months. This study focuses on our findings related to September sea ice predictions from 2012 to 2021, specifically examining how lead time and initial conditions influence the quantitative skill of seasonal predictability for Arctic sea ice. Our results confirm that improving model physics and achieving more realistic initial conditions are crucial for skillful seasonal to decadal climate predictions and for enhancing the representation of sea ice in climate models.

By integrating advanced observational data with high-resolution modeling as part of the High Latitude Application and Testing of Earth System Models (HiLAT)-RASM project, our work contributes to reducing uncertainties in climate predictions and enhancing our ability to forecast Arctic sea ice dynamics.