Many electrical grid transients can be described by the propagation of electromechanical (EM) waves that couple oscillations of power flows over transmission lines and the inertia of synchronous generators. These EM waves can take several forms: large-scale standing waves forming inter-area modes, localized oscillations of single or multi-machine modes, or traveling waves that spread quasi-circularly from major grid disturbances. The propagation speed and damping of these EM waves are potentially a powerful tool for assessing grid stability, e.g. small signal or rotor angle stability, however, EM wave properties have been mostly extracted from post-event analysis of major grid disturbances. Using a small set of data from the FNET sensor network, we show how the spatially resolved Green's function for EM wave propagation can be extracted from ambient frequency noise without the need for a major disturbance. If applied to an entire interconnection, an EM-wave Green's function map will enable a modelindependent method of predicting the propagation of grid disturbances and assessing stability.