[Taxacom] Digitizing Science Collections

[David Campbell]

> Unfortunately, original data labels are often extremely misleading,
> especially regarding distributional information. After georeferencing
> over 200,000 insect specimens from multiple institutions, the figures
> work out to over 20% of all original labels have significant
> omissions or ambiguities (the latter category being primarily things
> like abbreviations, uncertain reference points, uncertain date
> formats, or handwriting problems), and some 5% of all original labels
> have genuine errors of one sort or another (mostly typographical
> errors). I doubt such problems are unique to insect labels.

I can confirm that mollusk labels have similar issues with similarly
high frequency, though I have not tallied percentages. In fact,
insects may have better than average quality of average data because
fewer specimens survive from the "I picked these up somewhere when my
ship sailed around the world" era.  Having a misidentified specimen
with good locality data is far more useful than a correctly identified
specimen with bad locality data, as the identification can be fixed
from the specimen.  Someone knowledgeable about the taxon or locality
may be able to recognize that a record is supect, even without
examining the specimen.  But all data in a database needs competent
checking to be reliable.  And many larger collections will include
items from various sources, of varying reliability for their
localities and identifications.

The problem is compounded when online databases grab information from
varied sources without careful verification or good harmonization
protocols.  In particular, homonyms can lead to perfectly good
information from one database being presented wrongly in another one,
which claims that corrections should be referred back to the first
database.

-- 
Dr. David Campbell
Collections Assistant
The Paleontological Research Institution
1259 Trumansburg Road
Ithaca NY 14850