Bathymetric monitoring provides coastal scientists, engineers and managers with morphometric information of wide-ranging application. Even with access to such data, it may not have the temporal resolution, or the time-series continuity to span the duration of cyclic trends of relevance in documenting and predicting rates and magnitudes of change. For some coastal zones, certain archival data sets might yield such improvement to the derivable record, thereby extending its value. In these terms, we report here an approach of combining the data-integration power of Geographic Information System (GIS) with the "learning power" of Artificial Neural Networks (ANN) for deriving water depths from vertical real-colour aerial images. Calibration of spectral signatures for depth was achieved by identifying image pixels corresponding to spatial points of known water depth. The rules of mapping from color to depth were established and held by employment of an ANN training process. Applying these rules, pixel by pixel, for any new images covering the same spatial area the depth pattern at the image snap-shot time can be derived. This approach was tested with spatial data representing the Lakes Entrance sector of the coast of Victoria, Australia. Results demonstrated the potential of ANN in dealing with estimation problems, especially regarding those calling on data assembled via deployment of spatial and/or geographic science. It is a low-cost procedure offering potential to give significant advantage in time-series archive assembly for applications that do not require the accuracy that can now be obtained by GPS-controlled and very detailed sounding methods.