tmoritz at CAS.CALACADEMY.ORG
Wed Aug 17 02:00:29 CDT 1994
Seems to me there are several related problems/questions being raised
by Steve Shattuck and Robert Robbins.
When Steve says:
"It is very helpful (and more honest) to include some error estimate along
with the geocoded data." He is addresseing the crux of our problem...
But also raising a series of thorny issues...
Any use of "enhancement"/"extrapolation" techniques with respect to
locality data must always be explicit.
It is true that for rivers and streams, often the Lat./Long. coordinates
are taken from the mouth of the stream (or from some other arbitrary point).
(There are alternatives -- such as the river mile index system...)
It is also apparently true that at least in the past, some coordinates for
geographic features were typographically derived -- that is, the
coordinates for the feature were calculated from the first letter of the
feature name as it appeared on the base map. Thus whether print or
electronic, there are problems but at least, they are *standard* problems...
Then there is simply the issue of highly variable precision in the
description of the locality as it is *expressed* either on the specimen
label or in the related field notes. This is a structural problem which
pertains to all specimen data sets. (It may have a subjective aspect in
that some collectors are known to be more reliable (accurate) than others
-- irrespective of what is written down.) In collating such variable data,
how should we qualify these differences in resolution?
It also seems absolutely clear that the original provenance (in archival
terms) of the specimen data needs to be preserved regardless of how we may
end up "enhancing" those data. Community agreement on linking original
label data & field notes to records seems essential. (In most cases, this
should be implicit in the specimen management system.)
Robt. Robbins point conerning "qualitative" aspects of the data is good --
I believe it applies as well to features such as "slope" and "aspect" which
are more typical (I believe) of botanical data as well as to "altitude"/
"depth" or other site qualifiers (soil?/substrate?)
Ultimately, it seems to me that two key issues with respect to geo-referenced
specimen data are:
1) How do we express appropriate levels of confidence in the "resolution"
of our data? ( A simple triage system? Devising systems which allow graphic
depiction of data with "qualified resolution"? fuzzy locales? polygonal not
point expression of some data originally expressed as a points?)
2) For policy makers and others, how does the import of "resolution" vary
by taxa? Are different levels of confidence acceptable for different
purposes / "applications"? If yes, can we devise systems which clearly
and unambiguously express these differences...?
California Academy of Sciences
(from IUCN, Gland, Switzerland)
More information about the Taxacom