Catching lat lon in wrong country errors

[Robert Mesibov]

I'm not 100% sure, but I'm guessing that Doug Yanega's question and the
thread to which it's tied have to do with locality data NOT defined by
lat/lon or UTM coordinates supplied by the collector, and coming from the
pre-GPS era. My own experience with "pre-numerical" localities suggests that
human checking, and lots of it, is a very good idea.

A couple of years back I audited a Tasmanian digital gazeteer with about
1700 name + lat/lon entries. Some entries were copied (electronically) from
an "official" government-produced gazeteer. Most had been created by someone
finding a location on a printed map and carefully estimating the lat/lon
from lat/lon marks on the map's margin.

About 45% of the 1700 entries needed to be edited. There were many minor
data-entry problems of one kind or another (spelling errors, multiple
entries for the same place) which probably aren't really relevant to Doug's
query. However, a remarkably high number of the dubious entries, including
many from the "official" source, were long linear locations (rivers, roads)
or very large landscape features (e.g. mountain ranges). I have no idea how
the lat/lons were determined for such features. The error in using gazeteer
lat/lon as opposed to "X on map" lat/lon in these cases was up to 20 km.

For smaller features (individual hills, towns, etc) I used GIS to calculate
the distance discrepancy between the feature's location on a digital
map (with ca. 15m horizontal accuracy) and its gazeteer location. My cut-off
for being interested was 5+ km, and I was interested 60 times out of 1700,
with displacement errors of up to 90 km. I checked each of the 60 (both map
and gazeteer) to see what had happened. Some were obvious data entry errors
(like misreading a map) and some were placename ambiguities. A surprising
number of both "official" and homemade entries were just plain wrong.

I emphasise that these were errors in the digital gazeteer, which was the
spatial reference providing "look ups" for a collection database. At the
next level up, people entering data in the collection database often
generated additional errors by spatially misinterpreting text strings such
as "0.5 km downstream from Rickety Bridge". They could find Rickety Bridge
on the map but not in the gazeteer, so they entered the nearest named place
on the map which was also in the gazeteer. When I audited these database
entries I found some very large displacements. Nearest named places have a
strange attractive power! The text string, of course, was also entered in
the collection database, but spatial searches in this database are done on
placename and/or numerical coordinates. Should I mention that I also found
data entry errors for collector-supplied lat/lon and UTM?

This particular audit was carried out because the gazeteer and database
managers believed, as I think Doug does, that people always need to be part
of the
data-validation process.
---
Dr Robert Mesibov
Honorary Research Associate, Queen Victoria Museum and Art Gallery
and School of Zoology, University of Tasmania
Home contact: PO Box 101, Penguin, Tasmania, Australia 7316
(03) 6437 1195