Various speed--accuracy tradeoffs were made in the viewshed programs. With the very fast, but less accurate, Xdraw program, we calculated a fuzzy visibility for each point. R3, the slowest program, was most accuracte, while R2 was much faster and almost as good. In the visibility index programs, LOS calculated a very fast approximate weighted index for each possible observer. Proc-LOS tested how many rays needed to be fired from each observer, and founds that 32 was as good as 128, both for the mass visibility indices, and for identifying the best points. Proc-LOS was run on 30,000,000 points. Some of its results were encapsulated in a PC-based demo program to help in observer siting.
Massive visibility index and viewshed calculations on databases have produced some tantalizing results. The question is now whether they extend to other DEMs, to data at higher resolutions, and to nonterrestrial data, such as Mars and Venus. Also, since the available data has frequent errors, and even when correct, it's not always obvious what the elevation of the representative point in each 3''x3'' rectangle means, further investigation into the robustness of these results is indicated. We are now continuing this research and extending it automatically to site observers to cover some specified terrain.