Towards Optimizing Hierarchical Data Revisions

The revision of existing data must always be considered when new data are collected which have known relations with the old data, thereby taking into account that the two datasheets in question may belong to one and the same, or to two different hierarchical levels. In the first case, optimal data fusion would amount to a joint adjustment and, as a result, to modifications of the existing data which may then be checked for their significance. In the second case, the situation turns out to be somewhat trickier since, after the integration, the old dataset with a higher position in the hierarchy should still be unaffected, including the corresponding dispersion matrix. Here we shall explore the optimal procedure for the second case and present a unifying algorithm which would allow us to go ahead with the revisions until (only in the last step) we have to decide about the hierarchical behaviour. Although some of the more theoretical questions must be left unanswered at this point, we do include an example in which two photogrammetric networks of substantially different scales are to be integrated.