Estimate of Survey Positional Error

[Olly Betts]

On Wed, Apr 14, 2010 at 05:51:47PM +0100, Imperial College Caving Club wrote:
> We're trying to connect two large cave systems on our expedition in
> Slovenia this summer. To aid this, it would be very useful to have a
> best estimate of error in our survey, in particular the relative
> position of those two 'closest approach' stations in either cave. Of
> course, we have an approximate error from our loop closures, but we
> are wondering whether something with more mathematical sophistication
> might give a better grasp.
> 
> I believe that survex samples the error from the loop closure, and
> then distributes this according to an estimate for
> tape/compass/clino/positional error:

Survex essentially builds a model of the errors from all the instrument
readings (which as you suggest is used to distribute loop closure errors).
In theory you can calculate error ellipsoids for every station from this
data, but when I look into this (many years ago now) I think I found that
it wasn't easy to do due to how Survex does its calculations (which is
compact in memory usage because it does calculations in place in the
matrix).

It's worth noting that (ignoring information from loop closures) the
reliability of these ellipsoids is going to be entirely dependent on the
error estimates being used.  For distributing misclosures, only the relative
sizes matter, but here the absolute sizes do.

I agree it would be useful to know (and also interesting for being able to
give estimated error for cave depth), but I think it will probably require
substantial work on the loop closure code.

Cheers,
    Olly