The Missing Measurement Problem

The Mexican cave Gruta de Bustamante is primarily a 100+ meter wide passage whose floor is a steep slope of huge boulders -- some almost house-sized.  A detailed survey in progress, conducted by Jan and Orion Knox, consists of many measurements between station flags placed prominently on boulders throughout the passage.  For this kind of project the surveyors might reasonably ask: "If we take numerous azimuth and inclination measurements between the stations without measuring most of the corresponding distances, will Walls be able to handle the data and produce an optimal adjustment?"

The short answer is: Not very elegantly.  The only recourse for now is to substitute approximate values for the missing distances (such as those obtained in a compilation in which the corresponding vectors are floated) and run Walls in an iterative fashion, obtaining new approximations with each run.  The iterations would stop when further changes to the dummy values would not significantly improve consistency by lowering the UVEs.

The same problem can arise when specific measurements recorded in fieldbooks are either undecipherable or so obviously far off as to be worthless. Ideally, the remaining good measurements that make up a shot would still be given appropriate weight in a least-squares adjustment -- assuming loops are present, of course.