One of the principle obstacles to the utility of passive seismic data is its bulk. With several hundred to thousands of geophones, we are able to generate mountains of data in a very short time. The simplest method of trimming down this volume is to keep only the recorded wavefield around times when usable source energy is known to be present. In the case of teleseismic imaging, or when utilizing unconventional, but known sources, this is easily done. However, if one hopes to image with the truly ambient noise field, time windowing amounts to removing needed signal.
After cross-correlation of passive traces, one can produce shot-gathers equivalent to an active source experiment. Having done so, we realize that most of the recording time axis can be discarded and maintain only enough time/lag samples to keep the reflection from the deepest subsurface structure of interest. As most of the processing of seismic data is accomplished in the Fourier-domain, I will consider the procedure and ramifications of time windowing after data have been transformed to frequency.