While the presence of ghosts has been known for decades, the impact of these apparitions on remote sensing observations has gone unquantified, leaving atmospheric corrections susceptible to ghosting. In this work, we present the first spectral characterization of three common ghost types and provide a framework for incorporating these properties into atmospheric correction algorithms. We demonstrate the effect of this improved atmospheric ghost correction compared to an atmosphere-only implementation. Lastly, we discuss some preliminary spatiotemporal variations in haunting intensity and type.