Using the case of the rapid intensification (RI) of Hurricane Harvey (2017), convection-permitting initialization, analysis, and prediction from a cycling ensemble Kalman filter (EnKF) that assimilates all-sky infrared radiances from GOES-16 was conducted. Ensemble forecast experiments demonstrated that significant growth in ensemble spread was induced by the individual initial perturbations either in wind or moisture fields. Their nonlinear interactions further limited the predictability of the intensification process of Harvey by increasing the uncertainty in simulating the pouch-shaped wind and moisture distributions, and modifying the convective activity and its feedback on vortex flow. In addition to tropical cyclones, we have recently conducted a study of the contribution of multi-variable nonlinear interactions on the predictability of individual convective events using a meso-scale convective system that occurred during the Convective Processes Experiment (CPEX) field campaign. This study highlights the importance of better initializing the moisture and dynamic state variables simultaneously, and the potential contribution of all-sky satellite radiance assimilation to constrain convective activity that leads to the development of severe weather events.