Hi,
Yesterday I sent a mail to this list in reply to Peter Iles's earlier mentioning
of the Anytown barrow example, but I think somehow it didn't get through to the
list subscribers. Since today he again asked:
> Are there any Information Scientists out there who deal with
> this sort of thing and can give us pointers, theoretical and practical, to
> methods of dealing with these 'faulty' datasets?
I will give it another shot.
Firstly, to generalise from the examples given yesterday and today, what we want
is metadata fields which indicate the quality of our data fields. So, if we have
a point location for some monument, then there should also be at least one
database field containing an indication of the reliability of that location.
There are many ways in which this can be done, including simple radii ("the real
position of this monument is within 50 m of its stated position"), zonation
("the monument is somewhere within this administrative area (polygon) but we've
no idea where"), and probability functions ("fuzzy" locations). I'd guess that
all of these could be implemented on today's simple systems, eg. an Access RDBMS
and ArcView/MapInfo/Idrisi GIS, although if you're going to work with
probabilities you're probably (no pun intended) better off with raster GIS than
with vector GIS. However, en users (including planning officers) may well prefer
easier to interpret *presentations* of the metadadata, using some device such as
the Level 1, 2, etc suggested by Peter.
Secondly, it is obviously not just the location of archaeological remains which
may be uncertain. Everything from identification to chronology is to some
(large) extent a matter of interpretation, and can/will change over time. While
this means that an SMR database must be careful to separate observations from
interpretations, and to use metadata fields to record the degree of uncertainty
in the latter, for the purposes of this e-conference we should ask ourselves how
much of this should carry over into the GIS? For example, do you want the GIS to
be able to distinguish between 'reported' barrows, and 'confirmed' barrows?
Third and lastly, I'd like to bring up a point which hasn't been mentioned yet -
I believe that SMR records generally, and not just in Britain, represent only a
small portion of the extant archaeology, and that (for planning control
purposes) it is much more important to look the *likelihood* of archaeological
resources being present in an area. This takes us into the area of predictive
modelling which (in the Netherlands at least) is already deeply affecting
planning policy from the national down to the municipal level. Unfortunately,
here again we must deal with probabilistic maps rather than point or zone
(polygon) maps.
So the conclusion I would like to submit to you is, that GIS standards for CRM
must include a) standards for metadata for locational precision and accuracy, b)
standards for the representation and description of probabilistic data. I repeat
that there are no technical problems in implementing these, but I am aware that
current SMR do not contain such information at the moment, and users will also
not be familiar with such 'fuzzy' formats.
Martijn van Leusen
PhD student,
University of Groningen
