Floating Traverse Chains


Surveys of complex caves typically contain several traverse chains, or sets of two or more traverses that belong to the same set of closed loops no matter how those loops are defined. Because chained traverses have identical error statistics, it's sometimes impossible to narrow the location of a blunder to a specific chain member when it's obvious from the statistics that at least one of them is bad. For the time being we may want to remove the chain's participation in the overall least-squares adjustment, treating it exactly like we would an obviously bad traverse. That would mean floating the entire traverse chain, not just an arbitrary part of it.

(Note: In Walls the operation of floating a vector or traverse causes the corresponding data to be given zero weight in a least-squares adjustment (or readjustment) of the containing loop system. Although the traverse itself is adjusted to conform to the remaining data, the system's statistics and all other vector estimates are the same as if the traverse had been thrown out or disconnected. The horizontal and vertical components of a traverse can be floated independently.)

In the Traverses list box of the Walls Geometry dialog, the statistics for chained traverses are grouped together and displayed with a gray background. The F-ratios, UVEs after deletion, and best corrections of the grouped traverses are initially identical. As soon as we float one of them, however, <FLT-00n> replaces the statistics of the floated traverse while a lower-case <brg-00n> replaces the statistics of all other traverses in the chain. (The number "n" identifies a traverse as belonging to the loop system's n-th chain as found by Walls during the initial data compilation.) Technically those unfloated traverses in the chain have become bridges -- the only remaining "hard" connections between what are now separate loop systems.

Although bridges aren't floatable in the usual sense, the FloatH and FloatV buttons on the Geometry page are enabled for traverse chain bridges whereas they're disabled for normal bridges. So what exactly happens when we float a bridge that's part of a chain?

To demonstrate we'll operate on a simple 2-traverse chain -- not a particularly bad one but an easy one to visualize. The image below shows portions of the Geometry and Map pages after we've floated the horizontal components of just one chain member. Earlier versions of Walls would have given us only two choices for interactively removing the chain's effect. We could have floated the 15-vector traverse as shown here (red with yellow unadjusted version), or alternatively we could have floated the other chain member, a 5-vector traverse that's colored light blue:

chain01

The program now provides a third option. As long as at least one member of a traverse chain is already floated, we can redistribute parts of the best correction to other members of the chain by floating them also. In this example, selecting the <brg-014> item and clicking the FloatH button immediately produces the following result:

chain02

Unlike ordinary float operations, floating bridges will not reduce the loop count or change the loop system's overall statistics. Likewise, when multiple members of a traverse chain are floated, unfloating one of them will not increase the loop count. It just redistributes the best correction across the remaining floated members in accordance with their assigned weights. Such operations will necessarily affect the computed coordinates (but not the vectors and statistics) of other parts of the network. In this example the location of the hat-shaped loop system separating the two chain members will be affected. By distributing the chain's discrepancy across two traverses we've at least avoided the worst-case error in that system's location.

Of course we can hope to do better than leave any traverses floated, chained or not. A traverse chain is a special case only because the statistics on the Geometry page alone can't isolate a blunder to one of its members. If the chain has a large F-ratio, we first look for the blunder in the usual way by floating each member individually while examining the Traverse and Map review pages. If by using those tools we can find no obvious error in a specific vector or traverse, we might then consider floating the entire chain for the purpose of creating a working map. As illustrated above, the Walls preview map uses different colors to highlight all traverse chain co-members of a selected traverse. This can be helpful since a large cave survey might have numerous hard-to-see traverse chains, some with several members separated by hundreds of meters.

For more information on traverse chains see the help file topic, Network Terminology.