Biodiversity studies - computer solutions to identification
Wed May 21 12:28:38 CDT 1997
Dave Chesmore wrote:
> I see technological solutions to some areas:
> a) Computer-based identification keys (not on laptops but on good
> ergonomically designed hand-held computers that are rugged, easy
> to use in the field and are low cost).
In fact, identifications are done in the lab. Material as it comes out
of a trap has to be sorted, cleaned of debris, washed in fresh alcohol,
and generally prepared in some way to make the taxonomically useable
characters accessible. Not to mention the fact that a whopping
percentage is in the 1 mm to 5 mm range in maximum dimension, requiring
a microscope simply to sort the stuff. (I'm an entomologist, so I'm
talking about insects first off, of course). Plants are put into
newsprint folders and dried in presses. Sometimes, in the wet tropics,
you have to collect into plastic bags with formalin and bring them back
to the lab to dry. The point is that being able to take the id program
into the field is a non-starter because there are compelling reasons to
separate collection and identification.
> b) Automated identification systems that can identify to species or at
> least genus. Such systems could pre-sort large catches from traps to
> save time.
Specimens have to be prepared and examined in order to have the
characters to input. Often in insects the genitalia have to be dissected
out, cleaned, and slide-mounted. Often quite different critters, even
different families, can't be separated by shape factors ( I presume
you're talking about some sort of scanning process). In plants the
flower parts often have to be dissected and microscopic details of
pubescence and such are necessary.
> c) "Intelligent" traps capable of catching particular species or species
> that can't be identified by the system - all other species are logged
> and released, again saving time.
Venn kompt Messiach! (When comes the Messiah.) Also, the frequency and
distribution of taxa is often involved. Most importantly, expense is a
(enormous) factor. Standard pitfall traps consist of a plastic cup, some
sort of cover (eg. a ceramic tile) and a bit of antifreeze. You can put
out a lot of them for the cost of such a trap! Even if it was possible.
Again, it may be necessary to dissect something simply to distinguish
> The technology already exists for most of the solutions. It is obvious
> that such systems must be highly reliable and not mis-identify
> species. I foresee these systems to be time-savers but in no way
> replacements for taxonomists.
The problem is the enormous number of species. In North America north of
Mexico, there are 3187 (described) species and 315 genera in the single
beetle family Staphylinidae. The number of species to be discriminated
in Amazonia is between one and two orders of magnitude larger (note to
rain forest students: all right, so I'm a taxonomic conservative, so
make something of it) a large percentage of which (90%?) are
undescribed. Staphylinidae constitute roughly (VERY roughly) 3% of the
temperate North American fauna. Transpose that to the Amazonian fauna
and you get a feeling for the size of the problem. Very optimistically,
less than 30% of the world arthropod fauna has even been _described_.
And a lot of those undescribed taxa are common.
> I would like to receive comments on:
> 1. Experiences with computers in the field, especially if they have
> been used with identification keys.
I work mostly in the Mohave desert. Not too long ago I was present when
somebody's laptop died of the heat as she typed in the shade, where the
temperature was a quite modest 43 C. And how about 37 C and 100%
humidity (as in rain falling). Actual _field_ conditions aren't friendly
to equipment. Back in the field lab, maybe. Even in the lab, there are
problems. For one thing, I hate to use my keyboard when my hands are
dripping alcohol. Text on the screen is often hard to read, requiring
peering at the computer. Also, the angle of the screen is wrong for
using it while working at the microscope. There are important roles for
the computer, such as databases. And I'm working on a separate message
suggesting some possible computer based alleviations of the id problem.
But the actual physical box is a problem.
Much identification is done by "cheating", jumping ahead in the key to
sections you know by previous experience apply to the specimen at hand.
The problem isn't indentifying one taxon, but checking off something
like fifty a day. Day after day after day. I have seen some beautiful
examples of keys on screen, but most of them could be printed out and be
still more useful.
> 2. Thoughts on the pros and cons of automating at least some
I like it, I like it. But I don't hold out much hope. It would be
equivalent to automated writing of software. Not OLE, you understand,
but tell the computer "I need a program to do X" and sit back while it
generates code. The ids that could be automated already are: that is,
the investigator looks at the specimen and says "that's a Trimerotropis
pallidipennis" or something equivalent.
There are partial solutions using computer technology, though. (There
_are_ no complete solutions for anything in the sciences.)
Dr. William L. Pratt, Curator of Invertebrates
Marjorie Barrick Museum of Natural History
University of Nevada, Las Vegas Box 454009
Las Vegas, NV 89154-4009
(702)895-1403, fax (702)895-3094 e-mail prattw at nevada.edu
More information about the Taxacom