Despite its societal importance, accurately representing atmospheric blocking in climate models has proven to be a formidable challenge. Most generations of climate models (from CMIP3 to CMIP6) have shown significant negative biases in blocking frequency during boreal winter, particularly over the Euro-Atlantic sector. This underestimation of blocking frequency has often been attributed to an incorrect representation of the mean state that affects the propagation of Rossby waves. However, which physical processes responsible for bias in the mean state and eddy activity remain elusive and continue to be an active area of research.

In this study, we propose that cloud radiative effects (CREs) and their driving dynamics can influence wintertime blocking formation over the Euro-Atlantic Sector. To test this hypothesis, a suite of experiments is conducted in DOE’s E3SM atmospheric model, by either making clouds transparent to longwave radiation or by applying a cloud locking technique to disable interactive CREs. The results reveal that CREs exert a positive feedback to the formation of blocking events over the Euro-Atlantic sector. In particular, the frequency of occurrence of the blocking significantly reduces by up to ~20% in the cloud locking experiment and by up to ~35% in the experiment with clouds transparent to longwave radiation. Using a local finite-amplitude wave activity framework, we further find that the reduced blocking frequency without cloud radiative feedback does not arise from changes in the mean climate; instead, evidence is presented that cloud-radiative feedbacks enhance diabatic heating-induced wave source in the warm conveyor belt upstream of the block. This enhanced diabatic source, in turn, contributes critically to an enhanced zonal wave activity flux downstream, whose convergence drives the blocking formation during the blocking onset (akin to a blocking-traffic jam paradigm). These findings are supported by multi-model experiments within the Clouds Feedback Model Intercomparison Experiment Project (CFMIP), which show a notable ~10%-25% decrease in blocking frequency across models without CREs. Overall, the results highlight the necessity of correctly representing the distribution of CREs in order to accurately represent blocking frequency in climate models.