Slide 16 of 29
This approach was published only in 1995 ! Snapshot data is processed into Events which define when the data changed: changes, not states are stored. The research prototype stored only changes (so removing redundancy and consistency problems). A “component” is a pixel (raster cell).
A series of these events is required for each theme (layer) of the raster dataset.
This is probably a good minimum target to aim for when looking for a t-GIS to support land-cover analysis where you are going to write the algorithms yourself but want some simple repository support. You would also want the t-GIS to handle automatic registration of images acquired at the different times, projection transformations, tools to manage pixel skew and overlap between snapshots, and something to make it easy to deal with several distinct themes over the same area but with non-coincident snapshot times, e.g. radar and optical satellite, aerial photography.
It can’t handle areas which move (change location), you would need to do that yourself. It just does cells (pixels).
Example and diagram from May Yuan’s paper on NCGIA website
