In this paper we review past work and present new analysis of seafloor temporal variability and how this variability impacts synthetic aperture sonar coherent change detection. The new work presented here links the decorrelation of complex images taken at different times with the decorrelation of seafloor roughness spectral estimates using a model based on first-order perturbation theory. Results are assessed through a comparison of decorrelation values generated by processing seafloor roughness data recorded by a digital photogrammetry system and complex SAS image data acquired with a translating source/receiver rail assembly. These data sets were collected off the western coast of Florida as part of the US Office of Naval Research-sponsored Sediment Acoustics Experiment 2004 (SAX04). In both the past work and new analysis, decorrelation was found to be frequency dependent with decorrelation times of hours to days setting a limit on reasonable time-frames for successful repeat-pass CCD.