Coherent change detection (CCD) uses a reduction in the cross-correlation between two complex repeat-pass sonar images as a metric to identify changes between two surveys. This is made challenging by other sources of coherence loss that cannot be attributed to specific changes. This paper presents results from an automated coregistration method for the purpose of CCD applied to data obtained from an Autonomous Underwater Vehicle (AUV) equipped with a 100 kHz high-resolution synthetic aperture sonar (SAS). Two sites with varying complexity and seabed composition were repeatedly surveyed with and without objects with time intervals of 2/3/5/8 days. An automated method is presented that uses a series of increasingly fine control point matching and warping techniques in order to achieve sub-pixel accuracy in image co-registration. The coherence between a number of image pairs shows that it is possible for this environment to maintain coherence over time periods of operational relevance. It is also possible to detect changes using a drop in coherence, including very subtle changes not immediately detectable by visual inspection. A brief analysis of the sources of decorrelation, in particular the temporal decorrelation rate is also given.