The firebrand shower phenomenon, i.e., the generation and transport of firebrands away from the fire front and ignition of spot fires, is one of the major mechanisms of wildfire propagation, particularly in heterogeneous fuels. It is known to cause substantial losses in Wildland Urban Interface (WUI) zones. Despite the significance and its far-reaching impacts, most coupled fire atmosphere models and the operational models fail to accurately factor in the role of firebrands in wildfire spread. The primary reason is due to the inherent stochasticity and complex multi-scale physics of the problem that increase the computational costs for a detailed model. Also, most fire models do not have reliable and physics-based spot fire ignition feedback, and few, if any, use parametric models. Such limitation has hampered our ability to quantify the contribution of firebrand showers in the overall rate of spread and accurately estimate the risks from this phenomenon. In this talk, I review our efforts to understand the different phases of the firebrand shower and elaborate on our new multi-scale approach in numerical simulations. Finally, I discuss the challenges in creating a high-fidelity numerical model of the firebrand shower at large scales.