Abstract
Simulating and predicting Arctic sea ice accurately remains an academic focus due to the complex and unclear mechanisms of Arctic sea ice variability and model biases. Meanwhile, the relevant forecasting and monitoring authorities are searching for models to meet practical needs. Given the previous ideal performance of cGENIE model in other fields and notable features, we evaluated the model's skill in simulating Arctic sea ice using multiple methods and it demonstrates great potential and combined advantages. On this basis, we examined the direct drivers of sea-ice variability and predicted the future spatio-temporal changes of Arctic sea ice using the model under different Representative Concentration Pathways (RCP) scenarios. Further studies also found that Arctic sea ice concentration shows large regional differences under RCP 8.5, while the magnitude of the reduction in Arctic sea ice thickness is generally greater compared to concentration, showing a more uniform consistency of change.