This study uses a newly-developed firebrand spotting parameterization in simulations of the Marshall Fire (2021) to demonstrate that without fire spotting, wind-driven fire simulations cannot reproduce the behavior of some fires. The Marshall Fire, the most destructive in Colorado’s history, took mere hours to cause nearly half a billion dollars in damage and destroy over 1000 homes. In wind-driven events that occur in the wildland-urban interface, the model’s ability to spot is critical for modeling fire spread over water streams and urban features such as highways. Without ignition of fire spots, the simulated Marshall Fire cannot advance. In cases when spotting significantly contributes to fire spread, the process’ nonlinear nature is a source of uncertainty to modeling fire behavior that can broaden the model’s ensemble spread and possibly produce a more realistic probability of outcomes. The results in this study corroborate the importance of representing fire spotting in atmosphere-fire behavior coupled models, such as WRF-Fire.