Abstract
Marine heatwaves (MHWs) in the Arctic are increasingly influenced by rapid warming and sea ice decline, yet large uncertainties remain in projections due to substantial inter-model spread. This study evaluates 19 CMIP6 models against OISST in simulating MHW mean intensity, max intensity, mean frequency, and annual total days during 1985–2014, and projects future MHW changes based on model performance. Although all models reproduce large-scale SST patterns, substantial biases exist in MHW simulations. Spatial bias patterns of mean and max intensity are similar within individual models but differ across models, whereas those of mean frequency and annual total days are highly consistent across models. Based on a comprehensive evaluation, six best-performing models are selected to construct the multi-model mean based on the best-performing group (MMM-BPG). Under SSP2-4.5, no significant MHW trends are detected because of compensating regional signals and rising detection thresholds, whereas all MHW features increase significantly under SSP5-8.5. Compared with the unweighted multi-model mean of all models (MMM-All), the MMM-BPG projects weaker MHWs overall, since most models included in the MMM-BPG generally project lower mean values of MHW features. Additionally, MHWs exhibit pronounced seasonal and regional differences. High sea ice concentration suppresses MHWs, while the strongest increasing trends occur in regions with intermediate sea ice cover due to sea ice melt and retreat. MHW projections are also sensitive to baseline selection, affecting both the magnitude and direction of projected MHW trends. This study improves understanding of CMIP6 model performance and future Arctic MHW changes.
•CMIP6 models exhibit substantial inter-model spread in simulating Arctic MHWs.•Best-performing group mean generally shows weaker MHW features than all-model mean.•MHW projections lack significant trends under SSP2-4.5 but increase under SSP5-8.5.•Arctic MHWs show seasonal and regional differences, with high SIC suppressing MHWs.•MHW projections are sensitive to baseline choice, with notable differences in trends.