Because of the challenges in forecasting CAD along the eastern seaboard, we investigate the sources and scales of the predictability of CAD induced by synoptic forcing—in particular, the occurrence of a cold surface high north of our region of interest. In this research we utilize an objective CAD identification algorithm (Bailey et al. 2003), high-resolution mesoscale NWP forecasts, and a novel Bayesian multi-scale dynamical-statistical forecasting method building upon analogue post-processing techniques in the literature (Delle Monache et al. 2013, Frediani et al. 2017). In addition, our hybrid forecasting method is sensitive to the relationship of ensemble skill and spread, allowing us to explore how differing periods of predictability are reflected in the forecast ensemble’s variable dispersion.