Current GIS data structures do not support error bounds. This is one
of the major weaknesses of analysis carried out in a GIS-one does
not know how to asses the quality of the result. One of the major
challenges in GIS is a computationally efficient theory of spatial
error.
There has been much discussion on the
topic. In 1966, Perkal [118],
for example, proposed fattening boundaries by convolving them with a
disk of radius ; these are usually
called buffer zones [146], and queries with buffer zones are an
important type of overlay. See the survey [62].
Computation in floating point is one source of error.
GIS data sets are riddled with degeneracies, so robust computation can
be important.